Kristian Høgsberg [Wed, 18 Jul 2012 21:23:41 +0000 (17:23 -0400)]
compositor-drm: Remove cursor debug message
Kristian Høgsberg [Wed, 18 Jul 2012 16:09:58 +0000 (12:09 -0400)]
compositor-drm: Only update hw cursor contents when it changes
We avoid the memcpy, gbm_bo_write and the drmModeSetCursor ioctl per frame.
Kristian Høgsberg [Wed, 18 Jul 2012 16:08:37 +0000 (12:08 -0400)]
compositor-drm: Only call drmModeMoveCursor() if hw cursor position changed
Saves an ioctl per frame.
Kristian Høgsberg [Wed, 18 Jul 2012 16:02:51 +0000 (12:02 -0400)]
compositor: Don't record surface damage when we move or transform surfaces
surface->damage is for when the contents of the surface changes. Instead,
use weston_surface_damage_below() to repaint the damaged area. We avoid
unecessary uploading shm and hw cursor contents this way.
Kristian Høgsberg [Wed, 18 Jul 2012 15:39:05 +0000 (11:39 -0400)]
wcap: Just make wcap-decode dump YUV4MPEG2
Instead of having a custom fork of the vpxenc tool in weston, we can
just dump raw YUV data in the YUV4MPEG2 format and feed that into the
upstream vpxenc. This also works with theora_encoder and probably many
other encoders.
Kristian Høgsberg [Wed, 18 Jul 2012 13:54:04 +0000 (09:54 -0400)]
shell: Default to maximize/fullscreen on current output
Kristian Høgsberg [Tue, 17 Jul 2012 18:21:25 +0000 (14:21 -0400)]
compositor: Free the right EGLImage
We forgot to index the image array and freed image[0] multiple times
Kristian Høgsberg [Sat, 14 Jul 2012 05:24:58 +0000 (01:24 -0400)]
compositor-drm: Allocate hw cursor like other planes
We no longer reserve the hw cursor for just the pointer sprite. Any
surface that satisfies the requirements for the hw cursor can be used.
Of course that's more or less always the pointer sprite, but at least now
we don't hard-code the references to compositor->seat anymore.
Kristian Høgsberg [Sat, 14 Jul 2012 05:06:28 +0000 (01:06 -0400)]
compositor-drm: Combine drm_output_set_cursor and weston_output_set_cursor
weston_output_set_cursor was the left-over of a core compositor cursor
helper that we moved into compositor-drm.c.
Kristian Høgsberg [Sat, 14 Jul 2012 04:31:32 +0000 (00:31 -0400)]
compositor-drm: Assign primary framebuffer in assign_planes callout
Determining whether it was possible to use a surface buffer as the
primary framebuffer was always a bit of a hack. Just before rendering,
we look at the second top-most surface to see if it's the right size and
position and then use it if it is. If we have a hw cursor and a drm
plane on top, it's no longer the second top-most and it wouldn't be picked
even if it could work.
With this change, we inspect surfaces in drm_assign_planes and try to find
a candidate for the primary fb there. We track the overlap like for
overlays and if there is a surface that's not overlapped and fills the
output we pick it.
This commit also consolidate the logic to track damage as we move a surface
in and out of the primary plane, which was missing in the drm plane case.
Kristian Høgsberg [Fri, 13 Jul 2012 19:25:07 +0000 (15:25 -0400)]
compositor-drm: Support hardware scaling with drm planes
The kms planes support scaling, so try to detect transformations that
are just translations + scaling and program the kms plane accordingly.
In particular, this lets us fullscreen a yuv surfaces with the scale method
and have the compositor use a kms plane for scaling and color conversion.
Kristian Høgsberg [Mon, 16 Jul 2012 20:44:16 +0000 (16:44 -0400)]
compositor-drm: Follow gbm_bo_get_pitch rename to gbm_bo_get_stride
It was always a confusing mis-nomer, fix it while we can.
Kristian Høgsberg [Sat, 14 Jul 2012 02:54:43 +0000 (22:54 -0400)]
compositor-drm: Use gbm_bo_import instead of gbm_bo_create_from_egl_surface
This lets us create the gbm bo from the wl_buffer instead of the
EGLImage. With planar buffer, the EGLImage corresponds to a plane
and not the entire planar buffer.
More importantly, with some YUV formats (packed YUYV and similar) we use
several EGLImages to sample the different components, but KMS only
want one plane in that case.
Tiago Vignatti [Mon, 16 Jul 2012 16:15:37 +0000 (12:15 -0400)]
xwm: set the root window pointer to left_ptr
More pretty and also consistent with toytoolkit.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Tiago Vignatti [Mon, 16 Jul 2012 16:09:19 +0000 (12:09 -0400)]
xwm: set appropriate cursors for the frame window
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Tiago Vignatti [Mon, 16 Jul 2012 16:02:08 +0000 (12:02 -0400)]
xwayland: add simple xcb-cursor hook based on libXcursor
It's in fact based on the core of libXcursor, which doesn't bring any Xlib
dependency.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Kristian Høgsberg [Mon, 16 Jul 2012 15:57:08 +0000 (11:57 -0400)]
xwayland: Look up 32 depth rgba render format as well
Kristian Høgsberg [Mon, 16 Jul 2012 15:52:40 +0000 (11:52 -0400)]
xwayland: Use the right length for iterating through the formats array
Spotted by Tiago Vignatti, minimal fix to just use the right length.
Ander Conselvan de Oliveira [Mon, 16 Jul 2012 11:15:50 +0000 (14:15 +0300)]
desktop-shell: Don't get a shell surface for the grab surface
The grab surface does not depend on any of the shell surface
functionality, so don't allocated one.
Ander Conselvan de Oliveira [Mon, 16 Jul 2012 11:15:49 +0000 (14:15 +0300)]
window: Workaround a stuck frame callback on the cursor surface
It is possible that a client loses the focus between receiving a
pointer.enter event and sending a pointer.set_cursor request. In that
case, the cursor surface might not be mapped and the frame callback
requested on it will never trigger.
Work around this by trying to remap the cursor surface whenever there
is a frame callback and the serial for the enter event is higher than
the cursor serial.
Ander Conselvan de Oliveira [Mon, 16 Jul 2012 11:15:48 +0000 (14:15 +0300)]
shell: Don't ping the grab surface
If the grab surface happens to be unresponsive, the busy cursor grab
will be started and that will cause the grab surface to receive focus,
which in turn leads to it being pingged again. Break the cycle by not
sending pings to it. If the shell is unresponsive it won't be able to
set the busy cursor anyway.
Kristian Høgsberg [Thu, 12 Jul 2012 16:34:43 +0000 (12:34 -0400)]
compositor: Don't send wl_surface.enter/leave if we didn't find a resource
Kristian Høgsberg [Thu, 12 Jul 2012 16:32:31 +0000 (12:32 -0400)]
shell: Scale fullscreen surface without cropping them, center on output
Scott Moreau [Thu, 12 Jul 2012 02:57:15 +0000 (20:57 -0600)]
toytoolkit: Remove obsolete function window_set_custom().
Scott Moreau [Thu, 12 Jul 2012 02:57:14 +0000 (20:57 -0600)]
wscreensaver: Fix --demo mode.
Tiago Vignatti [Wed, 11 Jul 2012 21:46:10 +0000 (00:46 +0300)]
compositor: remove useless weston_compositor_repick call
compositor->focus was just unset, and weston_compositor_repick() has no
effect.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Philipp Brüschweiler [Wed, 11 Jul 2012 13:19:23 +0000 (15:19 +0200)]
Remove dead code
window->title is always NULL inside window_create_internal
Kristian Høgsberg [Wed, 11 Jul 2012 19:45:59 +0000 (15:45 -0400)]
image-loader: Use pixman_image_t, not struct pixman_image_t
Rafal Mielniczuk [Wed, 11 Jul 2012 16:48:25 +0000 (18:48 +0200)]
image-loader: don't leak pixman_image bits data
Set the pixman_image destroy function to free bits data
Rafal Mielniczuk [Wed, 11 Jul 2012 16:48:12 +0000 (18:48 +0200)]
clipboard: add seat destroy signal handler
Gwenole Beauchesne [Fri, 20 Apr 2012 09:44:06 +0000 (11:44 +0200)]
compositor: handle YUV buffers.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Kristian Høgsberg [Tue, 10 Jul 2012 19:51:13 +0000 (15:51 -0400)]
Update weston-egl-ext.h with EGL_WL_bind_wayland_display changes
Gwenole Beauchesne [Fri, 20 Apr 2012 09:22:16 +0000 (11:22 +0200)]
compositor: add YUV shaders.
Add shaders for NV12 (2 planes), YUV (3 planes) and YUYV (sampled as 2 planes)
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Gwenole Beauchesne [Fri, 20 Apr 2012 09:15:51 +0000 (11:15 +0200)]
compositor: factor out RGB shader.
Simplify RGB shader code and split off common code that could be reused.
This is preparatory work for YUV shaders.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Gwenole Beauchesne [Fri, 20 Apr 2012 09:07:06 +0000 (11:07 +0200)]
compositor: prepare for multi-planar surfaces.
Make weston_surface::texture and ::surface an array, while keeping
[0] for RGB surfaces.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Juan Zhao [Tue, 10 Jul 2012 05:24:09 +0000 (22:24 -0700)]
shell: Use the correct the member name when iterating panel_list
It does not get the correct panel height for now. The memeber of list
to read should be layer_link not link now.
Signed-off-by: Juan Zhao <juan.j.zhao@linux.intel.com>
Kristian Høgsberg [Tue, 10 Jul 2012 02:22:37 +0000 (22:22 -0400)]
window.c: Only set input region when we have a decoration
https://bugs.freedesktop.org/show_bug.cgi?id=50490
Kristian Høgsberg [Tue, 10 Jul 2012 02:17:03 +0000 (22:17 -0400)]
tests: Handle modifier event in test-client.c
We need the handle the newly added modifiers event.
https://bugs.freedesktop.org/show_bug.cgi?id=51567
Kristian Høgsberg [Tue, 10 Jul 2012 01:43:22 +0000 (21:43 -0400)]
Only position fullscreen surface in surface->configure callback
We end up calling shell_configure_fullscreen() from activate(), where
surface->buffer may be NULL. Instead, call it on initial map and
first buffer attach after surface type change. Both code path
come from the surface->configure callback where surface->buffer is know
to be non-NULL.
https://bugs.freedesktop.org/show_bug.cgi?id=51678
Jan Engelhardt [Mon, 9 Jul 2012 15:18:10 +0000 (17:18 +0200)]
wcap-snapshot: add return value
Distro build with rpmlint aborts due to missing exit status.
Daniel Stone [Mon, 9 Jul 2012 11:27:28 +0000 (12:27 +0100)]
toytoolkit: Remove redundant XKB free
We already unref these in fini_xkb().
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Martin Olsson [Mon, 9 Jul 2012 09:35:53 +0000 (11:35 +0200)]
Add cscope.out to .gitignore
Martin Olsson [Sun, 8 Jul 2012 09:11:18 +0000 (11:11 +0200)]
clients: Free resources on exit in eventdemo
Martin Olsson [Sun, 8 Jul 2012 01:03:47 +0000 (03:03 +0200)]
clients: Silence 2x gcc 4.6.3 "warn_unused_result" compiler warnings
window.c:1173:6: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]
desktop-shell.c:305:6: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]
Martin Olsson [Sun, 8 Jul 2012 01:03:46 +0000 (03:03 +0200)]
wcap: Plug memory leak in wcap_decoder_create()
Martin Olsson [Sun, 8 Jul 2012 01:03:45 +0000 (03:03 +0200)]
xwayland: Don't leak fd while reading lock file in src/xwayland/launcher.c
Martin Olsson [Sun, 8 Jul 2012 01:03:44 +0000 (03:03 +0200)]
compositor-x11: Add error handling in x11_compositor_create()
Martin Olsson [Sun, 8 Jul 2012 01:03:43 +0000 (03:03 +0200)]
compositor-wayland: Add error handling in wayland_compositor_create()
Martin Olsson [Sun, 8 Jul 2012 01:03:42 +0000 (03:03 +0200)]
compositor-android: Add proper error handling in android_compositor_create()
Martin Olsson [Sun, 8 Jul 2012 01:03:41 +0000 (03:03 +0200)]
shared: Don't leak file descriptor in shared/image-loader.c
Martin Olsson [Sun, 8 Jul 2012 01:03:40 +0000 (03:03 +0200)]
clients: Don't leak "menu" on error path in clients/window.c
Martin Olsson [Sun, 8 Jul 2012 01:03:38 +0000 (03:03 +0200)]
clients: Add clients/weston-screensaver to clients/.gitignore
Ander Conselvan de Oliveira [Wed, 4 Jul 2012 12:48:29 +0000 (15:48 +0300)]
compositor: Always empty the input region of the pointer surface
Otherwise a surface.attach request might cause the input region to be
reset to the default value (the entire surface) causing it to receive
focus.
Tiago ran into this problem with xwayland.
Martin Minarik [Mon, 2 Jul 2012 21:05:50 +0000 (23:05 +0200)]
weston: Clip window title text to the frame
Fixes an issue with text overflowing the available frame space.
Kristian Høgsberg [Tue, 3 Jul 2012 01:24:57 +0000 (21:24 -0400)]
compositor-x11: Set make and model so we don't send NULL strings
Rob Bradford [Mon, 2 Jul 2012 18:00:19 +0000 (19:00 +0100)]
shell: Use desired output when sending configure events when maximised
With shell_surface.set_maximised the caller can provide an output to maximise
to or the default output will be used. With the corresponding configure we
should use the dimensions from the chosen output not the output the surface
was currently on.
Similarly when calculating the position for the window in the map() function
we should use the desired output again.
There is no need to assign shsurf->output to es->output since that happens in
the map() function later.
Rob Bradford [Mon, 2 Jul 2012 16:33:40 +0000 (17:33 +0100)]
compositor: Correctly calculate changed sub image rectangle
When calling glTexSubImage2D for sub image updates for SHM surfaces the
changed rectangle was being wrongly calculated. This resulted in interesting
redraw artefacts for clients using SHM.
Kristian Høgsberg [Mon, 2 Jul 2012 01:31:46 +0000 (21:31 -0400)]
terminal: Extend selection to end of line when selecting past last char
Kristian Høgsberg [Mon, 2 Jul 2012 01:25:41 +0000 (21:25 -0400)]
terminal: Insert newlines when copying multiple lines
Rafal Mielniczuk [Sat, 30 Jun 2012 16:33:34 +0000 (18:33 +0200)]
clipboard: initialise clipboard memory after creation
In case we don't, weston segfault in clipboard_set_selection
on:
if (clipboard->source)
clipboard_source_unref(clipboard->source);
Kristian Høgsberg [Fri, 29 Jun 2012 15:35:24 +0000 (11:35 -0400)]
terminal: Handle selection for pointer positions outside widget correctly
That is, don't crash and select entire first/last line when the pointer
is above or below widget.
Kristian Høgsberg [Fri, 29 Jun 2012 02:04:06 +0000 (22:04 -0400)]
window.c: Don't allocate proxy manually
With the wayland change to automatically allocate the client side proxy
manually, we can now drop the code (and the FIXME) that did that and just
receive the proxy from the callback arguments.
Kristian Høgsberg [Thu, 28 Jun 2012 18:27:02 +0000 (14:27 -0400)]
terminal: Add a terminal enter handler for setting I-beam cursor
Now that the mod+click shortcuts change the cursor, this was suddenly obvious.
Ander Conselvan de Oliveira [Thu, 28 Jun 2012 15:08:06 +0000 (18:08 +0300)]
shell: Don't move fullscreen surfaces
Ander Conselvan de Oliveira [Thu, 28 Jun 2012 15:08:05 +0000 (18:08 +0300)]
shell: Show approprite cursors during move and resize grabs
This patch expands the busy cursor mechanism so that it is possible for
the desktop-shell client to show the appropriate cursor during grabs.
Ander Conselvan de Oliveira [Thu, 28 Jun 2012 15:08:04 +0000 (18:08 +0300)]
compositor: Always allow owner of pointer surface to change the hotspot
Make pointer_set_cursor() succeed if a client that doesn't have pointer
focus but is the owner of the current pointer surface calls it.
Kristian Høgsberg [Thu, 28 Jun 2012 18:13:10 +0000 (14:13 -0400)]
compositor: Move repaint debug binding to mod-shift-space
Alt-space or Super-space are too likely to conflict with application
bindings.
Kristian Høgsberg [Thu, 28 Jun 2012 17:46:09 +0000 (13:46 -0400)]
compositor: Update xkb state with key releases on focus out
This happens when vt-switching away from the compositor (drm) or
giving keyboard focus to a different X window. Release the modifiers
so we don't get stuck modifiers. We'll update with the new keys down
when we come back.
Kristian Høgsberg [Thu, 28 Jun 2012 15:28:15 +0000 (11:28 -0400)]
compositor-drm: Improve initial mode picking
We now pick the driver preferred mode for our initial mode. If no preferred
mode is available we default to the current mode. We also have a command
line option now to keep the current mode if it differs from the preferred.
This commit also drops the built-in mode and insteade uses the current mode
if the connector doesn't report any modes.
Kristian Høgsberg [Wed, 27 Jun 2012 23:29:41 +0000 (19:29 -0400)]
terminal: Scroll selection with terminal contents
Kristian Høgsberg [Wed, 27 Jun 2012 21:52:23 +0000 (17:52 -0400)]
terminal: Make font size configurable
Kristian Høgsberg [Wed, 27 Jun 2012 21:43:10 +0000 (17:43 -0400)]
terminal: Implement word and line based selection
Double-click to select by word, triple click to select by line.
Kristian Høgsberg [Wed, 27 Jun 2012 14:22:58 +0000 (10:22 -0400)]
shell: Take a wl_surface as parent for transient and popup windows
Kristian Høgsberg [Wed, 27 Jun 2012 14:22:15 +0000 (10:22 -0400)]
shell: Make input_panel just a regular wl_surface
Kristian Høgsberg [Wed, 27 Jun 2012 02:15:53 +0000 (22:15 -0400)]
shell: Make screensaver just a wl_surface
Kristian Høgsberg [Wed, 27 Jun 2012 01:44:35 +0000 (21:44 -0400)]
shell: Make the lock surface just a plain wl_surface too
Kristian Høgsberg [Wed, 27 Jun 2012 01:19:23 +0000 (21:19 -0400)]
shell: Make panel just a wl_surface too
Kristian Høgsberg [Tue, 26 Jun 2012 20:29:50 +0000 (16:29 -0400)]
shell: Don't make desktop_shell background a wl_shell_surface
We don't gain anything from taking a wl_shell_surface in
desktop_surface.set_background, except making wl_shell_surface
gratuitously dependent on wl_shell. In shell.c we can also handle
backgrounds in their own background_configure function which simplifies
the mapping and placement logic.
Kristian Høgsberg [Tue, 26 Jun 2012 15:57:40 +0000 (11:57 -0400)]
terminal: Tab should not output spaces, just move cursor
Emacs uses tab and backspace to move the cursor as well as the regular
cursor movement escape codes. When it's less bytes than the escape code,
emacs will use a tab or tab + backspace to move the cursor forward. The
effect is that as you're moving around in the buffer, emacs will
(seemingly) randomly insert spaces and overwrite what's in the terminal.
Making tab just move the cursor as it should fixes this.
Ander Conselvan de Oliveira [Tue, 26 Jun 2012 14:09:15 +0000 (17:09 +0300)]
compositor-drm: Don't leak FBs when the buffer of a sprite is detroyed
As a side effect, the overlay is disabled when the buffer is destroyed.
Ander Conselvan de Oliveira [Tue, 26 Jun 2012 14:09:14 +0000 (17:09 +0300)]
compositor-drm: Remove destroy listeners when disabling planes
When an unused plane is disabled, the destroy listener for a previously
used buffer needs to be removed. This fixes a crash when an overlay
would be reenabled using the same buffer as before, causing the destroy
listener to be inserted twice.
Ander Conselvan de Oliveira [Tue, 26 Jun 2012 14:09:13 +0000 (17:09 +0300)]
compositor-drm: Wait for vblank events before starting next frame
Besides the fact of the frame not being done, assigning planes before
the vblank handler is called will make the state inconsistent, leading
to a crash.
Ander Conselvan de Oliveira [Tue, 26 Jun 2012 14:09:12 +0000 (17:09 +0300)]
compositor-drm: Check surface's outputs when assigning overlays
Check if the output passed to drm_output_prepare_overlay_surface() does
actually contain the surface and fail if it doesn't. Also fail if a
surface spans multiple outputs since clearing the damage will cause the
portion that is not in an overlay to not be updated at all.
Ander Conselvan de Oliveira [Tue, 26 Jun 2012 14:09:11 +0000 (17:09 +0300)]
simple-egl: Make window moveable
Kristian Høgsberg [Mon, 25 Jun 2012 22:12:10 +0000 (18:12 -0400)]
configure.ac: Bump version to 0.94.90
Kristian Høgsberg [Tue, 26 Jun 2012 01:35:29 +0000 (21:35 -0400)]
shell: Fix inverted test for locked in workspace bindings
Kristian Høgsberg [Mon, 25 Jun 2012 22:05:08 +0000 (18:05 -0400)]
Makefile.am: Don't list wcap sub directory twice in DIST_SUBDIRS
Kristian Høgsberg [Mon, 25 Jun 2012 21:50:22 +0000 (17:50 -0400)]
tests: Add weston-test to EXTRA_DIST
Kristian Høgsberg [Mon, 25 Jun 2012 21:13:58 +0000 (17:13 -0400)]
Move text-cursor-position.c into zoom.c
Kristian Høgsberg [Mon, 25 Jun 2012 21:13:34 +0000 (17:13 -0400)]
wcap: Add vpx_config.h to Makefile.am
Kristian Høgsberg [Mon, 25 Jun 2012 20:00:48 +0000 (16:00 -0400)]
xwm: Add xwayland.h to Makefile.am
Kristian Høgsberg [Mon, 25 Jun 2012 18:06:45 +0000 (14:06 -0400)]
Install screensaver by default
Kristian Høgsberg [Mon, 25 Jun 2012 18:03:13 +0000 (14:03 -0400)]
shell: Hide workspace layer when locking screen
Kristian Høgsberg [Fri, 22 Jun 2012 20:57:21 +0000 (16:57 -0400)]
compositor: Send ping to keyboard focus surface for key press
We were sending to the pointer focus surface.
Kristian Høgsberg [Fri, 22 Jun 2012 16:18:56 +0000 (12:18 -0400)]
terminal: Ignore more modifier keys
Don't want these to leak into the terminal.
Kristian Høgsberg [Fri, 22 Jun 2012 16:17:17 +0000 (12:17 -0400)]
terminal: Only hide cursor for keypresses that generate data for the process
Daniel Stone [Fri, 22 Jun 2012 13:04:37 +0000 (14:04 +0100)]
compositor-drm: Add proper error handling to init
This way, if initialisation fails (say, udev or Mesa are broken, or we
couldn't find any devices), we'll at least take you back to where you
were, rather than leave you at a totally broken VT you can't escape
from.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Daniel Stone [Fri, 22 Jun 2012 13:04:36 +0000 (14:04 +0100)]
Split weston_compositor_init into base and _gl
weston_compositor_init is always called late because most
implementations can't initialise GL until fairly late in the game.
Split it into a base version with the same name, followed by
weston_compositor_init_gl which can be called later on.
This simplifies compositor-wayland, which no longer needs a separate
global handler just for wl_seat.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Daniel Stone [Fri, 22 Jun 2012 12:21:41 +0000 (13:21 +0100)]
compositor-wayland: Synchronise keyboard state with parent
Use the wl_keyboard::modifiers events our parent helpfully sends us to
make sure our views of the keyboard state are always identical, rather
than relying on key press events to do the right thing.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Daniel Stone [Fri, 22 Jun 2012 12:21:40 +0000 (13:21 +0100)]
compositor-x11: Attempt to track state without XKB
This isn't very pleasant, but it's pretty much the best we can do in the
absence of either XCB XKB support, or XCB XI2 bindings (argh!).
We get a state mask with most significant X11 events (which inexplicably
includes EnterNotify but not FocusIn), but unfortunately it's only a
single flat set of effective modifiers rather than the more granular
sets we want, so we still update the state with every key, but then also
use the core X11 state as a mask to make sure we don't get any stuck
modifiers.
Ugh.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>