gtk2/gdk/win32/gdkdnd-win32.c

3282 lines
93 KiB
C
Raw Normal View History

/* GDK - The GIMP Drawing Kit
* Copyright (C) 1995-1999 Peter Mattis, Spencer Kimball and Josh MacDonald
* Copyright (C) 2001 Archaeopteryx Software Inc.
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
* Copyright (C) 1998-2002 Tor Lillqvist
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
2012-02-27 13:01:10 +00:00
* License along with this library. If not, see <http://www.gnu.org/licenses/>.
*/
/*
* Modified by the GTK+ Team and others 1997-2000. See the AUTHORS
* file for a list of people on the GTK+ Team. See the ChangeLog
* files for a list of changes. These files are distributed with
* GTK+ at ftp://ftp.gtk.org/pub/gtk/.
*/
#include "config.h"
#include <string.h>
#include <io.h>
#include <fcntl.h>
/*
* Support for OLE-2 drag and drop added at Archaeopteryx Software, 2001
* For more information, contact Stephan R.A. Deibel (sdeibel@archaeopteryx.com)
*
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
* Notes on the implementation:
*
* Source drag context, IDragSource and IDataObject for it are created
* (almost) simultaneously, whereas target drag context and IDropTarget
* are separated in time - IDropTarget is created when a window is made
* to accept drops, while target drag context is created when a dragging
* cursor enters the window and is destroyed when that cursor leaves
* the window.
*
* There's a mismatch between data types supported by W32 (W32 formats)
* and by GTK+ (GDK targets).
* To account for it the data is transmuted back and forth. There are two
* main points of transmutation:
* * GDK convert selection: transmute W32 data to GTK+ data
* * GDK window property change: transmute GTK+ data to W32 data
*
* There are also two points where data formats are considered:
* * When source drag context is created, it gets a list of GTK+ targets
* that it supports, these are matched to the W32 formats they
* correspond to (possibly with transmutation). New W32 formats for
* GTK+-specific formats are also created here (see below).
* * When target drag context is created, it queries the IDataObject
* for the list of W32 formats it supports and matches these to
* corresponding GTK+ formats that it will be able to provide
* (possibly with transmutation) later. Missing GDK targets for
* W32-specific formats are also created here (see below).
*
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
* W32 formats and GTK+ targets are both integers (CLIPFORMAT and GdkAtom
* respectively), but cannot be used interchangeably.
*
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
* To accommodate advanced GTK+ applications the code allows them to
* register drop targets that accept W32 data formats, and to register
* drag sources that provide W32 data formats. To do that they must
* register either with the string name of the format in question
* (for example, "Shell IDList Array") or, for unnamed pre-defined
* formats, register with the stringified constant name of the format
* in question (for example, "CF_UNICODETEXT").
* If such target format is accepted/provided, GDK will not try to
* transmute it to/from something else. Otherwise GDK will do the following
* transmutation:
* * If GTK+ application provides image/png, image/gif or image/jpeg,
* GDK will claim to also provide "PNG", "GIF" or "JFIF" respectively,
* and will pass these along verbatim.
* * If GTK+ application provides any GdkPixbuf-compatible target,
* GDK will also offer "PNG" and CF_DIB W32 formats.
* * If GTK+ application provides UTF8_STRING, GDK will also offer
* CF_UNICODETEXT (UTF-16-encoded) and CF_TEXT (encoded with thread-
* and locale-depenant codepage), and will do the conversion when such
* data is requested.
* * If GTK+ application accepts image/png, image/gif or image/jpeg,
* GDK will claim to also accept "PNG", "GIF" or "JFIF" respectively,
* and will pass these along verbatim.
* * If GTK+ application accepts image/bmp, GDK will
* claim to accept CF_DIB W32 format, and will convert
* it, changing the header, when such data is provided.
* * If GTK+ application accepts UTF8_STRING, GDK will
* claim to accept CF_UNICODETEXT and CF_TEXT, and will do
* the conversion when such data is provided.
* * If GTK+ application accepts text/uri-list, GDK will
* claim to accept "Shell IDList Array", and will do the
* conversion when such data is provided.
*
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
* Currently the conversion from text/uri-list to Shell IDList Array is not
* implemented, so it's not possible to drag & drop files from GTK+
* applications to non-GTK+ applications the same way one can drag files
* from Windows Explorer.
*
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
* To accommodate GTK+ application compaibility the code allows
* GTK+ applications to register drop targets that accept GTK+-specific
* data formats, and to register drag sources that provide GTK+-specific
* data formats. This is done by simply registering target atom names
* as clipboard formats. This way two GTK+ applications can exchange
* data in their native formats (both well-known ones, such as UTF8_STRING,
* and special, known only to specific applications). This will work just
* fine as long as both applications agree on what kind of data is stored
* under such format exactly.
*
* Note that clipboard format space is limited, there can only be 16384
* of them for a particular user session. Therefore it is highly inadvisable
* to create and register such formats out of the whole cloth, dynamically.
* If more flexibility is needed, register one format that has some
* internal indicators of the kind of data it contains, then write the application
* in such a way that it requests the data and inspects its header before deciding
* whether to accept it or not. For details see GTK+ drag & drop documentation
* on the "drag-motion" and "drag-data-received" signals.
*/
/* The mingw.org compiler does not export GUIDS in it's import library. To work
* around that, define INITGUID to have the GUIDS declared. */
#if defined(__MINGW32__) && !defined(__MINGW64_VERSION_MAJOR)
#define INITGUID
#endif
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* For C-style COM wrapper macros */
#define COBJMACROS
#include "gdkdnd.h"
#include "gdkproperty.h"
Large changes to the Win32 backend, partially made necessary by the 2000-05-02 Tor Lillqvist <tml@iki.fi> Large changes to the Win32 backend, partially made necessary by the changes to the backend-independent internal structures. Attempts to implement similar backing store stuff as on X11. The current (CVS) version of the Win32 backend is *not* as stable as it was before the no-flicker branch was merged. A zipfile with that version is available from http://www.gimp.org/win32/. That should be use by "production" code until this CVS version is usable. (But note, the Win32 backend has never been claimed to be "production quality".) * README.win32: Add the above comment about versions. * gdk/gdkwindow.c: Don't use backing store for now on Win32. * gdk/gdk.def: Update. * gdk/gdkfont.h: Declare temporary Win32-only functions. Will presumably be replaced by some more better mechanism as 1.4 gets closer to release shape. * gdk/makefile.{cygwin,msc}: Update. * gdk/win32/*.c: Correct inclusions of the backend-specific and internal headers. Change code according to changes in these. Use gdk_drawable_*, not gdk_window_* where necessary. * gdk/win32/gdkdnd-win32.c: Use MISC selector for GDK_NOTE, not our old DND. * gdk/win32/gdkdrawable-win32.c (gdk_win32_draw_text): Don't try to interpret single characters as UTF-8. Thanks to Hans Breuer. Use correct function name in warning messages. * gdk/win32/gdkevents-win32.c: Use correct parameter lists for the GSourceFuncs gdk_event_prepare and gdk_event_check. (gdk_event_get_graphics_expose): Do implement, use PeekMessage. Thanks to Hans Breuer. (event_mask_string): Debugging function to print an GdkEventMask. (gdk_pointer_grab): Use it. * gdk/win32/gdkfont-win32.c: The Unicode subrange that the (old) book I used claimed was Hangul actually is CJK Unified Ideographs Extension A. Also, Hangul Syllables were missing. Improve logging. * gdk/win32/gdkgc-win32.c: Largish changes. * gdk/win32/gdkim-win32.c (gdk_set_locale): Use g_win32_getlocale() from GLib, and not setlocale() to get current locale name. * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkwin32.h: Move stuff from gdkprivate-win32.h to gdkwin32.h, similarily as in the X11 backend. * gdk/win32/gdkwindow-win32.c (gdk_propagate_shapes): Bugfix, assignment was used instead of equals in if test. Thanks to Hans Breuer. * gdk/win32/makefile.{cygwin,msc} * gtk/makefile.{cygwin,msc}: Updates. Better kludge to get the path to the Win32 headers that works also with the mingw compiler. * gtk/gtkstyle.c: Include <string.h>.
2000-05-01 22:06:49 +00:00
#include "gdkinternals.h"
#include "gdkprivate-win32.h"
#include "gdkwin32.h"
#include "gdkwin32dnd.h"
#include "gdkdisplayprivate.h"
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
#include "gdk/gdkdndprivate.h"
#include "gdkwin32dnd-private.h"
#include "gdkdisplay-win32.h"
#include "gdkselection-win32.h"
#include "gdkdeviceprivate.h"
#include <ole2.h>
#include <shlobj.h>
#include <shlguid.h>
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
#include <objidl.h>
#include <gdk/gdk.h>
#include <glib/gstdio.h>
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* from gdkselection-win32.c */
extern GdkAtom *known_pixbuf_formats;
extern int n_known_pixbuf_formats;
typedef enum {
GDK_DRAG_STATUS_DRAG,
GDK_DRAG_STATUS_MOTION_WAIT,
GDK_DRAG_STATUS_ACTION_WAIT,
GDK_DRAG_STATUS_DROP
} GdkDragStatus;
static GList *contexts;
static GdkDragContext *current_dest_drag = NULL;
static gboolean use_ole2_dnd = FALSE;
G_DEFINE_TYPE (GdkWin32DragContext, gdk_win32_drag_context, GDK_TYPE_DRAG_CONTEXT)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static void
move_drag_window (GdkDragContext *context,
guint x_root,
guint y_root)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
gdk_window_move (context_win32->drag_window,
x_root - context_win32->hot_x,
y_root - context_win32->hot_y);
gdk_window_raise (context_win32->drag_window);
}
static void
gdk_win32_drag_context_init (GdkWin32DragContext *context)
{
if (!use_ole2_dnd)
{
contexts = g_list_prepend (contexts, context);
}
else
{
}
GDK_NOTE (DND, g_print ("gdk_drag_context_init %p\n", context));
}
static void
gdk_win32_drag_context_finalize (GObject *object)
{
GdkDragContext *context;
GdkWin32DragContext *context_win32;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWindow *drag_window;
GDK_NOTE (DND, g_print ("gdk_drag_context_finalize %p\n", object));
g_return_if_fail (GDK_IS_WIN32_DRAG_CONTEXT (object));
context = GDK_DRAG_CONTEXT (object);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
if (!use_ole2_dnd)
{
contexts = g_list_remove (contexts, context);
if (context == current_dest_drag)
current_dest_drag = NULL;
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
drag_window = context_win32->drag_window;
g_array_unref (context_win32->droptarget_format_target_map);
G_OBJECT_CLASS (gdk_win32_drag_context_parent_class)->finalize (object);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (drag_window)
gdk_window_destroy (drag_window);
}
/* Drag Contexts */
static GdkDragContext *
gdk_drag_context_new (GdkDisplay *display)
{
GdkWin32DragContext *context_win32;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Display *win32_display = GDK_WIN32_DISPLAY (display);
GdkDragContext *context;
context_win32 = g_object_new (GDK_TYPE_WIN32_DRAG_CONTEXT,
"display", display,
NULL);
context = GDK_DRAG_CONTEXT(context_win32);
gdk_drag_context_set_device (context, gdk_seat_get_pointer (gdk_display_get_default_seat (display)));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (win32_display->has_fixed_scale)
context_win32->scale = win32_display->window_scale;
else
context_win32->scale = _gdk_win32_display_get_monitor_scale_factor (win32_display, NULL, NULL, NULL);
context_win32->droptarget_format_target_map = g_array_new (FALSE, FALSE, sizeof (GdkSelTargetFormat));
return context;
}
static GdkDragContext *
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
gdk_drag_context_find (gboolean is_source,
GdkWindow *source,
GdkWindow *dest)
{
GList *tmp_list = contexts;
GdkDragContext *context;
while (tmp_list)
{
context = (GdkDragContext *)tmp_list->data;
if ((!context->is_source == !is_source) &&
((source == NULL) || (context->source_window && (context->source_window == source))) &&
((dest == NULL) || (context->dest_window && (context->dest_window == dest))))
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
return context;
tmp_list = tmp_list->next;
}
return NULL;
}
#define PRINT_GUID(guid) \
g_print ("%.08lx-%.04x-%.04x-%.02x%.02x-%.02x%.02x%.02x%.02x%.02x%.02x", \
((gulong *) guid)[0], \
((gushort *) guid)[2], \
((gushort *) guid)[3], \
((guchar *) guid)[8], \
((guchar *) guid)[9], \
((guchar *) guid)[10], \
((guchar *) guid)[11], \
((guchar *) guid)[12], \
((guchar *) guid)[13], \
((guchar *) guid)[14], \
((guchar *) guid)[15]);
typedef struct {
IDropTarget idt;
GdkDragContext *context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gint ref_count;
GdkWindow *dest_window;
} target_drag_context;
typedef struct {
IDropSource ids;
GdkDragContext *context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gint ref_count;
} source_drag_context;
typedef struct {
IDataObject ido;
int ref_count;
GdkDragContext *context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GArray *formats;
} data_object;
typedef struct {
IEnumFORMATETC ief;
int ref_count;
int ix;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
data_object *dataobj;
} enum_formats;
static source_drag_context *pending_src_context = NULL;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static source_drag_context *current_src_context = NULL;
static data_object *current_src_object = NULL;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static enum_formats *enum_formats_new (data_object *dataobj);
/* map windows -> target drag contexts. The table
* owns a ref to both objects.
*/
static GHashTable* target_ctx_for_window = NULL;
static ULONG STDMETHODCALLTYPE
idroptarget_addref (LPDROPTARGET This)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
int ref_count = ++ctx->ref_count;
GDK_NOTE (DND, g_print ("idroptarget_addref %p %d\n", This, ref_count));
return ref_count;
}
static HRESULT STDMETHODCALLTYPE
idroptarget_queryinterface (LPDROPTARGET This,
REFIID riid,
LPVOID *ppvObject)
{
GDK_NOTE (DND, {
g_print ("idroptarget_queryinterface %p ", This);
PRINT_GUID (riid);
});
*ppvObject = NULL;
if (IsEqualGUID (riid, &IID_IUnknown))
{
GDK_NOTE (DND, g_print ("...IUnknown S_OK\n"));
idroptarget_addref (This);
*ppvObject = This;
return S_OK;
}
else if (IsEqualGUID (riid, &IID_IDropTarget))
{
GDK_NOTE (DND, g_print ("...IDropTarget S_OK\n"));
idroptarget_addref (This);
*ppvObject = This;
return S_OK;
}
else
{
GDK_NOTE (DND, g_print ("...E_NOINTERFACE\n"));
return E_NOINTERFACE;
}
}
static ULONG STDMETHODCALLTYPE
idroptarget_release (LPDROPTARGET This)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
int ref_count = --ctx->ref_count;
GDK_NOTE (DND, g_print ("idroptarget_release %p %d\n", This, ref_count));
if (ref_count == 0)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_object_unref (ctx->context);
g_clear_object (&ctx->dest_window);
g_free (This);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
return ref_count;
}
static GdkDragAction
get_suggested_action (DWORD grfKeyState)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
/* This is the yucky Windows standard: Force link action if both
* Control and Alt are down, copy if Control is down alone, move if
* Alt is down alone, or use default of move within the app or copy
* when origin of the drag is in another app.
*/
if (grfKeyState & MK_CONTROL && grfKeyState & MK_SHIFT)
return GDK_ACTION_LINK; /* Link action not supported */
else if (grfKeyState & MK_CONTROL)
return GDK_ACTION_COPY;
else if (grfKeyState & MK_ALT)
return GDK_ACTION_MOVE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
else if (sel_win32->dnd_source_state == GDK_WIN32_DND_DRAGGING)
return GDK_ACTION_MOVE;
else
return GDK_ACTION_COPY;
/* Any way to determine when to add in DROPEFFECT_SCROLL? */
}
/* Process pending events -- we don't want to service non-GUI events
* forever so do one iteration and then do more only if theres a
* pending GDK event.
*/
static void
process_pending_events (GdkDisplay *display)
{
g_main_context_iteration (NULL, FALSE);
while (_gdk_event_queue_find_first (display))
g_main_context_iteration (NULL, FALSE);
}
static DWORD
drop_effect_for_action (GdkDragAction action)
{
switch (action)
{
case GDK_ACTION_MOVE:
return DROPEFFECT_MOVE;
case GDK_ACTION_LINK:
return DROPEFFECT_LINK;
case GDK_ACTION_COPY:
return DROPEFFECT_COPY;
default:
return DROPEFFECT_NONE;
}
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static GdkDragAction
action_for_drop_effect (DWORD effect)
{
switch (effect)
{
case DROPEFFECT_MOVE:
return GDK_ACTION_MOVE;
case DROPEFFECT_LINK:
return GDK_ACTION_LINK;
case DROPEFFECT_COPY:
return GDK_ACTION_COPY;
default:
return 0;
}
}
static void
dnd_event_put (GdkEventType type,
GdkDragContext *context,
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gint pt_x,
gint pt_y,
gboolean to_dest_window)
{
2010-05-25 22:38:44 +00:00
GdkEvent *e;
e = gdk_event_new (type);
if (to_dest_window)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&e->dnd.window, context->dest_window);
else
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&e->dnd.window, context->source_window);
2010-05-25 22:38:44 +00:00
e->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&e->dnd.context, context);
2010-05-25 22:38:44 +00:00
e->dnd.time = GDK_CURRENT_TIME;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
e->dnd.x_root = pt_x;
e->dnd.y_root = pt_y;
2010-05-25 22:38:44 +00:00
gdk_event_set_device (e, gdk_drag_context_get_device (context));
gdk_event_set_seat (e, gdk_device_get_seat (gdk_drag_context_get_device (context)));
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (e));
gdk_event_put (e);
gdk_event_free (e);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static GdkContentFormats *
query_targets (LPDATAOBJECT pDataObj,
GArray *format_target_map)
{
IEnumFORMATETC *pfmt = NULL;
FORMATETC fmt;
GList *result = NULL;
HRESULT hr;
GdkContentFormatsBuilder *builder;
GdkContentFormats *result_formats;
GList *p;
if ((LPDATAOBJECT) current_src_object == pDataObj)
return gdk_content_formats_ref (current_src_object->context->formats);
hr = IDataObject_EnumFormatEtc (pDataObj, DATADIR_GET, &pfmt);
if (SUCCEEDED (hr))
hr = IEnumFORMATETC_Next (pfmt, 1, &fmt, NULL);
while (SUCCEEDED (hr) && hr != S_FALSE)
{
gboolean is_predef;
gchar *registered_name = _gdk_win32_get_clipboard_format_name (fmt.cfFormat, &is_predef);
if (registered_name && is_predef)
GDK_NOTE (DND, g_print ("supported built-in source format 0x%x %s\n", fmt.cfFormat, registered_name));
else if (registered_name)
GDK_NOTE (DND, g_print ("supported source format 0x%x %s\n", fmt.cfFormat, registered_name));
else
GDK_NOTE (DND, g_print ("supported unnamed? source format 0x%x\n", fmt.cfFormat));
g_free (registered_name);
_gdk_win32_add_format_to_targets (fmt.cfFormat, format_target_map, &result);
hr = IEnumFORMATETC_Next (pfmt, 1, &fmt, NULL);
}
if (pfmt)
IEnumFORMATETC_Release (pfmt);
builder = gdk_content_formats_builder_new ();
for (p = g_list_reverse (result); p; p = p->next)
gdk_content_formats_builder_add_mime_type (builder, (const gchar *) p->data);
result_formats = gdk_content_formats_builder_free (builder);
g_list_free (result);
return result_formats;
}
static void
set_data_object (LPDATAOBJECT *location, LPDATAOBJECT data_object)
{
if (*location != NULL)
IDataObject_Release (*location);
*location = data_object;
if (*location != NULL)
IDataObject_AddRef (*location);
}
static HRESULT STDMETHODCALLTYPE
idroptarget_dragenter (LPDROPTARGET This,
LPDATAOBJECT pDataObj,
DWORD grfKeyState,
POINTL pt,
LPDWORD pdwEffect)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
GdkDragContext *context;
GdkWin32DragContext *context_win32;
gint pt_x;
gint pt_y;
GDK_NOTE (DND, g_print ("idroptarget_dragenter %p @ %ld : %ld for dest window 0x%p S_OK\n", This, pt.x, pt.y, ctx->dest_window));
g_clear_object (&ctx->context);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context = gdk_drag_context_new (gdk_window_get_display (ctx->dest_window));
context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
ctx->context = context;
g_set_object (&context->dest_window, ctx->dest_window);
context->protocol = GDK_DRAG_PROTO_OLE2;
context->is_source = FALSE;
context->source_window = NULL;
/* OLE2 DnD does not allow us to get the source window,
* but we *can* find it if it's ours. This is needed to
* support DnD within the same widget, for example.
*/
if (current_src_context && current_src_context->context)
g_set_object (&context->source_window, current_src_context->context->source_window);
else
g_set_object (&context->source_window, NULL); /* FIXME: Root window used to be here. Find a substitute? */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&sel_win32->target_drag_context, context);
context->actions = GDK_ACTION_DEFAULT | GDK_ACTION_COPY | GDK_ACTION_MOVE;
context->suggested_action = GDK_ACTION_MOVE;
context->action = GDK_ACTION_MOVE;
g_array_set_size (context_win32->droptarget_format_target_map, 0);
context->formats = query_targets (pDataObj, context_win32->droptarget_format_target_map);
ctx->context->suggested_action = get_suggested_action (grfKeyState);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
set_data_object (&sel_win32->dnd_data_object_target, pDataObj);
pt_x = pt.x / context_win32->scale + _gdk_offset_x;
pt_y = pt.y / context_win32->scale + _gdk_offset_y;
dnd_event_put (GDK_DRAG_ENTER, ctx->context, pt_x, pt_y, TRUE);
dnd_event_put (GDK_DRAG_MOTION, ctx->context, pt_x, pt_y, TRUE);
context_win32->last_key_state = grfKeyState;
context_win32->last_x = pt_x;
context_win32->last_y = pt_y;
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
*pdwEffect = drop_effect_for_action (ctx->context->action);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("idroptarget_dragenter returns with action %d and drop effect %lu\n", ctx->context->action, *pdwEffect));
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
idroptarget_dragover (LPDROPTARGET This,
DWORD grfKeyState,
POINTL pt,
LPDWORD pdwEffect)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (ctx->context);
gint pt_x = pt.x / context_win32->scale + _gdk_offset_x;
gint pt_y = pt.y / context_win32->scale + _gdk_offset_y;
ctx->context->suggested_action = get_suggested_action (grfKeyState);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("idroptarget_dragover %p @ %ld : %ld, suggests %d action S_OK\n", This, pt.x, pt.y, ctx->context->suggested_action));
if (pt_x != context_win32->last_x ||
pt_y != context_win32->last_y ||
grfKeyState != context_win32->last_key_state)
{
dnd_event_put (GDK_DRAG_MOTION, ctx->context, pt_x, pt_y, TRUE);
context_win32->last_key_state = grfKeyState;
context_win32->last_x = pt_x;
context_win32->last_y = pt_y;
}
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
*pdwEffect = drop_effect_for_action (ctx->context->action);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("idroptarget_dragover returns with action %d and effect %lu\n", ctx->context->action, *pdwEffect));
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
idroptarget_dragleave (LPDROPTARGET This)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
GDK_NOTE (DND, g_print ("idroptarget_dragleave %p S_OK\n", This));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
dnd_event_put (GDK_DRAG_LEAVE, ctx->context, 0, 0, TRUE);
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_clear_object (&sel_win32->target_drag_context);
g_clear_object (&ctx->context);
set_data_object (&sel_win32->dnd_data_object_target, NULL);
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
idroptarget_drop (LPDROPTARGET This,
LPDATAOBJECT pDataObj,
DWORD grfKeyState,
POINTL pt,
LPDWORD pdwEffect)
{
target_drag_context *ctx = (target_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (ctx->context);
gint pt_x = pt.x / context_win32->scale + _gdk_offset_x;
gint pt_y = pt.y / context_win32->scale + _gdk_offset_y;
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
GDK_NOTE (DND, g_print ("idroptarget_drop %p ", This));
if (pDataObj == NULL)
{
GDK_NOTE (DND, g_print ("E_POINTER\n"));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_clear_object (&ctx->context);
return E_POINTER;
}
ctx->context->suggested_action = get_suggested_action (grfKeyState);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
dnd_event_put (GDK_DROP_START, ctx->context, pt_x, pt_y, TRUE);
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
/* Notify OLE of copy or move */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (sel_win32->dnd_target_state != GDK_WIN32_DND_DROPPED)
*pdwEffect = DROPEFFECT_NONE;
else
*pdwEffect = drop_effect_for_action (ctx->context->action);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_clear_object (&sel_win32->target_drag_context);
g_clear_object (&ctx->context);
set_data_object (&sel_win32->dnd_data_object_target, NULL);
GDK_NOTE (DND, g_print ("drop S_OK with effect %lx\n", *pdwEffect));
return S_OK;
}
static ULONG STDMETHODCALLTYPE
idropsource_addref (LPDROPSOURCE This)
{
source_drag_context *ctx = (source_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
int ref_count = ++ctx->ref_count;
GDK_NOTE (DND, g_print ("idropsource_addref %p %d\n", This, ref_count));
return ref_count;
}
static HRESULT STDMETHODCALLTYPE
idropsource_queryinterface (LPDROPSOURCE This,
REFIID riid,
LPVOID *ppvObject)
{
GDK_NOTE (DND, {
g_print ("idropsource_queryinterface %p ", This);
PRINT_GUID (riid);
});
*ppvObject = NULL;
if (IsEqualGUID (riid, &IID_IUnknown))
{
GDK_NOTE (DND, g_print ("...IUnknown S_OK\n"));
idropsource_addref (This);
*ppvObject = This;
return S_OK;
}
else if (IsEqualGUID (riid, &IID_IDropSource))
{
GDK_NOTE (DND, g_print ("...IDropSource S_OK\n"));
idropsource_addref (This);
*ppvObject = This;
return S_OK;
}
else
{
GDK_NOTE (DND, g_print ("...E_NOINTERFACE\n"));
return E_NOINTERFACE;
}
}
static ULONG STDMETHODCALLTYPE
idropsource_release (LPDROPSOURCE This)
{
source_drag_context *ctx = (source_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
int ref_count = --ctx->ref_count;
GDK_NOTE (DND, g_print ("idropsource_release %p %d\n", This, ref_count));
if (ref_count == 0)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
{
g_clear_object (&ctx->context);
if (current_src_context == ctx)
current_src_context = NULL;
g_free (This);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
}
return ref_count;
}
/* Emit GDK events for any changes in mouse events or control key
* state since the last recorded state. Return true if any events
* have been emitted and false otherwise.
*/
static gboolean
send_change_events (GdkDragContext *context,
DWORD key_state,
gboolean esc_pressed)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
POINT pt;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
POINT pt_client;
gboolean changed = FALSE;
HWND hwnd = GDK_WINDOW_HWND (context->source_window);
LPARAM lparam;
WPARAM wparam;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gint pt_x;
gint pt_y;
if (!API_CALL (GetCursorPos, (&pt)))
return FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
pt_client = pt;
if (!API_CALL (ScreenToClient, (hwnd, &pt_client)))
return FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
pt_x = pt.x / context_win32->scale + _gdk_offset_x;
pt_y = pt.y / context_win32->scale + _gdk_offset_y;
if (pt_x != context_win32->last_x || pt_y != context_win32->last_y ||
key_state != context_win32->last_key_state)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
lparam = MAKELPARAM (pt_client.x, pt_client.y);
wparam = key_state;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (pt_x != context_win32->last_x || pt_y != context_win32->last_y)
{
GDK_NOTE (DND, g_print ("Sending WM_MOUSEMOVE (%ld,%ld)\n", pt.x, pt.y));
SendMessage (hwnd, WM_MOUSEMOVE, wparam, lparam);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if ((key_state & MK_LBUTTON) != (context_win32->last_key_state & MK_LBUTTON))
{
if (key_state & MK_LBUTTON)
SendMessage (hwnd, WM_LBUTTONDOWN, wparam, lparam);
else
SendMessage (hwnd, WM_LBUTTONUP, wparam, lparam);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if ((key_state & MK_MBUTTON) != (context_win32->last_key_state & MK_MBUTTON))
{
if (key_state & MK_MBUTTON)
SendMessage (hwnd, WM_MBUTTONDOWN, wparam, lparam);
else
SendMessage (hwnd, WM_MBUTTONUP, wparam, lparam);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if ((key_state & MK_RBUTTON) != (context_win32->last_key_state & MK_RBUTTON))
{
if (key_state & MK_RBUTTON)
SendMessage (hwnd, WM_RBUTTONDOWN, wparam, lparam);
else
SendMessage (hwnd, WM_RBUTTONUP, wparam, lparam);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if ((key_state & MK_CONTROL) != (context_win32->last_key_state & MK_CONTROL))
{
if (key_state & MK_CONTROL)
SendMessage (hwnd, WM_KEYDOWN, VK_CONTROL, 0);
else
SendMessage (hwnd, WM_KEYUP, VK_CONTROL, 0);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if ((key_state & MK_SHIFT) != (context_win32->last_key_state & MK_SHIFT))
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (key_state & MK_SHIFT)
SendMessage (hwnd, WM_KEYDOWN, VK_SHIFT, 0);
else
SendMessage (hwnd, WM_KEYUP, VK_SHIFT, 0);
}
changed = TRUE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32->last_key_state = key_state;
context_win32->last_x = pt_x;
context_win32->last_y = pt_y;
}
if (esc_pressed)
{
GDK_NOTE (DND, g_print ("Sending a escape key down message to %p\n", hwnd));
SendMessage (hwnd, WM_KEYDOWN, VK_ESCAPE, 0);
changed = TRUE;
}
return changed;
}
static HRESULT STDMETHODCALLTYPE
idropsource_querycontinuedrag (LPDROPSOURCE This,
BOOL fEscapePressed,
DWORD grfKeyState)
{
source_drag_context *ctx = (source_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("idropsource_querycontinuedrag %p esc=%d keystate=0x%lx with state %d", This, fEscapePressed, grfKeyState, sel_win32->dnd_source_state));
if (send_change_events (ctx->context, grfKeyState, fEscapePressed))
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (sel_win32->dnd_source_state == GDK_WIN32_DND_DROPPED)
{
GDK_NOTE (DND, g_print ("DRAGDROP_S_DROP\n"));
return DRAGDROP_S_DROP;
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
else if (sel_win32->dnd_source_state == GDK_WIN32_DND_NONE)
{
GDK_NOTE (DND, g_print ("DRAGDROP_S_CANCEL\n"));
return DRAGDROP_S_CANCEL;
}
else
{
GDK_NOTE (DND, g_print ("S_OK\n"));
return S_OK;
}
}
static HRESULT STDMETHODCALLTYPE
idropsource_givefeedback (LPDROPSOURCE This,
DWORD dwEffect)
{
source_drag_context *ctx = (source_drag_context *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (ctx->context);
GdkDragAction suggested_action;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkEvent *e;
POINT pt;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("idropsource_givefeedback %p with drop effect %lu S_OK\n", This, dwEffect));
if (!API_CALL (GetCursorPos, (&pt)))
return S_OK;
suggested_action = action_for_drop_effect (dwEffect);
ctx->context->action = suggested_action;
if (dwEffect == DROPEFFECT_NONE)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_clear_object (&ctx->context->dest_window);
else if (ctx->context->dest_window == NULL)
ctx->context->dest_window = NULL; /* FIXME: Root window was here originally. Find a substitute? */
context_win32->last_x = pt.x / context_win32->scale + _gdk_offset_x;
context_win32->last_y = pt.y / context_win32->scale + _gdk_offset_y;
e = gdk_event_new (GDK_DRAG_STATUS);
g_set_object (&e->dnd.window, ctx->context->source_window);
e->dnd.send_event = FALSE;
g_set_object (&e->dnd.context, ctx->context);
e->dnd.time = GDK_CURRENT_TIME;
e->dnd.x_root = context_win32->last_x;
e->dnd.y_root = context_win32->last_y;
gdk_event_set_device (e, gdk_drag_context_get_device (ctx->context));
gdk_event_set_seat (e, gdk_device_get_seat (gdk_drag_context_get_device (ctx->context)));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (e));
gdk_event_put (e);
gdk_event_free (e);
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
GDK_NOTE (DND, g_print ("idropsource_givefeedback %p returns\n", This));
return S_OK;
}
static ULONG STDMETHODCALLTYPE
idataobject_addref (LPDATAOBJECT This)
{
data_object *dobj = (data_object *) This;
int ref_count = ++dobj->ref_count;
GDK_NOTE (DND, g_print ("idataobject_addref %p %d\n", This, ref_count));
return ref_count;
}
static HRESULT STDMETHODCALLTYPE
idataobject_queryinterface (LPDATAOBJECT This,
REFIID riid,
LPVOID *ppvObject)
{
GDK_NOTE (DND, {
g_print ("idataobject_queryinterface %p ", This);
PRINT_GUID (riid);
});
*ppvObject = NULL;
if (IsEqualGUID (riid, &IID_IUnknown))
{
GDK_NOTE (DND, g_print ("...IUnknown S_OK\n"));
idataobject_addref (This);
*ppvObject = This;
return S_OK;
}
else if (IsEqualGUID (riid, &IID_IDataObject))
{
GDK_NOTE (DND, g_print ("...IDataObject S_OK\n"));
idataobject_addref (This);
*ppvObject = This;
return S_OK;
}
else
{
GDK_NOTE (DND, g_print ("...E_NOINTERFACE\n"));
return E_NOINTERFACE;
}
}
static ULONG STDMETHODCALLTYPE
idataobject_release (LPDATAOBJECT This)
{
data_object *dobj = (data_object *) This;
int ref_count = --dobj->ref_count;
GDK_NOTE (DND, g_print ("idataobject_release %p %d\n", This, ref_count));
if (ref_count == 0)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
{
g_array_free (dobj->formats, TRUE);
g_free (This);
}
return ref_count;
}
static HRESULT
query (LPDATAOBJECT This,
LPFORMATETC pFormatEtc)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
data_object *ctx = (data_object *) This;
gint i;
if (!pFormatEtc)
return DV_E_FORMATETC;
if (pFormatEtc->lindex != -1)
return DV_E_LINDEX;
if ((pFormatEtc->tymed & TYMED_HGLOBAL) == 0)
return DV_E_TYMED;
if ((pFormatEtc->dwAspect & DVASPECT_CONTENT) == 0)
return DV_E_DVASPECT;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
for (i = 0; i < ctx->formats->len; i++)
if (pFormatEtc->cfFormat == g_array_index (ctx->formats, GdkSelTargetFormat, i).format)
return S_OK;
return DV_E_FORMATETC;
}
static HRESULT STDMETHODCALLTYPE
idataobject_getdata (LPDATAOBJECT This,
LPFORMATETC pFormatEtc,
LPSTGMEDIUM pMedium)
{
data_object *ctx = (data_object *) This;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *win32_sel = _gdk_win32_selection_get ();
HRESULT hr;
GdkEvent e;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gint i;
GdkAtom target;
gint64 loopend;
GDK_NOTE (DND, g_print ("idataobject_getdata %p %s ",
This, _gdk_win32_cf_to_string (pFormatEtc->cfFormat)));
/* Check whether we can provide requested format */
hr = query (This, pFormatEtc);
if (hr != S_OK)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
{
GDK_NOTE (DND, g_print ("Unsupported format, returning 0x%lx\n", hr));
return hr;
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* Append a GDK_SELECTION_REQUEST event and then hope the app sets the
* property associated with the _gdk_ole2_dnd atom
*/
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
win32_sel->property_change_format = pFormatEtc->cfFormat;
win32_sel->property_change_data = pMedium;
for (i = 0, target = NULL; i < ctx->formats->len; i++)
{
GdkSelTargetFormat *frec = &g_array_index (ctx->formats, GdkSelTargetFormat, i);
if (frec->format == pFormatEtc->cfFormat)
{
target = frec->target;
win32_sel->property_change_transmute = frec->transmute;
win32_sel->property_change_target_atom = frec->target;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
}
}
if (target == NULL)
{
GDK_NOTE (EVENTS, g_print ("(target not found)"));
return E_UNEXPECTED;
}
GDK_NOTE (DND, {
gchar *target_name = gdk_atom_name (target);
g_print ("idataobject_getdata will request target 0x%p (%s) ",
target, target_name);
g_free (target_name);
});
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
memset (&e, 0, sizeof (GdkEvent));
e.type = GDK_SELECTION_REQUEST;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&e.selection.window, ctx->context->source_window);
e.selection.send_event = FALSE; /* ??? */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* Both selection and property are OLE2_DND, because change_property()
* will only get the property and not the selection. Theoretically we
* could use two different atoms (SELECTION_OLE2_DND and PROPERTY_OLE2_DND),
* but there's little reason to do so.
*/
e.selection.selection = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_OLE2_DND);
e.selection.target = target;
/* Requestor here is fake, just to allow the event to be processed */
g_set_object (&e.selection.requestor, ctx->context->source_window);
e.selection.property = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_OLE2_DND);
e.selection.time = GDK_CURRENT_TIME;
GDK_NOTE (EVENTS, _gdk_win32_print_event (&e));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_put (&e);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* Don't hold up longer than one second */
loopend = g_get_monotonic_time () + 1000000000;
while (win32_sel->property_change_data != 0 &&
g_get_monotonic_time () < loopend)
process_pending_events (gdk_device_get_display (gdk_drag_context_get_device (ctx->context)));
if (pMedium->hGlobal == NULL) {
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print (" E_UNEXPECTED\n"));
return E_UNEXPECTED;
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print (" S_OK\n"));
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
idataobject_getdatahere (LPDATAOBJECT This,
LPFORMATETC pFormatEtc,
LPSTGMEDIUM pMedium)
{
GDK_NOTE (DND, g_print ("idataobject_getdatahere %p %s E_UNEXPECTED\n",
This, _gdk_win32_cf_to_string (pFormatEtc->cfFormat)));
return E_UNEXPECTED;
}
static HRESULT STDMETHODCALLTYPE
idataobject_querygetdata (LPDATAOBJECT This,
LPFORMATETC pFormatEtc)
{
HRESULT hr;
hr = query (This, pFormatEtc);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
#define CASE(x) case x: g_print (#x "\n"); break
GDK_NOTE (DND, {
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_print ("idataobject_querygetdata %p %s ",
This, _gdk_win32_cf_to_string (pFormatEtc->cfFormat));
switch (hr)
{
CASE (DV_E_FORMATETC);
CASE (DV_E_LINDEX);
CASE (DV_E_TYMED);
CASE (DV_E_DVASPECT);
CASE (S_OK);
default: g_print ("%#lx", hr);
}
});
return hr;
}
static HRESULT STDMETHODCALLTYPE
idataobject_getcanonicalformatetc (LPDATAOBJECT This,
LPFORMATETC pFormatEtcIn,
LPFORMATETC pFormatEtcOut)
{
GDK_NOTE (DND, g_print ("idataobject_getcanonicalformatetc %p E_UNEXPECTED\n", This));
return E_UNEXPECTED;
}
static HRESULT STDMETHODCALLTYPE
idataobject_setdata (LPDATAOBJECT This,
LPFORMATETC pFormatEtc,
LPSTGMEDIUM pMedium,
BOOL fRelease)
{
GDK_NOTE (DND, g_print ("idataobject_setdata %p %s E_UNEXPECTED\n",
This, _gdk_win32_cf_to_string (pFormatEtc->cfFormat)));
return E_UNEXPECTED;
}
static HRESULT STDMETHODCALLTYPE
idataobject_enumformatetc (LPDATAOBJECT This,
DWORD dwDirection,
LPENUMFORMATETC *ppEnumFormatEtc)
{
GDK_NOTE (DND, g_print ("idataobject_enumformatetc %p ", This));
if (dwDirection != DATADIR_GET)
{
GDK_NOTE (DND, g_print ("E_NOTIMPL\n"));
return E_NOTIMPL;
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
*ppEnumFormatEtc = &enum_formats_new ((data_object *) This)->ief;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print (" %p S_OK\n", *ppEnumFormatEtc));
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
idataobject_dadvise (LPDATAOBJECT This,
LPFORMATETC pFormatetc,
DWORD advf,
LPADVISESINK pAdvSink,
DWORD *pdwConnection)
{
GDK_NOTE (DND, g_print ("idataobject_dadvise %p E_NOTIMPL\n", This));
return E_NOTIMPL;
}
static HRESULT STDMETHODCALLTYPE
idataobject_dunadvise (LPDATAOBJECT This,
DWORD dwConnection)
{
GDK_NOTE (DND, g_print ("idataobject_dunadvise %p E_NOTIMPL\n", This));
return E_NOTIMPL;
}
static HRESULT STDMETHODCALLTYPE
idataobject_enumdadvise (LPDATAOBJECT This,
LPENUMSTATDATA *ppenumAdvise)
{
GDK_NOTE (DND, g_print ("idataobject_enumdadvise %p OLE_E_ADVISENOTSUPPORTED\n", This));
return OLE_E_ADVISENOTSUPPORTED;
}
static ULONG STDMETHODCALLTYPE
ienumformatetc_addref (LPENUMFORMATETC This)
{
enum_formats *en = (enum_formats *) This;
int ref_count = ++en->ref_count;
GDK_NOTE (DND, g_print ("ienumformatetc_addref %p %d\n", This, ref_count));
return ref_count;
}
static HRESULT STDMETHODCALLTYPE
ienumformatetc_queryinterface (LPENUMFORMATETC This,
REFIID riid,
LPVOID *ppvObject)
{
GDK_NOTE (DND, {
g_print ("ienumformatetc_queryinterface %p", This);
PRINT_GUID (riid);
});
*ppvObject = NULL;
if (IsEqualGUID (riid, &IID_IUnknown))
{
GDK_NOTE (DND, g_print ("...IUnknown S_OK\n"));
ienumformatetc_addref (This);
*ppvObject = This;
return S_OK;
}
else if (IsEqualGUID (riid, &IID_IEnumFORMATETC))
{
GDK_NOTE (DND, g_print ("...IEnumFORMATETC S_OK\n"));
ienumformatetc_addref (This);
*ppvObject = This;
return S_OK;
}
else
{
GDK_NOTE (DND, g_print ("...E_NOINTERFACE\n"));
return E_NOINTERFACE;
}
}
static ULONG STDMETHODCALLTYPE
ienumformatetc_release (LPENUMFORMATETC This)
{
enum_formats *en = (enum_formats *) This;
int ref_count = --en->ref_count;
GDK_NOTE (DND, g_print ("ienumformatetc_release %p %d\n", This, ref_count));
if (ref_count == 0)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
{
idataobject_release ((LPDATAOBJECT) en->dataobj);
g_free (This);
}
return ref_count;
}
static HRESULT STDMETHODCALLTYPE
ienumformatetc_next (LPENUMFORMATETC This,
ULONG celt,
LPFORMATETC elts,
ULONG *nelt)
{
enum_formats *en = (enum_formats *) This;
ULONG i, n;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
ULONG formats_to_get = celt;
GDK_NOTE (DND, g_print ("ienumformatetc_next %p %d %ld ", This, en->ix, celt));
n = 0;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
for (i = 0; i < formats_to_get; i++)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
UINT fmt;
if (en->ix >= en->dataobj->formats->len)
break;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
fmt = g_array_index (en->dataobj->formats, GdkSelTargetFormat, en->ix++).format;
/* skip internals */
if (fmt == 0 || fmt > 0xFFFF)
{
formats_to_get += 1;
continue;
}
elts[n].cfFormat = fmt;
elts[n].ptd = NULL;
elts[n].dwAspect = DVASPECT_CONTENT;
elts[n].lindex = -1;
elts[n].tymed = TYMED_HGLOBAL;
n++;
}
if (nelt != NULL)
*nelt = n;
GDK_NOTE (DND, g_print ("%s\n", (n == celt) ? "S_OK" : "S_FALSE"));
if (n == celt)
return S_OK;
else
return S_FALSE;
}
static HRESULT STDMETHODCALLTYPE
ienumformatetc_skip (LPENUMFORMATETC This,
ULONG celt)
{
enum_formats *en = (enum_formats *) This;
GDK_NOTE (DND, g_print ("ienumformatetc_skip %p %d %ld S_OK\n", This, en->ix, celt));
en->ix += celt;
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
ienumformatetc_reset (LPENUMFORMATETC This)
{
enum_formats *en = (enum_formats *) This;
GDK_NOTE (DND, g_print ("ienumformatetc_reset %p S_OK\n", This));
en->ix = 0;
return S_OK;
}
static HRESULT STDMETHODCALLTYPE
ienumformatetc_clone (LPENUMFORMATETC This,
LPENUMFORMATETC *ppEnumFormatEtc)
{
enum_formats *en = (enum_formats *) This;
enum_formats *new;
GDK_NOTE (DND, g_print ("ienumformatetc_clone %p S_OK\n", This));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
new = enum_formats_new (en->dataobj);
new->ix = en->ix;
*ppEnumFormatEtc = &new->ief;
return S_OK;
}
static IDropTargetVtbl idt_vtbl = {
idroptarget_queryinterface,
idroptarget_addref,
idroptarget_release,
idroptarget_dragenter,
idroptarget_dragover,
idroptarget_dragleave,
idroptarget_drop
};
static IDropSourceVtbl ids_vtbl = {
idropsource_queryinterface,
idropsource_addref,
idropsource_release,
idropsource_querycontinuedrag,
idropsource_givefeedback
};
static IDataObjectVtbl ido_vtbl = {
idataobject_queryinterface,
idataobject_addref,
idataobject_release,
idataobject_getdata,
idataobject_getdatahere,
idataobject_querygetdata,
idataobject_getcanonicalformatetc,
idataobject_setdata,
idataobject_enumformatetc,
idataobject_dadvise,
idataobject_dunadvise,
idataobject_enumdadvise
};
static IEnumFORMATETCVtbl ief_vtbl = {
ienumformatetc_queryinterface,
ienumformatetc_addref,
ienumformatetc_release,
ienumformatetc_next,
ienumformatetc_skip,
ienumformatetc_reset,
ienumformatetc_clone
};
static target_drag_context *
target_context_new (GdkWindow *window)
{
target_drag_context *result;
result = g_new0 (target_drag_context, 1);
result->idt.lpVtbl = &idt_vtbl;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->ref_count = 0;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->dest_window = g_object_ref (window);
idroptarget_addref (&result->idt);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("target_context_new: %p (window %p)\n", result, result->dest_window));
return result;
}
static source_drag_context *
source_context_new (GdkWindow *window,
GdkContentFormats *formats)
{
GdkDragContext *context;
GdkWin32DragContext *context_win32;
source_drag_context *result;
context = gdk_drag_context_new (gdk_window_get_display (window));
context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
result = g_new0 (source_drag_context, 1);
result->context = context;
result->ids.lpVtbl = &ids_vtbl;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->ref_count = 0;
result->context->protocol = GDK_DRAG_PROTO_OLE2;
result->context->is_source = TRUE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->context->source_window = g_object_ref (window);
result->context->dest_window = NULL;
result->context->formats = gdk_content_formats_ref (formats);
idropsource_addref (&result->ids);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("source_context_new: %p (drag context %p)\n", result, result->context));
if (current_src_context == NULL)
current_src_context = result;
return result;
}
static data_object *
data_object_new (GdkDragContext *context)
{
data_object *result;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GList *p;
const char * const *mime_types;
gsize n_mime_types, i;
result = g_new0 (data_object, 1);
result->ido.lpVtbl = &ido_vtbl;
result->ref_count = 1;
result->context = context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->formats = g_array_new (FALSE, FALSE, sizeof (GdkSelTargetFormat));
mime_types = gdk_content_formats_get_mime_types (context->formats, &n_mime_types);
for (i = 0; i < n_mime_types; i++)
{
gint added_count = 0;
gint j;
GDK_NOTE (DND, g_print ("DataObject supports target 0x%p\n", mime_types[i]));
added_count = _gdk_win32_add_target_to_selformats (mime_types[i], result->formats);
for (j = 0; j < added_count && result->formats->len - 1 - j >= 0; j++)
GDK_NOTE (DND, g_print ("DataObject will support format 0x%x\n", g_array_index (result->formats, GdkSelTargetFormat, j).format));
}
GDK_NOTE (DND, g_print ("data_object_new: %p\n", result));
return result;
}
static enum_formats *
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
enum_formats_new (data_object *dataobj)
{
enum_formats *result;
result = g_new0 (enum_formats, 1);
result->ief.lpVtbl = &ief_vtbl;
result->ref_count = 1;
result->ix = 0;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
result->dataobj = dataobj;
idataobject_addref ((LPDATAOBJECT) dataobj);
return result;
}
/* From MS Knowledge Base article Q130698 */
static gboolean
resolve_link (HWND hWnd,
wchar_t *link,
gchar **lpszPath)
{
WIN32_FILE_ATTRIBUTE_DATA wfad;
HRESULT hr;
IShellLinkW *pslW = NULL;
IPersistFile *ppf = NULL;
/* Check if the file is empty first because IShellLink::Resolve for
* some reason succeeds with an empty file and returns an empty
* "link target". (#524151)
*/
if (!GetFileAttributesExW (link, GetFileExInfoStandard, &wfad) ||
(wfad.nFileSizeHigh == 0 && wfad.nFileSizeLow == 0))
return FALSE;
/* Assume failure to start with: */
*lpszPath = 0;
/* Call CoCreateInstance to obtain the IShellLink interface
* pointer. This call fails if CoInitialize is not called, so it is
* assumed that CoInitialize has been called.
*/
hr = CoCreateInstance (&CLSID_ShellLink,
NULL,
CLSCTX_INPROC_SERVER,
&IID_IShellLinkW,
(LPVOID *)&pslW);
if (SUCCEEDED (hr))
{
/* The IShellLink interface supports the IPersistFile
* interface. Get an interface pointer to it.
*/
hr = pslW->lpVtbl->QueryInterface (pslW,
&IID_IPersistFile,
(LPVOID *) &ppf);
}
if (SUCCEEDED (hr))
{
/* Load the file. */
hr = ppf->lpVtbl->Load (ppf, link, STGM_READ);
}
if (SUCCEEDED (hr))
{
/* Resolve the link by calling the Resolve()
* interface function.
*/
hr = pslW->lpVtbl->Resolve (pslW, hWnd, SLR_ANY_MATCH | SLR_NO_UI);
}
if (SUCCEEDED (hr))
{
wchar_t wtarget[MAX_PATH];
hr = pslW->lpVtbl->GetPath (pslW, wtarget, MAX_PATH, NULL, 0);
if (SUCCEEDED (hr))
*lpszPath = g_utf16_to_utf8 (wtarget, -1, NULL, NULL, NULL);
}
if (ppf)
ppf->lpVtbl->Release (ppf);
if (pslW)
pslW->lpVtbl->Release (pslW);
return SUCCEEDED (hr);
}
#if 0
/* Check for filenames like C:\Users\tml\AppData\Local\Temp\d5qtkvvs.bmp */
static gboolean
filename_looks_tempish (const char *filename)
{
char *dirname;
char *p;
const char *q;
gboolean retval = FALSE;
dirname = g_path_get_dirname (filename);
p = dirname;
q = g_get_tmp_dir ();
while (*p && *q &&
((G_IS_DIR_SEPARATOR (*p) && G_IS_DIR_SEPARATOR (*q)) ||
g_ascii_tolower (*p) == g_ascii_tolower (*q)))
p++, q++;
if (!*p && !*q)
retval = TRUE;
g_free (dirname);
return retval;
}
static gboolean
close_it (gpointer data)
{
close (GPOINTER_TO_INT (data));
return FALSE;
}
#endif
static GdkFilterReturn
gdk_dropfiles_filter (GdkXEvent *xev,
GdkEvent *event,
gpointer data)
{
GdkDragContext *context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *context_win32;
GString *result;
MSG *msg = (MSG *) xev;
HANDLE hdrop;
POINT pt;
gint nfiles, i;
gchar *fileName, *linkedFile;
GPtrArray *formats;
formats = g_ptr_array_new ();
if (msg->message == WM_DROPFILES)
{
GDK_NOTE (DND, g_print ("WM_DROPFILES: %p\n", msg->hwnd));
context = gdk_drag_context_new (gdk_window_get_display (event->any.window));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
context->protocol = GDK_DRAG_PROTO_WIN32_DROPFILES;
context->is_source = FALSE;
context->source_window = NULL;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&context->dest_window, event->any.window);
/* WM_DROPFILES drops are always file names */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context->formats = gdk_content_formats_new ((const char *[2]) {
"text/uri-list",
NULL
}, 1);
context->actions = GDK_ACTION_COPY;
context->suggested_action = GDK_ACTION_COPY;
current_dest_drag = context;
event->dnd.type = GDK_DROP_START;
event->dnd.context = current_dest_drag;
2010-05-25 22:38:44 +00:00
gdk_event_set_device (event, gdk_drag_context_get_device (current_dest_drag));
gdk_event_set_seat (event, gdk_device_get_seat (gdk_drag_context_get_device (current_dest_drag)));
hdrop = (HANDLE) msg->wParam;
DragQueryPoint (hdrop, &pt);
ClientToScreen (msg->hwnd, &pt);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
event->dnd.x_root = pt.x / context_win32->scale + _gdk_offset_x;
event->dnd.y_root = pt.y / context_win32->scale + _gdk_offset_y;
gdk/win32/gdkprivate-win32.h Rename all global variables and functions to 2002-11-12 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkprivate-win32.h * gdk/win32/*.c: Rename all global variables and functions to start with underscore. Merge from stable: More work on the Win32 backend. The cause of some scrolling problems was that SetWindowPos() and ScrollWindowEx() don't blit those parts of the window they think are invalid. As we didn't keep Windows's update region in synch with GDK's, Windows thought those areas that in fact had been updated were invalid. Calling ValidateRgn() in _gdk_windowing_window_queue_antiexpose() seems to be an elegant and efficient solution, removing from Windows's update region those areas we are about to repaint proactively. In some cases garbage leftover values were used for the clip origin in GdkGCWin32. This showed up as odd blank areas around the pixmaps included in the Text Widget in gtk-demo. Having the clip region either as a GdkRegion or a HRGN in GdkGCWin32 was unnecessary, it's better to just use a HRGN. The translation and antiexpose queue handling in gdkgeometry-win32.c seems unnecessary (and not implementable in the same way as on X11 anyway, no serial numbers) on Windows, ifdeffed out. Don't (try to) do guffaw scrolling as there is no static window gravity on Windows. Guffaw scrolling would be unnecessary anyway, as there is the ScrollWindow() API. This improves the behaviour of the Text Widget demo in gtk-demo a lot. But I have no idea how the lack of static win gravity should be handled in other places where the X11 code uses it. Especially _gdk_window_move_resize_child(). There is still some problem in expose handling. By moving an obscuring window back and forth over testgtk's main window, for instance, every now and then you typically get narrow vertical or horizontal strips of pixels that haven't been properly redrawn after being exposed. A fencepost error somewhere? Otherwise, all of testgtk and gtk-demo except "big windows" now seem to work pretty well. Bug #79720 should be fixed now. * gdk/win32/gdkcolor-win32.c (gdk_win32_color_to_string, gdk_win32_print_paletteentries, gdk_win32_print_system_palette, gdk_win32_print_hpalette) * gdk/win32/gdkdrawable-win32.c (gdk_win32_drawable_description) * gdk/win32/gdkevents-win32.c (gdk_win32_message_name): Move all debugging helper functions to gdkmain-win32.c. * gdk/win32/gdkdrawable-win32.c (_gdk_win32_draw_tiles): Rewrite. Make static. Must take tile origin parameters, too. (gdk_win32_draw_rectangle): Pass the tile/stipple origin to _gdk_win32_draw_tiles(). Remove #if 0 code. (blit_inside_window): Don't call ScrollDC(), that didn't work at all like I thought. A simple call to BitBlt() is enough. * gdk/win32/gdkevents-win32.c (gdk_event_translate) Remove unused latin_locale_loaded variable. (_gdk_win32_get_next_tick): New function. Used to make sure timestamps of events are always increasing, both in events generated from the window procedure and in events gotten via PeekMessage(). Not sure whether this is actually useful, but it seemed as a good idea. (real_window_procedure): Don't use a local GdkEventPrivate variable. Don't attempt any compression of configure or expose events here, handled elsewhere. (erase_background): Accumulate window offsets when traversing up the parent chain for GDK_PARENT_RELATIVE_BG, in order to get correct alignment of background pixmaps. Don't fill with BLACK_BRUSH if GDK_NO_BG. (gdk_event_get_graphics_expose): A bit more verbose debugging output. (gdk_event_translate): Use _gdk_win32_get_next_tick(). In the WM_PAINT handler, don't check for empty update rect. When we get a WM_PAINT, the update region isn't empty. And if it for some strange reason is, that will be handled later anyway. Call GetUpdateRgn() before calling BeginPaint() and EndPaint() (which empty the update region). * gdk/win32/gdkdnd-win32.c * gdk/win32/gdkinput-win32.c: Use _gdk_win32_get_next_tick(). * gdk/win32/gdkfont-win32.c: Use %p to print HFONTs. (gdk_text_size): Remove, unused. * gdk/win32/gdkgc-win32.c: Set clip origins to zero when appropriate. (gdk_gc_copy): Increase refcount on colormap if present. (gdk_win32_hdc_get): Handle just hcliprgn. If we have a stipple, combine it with clip region after selecting into the DC. (_gdk_win32_bitmap_to_hrgn): Rename from _gdk_win32_bitmap_to_region. (_gdk_win3_gdkregion_to_hrgn): New function, code snippet extracted from gdk_win32_hdc_get(). * gdk/win32/gdkgeometry-win32.c: Ifdef out the translate_queue handling. (gdk_window_copy_area_scroll): Increase clipRect to avoid ScrollWindowEx() not scrolling pixels it thinks are invalid. Scroll also children with the ScrollWindowEx() call. No need to call gdk_window_move() on the children. (gdk_window_scroll): Don't do guffaw scrolling. (gdk_window_compute_position): Fix typo, used win32_y where x was intended. (gdk_window_premove, gdk_window_postmove, gdk_window_clip_changed): Add debugging output. (_gdk_windowing_window_queue_antiexpose): Just call ValidateRgn() on the region. (_gdk_window_process_expose): No use for the serial number parameter now. Instead of a rectangle, take a region parameter, as Windows gives us one in WM_PAINT. * gdk/win32/gdkmain-win32.c (_gdk_win32_lbstyle_to_string, _gdk_win32_pstype_to_string, _gdk_win32_psstyle_to_string, _gdk_win32_psendcap_to_string, _gdk_win32_psjoin_to_string, _gdk_win32_rect_to_string, _gdk_win32_gdkrectangle_to_string, _gdk_win32_gdkregion_to_string): New debugging functions. (static_printf): Helper function for the above. sprintfs into a static circular buffer, return value should be used "soon". * gdk/win32/gdkwindow-win32.c (gdk_propagate_shapes): Plug memory leak, free list after use. (gdk_window_gravity_works): Remove, we know that there is no such thing on Windows. (gdk_window_set_static_bit_gravity, gdk_window_set_static_win_gravity): Ditto, remove, they didn't do anything anyway. (_gdk_windowing_window_init, gdk_window_foreign_new): Call _gdk_window_init_position() like in the X11 backend. (gdk_window_reparent): Don't call the now nonexistent gdk_window_set_static_win_gravity(). No idea what should be done instead. (gdk_window_get_geometry): The returned x and y should be relative to parent. Used to be always zero.. (gdk_window_set_static_gravities): Return FALSE if trying to set static gravity. * gdk/win32/gdkprivate-win32.h: Drop the clip_region field from GdkGCWin32. Only use the HRGN hcliprgn. Declare new functions. * gdk/win32/*.c: Use new debugging functions. * gdk/win32/rc/gdk.rc.in: Update copyright year.
2002-11-12 22:17:48 +00:00
event->dnd.time = _gdk_win32_get_next_tick (msg->time);
nfiles = DragQueryFile (hdrop, 0xFFFFFFFF, NULL, 0);
result = g_string_new (NULL);
for (i = 0; i < nfiles; i++)
{
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
gchar *uri;
wchar_t wfn[MAX_PATH];
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
DragQueryFileW (hdrop, i, wfn, MAX_PATH);
fileName = g_utf16_to_utf8 (wfn, -1, NULL, NULL, NULL);
/* Resolve shortcuts */
if (resolve_link (msg->hwnd, wfn, &linkedFile))
{
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
uri = g_filename_to_uri (linkedFile, NULL, NULL);
if (uri != NULL)
{
g_string_append (result, uri);
GDK_NOTE (DND, g_print ("... %s link to %s: %s\n",
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
fileName, linkedFile, uri));
g_free (uri);
}
g_free (fileName);
fileName = linkedFile;
}
else
{
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
uri = g_filename_to_uri (fileName, NULL, NULL);
if (uri != NULL)
{
g_string_append (result, uri);
GDK_NOTE (DND, g_print ("... %s: %s\n", fileName, uri));
Massive changes. Too many to list here, but I'll try a summary: 2002-02-17 Tor Lillqvist <tml@iki.fi> * gdk/win32/*.c: Massive changes. Too many to list here, but I'll try a summary: 1) Unify GdkPixmap and GdkImage implementation: For each GdkPixmap, allocate a GdkImage, and vice versa. GdkPixmapImplWin32Data has a pointer to the GdkImage. GdkImage::windowing_data is a pointer to the GdkPixmap. This simplifies many pixmap and image related functions a lot, and reduces duplicated code snippets. For instance, there is only one place in gdk/win32 where CreateDIBSection() is called, in the function _gdk_win32_new_pixmap(). Converting a bitmap (GdkPixmap) to a Windows region is almost trivial, with the bitmap bits being readily accessible in the associated GdkImage. All blitting between GdkPixmaps, GdkWindows and GdkImages goes through handled the _gdk_win32_blit() function, which calls different functions to handle the cases of blitting from pixmaps, inside windows (scrolling), or from windows, which all require somewhat different handling. 2) Support 256-color mode. This has long been very broken, now it works more or less OK. Keep the logical palette for each colormap as small as possible while allocating and freeing colors. Select and realize the logical palette associated with a GdkColormap into a DC before drawing or blitting. When the display is in 256-color mode, make it possible for the user to override the size of the palette(s) used with either the GDK_WIN32_MAX_COLORS environment variable, or a -max-colors command line option. It is possible to reduce the palette size all the way down to using just the 16 static colors (which causes the system visual to be of type GDK_VISUAL_STATIC_COLOR. This could possibly be useful if one desperately wants to avoid color flashing. (Note that in order for this to work properly, an as of yet not commited fix to gdkrgb.c is needed.) Handle the palette messages. On WM_PALETTECHANGED, call UpdateColors() for the given window hierarchy. Do this only if a window in some other top-level window hierarchy caused the palette change (realized a palette). Do this max five times in a row (an arbitrarily chosen limit), though, otherwise redraw by generating expose events. On WM_QUERYNEWPALETTE, cause a redraw of the whole window hierarchy by generating GDK_EXPOSE events. 3) Code cleanup in general. For instance, remove the "emulated" X11 structs ColormapStruct, Visual and XStandardColormap. Use the new GDK_DEBUG_* flags for debugging output in the relevant source files. Remove the unused colormap hash table in gdkcolor-win32.c 4) Plug some resource leaks. 2002-02-14 Tor Lillqvist <tml@iki.fi> * gdk/win32/gdkdnd-win32.c (gdk_dropfiles_filter): Use g_filename_to_uri() to actually create legal URIs in the text/uri-list data.
2002-02-17 00:25:05 +00:00
g_free (uri);
}
}
#if 0
/* Awful hack to recognize temp files corresponding to
* images dragged from Firefox... Open the file right here
* so that it is less likely that Firefox manages to delete
* it before the GTK+-using app (typically GIMP) has opened
* it.
*
* Not compiled in for now, because it means images dragged
* from Firefox would stay around in the temp folder which
* is not what Firefox intended. I don't feel comfortable
* with that, both from a geenral sanity point of view, and
* from a privacy point of view. It's better to wait for
* Firefox to fix the problem, for instance by deleting the
* temp file after a longer delay, or to wait until we
* implement the OLE2_DND...
*/
if (filename_looks_tempish (fileName))
{
int fd = g_open (fileName, _O_RDONLY|_O_BINARY, 0);
if (fd == -1)
{
GDK_NOTE (DND, g_print ("Could not open %s, maybe an image dragged from Firefox that it already deleted\n", fileName));
}
else
{
GDK_NOTE (DND, g_print ("Opened %s as %d so that Firefox won't delete it\n", fileName, fd));
g_timeout_add_seconds (1, close_it, GINT_TO_POINTER (fd));
}
}
#endif
g_free (fileName);
g_string_append (result, "\015\012");
}
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
_gdk_dropfiles_store (result->str);
g_string_free (result, FALSE);
DragFinish (hdrop);
return GDK_FILTER_TRANSLATE;
}
else
return GDK_FILTER_CONTINUE;
}
void
Make gdkx.h the only installed header from gdk/x11. All structures in Fri Sep 7 11:51:44 2001 Owen Taylor <otaylor@redhat.com> Make gdkx.h the only installed header from gdk/x11. All structures in gdk/x11 are opaque. * gdk/x11/Makefile.am gdk/x11/gdkx.h gdk/x11/gdkprivate-x11.h: Don't install gdk{drawable,pixmap,window}-x11.h. * gdk/x11/{gdkcolormap-x11.c, gdkfont-x11.c, gdkx.h, gdkvisual-x11.c: Move GdkColormapPrivateX11, GdkFontPrivateX GdkImagePrivateX11, GdkVisualClass into C files. * gdk/gdkpixmap-x11.[ch]: Make gdk_pixmap_impl_get_type() static. * gdk/x11/{gdkcolor-x11.c, gdkcursor-x11.c, gdkdrawable-x11.c, gdkfont-x11.c, gdkgc-x11.c, gdkx.h, gdkimage-x11,gdkvisual-x11.c} Add public functions to replace previously exported direct structure access. gdk_x11_colormap_get_{xdisplay,xcolormap} gdk_x11_cursor_get_{xdisplay,xcursor}, gdk_x11_drawable_get_{xdisplay,xcursor,gdk_x11_visual_get_xvisual, gdk_x11_font_get_{xdisplay,xfont}, gdk_x11_image_get_{xdisplay,ximage}, gdk_x11_gc_get_{xdisplay,ximage} * gdk/gdkprivate.h gdk/gdkinternals.h: Move GdkColorInfo, GdkEventFilter, GdkClientFilter, GdkFontPrivate to gdkinternals. Fix a number of variables and functions that were exported "accidentally" from GDK. * gdk/**.[ch]: gdk => _gdk for gdk_visual_init, gdk_events_init, gdk_input_init, gdk_dnd_init, gdk_image_exit, gdk_input_exit, gdk_windowing_exit, gdk_event_func, gdk_event_data, gdk_event_notify, gdk_queued_events, gdk_queued_tail, gdk_event_new, gdk_events_queue, gdk_events_unqueue, gdk_event_queue_find_first, gdk_event_queue_remove_link, gdk_event_queue_append, gdk_event_button_generate, gdk_debug_flags, gdk_default_filters, gdk_parent_root. * gdk/x11/{gdkevents-x11.c, gdkglobals-x11.c, gdkimage-x11.c, gdkmain-x11.c, gdkprivate-x11.h, gdk/x11/gdkwindow-x11.c}: gdk => _gdk for gdk_event_mask_table, gkd_nevent_masks, gdk_wm_window_protocols, gdk_leader_window, gdk_xgrab_window, gdk_use_xshm, gdk_input_ignore_core. * gdk/x11/xsettings-common.h (xsettings_list_insert): Add #defines to namespace functions into the private _gdk_ namespace. * gdk/gdkwindow.[ch] gdk/x11/gdkx.h: Add gdk_get_default_root_window () to replace gdk_parent_root exported variable. Adjust and deprecate GDK_ROOT_PARENT(). * demos/{testpixbuf-drawable.c,testpixbuf-save.c}: Fix GDK_ROOT_PARENT usage, remove includes of port-specific headers. * gdk/{win32,x11,fb}/gdkinput*.[ch]: s/gdk/_gdk/ for _gdk_input_gxid_host, _gdk_input_gxid_port, _gdk_input_ignore_core, gdk_input_devices, _gdk_input_windows, gdk_init_input_core. * gdk/x11/{gdkevents-x11.,c gdkglobals-x11.c, gdkmain-x11.c} docs/Changes-2.0.txt: Remove gdk_wm_protocols, gdk_wm_delete_window functions, gdk_wm_take_focus, use gdk_atom_intern() instead. * gdk/linux-fb/{gdkselection-fb.c, gdkmain-fb.c, gdkprivatefb.h} gdk/win32/{gdkselection-win32.c, gdkmgdkwin32.h, gdkprivate-win32.h} gdk/x11/{gdkselection-x11.c gdkx.h, gtkprivate-x11.h} gtk/gtkselection.c Unexport gdk_selection_property, just use gdk_atom_intern ("GDK_SELECTION"). * gdk/x11/{gdkprivate-x11.h,gdkdrawable-x11h,gdkgc-x11.c,gdkx.h}: Unexport gdk_drawable_impl_x11_get_type, gdk_gc_x11_get_type, GDK_GC_X11 cast macros, GdkGCX11 structures, GdkCursorPrivate, GdkVisualprivate, gdk_x11_gc_flush. Make a number of public exports of variables into functions to increase encapsulation. * gdk/gdkinternals.h gdk/gdkinput.h gdk/gdkevents.h gdk/linux-fb/gdkmouse-fb.c: gdk_core_pointer => _gdk_core_pointer, move to gdkinternals.h. Add gdk_device_get_core_pointer (). * gdk/gdkprivate.h gdk/gdkpango.c gdk/gdkinternals.h docs/Changes-2.0.txt: Unexport gdk_parent_root, gdk_error_code, gdk_error_warnings. * gdk/x11/{gdkcolormap-x11.c, gdkmain-x11.c, gdkx.h} docs/Changes-2.0.txt: s/gdk_screen/_gdk_screen/, add gdk_x11_get_default_screen() s/gdk_root_window/_gdk_root_window/, add gdk_x11_get_default_root_xwindow() Add gdk_x11_get_default_xdisplay(). * gdk/gdk.h gdk/gdk.c linux-fb/gdkfb.h linux-fb/gdkglobals-fb.c win32/gdkwin32.h x11/gdkglobals-x11.c gdk/x11/gdkmain-x11.c gdk/x11/gdkx.h: gdk/gdk.def: Add gdk_get/set_program_class, Don't export gdk_progclass, move --class command line option and handling to common portion of GDK. Miscellaneous fixes: * gdk/x11/gdkwindow-x11.c (gdk_window_set_icon_list): Fix g_return_val_if_fail that should have been g_return_if_fail. * gdk/gdkinternals.h gdk/gdkprivate.h: Move gdk_synthesize_window_state() to the semi-public gdkprivate.h. * gtk/gtkdnd.c (_gtk_drag_source_handle_event): Remove uneeded X11 dependency. * gdk/linux-fb/gdkmain-fb.c gdk/win32/gdkmain-win32.c gdk/TODO: Remove unused gdk_key_repeat_disable/restore. * linux-fb/gdkglobals-fb.c win32/gdkglobals-win32.c x11/gdkglobals-x11.c x11/gdkprivate-x11.h gdk/gdk.def: Remove unused gdk_null_window_warnings variable. * gdk/Makefile.am (DIST_SUBDIRS) nanox/*: cvs remove nanox; it can be retrieved from the repository; it is too far from functional to be worth having people check out; it would be easier to start from scratch, I suspect. * gdk/x11/gdkpixmap-x11.c: Fix lvalue usage of GDK_PIXMAP_XID(). * gdk/x11/gdkkeys-x11.c gdk/gdkrgb.c gdk/gdkwindow.c gdk/x11/gdkpango-x11.c gdk/x11/gdkselection-x11.c: Fix some accidentally global variables and unused global variables. * gdk/x11/gdkkeys-x11.c gdk/gdkrgb.c gdk/gdkwindow.c gdk/x11/gdkpango-x11.c gdk/x11/gdkselection-x11.c: Fix some accidentally global variables and unused global variables. Add some space for future expansion to multihead. * gdk/gdkdrawable.h: Add four reserved function pointers for future expansion of GdkDrawableClass. * gtk/gtkwindow.h gtk/gtkinvisible.h: Add reserved pointer where we can put a GdkScreen * later.
2001-09-07 21:50:20 +00:00
_gdk_dnd_init (void)
{
CoInitializeEx (NULL, COINIT_APARTMENTTHREADED);
if (getenv ("GDK_WIN32_USE_EXPERIMENTAL_OLE2_DND"))
use_ole2_dnd = TRUE;
if (use_ole2_dnd)
{
HRESULT hr;
hr = OleInitialize (NULL);
if (! SUCCEEDED (hr))
g_error ("OleInitialize failed");
target_ctx_for_window = g_hash_table_new (g_direct_hash, g_direct_equal);
}
}
void
_gdk_win32_dnd_exit (void)
{
if (use_ole2_dnd)
{
OleUninitialize ();
}
CoUninitialize ();
}
/* Source side */
static void
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
local_send_leave (GdkDragContext *context,
guint32 time)
{
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
GDK_NOTE (DND, g_print ("local_send_leave: context=%p current_dest_drag=%p\n",
context,
current_dest_drag));
if ((current_dest_drag != NULL) &&
(current_dest_drag->protocol == GDK_DRAG_PROTO_LOCAL) &&
(current_dest_drag->source_window == context->source_window))
{
2010-05-25 22:38:44 +00:00
tmp_event = gdk_event_new (GDK_DRAG_LEAVE);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.window, context->dest_window);
/* Pass ownership of context to the event */
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, current_dest_drag);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = GDK_CURRENT_TIME; /* FIXME? */
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (context));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (context)));
current_dest_drag = NULL;
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
gdk_event_free (tmp_event);
}
}
static void
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
local_send_enter (GdkDragContext *context,
guint32 time)
{
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
GdkDragContext *new_context;
GDK_NOTE (DND, g_print ("local_send_enter: context=%p current_dest_drag=%p\n",
context,
current_dest_drag));
if (current_dest_drag != NULL)
{
g_object_unref (G_OBJECT (current_dest_drag));
current_dest_drag = NULL;
}
new_context = gdk_drag_context_new (gdk_window_get_display (context->source_window));
new_context->protocol = GDK_DRAG_PROTO_LOCAL;
new_context->is_source = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&new_context->source_window, context->source_window);
g_set_object (&new_context->dest_window, context->dest_window);
new_context->formats = gdk_content_formats_ref (context->formats);
gdk_window_set_events (new_context->source_window,
gdk_window_get_events (new_context->source_window) |
GDK_PROPERTY_CHANGE_MASK);
new_context->actions = context->actions;
2010-05-25 22:38:44 +00:00
tmp_event = gdk_event_new (GDK_DRAG_ENTER);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.window, context->dest_window);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, new_context);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = GDK_CURRENT_TIME; /* FIXME? */
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (context));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (context)));
current_dest_drag = new_context;
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
gdk_event_free (tmp_event);
}
static void
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
local_send_motion (GdkDragContext *context,
gint x_root,
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
gint y_root,
GdkDragAction action,
guint32 time)
{
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
GDK_NOTE (DND, g_print ("local_send_motion: context=%p (%d,%d) current_dest_drag=%p\n",
context, x_root, y_root,
current_dest_drag));
if ((current_dest_drag != NULL) &&
(current_dest_drag->protocol == GDK_DRAG_PROTO_LOCAL) &&
(current_dest_drag->source_window == context->source_window))
{
GdkWin32DragContext *current_dest_drag_win32;
2010-05-25 22:38:44 +00:00
tmp_event = gdk_event_new (GDK_DRAG_MOTION);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.window, current_dest_drag->dest_window);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, current_dest_drag);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = time;
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (current_dest_drag));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (current_dest_drag)));
current_dest_drag->suggested_action = action;
2010-05-25 22:38:44 +00:00
tmp_event->dnd.x_root = x_root;
tmp_event->dnd.y_root = y_root;
current_dest_drag_win32 = GDK_WIN32_DRAG_CONTEXT (current_dest_drag);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
current_dest_drag_win32->last_x = x_root;
current_dest_drag_win32->last_y = y_root;
context_win32->drag_status = GDK_DRAG_STATUS_MOTION_WAIT;
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
gdk_event_free (tmp_event);
}
}
static void
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
local_send_drop (GdkDragContext *context,
guint32 time)
{
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
GDK_NOTE (DND, g_print ("local_send_drop: context=%p current_dest_drag=%p\n",
context,
current_dest_drag));
if ((current_dest_drag != NULL) &&
(current_dest_drag->protocol == GDK_DRAG_PROTO_LOCAL) &&
(current_dest_drag->source_window == context->source_window))
{
GdkWin32DragContext *context_win32;
/* Pass ownership of context to the event */
2010-05-25 22:38:44 +00:00
tmp_event = gdk_event_new (GDK_DROP_START);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.window, current_dest_drag->dest_window);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, current_dest_drag);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = GDK_CURRENT_TIME;
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (current_dest_drag));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (current_dest_drag)));
context_win32 = GDK_WIN32_DRAG_CONTEXT (current_dest_drag);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
tmp_event->dnd.x_root = context_win32->last_x;
tmp_event->dnd.y_root = context_win32->last_y;
current_dest_drag = NULL;
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
gdk_event_free (tmp_event);
}
}
static void
gdk_drag_do_leave (GdkDragContext *context,
guint32 time)
{
if (context->dest_window)
{
GDK_NOTE (DND, g_print ("gdk_drag_do_leave\n"));
if (!use_ole2_dnd)
{
if (context->protocol == GDK_DRAG_PROTO_LOCAL)
local_send_leave (context, time);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_clear_object (&context->dest_window);
}
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static GdkWindow *
create_drag_window (GdkDisplay *display)
{
GdkWindow *window;
window = gdk_window_new_popup (display, &(GdkRectangle) { 0, 0, 100, 100 });
gdk_window_set_type_hint (window, GDK_WINDOW_TYPE_HINT_DND);
return window;
}
GdkDragContext *
_gdk_win32_window_drag_begin (GdkWindow *window,
GdkDevice *device,
GdkContentFormats *formats,
gint dx,
gint dy)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkDragContext *new_context;
GdkWin32DragContext *context_win32;
BYTE kbd_state[256];
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
int x_root, y_root;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (!use_ole2_dnd)
{
g_return_val_if_fail (window != NULL, NULL);
new_context = gdk_drag_context_new (gdk_window_get_display (window));
2010-05-25 22:38:44 +00:00
new_context->is_source = TRUE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&new_context->source_window, window);
new_context->formats = gdk_content_formats_ref (formats);
new_context->actions = 0;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32 = GDK_WIN32_DRAG_CONTEXT (new_context);
}
else
{
source_drag_context *ctx;
g_return_val_if_fail (window != NULL, NULL);
GDK_NOTE (DND, g_print ("gdk_drag_begin\n"));
ctx = source_context_new (window, formats);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
sel_win32->dnd_source_state = GDK_WIN32_DND_PENDING;
pending_src_context = ctx;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
new_context = g_object_ref (ctx->context);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32 = GDK_WIN32_DRAG_CONTEXT (new_context);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_device_get_position (device, &x_root, &y_root);
x_root += dx;
y_root += dy;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32->start_x = x_root;
context_win32->start_y = y_root;
context_win32->last_x = context_win32->start_x;
context_win32->last_y = context_win32->start_y;
context_win32->last_key_state = 0;
API_CALL (GetKeyboardState, (kbd_state));
if (kbd_state[VK_CONTROL] & 0x80)
context_win32->last_key_state |= MK_CONTROL;
if (kbd_state[VK_SHIFT] & 0x80)
context_win32->last_key_state |= MK_SHIFT;
if (kbd_state[VK_LBUTTON] & 0x80)
context_win32->last_key_state |= MK_LBUTTON;
if (kbd_state[VK_MBUTTON] & 0x80)
context_win32->last_key_state |= MK_MBUTTON;
if (kbd_state[VK_RBUTTON] & 0x80)
context_win32->last_key_state |= MK_RBUTTON;
context_win32->drag_window = create_drag_window (gdk_window_get_display (window));
return new_context;
}
void
_gdk_win32_dnd_do_dragdrop (void)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkDragContext* drag_ctx;
data_object *dobj;
HRESULT hr;
DWORD dwEffect;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (!use_ole2_dnd)
return;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (pending_src_context == NULL)
return;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
drag_ctx = pending_src_context->context;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
dobj = data_object_new (drag_ctx);
current_src_object = dobj;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* Start dragging with mainloop inside the OLE2 API. Exits only when done */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("Calling DoDragDrop\n"));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
_gdk_win32_begin_modal_call (GDK_WIN32_MODAL_OP_DND);
hr = DoDragDrop (&dobj->ido, &pending_src_context->ids,
DROPEFFECT_COPY | DROPEFFECT_MOVE,
&dwEffect);
_gdk_win32_end_modal_call (GDK_WIN32_MODAL_OP_DND);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("DoDragDrop returned %s\n",
(hr == DRAGDROP_S_DROP ? "DRAGDROP_S_DROP" :
(hr == DRAGDROP_S_CANCEL ? "DRAGDROP_S_CANCEL" :
(hr == E_UNEXPECTED ? "E_UNEXPECTED" :
g_strdup_printf ("%#.8lx", hr))))));
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
/* Delete dnd selection after successful move */
if (hr == DRAGDROP_S_DROP && dwEffect == DROPEFFECT_MOVE)
{
GdkWin32Selection *win32_sel = _gdk_win32_selection_get ();
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkEvent tmp_event;
memset (&tmp_event, 0, sizeof (tmp_event));
tmp_event.type = GDK_SELECTION_REQUEST;
g_set_object (&tmp_event.selection.window, drag_ctx->source_window);
tmp_event.selection.send_event = FALSE;
tmp_event.selection.selection = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_OLE2_DND);
tmp_event.selection.target = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_DELETE);
win32_sel->property_change_target_atom = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_DELETE);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
tmp_event.selection.property = _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_OLE2_DND);
g_set_object (&tmp_event.selection.requestor, drag_ctx->source_window);
tmp_event.selection.time = GDK_CURRENT_TIME; /* ??? */
GDK_NOTE (EVENTS, _gdk_win32_print_event (&tmp_event));
gdk_event_put (&tmp_event);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
{
GdkEvent *tmp_event;
tmp_event = gdk_event_new (GDK_DROP_FINISHED);
g_set_object (&tmp_event->dnd.window, drag_ctx->source_window);
tmp_event->dnd.send_event = FALSE;
g_set_object (&tmp_event->dnd.context, drag_ctx);
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (drag_ctx));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (drag_ctx)));
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
gdk_event_free (tmp_event);
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
current_src_object = NULL;
dobj->ido.lpVtbl->Release (&dobj->ido);
if (pending_src_context != NULL)
{
pending_src_context->ids.lpVtbl->Release (&pending_src_context->ids);
pending_src_context = NULL;
}
}
typedef struct {
gint x;
gint y;
HWND ignore;
HWND result;
} find_window_enum_arg;
static BOOL CALLBACK
find_window_enum_proc (HWND hwnd,
LPARAM lparam)
{
RECT rect;
POINT tl, br;
find_window_enum_arg *a = (find_window_enum_arg *) lparam;
if (hwnd == a->ignore)
return TRUE;
if (!IsWindowVisible (hwnd))
return TRUE;
tl.x = tl.y = 0;
ClientToScreen (hwnd, &tl);
GetClientRect (hwnd, &rect);
br.x = rect.right;
br.y = rect.bottom;
ClientToScreen (hwnd, &br);
if (a->x >= tl.x && a->y >= tl.y && a->x < br.x && a->y < br.y)
{
a->result = hwnd;
return FALSE;
}
else
return TRUE;
}
static GdkWindow *
gdk_win32_drag_context_find_window (GdkDragContext *context,
GdkWindow *drag_window,
gint x_root,
gint y_root,
GdkDragProtocol *protocol)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
GdkWindow *dest_window, *dw;
find_window_enum_arg a;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
a.x = x_root * context_win32->scale - _gdk_offset_x;
a.y = y_root * context_win32->scale - _gdk_offset_y;
a.ignore = drag_window ? GDK_WINDOW_HWND (drag_window) : NULL;
a.result = NULL;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND,
g_print ("gdk_drag_find_window_real: %p %+d%+d\n",
(drag_window ? GDK_WINDOW_HWND (drag_window) : NULL),
a.x, a.y));
EnumWindows (find_window_enum_proc, (LPARAM) &a);
if (a.result == NULL)
dest_window = NULL;
else
{
dw = gdk_win32_handle_table_lookup (a.result);
if (dw)
{
dest_window = gdk_window_get_toplevel (dw);
g_object_ref (dest_window);
}
else
dest_window = gdk_win32_window_foreign_new_for_display (context->display, a.result);
if (use_ole2_dnd)
*protocol = GDK_DRAG_PROTO_OLE2;
else if (context->source_window)
*protocol = GDK_DRAG_PROTO_LOCAL;
else
*protocol = GDK_DRAG_PROTO_WIN32_DROPFILES;
}
GDK_NOTE (DND,
g_print ("gdk_drag_find_window: %p %+d%+d: %p: %p %s\n",
(drag_window ? GDK_WINDOW_HWND (drag_window) : NULL),
x_root, y_root,
a.result,
(dest_window ? GDK_WINDOW_HWND (dest_window) : NULL),
_gdk_win32_drag_protocol_to_string (*protocol)));
return dest_window;
}
static gboolean
gdk_win32_drag_context_drag_motion (GdkDragContext *context,
GdkWindow *dest_window,
GdkDragProtocol protocol,
gint x_root,
gint y_root,
GdkDragAction suggested_action,
GdkDragAction possible_actions,
guint32 time)
{
GdkWin32DragContext *context_win32;
g_return_val_if_fail (context != NULL, FALSE);
context->actions = possible_actions;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GDK_NOTE (DND, g_print ("gdk_drag_motion: @ %+d:%+d %s suggested=%s, possible=%s\n"
" context=%p:{actions=%s,suggested=%s,action=%s}\n",
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
x_root, y_root,
_gdk_win32_drag_protocol_to_string (protocol),
_gdk_win32_drag_action_to_string (suggested_action),
_gdk_win32_drag_action_to_string (possible_actions),
context,
_gdk_win32_drag_action_to_string (context->actions),
_gdk_win32_drag_action_to_string (context->suggested_action),
_gdk_win32_drag_action_to_string (context->action)));
context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (context_win32->drag_window)
move_drag_window (context, x_root, y_root);
if (!use_ole2_dnd)
{
if (context->dest_window == dest_window)
{
GdkDragContext *dest_context;
dest_context = gdk_drag_context_find (FALSE,
context->source_window,
dest_window);
if (dest_context)
dest_context->actions = context->actions;
context->suggested_action = suggested_action;
}
else
{
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
/* Send a leave to the last destination */
gdk_drag_do_leave (context, time);
context_win32->drag_status = GDK_DRAG_STATUS_DRAG;
/* Check if new destination accepts drags, and which protocol */
if (dest_window)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&context->dest_window, dest_window);
context->protocol = protocol;
switch (protocol)
{
case GDK_DRAG_PROTO_LOCAL:
local_send_enter (context, time);
break;
default:
break;
}
context->suggested_action = suggested_action;
}
else
{
context->dest_window = NULL;
context->action = 0;
}
/* Push a status event, to let the client know that
* the drag changed
*/
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
tmp_event = gdk_event_new (GDK_DRAG_STATUS);
g_set_object (&tmp_event->dnd.window, context->source_window);
/* We use this to signal a synthetic status. Perhaps
* we should use an extra field...
*/
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = TRUE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, context);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = time;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (context));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (context)));
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_free (tmp_event);
}
/* Send a drag-motion event */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_win32->last_x = x_root;
context_win32->last_y = y_root;
if (context->dest_window)
{
if (context_win32->drag_status == GDK_DRAG_STATUS_DRAG)
{
switch (context->protocol)
{
case GDK_DRAG_PROTO_LOCAL:
local_send_motion (context, x_root, y_root, suggested_action, time);
break;
case GDK_DRAG_PROTO_NONE:
g_warning ("GDK_DRAG_PROTO_NONE is not valid in gdk_drag_motion()");
break;
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
default:
break;
}
}
else
{
GDK_NOTE (DND, g_print (" returning TRUE\n"
" context=%p:{actions=%s,suggested=%s,action=%s}\n",
context,
_gdk_win32_drag_action_to_string (context->actions),
_gdk_win32_drag_action_to_string (context->suggested_action),
_gdk_win32_drag_action_to_string (context->action)));
return TRUE;
}
}
}
GDK_NOTE (DND, g_print (" returning FALSE\n"
" context=%p:{actions=%s,suggested=%s,action=%s}\n",
context,
_gdk_win32_drag_action_to_string (context->actions),
_gdk_win32_drag_action_to_string (context->suggested_action),
_gdk_win32_drag_action_to_string (context->action)));
return FALSE;
}
static void
gdk_win32_drag_context_drag_drop (GdkDragContext *context,
guint32 time)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
g_return_if_fail (context != NULL);
GDK_NOTE (DND, g_print ("gdk_drag_drop\n"));
if (!use_ole2_dnd)
{
if (context->dest_window &&
context->protocol == GDK_DRAG_PROTO_LOCAL)
local_send_drop (context, time);
}
else
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
sel_win32->dnd_source_state = GDK_WIN32_DND_DROPPED;
}
}
static void
gdk_win32_drag_context_drag_abort (GdkDragContext *context,
guint32 time)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
g_return_if_fail (context != NULL);
GDK_NOTE (DND, g_print ("gdk_drag_abort\n"));
if (use_ole2_dnd)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
sel_win32->dnd_source_state = GDK_WIN32_DND_NONE;
}
/* Destination side */
static void
gdk_win32_drag_context_drag_status (GdkDragContext *context,
GdkDragAction action,
guint32 time)
{
GdkDragContext *src_context;
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
g_return_if_fail (context != NULL);
GDK_NOTE (DND, g_print ("gdk_drag_status: %s\n"
" context=%p:{actions=%s,suggested=%s,action=%s}\n",
_gdk_win32_drag_action_to_string (action),
context,
_gdk_win32_drag_action_to_string (context->actions),
_gdk_win32_drag_action_to_string (context->suggested_action),
_gdk_win32_drag_action_to_string (context->action)));
context->action = action;
if (!use_ole2_dnd)
{
src_context = gdk_drag_context_find (TRUE,
context->source_window,
context->dest_window);
if (src_context)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32DragContext *src_context_win32 = GDK_WIN32_DRAG_CONTEXT (src_context);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
if (src_context_win32->drag_status == GDK_DRAG_STATUS_MOTION_WAIT)
src_context_win32->drag_status = GDK_DRAG_STATUS_DRAG;
2010-05-25 22:38:44 +00:00
tmp_event = gdk_event_new (GDK_DRAG_STATUS);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.window, context->source_window);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, src_context);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.time = GDK_CURRENT_TIME; /* FIXME? */
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (src_context));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (src_context)));
if (action == GDK_ACTION_DEFAULT)
action = 0;
src_context->action = action;
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_free (tmp_event);
}
}
}
static void
gdk_win32_drag_context_drop_reply (GdkDragContext *context,
gboolean ok,
guint32 time)
{
g_return_if_fail (context != NULL);
GDK_NOTE (DND, g_print ("gdk_drop_reply\n"));
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
if (!use_ole2_dnd)
if (context->dest_window)
{
if (context->protocol == GDK_DRAG_PROTO_WIN32_DROPFILES)
_gdk_dropfiles_store (NULL);
}
}
static void
gdk_win32_drag_context_drop_finish (GdkDragContext *context,
gboolean success,
guint32 time)
{
GdkDragContext *src_context;
2010-05-25 22:38:44 +00:00
GdkEvent *tmp_event;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
g_return_if_fail (context != NULL);
GDK_NOTE (DND, g_print ("gdk_drop_finish\n"));
if (!use_ole2_dnd)
{
src_context = gdk_drag_context_find (TRUE,
context->source_window,
context->dest_window);
if (src_context)
{
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
tmp_event = gdk_event_new (GDK_DROP_FINISHED);
g_set_object (&tmp_event->dnd.window, src_context->source_window);
2010-05-25 22:38:44 +00:00
tmp_event->dnd.send_event = FALSE;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
g_set_object (&tmp_event->dnd.context, src_context);
gdk_event_set_device (tmp_event, gdk_drag_context_get_device (src_context));
gdk_event_set_seat (tmp_event, gdk_device_get_seat (gdk_drag_context_get_device (src_context)));
2010-05-25 22:38:44 +00:00
GDK_NOTE (EVENTS, _gdk_win32_print_event (tmp_event));
gdk_event_put (tmp_event);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_event_free (tmp_event);
}
}
else
{
gdk_drag_do_leave (context, time);
if (success)
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
sel_win32->dnd_target_state = GDK_WIN32_DND_DROPPED;
else
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
sel_win32->dnd_target_state = GDK_WIN32_DND_FAILED;
}
}
#if 0
static GdkFilterReturn
gdk_destroy_filter (GdkXEvent *xev,
GdkEvent *event,
gpointer data)
{
MSG *msg = (MSG *) xev;
if (msg->message == WM_DESTROY)
{
IDropTarget *idtp = (IDropTarget *) data;
GDK_NOTE (DND, g_print ("gdk_destroy_filter: WM_DESTROY: %p\n", msg->hwnd));
#if 0
idtp->lpVtbl->Release (idtp);
#endif
RevokeDragDrop (msg->hwnd);
CoLockObjectExternal ((IUnknown*) idtp, FALSE, TRUE);
}
return GDK_FILTER_CONTINUE;
}
#endif
void
_gdk_win32_window_register_dnd (GdkWindow *window)
{
target_drag_context *ctx;
HRESULT hr;
g_return_if_fail (window != NULL);
if (g_object_get_data (G_OBJECT (window), "gdk-dnd-registered") != NULL)
return;
else
g_object_set_data (G_OBJECT (window), "gdk-dnd-registered", GINT_TO_POINTER (TRUE));
Apply the same fixes and improvements as to the gtk-1-3-win32-production 2002-01-10 Tor Lillqvist <tml@iki.fi> Apply the same fixes and improvements as to the gtk-1-3-win32-production branch: Bug fixes and cleanup of selection and DND functionality. Still doesn't work as well as the win32-production branch, though, but getting closer. After this, need to add Archaeopteryx Software's OLE2 DND support. * gdk/win32/gdkselection-win32.c (gdk_selection_owner_set, gdk_selection_send_notify, generate_selection_notify): Don't use SendMessage() to generate events for the same app, instead use gdk_event_put(). * gdk/win32/gdkprivate-win32.h * gdk/win32/gdkglobals-win32.c * gdk/win32/gdkmain-win32.c * gdk/win32/gdkevents-win32.c: Thus, remove declaration, definition, initialisation and handling of gdk_selection_notify_msg, gdk_selection_request_msg and gdk_selection_clear_msg. * gdk/win32/gdkselection-win32.c (gdk_text_property_to_text_list, gdk_free_text_list, gdk_string_to_compound_text, gdk_free_compound_text): Implement trivially, witrh a text_list always having a single element, and a compound text always consisting of just a single (UTF-8!) string. Let's see how well this works. * gdk/win32/gdkselection-win32.c (gdk_selection_convert): Fix non-ASCII paste from the clipboard: Try getting the same formats from the Windows clipboard that gdk_property_change() puts there: CF_UNICODETEXT, UTF8_STRING or CF_TEXT+CF_LOCALE. * gdk/win32/gdkproperty-win32.c (gdk_property_change): When storing text on the clipboard, handle non-ASCII text correctly. The logic is as follows: If we have only ASCII characters, use CF_TEXT. Else, if we are on NT, use CF_UNICODETEXT. Else (we are on Win9x), if all the characters are present in the code page of some installed locale, use CF_TEXT and also set CF_LOCALE to that locale. Else (still on Win9x) store as RTF. We use a very simple RTF string, just the text, no fonts or other crap, with the non-ASCII characters as Unicode \uN keywords. Additionally, also store the UTF-8 string as such, under the format "UTF8_STRING", so that GDK can also paste from the Clipboard what it has copied there. (Thus no need to implement any RTF parser.) (find_common_locale): New function, implements the search for a locale for case 3 above. * gdk/win32/gdkglobals-win32.c: New global variables compound_text, text_uri_list, utf8_string, cf_rtf and cf_utf8_string. * gdk/win32/gdkim-win32.c (_gdk_ucs2_to_utf8): New function, converts from a wchar_t string to UTF-8. (_gdk_utf8_to_ucs2): Rename from _gdk_win32_nmbstowchar_ts. (_gdk_utf8_to_wcs): Rename from gdk_nmbstowchar_ts. * gdk/win32/gdkevents-win32.c (build_keypress_event): Use _gdk_ucs2_to_utf8(). * gdk/win32/gdkselection-win32.c: Remove some unnecessary logging. * gdk/win32/gdkdnd-win32.c: Plug memory leaks, the gdk_drag_context_ref() was called unnecessarily in a couple of places, meaning drag contexts were never freed. The same memory leaks seem to be present in gdk/linux-fb/gdkselection-fb.c, BTW. (gdk_drop_reply): For WIN32_DROPFILES drops, free the temporarily stored file list. * gdk/win32/gdkselection-win32.c: Clarify the use of the sel_prop_table. Now it is used only for storing the GDK_SELECTION "properties". The file names dropped with WM_DROPFILES -style DND is stored temporarily (between the drop and the target picking them up) in a separate place. Have a separate hash table to map selection atoms to owner windows. This used to be quite mixed up. (_gdk_dropfiles_store): New function, to store the dropped file list for the drop target to possibly fetch, and clear it afterwards, from gdk_drop_reply(). (gdk_selection_owner_get): Much simplified now.
2002-01-10 00:53:39 +00:00
GDK_NOTE (DND, g_print ("gdk_window_register_dnd: %p\n", GDK_WINDOW_HWND (window)));
if (!use_ole2_dnd)
{
/* We always claim to accept dropped files, but in fact we might not,
* of course. This function is called in such a way that it cannot know
* whether the window (widget) in question actually accepts files
* (in gtk, data of type text/uri-list) or not.
*/
gdk_window_add_filter (window, gdk_dropfiles_filter, NULL);
DragAcceptFiles (GDK_WINDOW_HWND (window), TRUE);
}
else
{
/* Return if window is already setup for DND. */
if (g_hash_table_lookup (target_ctx_for_window, GDK_WINDOW_HWND (window)) != NULL)
return;
ctx = target_context_new (window);
hr = CoLockObjectExternal ((IUnknown *) &ctx->idt, TRUE, FALSE);
if (!SUCCEEDED (hr))
OTHER_API_FAILED ("CoLockObjectExternal");
else
{
hr = RegisterDragDrop (GDK_WINDOW_HWND (window), &ctx->idt);
if (hr == DRAGDROP_E_ALREADYREGISTERED)
{
g_print ("DRAGDROP_E_ALREADYREGISTERED\n");
CoLockObjectExternal ((IUnknown *) &ctx->idt, FALSE, FALSE);
}
else if (!SUCCEEDED (hr))
OTHER_API_FAILED ("RegisterDragDrop");
else
{
g_object_ref (window);
g_hash_table_insert (target_ctx_for_window, GDK_WINDOW_HWND (window), ctx);
}
}
}
}
static gboolean
gdk_win32_drag_context_drop_status (GdkDragContext *context)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
return ! context_win32->drop_failed;
}
static GdkAtom
gdk_win32_drag_context_get_selection (GdkDragContext *context)
{
switch (context->protocol)
{
case GDK_DRAG_PROTO_LOCAL:
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
return _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_LOCAL_DND_SELECTION);
case GDK_DRAG_PROTO_WIN32_DROPFILES:
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
return _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_DROPFILES_DND);
case GDK_DRAG_PROTO_OLE2:
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
return _gdk_win32_selection_atom (GDK_WIN32_ATOM_INDEX_OLE2_DND);
default:
2017-11-15 17:13:31 +00:00
return NULL;
}
}
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
static void
gdk_win32_drag_context_set_cursor (GdkDragContext *context,
GdkCursor *cursor)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
GDK_NOTE (DND, g_print ("gdk_drag_context_set_cursor: 0x%p 0x%p\n", context, cursor));
if (!g_set_object (&context_win32->cursor, cursor))
return;
if (context_win32->grab_seat)
{
G_GNUC_BEGIN_IGNORE_DEPRECATIONS;
gdk_device_grab (gdk_seat_get_pointer (context_win32->grab_seat),
context_win32->ipc_window,
GDK_OWNERSHIP_APPLICATION, FALSE,
GDK_POINTER_MOTION_MASK | GDK_BUTTON_RELEASE_MASK,
cursor, GDK_CURRENT_TIME);
G_GNUC_END_IGNORE_DEPRECATIONS;
}
}
static double
ease_out_cubic (double t)
{
double p = t - 1;
return p * p * p + 1;
}
#define ANIM_TIME 500000 /* half a second */
typedef struct _GdkDragAnim GdkDragAnim;
struct _GdkDragAnim {
GdkWin32DragContext *context;
GdkFrameClock *frame_clock;
gint64 start_time;
};
static void
gdk_drag_anim_destroy (GdkDragAnim *anim)
{
g_object_unref (anim->context);
g_slice_free (GdkDragAnim, anim);
}
static gboolean
gdk_drag_anim_timeout (gpointer data)
{
GdkDragAnim *anim = data;
GdkWin32DragContext *context = anim->context;
GdkFrameClock *frame_clock = anim->frame_clock;
gint64 current_time;
double f;
double t;
if (!frame_clock)
return G_SOURCE_REMOVE;
current_time = gdk_frame_clock_get_frame_time (frame_clock);
f = (current_time - anim->start_time) / (double) ANIM_TIME;
if (f >= 1.0)
return G_SOURCE_REMOVE;
t = ease_out_cubic (f);
gdk_window_show (context->drag_window);
gdk_window_move (context->drag_window,
context->last_x + (context->start_x - context->last_x) * t - context->hot_x,
2017-11-25 17:04:48 +00:00
context->last_y + (context->start_y - context->last_y) * t - context->hot_y);
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
gdk_window_set_opacity (context->drag_window, 1.0 - f);
return G_SOURCE_CONTINUE;
}
static void
gdk_win32_drag_context_drop_done (GdkDragContext *context,
gboolean success)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GdkDragAnim *anim;
cairo_surface_t *win_surface;
cairo_surface_t *surface;
cairo_t *cr;
guint id;
GDK_NOTE (DND, g_print ("gdk_drag_context_drop_done: 0x%p %s\n",
context,
success ? "dropped successfully" : "dropped unsuccessfully"));
/* FIXME: This is temporary, until the code is fixed to ensure that
* gdk_drop_finish () is called by GTK.
*/
if (use_ole2_dnd)
{
GdkWin32Selection *sel_win32 = _gdk_win32_selection_get ();
if (success)
sel_win32->dnd_source_state = GDK_WIN32_DND_DROPPED;
else
sel_win32->dnd_source_state = GDK_WIN32_DND_NONE;
}
if (success)
{
gdk_window_hide (win32_context->drag_window);
return;
}
win_surface = _gdk_window_ref_cairo_surface (win32_context->drag_window);
surface = gdk_window_create_similar_surface (win32_context->drag_window,
cairo_surface_get_content (win_surface),
gdk_window_get_width (win32_context->drag_window),
gdk_window_get_height (win32_context->drag_window));
cr = cairo_create (surface);
cairo_set_source_surface (cr, win_surface, 0, 0);
cairo_paint (cr);
cairo_destroy (cr);
cairo_surface_destroy (win_surface);
/*
pattern = cairo_pattern_create_for_surface (surface);
gdk_window_set_background_pattern (win32_context->drag_window, pattern);
cairo_pattern_destroy (pattern);
*/
cairo_surface_destroy (surface);
anim = g_slice_new0 (GdkDragAnim);
g_set_object (&anim->context, win32_context);
anim->frame_clock = gdk_window_get_frame_clock (win32_context->drag_window);
anim->start_time = gdk_frame_clock_get_frame_time (anim->frame_clock);
id = gdk_threads_add_timeout_full (G_PRIORITY_DEFAULT, 17,
gdk_drag_anim_timeout, anim,
(GDestroyNotify) gdk_drag_anim_destroy);
g_source_set_name_by_id (id, "[gtk+] gdk_drag_anim_timeout");
}
static gboolean
drag_context_grab (GdkDragContext *context)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
GdkSeatCapabilities capabilities;
GdkSeat *seat;
GdkCursor *cursor;
if (!context_win32->ipc_window)
return FALSE;
seat = gdk_device_get_seat (gdk_drag_context_get_device (context));
capabilities = GDK_SEAT_CAPABILITY_ALL;
cursor = gdk_drag_get_cursor (context, gdk_drag_context_get_selected_action (context));
g_set_object (&context_win32->cursor, cursor);
if (gdk_seat_grab (seat, context_win32->ipc_window,
capabilities, FALSE,
context_win32->cursor, NULL, NULL, NULL) != GDK_GRAB_SUCCESS)
return FALSE;
g_set_object (&context_win32->grab_seat, seat);
/* TODO: Should be grabbing keys here, to support keynav. SetWindowsHookEx()? */
return TRUE;
}
static void
drag_context_ungrab (GdkDragContext *context)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
if (!context_win32->grab_seat)
return;
gdk_seat_ungrab (context_win32->grab_seat);
g_clear_object (&context_win32->grab_seat);
/* TODO: Should be ungrabbing keys here */
}
static gboolean
gdk_win32_drag_context_manage_dnd (GdkDragContext *context,
GdkWindow *ipc_window,
GdkDragAction actions)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
if (context_win32->ipc_window)
return FALSE;
if (use_ole2_dnd)
context->protocol = GDK_DRAG_PROTO_OLE2;
else
context->protocol = GDK_DRAG_PROTO_LOCAL;
g_set_object (&context_win32->ipc_window, ipc_window);
if (drag_context_grab (context))
{
context_win32->actions = actions;
move_drag_window (context, context_win32->start_x, context_win32->start_y);
return TRUE;
}
else
{
g_clear_object (&context_win32->ipc_window);
return FALSE;
}
}
static void
gdk_win32_drag_context_cancel (GdkDragContext *context,
GdkDragCancelReason reason)
{
const gchar *reason_str = NULL;
switch (reason)
{
case GDK_DRAG_CANCEL_NO_TARGET:
reason_str = "no target";
break;
case GDK_DRAG_CANCEL_USER_CANCELLED:
reason_str = "user cancelled";
break;
case GDK_DRAG_CANCEL_ERROR:
reason_str = "error";
break;
default:
reason_str = "<unknown>";
break;
}
GDK_NOTE (DND, g_print ("gdk_drag_context_cancel: 0x%p %s\n",
context,
reason_str));
drag_context_ungrab (context);
gdk_drag_drop_done (context, FALSE);
}
static void
gdk_win32_drag_context_drop_performed (GdkDragContext *context,
guint32 time_)
{
GDK_NOTE (DND, g_print ("gdk_drag_context_drop_performed: 0x%p %u\n",
context,
time_));
gdk_drag_drop (context, time_);
drag_context_ungrab (context);
}
#define BIG_STEP 20
#define SMALL_STEP 1
static void
gdk_drag_get_current_actions (GdkModifierType state,
gint button,
GdkDragAction actions,
GdkDragAction *suggested_action,
GdkDragAction *possible_actions)
{
*suggested_action = 0;
*possible_actions = 0;
if ((button == GDK_BUTTON_MIDDLE || button == GDK_BUTTON_SECONDARY) && (actions & GDK_ACTION_ASK))
{
*suggested_action = GDK_ACTION_ASK;
*possible_actions = actions;
}
else if (state & (GDK_SHIFT_MASK | GDK_CONTROL_MASK))
{
if ((state & GDK_SHIFT_MASK) && (state & GDK_CONTROL_MASK))
{
if (actions & GDK_ACTION_LINK)
{
*suggested_action = GDK_ACTION_LINK;
*possible_actions = GDK_ACTION_LINK;
}
}
else if (state & GDK_CONTROL_MASK)
{
if (actions & GDK_ACTION_COPY)
{
*suggested_action = GDK_ACTION_COPY;
*possible_actions = GDK_ACTION_COPY;
}
}
else
{
if (actions & GDK_ACTION_MOVE)
{
*suggested_action = GDK_ACTION_MOVE;
*possible_actions = GDK_ACTION_MOVE;
}
}
}
else
{
*possible_actions = actions;
if ((state & (GDK_MOD1_MASK)) && (actions & GDK_ACTION_ASK))
*suggested_action = GDK_ACTION_ASK;
else if (actions & GDK_ACTION_COPY)
*suggested_action = GDK_ACTION_COPY;
else if (actions & GDK_ACTION_MOVE)
*suggested_action = GDK_ACTION_MOVE;
else if (actions & GDK_ACTION_LINK)
*suggested_action = GDK_ACTION_LINK;
}
}
static void
gdk_drag_update (GdkDragContext *context,
gdouble x_root,
gdouble y_root,
GdkModifierType mods,
guint32 evtime)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GdkDragAction action, possible_actions;
GdkWindow *dest_window;
GdkDragProtocol protocol;
gdk_drag_get_current_actions (mods, GDK_BUTTON_PRIMARY, win32_context->actions,
&action, &possible_actions);
gdk_drag_find_window (context,
win32_context->drag_window,
x_root, y_root, &dest_window, &protocol);
gdk_drag_motion (context, dest_window, protocol, x_root, y_root,
action, possible_actions, evtime);
}
static gboolean
gdk_dnd_handle_motion_event (GdkDragContext *context,
const GdkEventMotion *event)
{
GdkModifierType state;
if (!gdk_event_get_state ((GdkEvent *) event, &state))
return FALSE;
GDK_NOTE (DND, g_print ("gdk_dnd_handle_motion_event: 0x%p\n",
context));
gdk_drag_update (context, event->x_root, event->y_root, state,
gdk_event_get_time ((GdkEvent *) event));
return TRUE;
}
static gboolean
gdk_dnd_handle_key_event (GdkDragContext *context,
const GdkEventKey *event)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GdkModifierType state;
GdkWindow *root_window;
GdkDevice *pointer;
gint dx, dy;
GDK_NOTE (DND, g_print ("gdk_dnd_handle_key_event: 0x%p\n",
context));
dx = dy = 0;
state = event->state;
pointer = gdk_device_get_associated_device (gdk_event_get_device ((GdkEvent *) event));
if (event->type == GDK_KEY_PRESS)
{
switch (event->keyval)
{
case GDK_KEY_Escape:
gdk_drag_context_cancel (context, GDK_DRAG_CANCEL_USER_CANCELLED);
return TRUE;
case GDK_KEY_space:
case GDK_KEY_Return:
case GDK_KEY_ISO_Enter:
case GDK_KEY_KP_Enter:
case GDK_KEY_KP_Space:
if ((gdk_drag_context_get_selected_action (context) != 0) &&
(gdk_drag_context_get_dest_window (context) != NULL))
{
g_signal_emit_by_name (context, "drop-performed",
gdk_event_get_time ((GdkEvent *) event));
}
else
gdk_drag_context_cancel (context, GDK_DRAG_CANCEL_NO_TARGET);
return TRUE;
case GDK_KEY_Up:
case GDK_KEY_KP_Up:
dy = (state & GDK_MOD1_MASK) ? -BIG_STEP : -SMALL_STEP;
break;
case GDK_KEY_Down:
case GDK_KEY_KP_Down:
dy = (state & GDK_MOD1_MASK) ? BIG_STEP : SMALL_STEP;
break;
case GDK_KEY_Left:
case GDK_KEY_KP_Left:
dx = (state & GDK_MOD1_MASK) ? -BIG_STEP : -SMALL_STEP;
break;
case GDK_KEY_Right:
case GDK_KEY_KP_Right:
dx = (state & GDK_MOD1_MASK) ? BIG_STEP : SMALL_STEP;
break;
}
}
/* The state is not yet updated in the event, so we need
* to query it here.
*/
_gdk_device_query_state (pointer, NULL, NULL, NULL, NULL, NULL, NULL, &state);
if (dx != 0 || dy != 0)
{
win32_context->last_x += dx;
win32_context->last_y += dy;
gdk_device_warp (pointer, win32_context->last_x, win32_context->last_y);
}
gdk_drag_update (context, win32_context->last_x, win32_context->last_y, state,
gdk_event_get_time ((GdkEvent *) event));
return TRUE;
}
static gboolean
gdk_dnd_handle_grab_broken_event (GdkDragContext *context,
const GdkEventGrabBroken *event)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GDK_NOTE (DND, g_print ("gdk_dnd_handle_grab_broken_event: 0x%p\n",
context));
/* Don't cancel if we break the implicit grab from the initial button_press.
* Also, don't cancel if we re-grab on the widget or on our IPC window, for
* example, when changing the drag cursor.
*/
if (event->implicit ||
event->grab_window == win32_context->drag_window ||
event->grab_window == win32_context->ipc_window)
return FALSE;
if (gdk_event_get_device ((GdkEvent *) event) !=
gdk_drag_context_get_device (context))
return FALSE;
gdk_drag_context_cancel (context, GDK_DRAG_CANCEL_ERROR);
return TRUE;
}
static gboolean
gdk_dnd_handle_button_event (GdkDragContext *context,
const GdkEventButton *event)
{
GDK_NOTE (DND, g_print ("gdk_dnd_handle_button_event: 0x%p\n",
context));
#if 0
/* FIXME: Check the button matches */
if (event->button != win32_context->button)
return FALSE;
#endif
if ((gdk_drag_context_get_selected_action (context) != 0) &&
(gdk_drag_context_get_dest_window (context) != NULL))
{
g_signal_emit_by_name (context, "drop-performed",
gdk_event_get_time ((GdkEvent *) event));
}
else
gdk_drag_context_cancel (context, GDK_DRAG_CANCEL_NO_TARGET);
return TRUE;
}
gboolean
gdk_dnd_handle_drag_status (GdkDragContext *context,
const GdkEventDND *event)
{
GdkWin32DragContext *context_win32 = GDK_WIN32_DRAG_CONTEXT (context);
GdkDragAction action;
GDK_NOTE (DND, g_print ("gdk_dnd_handle_drag_status: 0x%p\n",
context));
if (context != event->context)
return FALSE;
action = gdk_drag_context_get_selected_action (context);
if (action != context_win32->current_action)
{
context_win32->current_action = action;
g_signal_emit_by_name (context, "action-changed", action);
}
return TRUE;
}
static gboolean
gdk_dnd_handle_drop_finished (GdkDragContext *context,
const GdkEventDND *event)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GDK_NOTE (DND, g_print ("gdk_dnd_handle_drop_finihsed: 0x%p\n",
context));
if (context != event->context)
return FALSE;
g_signal_emit_by_name (context, "dnd-finished");
gdk_drag_drop_done (context, !win32_context->drop_failed);
gdk_win32_selection_clear_targets (gdk_display_get_default (),
gdk_win32_drag_context_get_selection (context));
return TRUE;
}
gboolean
gdk_win32_drag_context_handle_event (GdkDragContext *context,
const GdkEvent *event)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
if (!context->is_source)
return FALSE;
if (!win32_context->grab_seat && event->type != GDK_DROP_FINISHED)
return FALSE;
switch (event->type)
{
case GDK_MOTION_NOTIFY:
return gdk_dnd_handle_motion_event (context, &event->motion);
case GDK_BUTTON_RELEASE:
return gdk_dnd_handle_button_event (context, &event->button);
case GDK_KEY_PRESS:
case GDK_KEY_RELEASE:
return gdk_dnd_handle_key_event (context, &event->key);
case GDK_GRAB_BROKEN:
return gdk_dnd_handle_grab_broken_event (context, &event->grab_broken);
case GDK_DRAG_STATUS:
return gdk_dnd_handle_drag_status (context, &event->dnd);
case GDK_DROP_FINISHED:
return gdk_dnd_handle_drop_finished (context, &event->dnd);
default:
break;
}
return FALSE;
}
void
gdk_win32_drag_context_action_changed (GdkDragContext *context,
GdkDragAction action)
{
GdkCursor *cursor;
cursor = gdk_drag_get_cursor (context, action);
gdk_drag_context_set_cursor (context, cursor);
}
static GdkWindow *
gdk_win32_drag_context_get_drag_window (GdkDragContext *context)
{
return GDK_WIN32_DRAG_CONTEXT (context)->drag_window;
}
static void
gdk_win32_drag_context_set_hotspot (GdkDragContext *context,
gint hot_x,
gint hot_y)
{
GdkWin32DragContext *win32_context = GDK_WIN32_DRAG_CONTEXT (context);
GDK_NOTE (DND, g_print ("gdk_drag_context_set_hotspot: 0x%p %d:%d\n",
context,
hot_x, hot_y));
win32_context->hot_x = hot_x;
win32_context->hot_y = hot_y;
if (win32_context->grab_seat)
{
/* DnD is managed, update current position */
move_drag_window (context, win32_context->last_x, win32_context->last_y);
}
}
static void
gdk_win32_drag_context_class_init (GdkWin32DragContextClass *klass)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
GdkDragContextClass *context_class = GDK_DRAG_CONTEXT_CLASS (klass);
object_class->finalize = gdk_win32_drag_context_finalize;
context_class->find_window = gdk_win32_drag_context_find_window;
context_class->drag_status = gdk_win32_drag_context_drag_status;
context_class->drag_motion = gdk_win32_drag_context_drag_motion;
context_class->drag_abort = gdk_win32_drag_context_drag_abort;
context_class->drag_drop = gdk_win32_drag_context_drag_drop;
context_class->drop_reply = gdk_win32_drag_context_drop_reply;
context_class->drop_finish = gdk_win32_drag_context_drop_finish;
context_class->drop_status = gdk_win32_drag_context_drop_status;
context_class->get_selection = gdk_win32_drag_context_get_selection;
W32: Massive W32 DnD fix Massive changes to OLE2 DnD protocol, which was completely broken before: * Keep GdkDragContext and OLE2 objects separate (don't ref/unref them together, don't necessarily create them together). * Keep IDataObject formats in the object itself, not in a global variable. * Fix getdata() to look up the request target in its format list, not in the global hash table * Create target GdkDragContext on each drag_enter, destroy it on drag_leave, whereas IDropTarget is created when a window becomes a drag destination and is re-used indefinitely. * Query the source IDataObject for its supported types, cache them in the target (!) context. This is how GTK+ works, honestly. * Remember current_src_object when we initiate a drag, to be able to detect later on that the data object is ours and use a shortcut when querying targets * Make sure GDK_DRAG_MOTION is only sent when something changes * Support GTK drag cursors * Ensure that exotic GTK clipboard formats are registered (but try to avoid registering formats that can't be used between applications). * Don't enumerate internal formats * Ensure that DnD indicator window can't accept drags or receive any kind of input (use WS_EX_TRANSPARENT). * Remove unneeded indentation in _gdk_win32_dnd_do_dragdrop() * Fix indentation in gdk_win32_drag_context_drop_finish() * Remove obsolete comments in _gdk_win32_window_register_dnd() * Check for DnD in progress when processing WM_KILLFOCUS, don't emit a grab break event in such cases (this allows alt-tabbing while DnD is in progress, though there may be lingering issues with focus after dropping...) * Support Shell ID List -> text/uri-list conversion, now it's possible to drop files (dragged from Explorer) on GTK+ applications * Explicitly use RegisterClipboardFormatA() when we know that the string is not in unicode. Otherwise explicitly use RegisterClipboardFormatW() with a UTF8->UTF16 converted string * Fix _gdk_win32_display_get_selection_owner() to correctly bail when selection owner HWND is NULL (looking up GdkWindow for NULL HWND always succeeds and returns the root window - not the intended effect) * More logging * Send DROP_FINISHED event after DnD loop ends * Send STATUS event on feedback * Move GetKeyboardState() and related code into _gdk_win32_window_drag_begin(), so that it's closer to the point where last_pt and start_pt are set * Use & 0x80 to check for the key being pressed. Windows will set low-order bit to 1 for all mouse buttons to indicate that they are toggled, so simply checking for the value not being 0 is not enough anymore. This is probably a new thing in modern W32 that didn't exist before (OLE2 DnD code is old). * Fixed (hopefully) and simplified HiDPI parts of the code. Also adds managed DnD implementation for W32 GDK backend (for both OLE2 and LOCAL protocols). Mostly a copy of the X11 backend code, but there are some minor differences: * doesn't use drag_window field in GdkDragContext, uses the one in GdkWin32DragContext exclusively * subtracts hotspot offset from the window coordinates when showing the dragback animation * tries to consistently support scaling and caches the scale in the context * Some keynav code is removed (places where grabbing/ungrabbing should happen is marked with TODOs), and the rest is probably inert. Also significantly changes the way selection (and clipboard) is handled (as MSDN rightly notes, the handling for DnD and Clipboard formats is virtually the same, so it makes sense to handle both with the same code): * Don't spam GDK_OWNER_CHANGE, send them only when owner actually changes * Open clipboard when our process becomes the clipboard owner (we are doing it anyway, to empty the clipboard and *become* the owner), and then don't close it until a scheduled selection request event (with TARGETS target) is received. Process that event by announcing all of our supported formats (by that time add_targets() should have been called up the stack, thus the formats are known; just in case, add_targets() will also schedule a selection request, if one isn't scheduled already, so that late-coming formats can still be announced). * Allow clipboard opening for selection_convert() to be delayed if it fails initially. * The last two points above should fix all the bugs about GTK+ rising too much ruckus over OpenClipboard() failures, as owner change *is allowed* to fail (though not all callers currently handle that case), and selection_convert() is asynchronous to begin with. Still, this is somewhat risky, as there's a possibility that the code will work in unexpected ways and the clipboard will remain open. There's now logging to track the clipboard being opened and closed, and a number of failsafes that try to ensure that it isn't kept open for no reason. * Added copious notes on the way clipboard works on X11, Windows and GDK-W32, also removed old comments in DnD implementation, replaced some of them with the new ones * A lot of crufty module-global variables are stuffed into a singleton object, GdkWin32Selection. It's technically possible to make it a sub-object of the Display object (the way Wayland backend does), but since Display object on W32 is a singleton anyway... why bother? * Fixed the send_change_events() a bit (was slightly broken in one of the previous iterations) * Ensure that there's no confusion between selection conversion (an artifact term from X11) and selection transmutation (changing the data to be W32-compatible) * Put all the transmutation code and format-target-matching code into gdkselection-win32.c, now this code isn't spread across multiple files. * Consequently, moved some code away from gdkproperty-win32.c and gdkdnd-win32.c * Extensive format transmutation checks for OLE2 DnD and clipboard. We now keep track of which format mappings are for transmutations, and which aren't (for example, when formats are passed as-is, or when a registered name is just an alias) * Put transmutation code into separate functions * Ensure that drop target keeps a format->target map for supported formats, this is useful when selection_convert() is called, as it only receives a single target and no hints on the format from which the data should be transmuted into this target. * Add clear_targets() on W32, to de called by GTK * Use g_set_object() instead of g_ref_object() where it is allowed. * Fix indentation (and convert tabs to spaces), remove unused variables (This commit is cherry-picked from the gtk-3-22 branch) https://bugzilla.gnome.org/show_bug.cgi?id=786509
2017-08-19 12:06:27 +00:00
context_class->get_drag_window = gdk_win32_drag_context_get_drag_window;
context_class->set_hotspot = gdk_win32_drag_context_set_hotspot;
context_class->drop_done = gdk_win32_drag_context_drop_done;
context_class->manage_dnd = gdk_win32_drag_context_manage_dnd;
context_class->set_cursor = gdk_win32_drag_context_set_cursor;
context_class->cancel = gdk_win32_drag_context_cancel;
context_class->drop_performed = gdk_win32_drag_context_drop_performed;
context_class->handle_event = gdk_win32_drag_context_handle_event;
context_class->action_changed = gdk_win32_drag_context_action_changed;
}