An xdg-popup requires a serial that the compositor will compare against
its own serial and will dismiss the popup if it doesn't match.
gtk+ uses either a pointer or touch serial for its helper function
_gdk_wayland_seat_get_last_implicit_grab_serial() but if the menu is
triggered before the user has had any pointer or touch interaction with
the client, using a keyboard shortcut, there is neither pointer nor
touch serial available, and gtk+ will use 0 as the default.
As a result, the compositor will instantly dismiss the xdg-popup. In
this case, gtk+ should use the keyboard serial instead.
Track keyboard serial as well and use the keyboard serial as the value
if there is no newer pointer or touch serial available.
https://bugzilla.gnome.org/show_bug.cgi?id=768017
At the time of move_to_rect() is called, not all state may have been set
up on the impl gdk window, causing the position to sometimes be
slightly offset due to drap shadow margins. For now, work around this
by postponing the processing of the move_to_rect() parameters until
showing, when its more likely that all state (such as shadow margin)
has been set correctly.
https://bugzilla.gnome.org/show_bug.cgi?id=769402
The Wayland backend manages a set of fake root coordinate spaces, where
each non-relative positioned toplevel (i.e. not popups, popovers,
tooltips etc) make up the basis of separate fake root coordinate spaces.
This means that the Wayland backend doesn't have the abilitiy get a
proper root coordinate when querying on a non-toplevel GdkWindow. To
avoid this issue, first find the toplevel, while translating the anchor
rect coordinates so that they are in the toplevel window coordinate
space. Then use this toplevel to translate the coordinates to root
window coordinate space.
https://bugzilla.gnome.org/show_bug.cgi?id=769402
The position of each transient-of will be in fake-root coordinate
space; thus we should not accumulate all the positions making it an
offset; each window is already in fake root coordinate space.
https://bugzilla.gnome.org/show_bug.cgi?id=769402
When using the set transient-for as a popup parent, fetch the effective
toplevel instead, otherwise we will position against the wrong
coordinate.
https://bugzilla.gnome.org/show_bug.cgi?id=769402
The order in which tooltips are created, drawn, shown and then positioned,
always requires repositioning the surface. The tooltip window type only has
limited capability to do so. An alternative could be to use bufferstreams.
https://bugzilla.gnome.org/show_bug.cgi?id=768138
When a dialog is created, the mir event source is already executed on the
call stack. So without the recurse flag it will not be run in the main loop
used for the dialog.
https://bugzilla.gnome.org/show_bug.cgi?id=768138
The event will be sent when the surface becomes visible on an output. With this change the GdkMirWindowImpl keeps track of the scale value and sends a configure event on change.
Signed-off-by: Andreas Pokorny <andreas.pokorny@canonical.com>
The event code could potentially dereference pointer_info if the
invariant that ENTER_NOTIFY and LEAVE_NOTIFY events are only emitted on
devices which have pointers is violated elsewhere.
Found with scan-build.
https://bugzilla.gnome.org/show_bug.cgi?id=712760
Pick the W32 API for possible deadkey+<something> combinations
and prefer these to other sources of deadkey combos.
Specifically, if W32 API supports at least one combo for a particular
deadkey, only use that data and do not attempt to do other, unsupported
combinations, even if they make sense otherwise.
This is needed to, for example, correctly support US-International
keyboard layout, which produces a combined character for <' + a>
combo, but not for <' + s>, for example.
This is achieved by stashing all the deadkeys that we find in
an array, then doing extra loop through all virtual key codes and
trying to combine them with each of these deadkeys. Any combinations
that produce a single character are cached for later use.
In GTK Simple IM context, call a new GDK W32 function to do a lookup
on that cached combination table early on, among the "special cases"
(which are now partially obsolete).
A limitation of this code is that combinations with more than
one deadkey are not supported, except for combinations that consist
entirely of 2 known deadkeys. The upshot is that lookups should
be relatively fast, as deadkey array stays small and the combination
tree stays shallow.
Note that the use of ToUnicodeEx() seems suboptimal, as it should
be possible to just load a keyboard library (KBD*.DLL) manually
and obtain and use its key table directly. However, that is much more
complicated and would result in a significant rewrite of gdkkeys-win32.
The code from this commit, though hacky, is a direct addition to
existing code and should cover vast majority of the use-cases.
https://bugzilla.gnome.org/show_bug.cgi?id=569581
This changes the group/level semantic.
Previously W32 backend used "group 0/1" to denote "AltGr OFF/ON"
and "level 0/1" to denote "Shift is OFF/ON".
Now "group" means "keyboard layout" and there can be up to 255 groups,
while AltGr and Shift are combined into a single level enum that
takes values between 0 and 4.
Unlike X, W32 doesn't do effective group overriding, meaning that
it will never tell the caller that a different group was actually
used (even for universal keys, such as Enter), because key symbol
table is completely fabricated and there's no point in trying to
save a few of kilobytes of RAM by not duplicating universal key
records for all groups.
Also contains many whitespace changes (tab elimination, fixed
indentation) and cleanup (axed a few global variables, these are
now accessed via the default keymap).
https://bugzilla.gnome.org/show_bug.cgi?id=768722
This matches the behaviour of Mutter, Metacity and traditional X11
window managers on the window manager side, and is what we want
for at least gnome-terminal. I can't think of any reason why we'd
want incremental resize in any other tiled window.
Signed-off-by: Simon McVittie <smcv@debian.org>
Bug: https://bugzilla.gnome.org/show_bug.cgi?id=760944https://bugzilla.gnome.org/show_bug.cgi?id=755947
Only update to using v2 headers/structs. The incompatible changes
to tool events are dealt with in the next commit. Pads aren't handled
in this commit either.
The sizes passed are in device pixels and do not take into account the
scaling factor of the window itself. We cannot change the semantics of
the function, so let's at least add a warning for this trap door.
On X11, device_query_state() uses XIQueryPointer() which will return a
child window only if the pointer is within an actual child of the given
window.
Wayland backend would return the pointer->focus window independently of
the given window, but that breaks the logic in get_device_state() and
later in gdk_window_get_device_position_double() because the window is
searched based on coordinates from another window without sibling
relationship, breaking gtkmenu sub-menus further down the line.
Fix the Wayland backend to mimic X11's XIQueryPointer() to return a
child only if really a child of the given window.
That's the most sensible thing to do to fix the issue, but the API here
seems to be modeled after the X11 implementation and the description of
gdk_window_get_device_position_double() is not entirely accurate.
https://bugzilla.gnome.org/show_bug.cgi?id=768016
This has most notably impact in selection buffers, because those were
shared across all selection atoms. This turned out wrong on 2 situations:
- Because the selection atom was set at SelectionBuffer creation time, the
GDK_SELECTION_NOTIFY events generated will have unexpected info if the
buffer is attempted to be reused for another selection.
- Anytime different selections imply different stored content for the same
target.
This is better separated into per-selection buffers, so it's not possible
to get collisions if a same target is used across different selections.
https://bugzilla.gnome.org/show_bug.cgi?id=768177
The sanitize_utf8() function has been copied from X11 so both
backends behave the same. This allows interaction with older clients
(mainly through Xwayland, and the STRING selection target) that
request non-utf8 text.
https://bugzilla.gnome.org/show_bug.cgi?id=768082
Debian stable currently ships with a 3.16 kernel, so
it doesn't have memfd available.
This commit adds shm_open fall back code for that case
(for now).
https://bugzilla.gnome.org/show_bug.cgi?id=766341
We currently use syscall() directly to invoke memfd_create,
since the function isn't available in libc headers yet.
The code, though, mishandles how errors are passed from syscall().
It assumes syscall returns the error code directly (but negative),
when in fact, syscall() uses errno.
Also, the code fails to retry on EINTR.
This commit moves the handling of memfd create to a helper function,
and changes the code to use errno and handle EINTR.
https://bugzilla.gnome.org/show_bug.cgi?id=766341
When disposing a GdkDrawingContext we should unset the association
between the instance and the Cairo context; this avoids stale pointers
in case a reference that has acquired on the Cairo context survives the
lifetime of the GdkDrawingContext.
Instead of associating the GdkWindow that created the GdkDrawingContext
we can directly bind the Cairo context to the GDK drawing context.
Cairo contexts created via gdk_cairo_create() go back to not having a
GdkWindow associated to them, like they did before we introduced the
gdk_window_begin_draw_frame() API.
https://bugzilla.gnome.org/show_bug.cgi?id=766675
Instead of giving out Cairo contexts, GdkWindow should provide a
"drawing context", which can then create Cairo contexts on demand; this
allows us to future proof the API for when we're going to use a
different rendering pipeline, like OpenGL.
https://bugzilla.gnome.org/show_bug.cgi?id=766675
Existing code drawing on a GDK window has to handle the direct drawing
and the buffered drawing by itself, by checking the window type and
whether or not the window is backed by a native windowing surface. After
that, the calling code has to create a Cairo context from the window and
keep an association between the context and the window itself.
This is completely unnecessary: GDK can determine whether or not it
should use a backing store to draw on a GdkWindow as well as create a
Cairo context, and keep track of it.
This allows to simplify the calling code, and enforce some of the
drawing behavior we want to guarantee to users.
https://bugzilla.gnome.org/show_bug.cgi?id=766675
A wl_buffer has a max size of 4096 bytes, of which 8 are needed for the
header and another 4 for the string argument length (in this case), so
make sure the we only save the first 4083 bytes that are still valid
UTF8.
https://bugzilla.gnome.org/show_bug.cgi?id=767241
xdg-shell allows desktop environments to extend the list of states
within a given range.
Use this possibility to add a new state for tiled so that gtk+ can
benefit from this.
https://bugzilla.gnome.org/show_bug.cgi?id=766860
The Wayland protocol does not share XI2's wealth of information
about individual devices, but it does provide discriminating
information about the source for scroll events. Pass this on to
the application by creating separate slave devices for these,
and setting them as source device on the scroll events.
These devices can be discriminated by their input-source property:
wheel - GDK_SOURCE_MOUSE
finger - GDK_SOURCE_TOUCHPAD
continuous - GDK_SOURCE_TRACKPOINT
https://bugzilla.gnome.org/show_bug.cgi?id=767093
fix up
Calling gdk_gl_context_realize() should always result in a valid result,
so we need to provide a default implementation, to avoid a call to a
NULL function pointer.
Given that Wayland has no global coordinate, the only way for gdk to
retrieve the monitor a window last entered is to retrieve it from the
GdkWaylandWindow itself.
Implement the backend specific get_monitor_at_window() to return the
monitor that was last entered by the window.
https://bugzilla.gnome.org/show_bug.cgi?id=766566
In Wayland, surfaces get an enter/leave notification each time they
enter or leave an output.
Add an API to GdkWaylandWindow to retrieve the output the window has
last entered.
https://bugzilla.gnome.org/show_bug.cgi?id=766566
Some backends (namely Wayland) do not support global coordinates so
using the window position to determine the monitor will always fail on
such backends.
In such cases, the backend itself might be better suited to identify
the monitor a given window resides on.
Add a vfunc get_monitor_at_window() to the display class so that we can
use the backend to retrieve the monitor, if the backend implements it.
https://bugzilla.gnome.org/show_bug.cgi?id=766566
The GdkDragContext should only listen to GDK_GRAB_BROKEN events sent to
its own pointer device. It turns out that the passive key grabs mistake
GDK into sending a GdkEventGrabBroken on the master keyboard, which the
DnD machinery mistakes as a signal to cancel the operation.
https://bugzilla.gnome.org/show_bug.cgi?id=766442
Only generate crossing events on wl_touch.down for the virtual master
device used for touch events, and only whenever this virtual device
actually moves across surfaces. This behavior resembles better what is
expected in X11, where the pointer is warped to the touch position
on XITouchBegin.
This avoids the double emission of leave events when the pointer
emulating touch is lifted, that crossing event will be instead
generated when/if the focus surface changes.
https://bugzilla.gnome.org/show_bug.cgi?id=766314
This is required for proper integration with any other library/application that
may perform wayland API calls and poll() the wayland fd from multiple threads.
Using wl_display_dispatch{_queue}() is thread-safe if not mixed with custom
poll() usage, which GSource/GMainContext does.
Essentially, the problem is that multiple threads polling and reading
the same fd is extremely racy. Use the wayland provided API for allowing
concurrent access to the wayland display fd.
See the wayland man pages for wl_display_prepare_read(),
wl_display_cancel_read() and wl_display_read_events() for more details.
https://bugzilla.gnome.org/show_bug.cgi?id=763852
WINBOOL is MinGW-specific, so change it to BOOL, which is universally
available.
Also, Visua Studio is more picky on where __stdcall (WINAPI) is placed, so
fix that to be in-sync with what is done in the other sources.
Use of g_signal_handlers_disconnect_by_func() needs to do more work than
necessary to find all the matching handlers. Instead, just hold on to the
signal identifier and remove it directly so we hit the fast path.
Not terribly ground breaking in terms of performance gains, but its done
enough to be worthwhile.
https://bugzilla.gnome.org/show_bug.cgi?id=766049
The active keyboard grab can be spared then. This way the passive
key grabs allow other key combinations (eg. alt-tab) that are not
mandatory to grab here.
The wayland specification for discrete step information for scroll and
other axes reads:
| The discrete value carries the directional information. e.g. a
| value of -2 is two steps towards the negative direction of this axis.
mutter sets a value of 1 for SCROLL_DOWN events and -1 for SCROLL_UP
events.
gdkdevice Wayland backend does the opposite, it translates a positive
discrete value as SCROLL_UP and a negative value as SCROLL_DOWN, which
ends up inverting the scrolling direction.
Fix the logic in gdkdevice Wayland to use a positive value as
SCROLL_DOWN and a negative value as SCROLL_UP so that it matches mutter
and weston logic.
https://bugzilla.gnome.org/show_bug.cgi?id=765907
If the monitor vfuncs are not implemented in a display class,
fall back to providing a single monitor object representing
the entire screen. This is not meant to be 'good enough', it
is just to provide some implementation until all backends
implement the monitor vfuncs. When that is the case, the
fallback should be removed.
NoSymbol is not a valid GDK symbol (it only has the concept of
VoidSymbol, for some reason, which is neither the same thing nor
produced by any sane keymap). Passing NoSymbol events through to GTK+
apps is unlikely to produce anything useful.
In particular, this meant VTE would scroll to the end of the buffer when
pressing Fn (required for Page Up/Down on Macs), as it was receiving a
keypress that wasn't a modifeir. This does not happen on X11, as the
KEY_FN keycode is above 255, so does not get sent to clients.
https://bugzilla.gnome.org/show_bug.cgi?id=764825
We don't care about the specific (possibly client-side) window that
requested the focus here, only the toplevel. Fixes mistakenly sent
focus events when the grab happens inside the current focus window.
https://bugzilla.gnome.org/show_bug.cgi?id=762756
Always associate a drag context with a GdkDisplay and use that when
getting a cursor for a given action.
If we don't do this, dragging on a window that doesn't use the default
display will make us use cursors from the wrong display.
https://bugzilla.gnome.org/show_bug.cgi?id=765565
Typically, there won't be any references on old frame timings except for
the most recent timing. So instead of discarding these and re-entering
gslice twice, just steal the old frame timing and reuse it.
https://bugzilla.gnome.org/show_bug.cgi?id=765592
Don't track all orphaned dialogs globally, as mixing them up with each
other would in most cases trigger errors when we try to pass bogus
values to Wayland requests.
https://bugzilla.gnome.org/show_bug.cgi?id=765474
The naming of pointers to GdkWaylandDisplay's were inconsistent.
Running the following commands in gtk+/gdk/wayland illustrate the
inconsistency:
$ grep -r '\<display_wayland\>' *.[ch] | wc -l
195
$ grep -r '\<wayland_display\>' *.[ch] | wc -l
81
This patch renames all occurrences of "wayland_display" to
"display_wayland". This is also consistent with naming in the X11
backend. A couple of whitespace changes were done as well in places
where the rename was already done, that added line breaks to long lines
that stood out.
https://bugzilla.gnome.org/show_bug.cgi?id=765474
When synthesizing keyboard repeat, we can potentially drift further from
the mark depending on the timing of the frame callback and how long it
took to deliver the event.
This patch attempts to reduce this by tracking from a stable epoch the
time of our next keyboard repeat.
https://bugzilla.gnome.org/show_bug.cgi?id=765567
This allows us to decide when the R and B color channels should be
flipped with a much better granularity.
For instance, when using GLX_EXT_texture_from_pixmap to create a GL
texture from a surface we don't need to swap the R and B channels, as
the internal representation of the texture data will already have the
appropriate colors.
We also don't need to flip color channels when blitting from a texture.
Cairo surfaces are in BGRA format, but we upload them as RGBA buffers on
GLES; this means that the R and B channels are flipped in the texture
data.
Instead of doing a costly channel flip before putting them on the GPU,
we can flip the values inside the GLSL shader we use specifically for
GLES.
We cannot use GL_BGRA and GL_UNSIGNED_INT_8_8_8_8_REV when reading back
from the frame buffer in the software fallback path, as they do not
exist on OpenGL ES.
This will make the slow path a bit slower, because of the implicit
format conversion.
When uploading a Cairo image surface to a GL texture we cannot use
GL_BGRA and GL_UNSIGNED_INT_8_8_8_8_REV on OpenGL ES, as they do not
exist in the core spec.
On some platforms we can ask the GL context machinery to create a GLES
context, instead of a GL one.
In order to ask for a GLES context at GdkGLContext realization time, we
use a bit field like we do for forward compatible, or debug contexts.
The 'use-es' bit also changes the way we select a default version,
because OpenGL and OpenGLES versions differ.
https://bugzilla.gnome.org/show_bug.cgi?id=743746
Windows save in hardware_keycode an information which is not so low
level and some application require the hardware scancode.
As Windows provides this information save it in GdkEventPrivate
and provide a function to get this information.
For no Windows system the function return the hardware_keycode instead.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
https://bugzilla.gnome.org/show_bug.cgi?id=765259
