If buffer age is undefined and the updated area is not the whole
window then we use bit-blits instead of swap-buffers to end the
frame.
This allows us to not repaint the entire window unnecessarily if
buffer_age is not supported, like e.g. with DRI2.
This moves the GDK_ALWAYS_USE_GL env var to GDK_GL=always.
It also changes GDK_DEBUG=nogl to GDK_GL=disable, as GDK_DEBUG
is really only about debug loggin.
It also adds some completely new flags:
software-draw-gl:
Always use software fallback for drawing gl content to a cairo_t.
This disables the fastpaths that exist for drawing directly to
a window and instead reads back the pixels into a cairo image
surface.
software-draw-surface:
Always use software fallback for drawing cairo surfaces onto a
gl-using window. This disables e.g. texture-from-pixmap on X11.
software-draw:
Enables both the above.
The Mir backend was checking for button mask changes to generate the appropriate
GDK event. When Mir generates a touch event it has no button mask. In this case
we'll just generate a primary button event.
This was unnecessarily creating a framebuffer in the texture case,
and it was not properly setting up a framebuffer with the texture
as source in the software fallback w/ texture source case.
Commit afd9709aff made us keep impl window
cairo surfaces around across changes of window scale. But the
window scale setter forgot to update the size and scale of the
surface. The effect of this was that toggling the window scale
from 1 to 2 in the inspector was not causing the window to draw
at twice the size, although the X window was made twice as big,
and input was scaled too. Fix this by updating the surface when
the window scale changes.
We need to use this in the code path where we make the context
non-current during destroy, because at that point the window
could be destroyed and gdk_window_get_display() would return
NULL.
This moves the code related to the frame sync code into
the is_attached check, which means we don't have to ever
run this when making non-window-paint contexts current.
This is a minior speed thing, but the main advantage
is that it makes making a non-paint context current
threadsafe.
This is not really needed. The gl context is totally tied to the
window it is created from by virtue of sharing the context with the
paint context of that window and that context always has the visual
of the window (which we already can get).
Also, all user visible contexts are essentially offscreen contexts, so
a visual doesn't make sense for them. They only use FBOs which have
whatever format that the users sets up.
To properly support multithreaded use we use a global GPrivate
to track the current context. Since we also don't need to track
the current context on the display we move gdk_display_destroy_gl_context
to GdkGLContext::discard.
We used to have a weak ref to the cairo surface and it was keep
alive by the references in the normal windows, but that reference
was removed by d48adf9cee, causing
us to constantly create and destroy the surface.
https://bugzilla.gnome.org/show_bug.cgi?id=738648
We want to create windows with the default visuals such that we then
have the right visual for GLX when we want to create the paint GL
context for the window.
For instance, (in bug 738670) the default rgba visual we picked for the
NVidia driver had an alpha size of 0 which gave us a BadMatch when later
trying to initialize a gl context on it with a alpha FBConfig.
Instead of just picking what the Xserver likes for the default, and just
picking the first rgba visual we now actually call into GLX to pick
an appropriate visual.
The visuals are typically sorted by some sort of "most useful first"
order. And picking the last one is likely to give us the weirdest
matching glx visual.
It is not possible to successfully build GTK+ on OS X 10.6 and below
since NSFullScreenWindowMask is only available starting with 10.7. Add
ifdef guards around setStyleMask: in order to allow it to build on
earlier OS X releases.
https://bugzilla.gnome.org/show_bug.cgi?id=737561
Commits 314b6abbe8 and eb9223c008 were ignoring
the fact that the code where found is set to 1 was modifying
col - which was an ok thing to do when that part of the code
was still breaking out of the loop, but it is no longer doing
that (since 2003 !). Fix things up by storing the final col
value in a separate variable and using that after the loop.
https://bugzilla.gnome.org/show_bug.cgi?id=738886
In various places, the broadway backend was just using
the default display and assumed that it is the broadway
display. That may not be the case in a multi-backend world,
so instead iterate over all displays and use the first
broadway display - still not perfect, but enough to survive
for now.
The current implementation of this script generate headers with \x-escaped
strings that can become too long (> 65535 characters) for Visual Studio
to consume, hence the build of broadwayd would break on Visual Studio.
This changes the script to instead format the string as an array of hex
characters, not unlike what GResource does, so that builds can continue as
normal on Visual Studio builds as well.
https://bugzilla.gnome.org/show_bug.cgi?id=739001
wayland doesn't strictly follow the XDG_RUNTIME_DIR spec by falling back
to another directory in case the runtime dir is not properly set.
When this variable is unset, wayland will log an error to us, which we
treat as fatal, aborting the entire program.
Skip ourselves all the trouble and don't try to bring up the wayland
backend when we know it will fail in this way.
https://bugzilla.gnome.org/show_bug.cgi?id=738873
This is mostly useful for fallback testing.
I suppose if people want finer grained GL ability testing, they can use
Mesa environment variables to tune things.
Its not really reasonable to handle failures to make_current, it
basically only happens if you pass invalid arguments to it, and
thats not something we trap on similar things on the X drawing side.
If GL is not supported that should be handled by the context creation
failing, and anything going wrong after that is essentially a critical
(or an async X error).
We make user facing gl contexts not attached to a surface if possible,
or attached to dummy surfaces. This means nothing can accidentally
read/write to the toplevel back buffer.
This adds the new type GdkGLContext that wraps an OpenGL context for a
particular native window. It also adds support for the gdk paint
machinery to use OpenGL to draw everything. As soon as anyone creates
a GL context for a native window we create a "paint context" for that
GdkWindow and switch to using GL for painting it.
This commit contains only an implementation for X11 (using GLX).
The way painting works is that all client gl contexts draw into
offscreen buffers rather than directly to the back buffer, and the
way something gets onto the window is by using gdk_cairo_draw_from_gl()
to draw part of that buffer onto the draw cairo context.
As a fallback (if we're doing redirected drawing or some effect like a
cairo_push_group()) we read back the gl buffer into memory and composite
using cairo. This means that GL rendering works in all cases, including
rendering to a PDF. However, this is not particularly fast.
In the *typical* case, where we're drawing directly to the window in
the regular paint loop we hit the fast path. The fast path uses opengl
to draw the buffer to the window back buffer, either by blitting or
texturing. Then we track the region that was drawn, and when the draw
ends we paint the normal cairo surface to the window (using
texture-from-pixmap in the X11 case, or texture from cairo image
otherwise) in the regions where there is no gl painted.
There are some complexities wrt layering of gl and cairo areas though:
* We track via gdk_window_mark_paint_from_clip() whenever gtk is
painting over a region we previously rendered with opengl
(flushed_region). This area (needs_blend_region) is blended
rather than copied at the end of the frame.
* If we're drawing a gl texture with alpha we first copy the current
cairo_surface inside the target region to the back buffer before
we blend over it.
These two operations allow us full stacking of transparent gl and cairo
regions.
This is a new function that gets called every time we're drawing
some area in the Gtk paint machinery. It is a no-op right now, but
it will be required later to keep track of what areas which
we previously rendered with GL was overwritten with cairo contents.
First of all we track the current update area during an
update in window->active_update_area. This will be used later
in end_paint to know the damaged area.
Secondly we keep track of old update areas for the last 2
frames. This will later allow us to reuse old framebuffer
contents in double or tripple buffer setups, only painting
what has changed since then.
Before 5e325c4, the default BitGravity was NorthWestGravity.
When static gravities were removed in 5e325c4, the BitGravity regressed
to the X11 default, Forget. Forget causes giant graphical glitches and
black flashes when resizing, especially in some environments that aren't
synchronized to a paint clock yet, like XWayland.
I'm assuming that the author assumed that the default of BitGravity was
NorthWestGravity, which is the default of WinGravity. Just go ahead and
fix this regression to make resizing look smooth again.
Currently writing wl_data_offer data into the fd is 1) synchronous, which
is noticeable when transferring large amounts of data, and 2) buggy, write()
error checking is done on the accumulator, breaking both the written data
accounting and error checking itself.
Fix both by making writes asynchonous through GOutputStream, the operation
is spun off and either finished, or cancelled if new data is stored in the
selection while the transfer is active.
The documentation explicitly states that 0 is an allowed value for using
the same scale as the window. This 0 value is also explicitly checked
down in the call chain and handled.
There is no need for a critical warning just because somebody
asked for a property that is not meaningful for the device.
Just document it as not useful for keyboard devices.
Parent is guaranteed to not be NULL. It can only ever be NULL for root
windows and root windows cannot be created with gdk_window_new() and
gdk_window_ensure_native() will exit early because they already are
native.
Also, both functions would crash a few lines below where parent gets
dereferenced.
Remove checks for NULL before g_free() and g_clear_object().
Merge check for NULL, freeing of pointer and its setting
to NULL by g_clear_pointer().
https://bugzilla.gnome.org/show_bug.cgi?id=733157
The warning may have had some value at some point, but if
people uninstall large icons just to make the warning go
away, it does more harm than good. So just remove it.
If we have a fullscreen window that covers a monitor, desktop
chrome is not relevant for placing of menus and other popups.
Therefore, return the full monitor geometry instead of the
workarea in this case.
https://bugzilla.gnome.org/show_bug.cgi?id=737251
If !owner_events, the pointer window has been usually set to NULL if
the pointer fell outside the grabbing widget, but it was not being
checked that the pointer_window is actually a child of the grab
window, in which case it should be obtained as if ungrabbed.
https://bugzilla.gnome.org/show_bug.cgi?id=735749
The list of devices was being scanned over incorrectly, causing us to
never actually fetch the keymap from the keyboard, as the keyboard was
the second device in the list, not the first.
This causes us to create a new temporary keymap every time, which is
quite expensive, because it involves parsing the entire XKB
file. Scanning the list correctly will cause us to use the XKB rules
file that was passed to us.
When recursing the update area down into native subwindows we forgot
to apply the native window position. This caused us to repaint the
wrong thing in certain cases. I noticed this when playing with the
wip/gdk-gl branch, because it was triggering unnecessary repaints
of the (native window) gl widgets.
A surface may be hidden when a frame is already scheduled, which may cause
crashes on on_frame_clock_after_paint() when calling commit() on a NULL
surface. To fix this, ensure commit_pending is also set to FALSE when the
surface is gone.
https://bugzilla.gnome.org/show_bug.cgi?id=735226
Only static cursors are supported in gdk_device_grab() so far. Obey the
cursor that gdk_device_grab() specifies, which may be different to
the pointer window one. As soon as the grab is gone, the pointer window
cursor will be restored as usual.
https://bugzilla.gnome.org/show_bug.cgi?id=735831
On DnD, pointer_handle_leave may be triggered without the pointer actually
leaving the window, and pointer_handle_enter() happening after intra-window
DnD won't actually manage to update the cursor (it does nothing directly,
and to the upper layers the cursor is still the same and consistent, so no
attempt will happen).
To fix this, keep the pointer cursor on leave, and ensure it is updated
on enter. The pointer cursor will be updated to any current new one through
the enter/motion events generated if it needs be.
https://bugzilla.gnome.org/show_bug.cgi?id=735831
cairo_surface_destroy() is called after the buffer is released, for every
wl_buffer. Windows usually reference their cairo surface before rendering,
so that extra reference is consumed after the buffer is released, so do
the same with cursor surfaces and add an extra reference whenever a cursor
surface change is about to be scheduled.
Otherwise, the GdkWaylandCursor is left with an invalid cairo_surface_t,
which causes crashes the next time it is used.
https://bugzilla.gnome.org/show_bug.cgi?id=735830
On wayland the DnD surface must be created early when starting the drag
operation, so offer API for GTK+ to get the GdkWindow used as a DnD
surface on the drag operation.
https://bugzilla.gnome.org/show_bug.cgi?id=697855
The wl_data_source is retrieved from the selection object for the DnD
selection, and used to initiate a drag. When the drag is finished, a
button release or touch end event is synthesized to finish the DnD
operation after the compositor grab is gone.
https://bugzilla.gnome.org/show_bug.cgi?id=697855
The wayland specific clipboard functions have been replaced by something
more similar to the hooking the win32 backend does, which allows for just
using the default GtkClipboard code in GTK+. As a consequence, the
wayland-specific GtkClipboard implementation is now gone.
https://bugzilla.gnome.org/show_bug.cgi?id=697855
This has been made to work similarly to X11, requests for the data device
contents are notified through GDK_SELECTION_REQUEST events, the data stored
in the GDK_SELECTION property as a reaction to that event is then stored
into the wayland selection implementation, and written to the fd when
requested/available.
https://bugzilla.gnome.org/show_bug.cgi?id=697855
This implementation makes the destination side of selections work
similarly to X11's, gdk_selection_convert() triggers data fetching,
which is notified through GDK_SELECTION_NOTIFY events on arrival,
the buffered data is then available through gdk_selection_property_get().
https://bugzilla.gnome.org/show_bug.cgi?id=697855
Subsurface position is deemed part of the state of the parent surface, so
ensure wl_surface_commit() happens on the parent surface if none is
scheduled, so the repositioning takes place.
Since GLib ≥ 2.41, attempting to release an unlocked mutex will abort(),
as it happens on most systems already.
Given the lack of proper documentation on how to use GDK with threads,
there is code in the wild that does:
gdk_threads_init ();
gdk_init ();
...
gtk_main ();
instead of the idiomatically correct:
gdk_threads_init ();
gdk_threads_enter ();
gtk_init ();
...
gtk_main ();
...
gdk_threads_leave ();
Which means that gtk_main() will try to release the GDK lock, and thus
trigger an error from GLib.
we cannot really fix all the wrong code everywhere, and since it does
not cost us anything, we can work around the issue inside GDK itself, by
trying to acquire the GDK lock inside gdk_threads_leave() with
trylock().
https://bugzilla.gnome.org/show_bug.cgi?id=735428
The latest implicit grab serial is used in order to start the compositor
grab, If it belongs to a touch event, remove that touch sequence, as the
rest of the sequence will be gone for good.
This avoids stale sequences (and implicit grab info) after a window is
moved/resized.
https://bugzilla.gnome.org/show_bug.cgi?id=731380
_gdk_wayland_device_get_button_press_serial() has been replaced by
_gdk_wayland_device_get_implicit_grab_serial(), which takes a touch/pointer
event and figures out the relevant serial, and
_gdk_wayland_device_get_last_implicit_grab_serial() which returns
the most recent serial.
The button press serial was currently used when operating popping up
xdg_shell/surface popups and window menus, so this is now touch aware, of
some sort.
https://bugzilla.gnome.org/show_bug.cgi?id=734374
If the compositor sends a keymap that fails on "compilation",
xkb_keymap_new_from_string() returns NULL, which makes xkb_state_new()
crash when assuming there is a keymap.
In these cases, gdk must remain with a xkb_state to handle modifiers/keys
properly, so warn about the invalid keymap string, and keep the previous
keymap (currently initialized to "us")
https://bugzilla.gnome.org/show_bug.cgi?id=735389
To all effects each window has its own "root" coordinates system, so set
toplevels at 0,0 in that coordinate system, so root coordinate calculations
are locally right.
https://bugzilla.gnome.org/show_bug.cgi?id=729215
This implements native fullscreen mode on OS X >= 10.7. This patch
adds tracking of the fullscreen mode if externally modified, as well
as toggling the native fullscreen mode of the window as needed.
At the present time broadway listens only for TCP/IP incoming
display connections. This patch implements the support for listening
on unix domain sockets too, adding the broadway_server_on_unix_socket_new()
constructor and the commandline option --unixsocket [path] to broadwayd.
https://bugzilla.gnome.org/show_bug.cgi?id=734420
X11 backend doesn't, and for good reason - main code body does not check
that the window it sets opacity for is, in fact, toplevel.
Just silently fail to do anything for non-toplevel windows.
https://bugzilla.gnome.org/show_bug.cgi?id=733769
Support environment variable GDK_WIN32_FONT_RESOLUTION that can be set to
a desired dpi (72, 96, 130, etc) to override system settings. Useful for
debugging, since changing system font scaling requires the user to log off
and log on again.
https://bugzilla.gnome.org/show_bug.cgi?id=734038
Use (cairo) input shape of the window to check whether a point is inside or not
inside the window.
If it is, let the default window procedure do its thing (which seems to be
working all right in all known cases).
If it isn't, override the default window procedure and tell WM what we think.
Don't do any of the above if the window has CSD-incompatible styles (WS_BORDER
or WS_THICKFRAME).
This is a crude kind of substitute for window input shape support (which W32
does not seem to have). Still probably enough to be positive about input shapes
support.
https://bugzilla.gnome.org/show_bug.cgi?id=733679
Delay the keyboard settings creation until we're delivering the key
press. This means we don't have to create the settings for a server that
sends us repeat information.
This function currently calls gdk_win32_window_shape_combine_region(),
which is wrong, because it leads to SetWindowRgn() being called with
non-NULL region, which makes W32 disable theming (particularly - decoration
theming), which makes decorations revert back to old GDI-drawn Windows 2000
variant, which looks out of place and interacts *badly* with alpha channel
(because GDI).
https://bugzilla.gnome.org/show_bug.cgi?id=733671
xdg-shell has moved on and replaced set_margin with set_window_geometry.
To properly support set_window_geometry requires a full rewrite of how
we've been dealing with toplevel windows for now, so just don't set any
margin until we can have a proper toplevel window abstraction in GTK+.
Some windows, like GtkWindow and some other apps, set a transparent
background. The guarantee for begin_paint_region is that there should
be a full clear to the specified background color, not a composite
against what was there before.
This fixes repaint artifacts in Wayland and Weston in a better way than
76922c169f.
gdk_x11_display_set_window_scale() affects the interpretation of the
Xft/DPI XSETTING - it is substituted inside GDK with the value of
Gdk/UnscaledDPI xsetting. However, this change is not propagated to
GTK+ and from GTK+ back to gdk_screen_set_resolution() until the
main loop is run.
Fix this by handling the screen resolution directly in gdk/x11.
This requires duplication of code between GDK and GTK+ since we still
have to handle DPI in GTK+ in the case that GdkSettings:gtk-xft-dpi
is set by the application.
https://bugzilla.gnome.org/show_bug.cgi?id=733076
The way that GtkTextView et al pop up their context menu is to first
query to see if the clipboard has some text, and if so, enable the Paste
menu item. But since the Wayland backend hasn't had the greatest
selection and clipboard code, the callback for the clipboard got dropped
on the floor.
Add some simple code to respond to the TARGETS selection.
This makes right-clicking on a GtkTextView work fine.
The resize grip code in GTK+ likes to call gdk_window_raise a lot. The
unfortunate side effect of gdk_window_raise is that it queues an
invalidation on the entire window, even if it's already the topmost
child.
Add a return value to gdk_window_raise_internal, and only queue the
invalidation if the raise had an effect.
Otherwise, a user that calls gdk_window_resize (window, 0, 0); over and
over won't properly fizzle out, and will queue a redraw. Clipped, but
still. These redraws can be chatty on some platforms like Wayland, and
there's no good reason to not avoid them.
This was the case for resize grips.
This reverts commit b875572f2a.
Apps like Abiword, gnumeric and gnome-chess, and toolkits like
ClutterGTK were all using this for various purposes, and this made them
break. Bring back this feature for now.
It still won't work under Wayland.
Apps had quite a bit of difficulty getting used to the new rules.
While we weren't expecting anything too deadly, it seemed that
gnumeric and Abiword both used gdk_cairo_create like this.
If a window both has an impl and its impl_window is of type offscreen,
that must mean that it is the offscreen window, and the impl window is
itself. We can reduce the indirection here and make it more clear.
Since the Win32 code never actually called InvalidateRgn or used the
Win32 update area at all, that meant the only thing that could possibly
invalidate the window was the Win32 window manager as part of scrolling
or resizing, which would also send it a WM_PAINT message.
But the WM_PAINT handling called BeginPaint / EndPaint, which clears the
update area completely! We also draw out-of-band, not directly when
handling WM_PAINT, so there's no way that the update area inside the
Win32 WM would match our local one.
There is no possible way that this queue_antiexpose implementation could
do anything. Remove it.
We removed the parameter from callers and from the implementation, but
not from the signature up top. I didn't notice because the branch I was
working on removed the signature entirely.
This code is only called with the current paint region as its argument.
Instead of having to copy it and do a no-op intersect against itself,
just use the current paint directly.
cairo_surface_create_for_rectangle takes a ref on the parent surface,
so we need to drop ours.
Rename get_window_surface to ref_window_surface to make the code more
clear and to stop this error from happening again.
Previously, each begin_paint_region added to a stack of current paints,
and when end_paint was called, the paint was popped off of the stack and
the surface was composited into the parent paint surface.
However, the code was broken in the case of a backend like Wayland which
didn't keep track of nested calls and simply wiped and returned the
native impl backing surface every time.
Since this feature is flat out unused by GTK+ and we don't want to
really support tricksy things like these for other clients, just remove
the feature. If somebody does call begin_paint_region more than once,
warn and return without doing anything.
Traditionally, the way painting was done in GTK+ was with the
"expose-event" handler, where you'd use GDK methods to do drawing on
your surface. In GTK+ 2.24, we added cairo support with gdk_cairo_create,
so you could paint your graphics with cairo.
Since then, we've added client-side windows, double buffering, the paint
clock, and various other enhancements, and the modern way to do drawing
is to connect to the "draw" signal on GtkWidget, which hands you a
cairo_t. To do double-buffering, the cairo_t we hand you is actually on
a secret surface, not the actual backing store of the window, and when
the draw handler completes we blit it into the main backing store
atomically.
The code to do this is with the APIs gdk_window_begin_paint_region,
which creates the temporary surface, and gdk_window_end_paint which
blits it back into the backing store. GTK+'s implementation of the
"draw" signal uses these APIs.
We've always sort-of supported people calling gdk_cairo_create
"outside" of a begin_paint / end_paint like old times, but then you're
not getting the benefit of double-buffering, and it's harder for GDK to
optimize.
Additionally, newer backends like Mir and Wayland can't actually support
this model, since they're based on double-buffering and swapping buffers
at various points in time. If we hand you a random cairo_t, we have no
idea when is a good time to swap.
Remove support for this.
This is technically a GDK API break: a warning is added in cases where
gdk_cairo_create is called outside of a paint cycle, and the returned
surface is a dummy that won't ever be composited back onto the main
surface. Testing with complex applications like Ardour didn't produce
any warnings.
gtk_widget_set_double_buffered is now deprecated, and we don't support
non-double-buffered widgets. This means that under normal circumstances,
paints are never outside of a begin_paint / end_paint sequence, which
natively-double-buffered backends like Wayland can't possibly support.
Weston releases buffers almost immediately after they're done, which
means that GTK+ doesn't use a temporary surface and instead paints
directly onto the SHM backing store that Weston will use.
Normally, after painting to the temporary surface, GTK+ *replaces*
the existing backing surface with CAIRO_OPERATOR_SOURCE. However,
if we immediately paint to the backing surface, it might have junk
from the last paint in it. So clear out the backing surface whenever
somebody calls begin_paint_region().
Maybe we should just always use the temporary surface like the X11
codepath, since that prevents us from having to do weird things like
this, but oh well.
wl_surfaces can't switch roles, so destroying the xdg_surface but not
the wl_surface means that we could get an error when trying to re-map
the surface.
We could fix this by not destroying the xdg resource and only do it at
finalization time, but it's just as easy to just create a new wl_surface.
Since the xdg roles are a special case of the surface, some compositors
like Weston destroy them automatically when the wl_surface is destroyed.
Thus, we need to destroy these first.
