This commit makes the wayland backend search through the config for
[output] sections with names starting with "WL" and create outputs
accordingly. Outputs created due to the config file support mode, scale,
and transform parameters. It also listens for the --output-count
command-line option.
This brings the wayland backend almost up to par, in terms of functionality
with the X11 backend.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Both the Pixman renderer and the X11 backend contained effectively the same
region transformation code. This commit adds a weston_transformed_region
function and changes pixman-renderer and compositor-x11 to use it.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
The output is renamed "output" from "x11_output" and the input coordinates
are changed to wl_fixed_t from int. This way it is useable in
compositor-wayland as well as compositor-x11 and evdev.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
The only user for this was the wayland backend with the GL renderer. It is
not needed in the Pixman renderer because you can easily create subimages.
All of the fancy output matrix calculations can be replaced by a single
glViewport call. Also, it didn't work with outputs located anywhere but
(0, 0) and I'm pretty sure output transformed outputs would break it too.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
This adds a window frame with a close button. Similar to the X11 backend,
The window supports dragging but not resizing.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
The first advantage of this new API is that it is per-output instead of
global to the gl_renderer instance. This means that different windows can
have different titles, different button states, etc. The new api also uses
four textures (one for each side) instead of one. This allows you to draw
real borders with text and buttons in them instead of a simple image that
gets streached.
Images will be scaled as needed, so the right and left can be one pixel
tall if desired.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
The logic here broke at some point so that we would only update the
input region for non-fullscreen windows. Thus, a fullscreen window would
be stuck with whatever size the most recent non-fullscreen size was.
https://bugs.freedesktop.org/show_bug.cgi?id=69219
This fixes the problem where animations will wait to play until input is
received. In general, it also makes the backend far more responsive.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
https://bugs.freedesktop.org/show_bug.cgi?id=68221
On shutdown, we can risk having a pending vt switch that we normally
handle in the vt signal handler. However as we put the vt back in VT_AUTO
mode, the pending VT switch will go through and if we haven't dropped
drm master at that point, we could switch to another display server
without dropping drm master. That will typically crash the other server,
so let's try to make sure we don't do that.
https://bugs.freedesktop.org/show_bug.cgi?id=70877
If the seat or tty doesn't match we return with r == 0, which looks like
success to weston_launcher_connect(), which then fails to fall back
to the legacy path.
https://bugs.freedesktop.org/show_bug.cgi?id=70876
Otherwise we don't repaint with the final state of the surface and
we're stuck with the second-to-last frame of the animation until
something else (moving the mouse or such) triggers a redraw.
https://bugs.freedesktop.org/show_bug.cgi?id=70930
get_animation_type() parses "none", "zoom" and "fade" but for the
startup animation, we only support "none" and "fade". If we get "zoom"
just fall back to "none" like we do for all unrecognized strings.
https://bugs.freedesktop.org/show_bug.cgi?id=71006
A grab can potentially allocate memory and would normally end the grab
itself, freeing the allocated memory in the process. However at in some
situations the compositor may want to abort a grab. The grab owner still
needs to free some memory and abort the grab properly. To do this a new
function 'cancel' is introduced in all the grab interfaces instructing
the grabs owner to abort the grab.
This patch also hooks up grab cancelling to seat device releasing and
when the compositor looses focus, which would potentially leak memory
before.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
We used to leak the input fds, as weston would reopen all fds on vt
enter. We could just close them after sending the open fd through the
socket, but this patch also adds support for the new EVIOCREVOKE evdev
ioctl, that revokes the fd in question (including the copy that we
sent to the compositor).
https://bugs.freedesktop.org/show_bug.cgi?id=70849
Previously the renderers destroy function assumed they are only called
when the compositor is shutting down and that the compositor had
already destroyed all the surfaces. However, if a runtime renderer
switch would be done, the surface state would be leaked.
This patch adds a destroy_signal to the pixman and gl renderers. The
surface state objects will listen for that signal and destroy
themselves if needed.
This is a step towards runtime switchable renderers.
Remove create_surface() and destroy_surface() from the renderer
interface and change the renderers to create surface state on demand
and destroy it using the weston_surface's destroy signal.
Also make sure the surfaces' renderer state is reset to NULL on
destruction.
This is a step towards runtime switchable renderers.
(rpi-renderer changes are only compile-tested)
Also make sure backends destroy the renderers before shutting down the
compositor to avoid a double call to weston_binding_destroy().
This is a step towards making renderers switchable during runtime.
The compositor will check if the client destroyed the wl_buffer
while it was in use in a display update, and delete the resource
itself once the update has finished.
It was possible to start a move or resize with a pointer and then
start another move for the same surface with touch or a pointer from
another seat, which would make the window flicker around. With this
commit, we now reject any attempts to move or resize a surface that is
currently grabbed.
Most of the shell.c bindings take effect on the current pointer or
touch focus. That's now a view, but we need a surface in these cases.
Just get the surface from view->surface instead of trying to cast a
view to a surface.
setup_output_destroy_handler() deal with output created at
drm backend initialize time.
handle_output_create() deal with output created by hot plug handler
output_destroy_handler is removed when output was unplugged or
shell is destroyed.
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
In drm backend, the cursor_surface->plane point to
drm_output->cursor_plane.when this output is removed,
drm_output->cursor_plane is destroyed, butcursor_surface->plane
still point to destroyed plane. So once mouse move to this
cursor_surface and system will repaint this cursor_surface,
segment fault will generate in weston_surface_damage_below() function.
V2:
-set surface->plane to NULL whose plane point to unplugged output,
then change weston_surface_damage_below() to do nothing if
surface->plane is NULL (Kristian)
-set surface->plane to NULL in weston_surface_unmap(),
so that all surfaces that have a non-NULL plane pointer wil be
on compositor->surface_list (Kristian).
bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=69777
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
How and when to update the keymap is left to each backend.
The new keymap only becomes effective when no keys are pressed and we
keep latched and locked modifiers from the previous state.
The weston_surface structure is split into two structures:
* The weston_surface structure storres everything required for a
client-side or server-side surface. This includes buffers; callbacks;
backend private data; input, damage, and opaque regions; and a few other
bookkeeping bits.
* The weston_view structure represents an entity in the scenegraph and
storres all of the geometry information. This includes clip region,
alpha, position, and the transformation list as well as all of the
temporary information derived from the geometry state. Because a view,
and not a surface, is a scenegraph element, the view is what is placed
in layers and planes.
There are a few things worth noting about the surface/view split:
1. This is *not* a modification to the protocol. It is, instead, a
modification to Weston's internal scenegraph to allow a single surface
to exist in multiple places at a time. Clients are completely unaware
of how many views to a particular surface exist.
2. A view is considered a direct child of a surface and is destroyed when
the surface is destroyed. Because of this, the view.surface pointer is
always valid and non-null.
3. The compositor's surface_list is replaced with a view_list. Due to
subsurfaces, building the view list is a little more complicated than
it used to be and involves building a tree of views on the fly whenever
subsurfaces are used. However, this means that backends can remain
completely subsurface-agnostic.
4. Surfaces and views both keep track of which outputs they are on.
5. The weston_surface structure now has width and height fields. These
are populated when a new buffer is attached before surface.configure
is called. This is because there are many surface-based operations
that really require the width and height and digging through the views
didn't work well.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
If we are about to finish a frame, but a redraw is pending and we let the
compositor redraw, we need to check for errors. If the redraw fails and
the backend cannot schedule a page-flip, we need to finish the frame,
anyway.
All backends except DRM use a timer to schedule frames. Hence, they cannot
fail. But for DRM, we need to be able to handle drmModePageFlip() failures
in case access got revoked.
This fixes a bug where logind+drm caused keyboard input to be missed as we
didn't reenable it after a failed page-flip during deactivation.
If the initial page-flip fails, immediately finish the frame to avoid
being stuck in the given frame. We already do this if we have no fbo
available. Now we do the same if the page-flip fails.
Instead of connecting to weston-launch from launcher-util, we now try to
connect to logind first. If logind provides session-devices, we use them.
If not, we fall back to the old weston-launch facility.
If you request a device via weston_launcher_open(), you should now release
it via weston_launcher_close() instead of close(). This is currently not
needed but will be required for logind devices.
This adds optional libdbus integration for weston. If libdbus is available
and not disabled via --disable-dbus during weston build, we now provide
basic DBusConnection main-loop integration for weston.
The dbus.c file provides a new helper to integrate any DBusConnection
object into a wl_event_loop object. This avoids any glib/qt/..
dependencies but instead only uses the low-level libdbus library.
Note that we do not provide dummy fallbacks for dbus helpers in case
dbus-support is disabled. The reason for that is that you need dbus/dbus.h
for nearly any operation you want to do via dbus. Therefore, only the most
basic helpers which can be used independently provide a "static inline"
dummy fallback to avoid #ifdef all over the code.
When the last input device with a certain capability is removed, unset
the focus of the seat device associated with the capability.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
This patch adds a new weston.ini key, gbm-format, to the [core] section.
This new key can be rgb565, xrgb8888 or xrgb2101010, and makes the
compositor use the corresponding GBM format for the framebuffer.
This used to work only for ARGB8888 and XRGB8888 buffers, but this lets
us support pageflipping to any client buffer as long as it matches the
framebuffer format.
The new key startup-animation in the [shell] section lets you
control the startup animation. Default is fade, but with this patch
we can also do none, which just show the desktop as fast as possible.
We check the state when we schedule a repaint, but we will still repaint
in weston_output_finish_frame() if a repaint is needed. Now we check
whether we went to sleep while waiting for the page flip and stop repainting
in that case.
https://bugs.freedesktop.org/show_bug.cgi?id=65802
If read() fails without EAGAIN/EINTR, the device is very likely dead.
However, we must not remove the device as it might be muted/revoked. So we
simply remove the event-source to avoid polling the device and simply wait
for the udev-remove signal now.
Note that we cannot call evdev_device_destroy() as the caller created the
FD and might need custom code to close it (like weston_launcher_close()).
We used to test for abs | rel | button, which inits a pointer device for
a device with just rel or abs capabilities. We now make sure we have either
rel or abs as well as button.
The time spent loading EGL and GLES libraries from disk can be a
considerable hit in some embedded use cases. If Weston is compiled
with EGL support, the binary will depend on those libraries, even if
a software renderer is in use.
This patch splits the GL renderer into a separate loadable module,
and moves the dependency on EGL and GLES to it. The backends still
need the EGL headers for the native types and EGLint. The function
load_module() is renamed to weston_load_module() and exported, so
that it can be used by the backends.
The gl renderer interface is changed so that there is only one symbol
that needs to be dlsym()'d. This symbol contains pointers to all the
functions and data necessary to interact with the renderer. As a side
effect, this change simplifies gl-renderer.h a great deal.
Previously if you move a window around and temporarily add a second
finger then it will cancel the grab even though the original finger is
still held on the screen. It seems more robust to avoid cancelling the
grab until all fingers have been removed.
Previously if you add a second finger while moving a window with a
touch grab then the position will keep jumping between the position of
each finger as you move them around. This patch changes it so that it
keeps track of the first touch id that starts the grab and only
updates the grab position when that finger moves.
Adds a new binding type for touch events via the new function
weston_compositor_add_touch_binding. The binding can only be added for
a touch down with the first finger. The shell now uses this to install
a binding to activate the current surface.
EGLInt is not always uint32_t so we need
to make sure we use the right int size for the format.
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
We get the child position but never use this information here. Just remove it.
Spotted by Christopher Michael.
Signed-off-by: Stefan Schmidt <s.schmidt@samsung.com>
Currently there is no guarentee that output remove event always happend after
page_flip event on the same output. So if the following situation occur:
first: unplug a output
second: output remove event arrive, output_destrory called adn free output
third: page_flip event arrive on the destroyed output
the segment fault will happpen in page_flip_handler().
This patch add a variable drm_compositor->destroy_pending, if page flip
event is pending when output remove event arrive, output_destroy will be
delayed until page flip finished.
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
This fixes an uninitialized serial error, were we could send out the
modifier event even if the client didn't have a pointer resource. We
send out the modifier event to let clients know the modifer mask when
they receive a pointer button event. Thus, if the client doesn't have
a pointer we don't need to send the modifier event.
Additionally we would send out the modifier event with an
uninitialized serial number.
Finally, this commit restores the order of sending the modifier event
before the enter, like it used to be. Not likely to be an issue,
since the client will always receive the modifier event before any
button event, but it's a little nicer to give the client the modifier
events before it receives any pointer events.
weston-launch has two supported use cases now: either launch from
and VT login as a regular user (running within that session) or
from an init script (systemd or such) with the -u option to create
a session for the specified user. Running from within X or ssh is
not possible. It's still possible to run weston as root from X or ssh
but that's strictly a development/debug/test feature.
https://bugs.freedesktop.org/show_bug.cgi?id=69727
This is only available when running weston directly as root typically for
ssh logins. It's a somewhat destructive option, as it will take over any
existing VT completely, unless there's already an display server running
there.
https://bugs.freedesktop.org/show_bug.cgi?id=69731
The wake handler set up by the shell will try to unlock the screen
which works by setting up an animation which fades in the display. The
animation is started by first scheduling a repaint. Subsequent
repaints are scheduled whenever the previous frame is finished.
However in the case of the wake up signal the state is still
WESTON_COMPOSITOR_SLEEPING when the animation is started.
weston_output_schedule_repaint() ignores attempts to schedule a
repaint if the compositor is sleeping which means the animation never
gets run and will never complete.
The animation gets unstuck and continues if anything else schedules a
repaint after the state has been changed so the bug only gets hit in
certain conditions. The first wake up creates the lock surface which
causes a redraw when the first buffer is attached so the first wake up
is always ok. A redraw can be triggered in the subsequent wake ups
just by moving the mouse.
A good way to trigger the bug is to try to wake up the compositor by
pressing the shift key. If you let the compositor go back to sleep
after waking it up without unlocking it, the second press of the shift
key will not cause a redraw so the animation will not run and it won't
fade in.
https://bugs.freedesktop.org/show_bug.cgi?id=69719
Add drm_set_master and drm_drop_master
as wrappers for drm(Set|Drop)Master, when building compositor-drm
or as empty functions otherwise.
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
Adding a check in weston_config_full_path so that
we don't crash if we started without a config file.
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Checking the return value from seteuid in
order to not launch clients with the wrong effective uid.
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
We'll add the GBM format code as the native visual ID for EGLConfigs
when running on GBM. This patch lets the drm backend pass in the
format code it's using with KMS and make sure we get a confing that
matches. In the future, mesa will add support for 10 bpc configs
which will match the "at least 8 color bits" requirement. By also
matching the native visual ID, we avoid rendering XRGB2101010 into a
XRGB8888 KMS framebuffer.
Previously only the touch up key event was used for single-touch
devices and the touch down event was generated on the first motion
event. This was breaking if the touch up and down events were sent
without a motion in-between because the evdev driver wouldn't generate
a touch down event and Weston would lose track of the number of touch
points that are down. This patch changes it to track the up and down
key events as pending events similar to how it does for multi-touch
devices.
https://bugs.freedesktop.org/show_bug.cgi?id=69759
Instead of having a mask of pending events there is now an enum with a
single value to represent the one pending event. The event gets
flushed explicitly as part of the handling code for each event type
rather than in the outer event reading loop. The pending event is used
so that we can combine multiple motion events into one and to make
sure that we have recieved the latest position before sending a touch
up or down event. This should fix the following problems with the old
approach:
• If you release a finger and press it down again quickly you could
get the up and down events in the same batch. However the pending
events were always processed in the order down then up so it would
end up notifying two down events and then an up. The pending event
is now always flushed when there is a new up or down event so they
will always be in the right order.
• When it got a slot event it would immediately change the slot number
and then set the pending event. Then when it flushed the events it
would use the new slot number to flush the old pending event so the
events could have the wrong finger. The pending event is now
immediately flushed when a slot event is received so it will have
the right finger.
• If you get more than 32 events in one read then it was resetting the
pending events before processing the next batch in
evdev_process_events. If four fingers were pressed down at once then
it ended up with more than 32 events and the sync message would be
in the second batch. The pending flag for the last finger was
getting cleared so it never got emitted. In this patch the pending
event is no longer reset after reading nor is it explicitly flushed.
Instead it is flushed when we receive a EV_SYN event or a different
pending event needs to replace it.
The touchpad handling code was trying to use the pending event
mechanism to notify the relative motion events. I'm not sure why it
was doing this because it looks the event would effectively get
emitted as soon as the touchpad_process function is finished anyway
and it wasn't accumulating the values. Instead I've just changed it to
emit the event directly.
https://bugs.freedesktop.org/show_bug.cgi?id=67563
Here too we must make sure the surface has a valid resource, as
there are some (xwayland, surfaces created by the shell) that
don't have it.
Fix a Weston crash when setting a mpv window fullscreen on drm.
If two fingers are released almost simultaneously then evdev can send
the touch up events in one bunch without sending a sync event
in-between. However, the evdev_device struct only keeps track of one
pending touch up event so in this case the second touch up event would
override the first and it would be lost. This patch changes it to also
flush the events whenever the slot changes so that it will flush the
previous touch up event before trying to queue the next one.
https://bugs.freedesktop.org/show_bug.cgi?id=67563
compositor.surface_list is recreated every redraw with the mapped
surfaces, but if a surface gets unmapped and then in the same frame
weston_compositor_pick_surface() is called we must make sure it
does not pick the unmapped surface, since it traverses the
surface_list to find one.
If after the unmap the surface gets also deleted it's even more
important, as it must not pick a destroyed surface.
When unmap and destroy a surface we need to make sure we don't pick it
before we rebuild the new surface list. Currently we ensure this
by rebuilding the surface list when destroying a surface, but just
removing the surface should be enough.
The Wayland protocol permits a client to request the pointer, keyboard
and touch multiple times from the seat global. This is very useful in a
component like Clutter-GTK where we are combining two libraries that use
Wayland together.
This change migrates the weston input handling code to emit the
events for all the resources for the client by using the newly added
wl_resource_for_each macro to iterate over the resources that are
associated with the focused surface's client.
We maintain a list of focused resources on the pointer and keyboard
which is updated when the focus changes. However since we can have
resources created after the focus has already been set we must add the
resources to the right list and also update any state.
Additionally when setting the pointer focus it will now send the
keyboard modifiers regardless of whether the focused client has a
pointer resource. This is important because otherwise if the client
gets the pointer later than you getting the keyboard then the
modifiers might not be up-to-date.
Co-author: Neil Roberts <neil@linux.intel.com>
This patch implements the notification of clients during mode_switch.
As discussed on IRC, clients are notified of mode_switch only when the
"native" mode is changed and activated. That means that if the native
mode is changed and the compositor had activated a temporary mode for
a fullscreen surface, the clients will be notified only when the native
mode is restored.
The scaling factor is treated the same way as modes.
the unmap event will be followed by the deletion of the weston_surface,
so the shell_surface will also be deleted by the shell. Having removed
the surface_destroy_listener, the surface_destroy callback doesn't
get called, so reset the value of shsurf here.
After the fade or zoom effects, alpha could not have been 1.0, preventing
not redrawing behind opaque windows.
This patch add a reset function in weston_surface_animation to reset
some variables the effects affect.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
The struct weston_launcher object will now either handle tty and vt switching
details in-process (when running weston directly as root) or talk to
the weston-launch process.
Previously, vaapi_recorder_frame() would wait until the encoded
contents for a frame is written to the output file descriptor. This
delayed the repainting of the next frame, and affected frame rate
when capturing with high resolutions. Instead, wait only if there is
and attempted to encode two frames at the same time.
Increases framerate from 30 to 60 fps when capturing at 1920x1200 on
my SandryBridge system, although there are periodic slowdowns due to
disk writes.
it uses the Android fbdev HAL[1] (through libhybris[2])
and the libhybris implementation of wayland-egl.
Configure flags:
cairo:
--enable-glesv2=yes --enable-egl=yes
weston:
--with-cairo-glesv2 --enable-fbdev-compositor
hybris:
--enable-wayland --enable-arch=x86
--with-android-headers=<android-headers> --enable-alinker=jb
The android headers are extracted from an AOSP tree,
using hybris/utils/extract-headers.sh
[1]:
https://github.com/android/platform_hardware_libhardware/blob/master/include/hardware/fb.h
[2]: https://github.com/libhybris/libhybris
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
Hi Kristian,
Here's a new patch for ref counting weston_xkb_info, as suggested.
So a seat created with a NULL keymap will now point to the global xkb_info.
Typically we can write it immediately without blocking, so save the overhead
of setting up an fd watch and writing the data in a callback. For the
case where the immediate write doesn't write all data, we fallback and
set up the fd watch as usual.
This patch also consolidates setting up the async write a bit.
This makes the drag-and-drop code available to in-weston data sources,
similar to how we can set a selection data source internally. The
wl_data_device.start_drag entry point now calls this function after
validating protocol arguments.
Another silent regression from the wl_resource opaquify effort. This was
causing our pageflip-to-client-buffer and sprites optimizations to
not kick in.
We used to destroy the frame window and reparent the client window to
wm_window. That means that we lose the destroy_notify event when the
client window is destroyed later, since we don't select for
substructure_notify on wm_window.
Instead of destroying and reparenting, just unmap the frame window.
This patch adds a feature to the DRM backend that uses libva for
encoding the screen contents in H.264. Screen recording can be
activated by pressing mod-shift-space q. A file named capture.h264
will be created in the current directory, which can be muxed into
an MP4 file with gstreamer using
gst-launch filesrc location=capture.h264 ! h264parse ! mp4mux ! \
filesink location=file.mp4
This is limitted to the DRM compositor in order to avoid a copy when
submitting the front buffer to libva. The code in vaapi-recorder.c
takes a dma_buf fd referencing it, does a colorspace conversion using
the video post processing pipeline and then uses that as input to the
encoder.
I'm sending this now so I get comments, but this is not ready for
prime time yet. I have a somewhat consistent GPU hang when using
i915 with SandyBridge. Sometimes a page flip never completes. If you
want to try this anyway and your system get stuck, you might need to
run the following:
# echo 1 > /sys/kernel/debug/dri/0/i915_wedged
After that, alt-sysrq [rv] should work.
Once that's fixed it would also be good to make the parameters used by
the encoder more flexible. For now the QP parameter is hardcoded to 0
and we have only I and P frames (no B frames), which causes the
resulting files to be very large.
GL_UNSIGNED_INT is only supported when GL_OES_element_index_uint is
available (mesa implements that extension). We don't need 32-bit
indices, so just use GL_UNSIGNED_SHORT.
We don't always get both an X and an Y event in a SYN report, so we end
up transforming the coordinate we don't get twice. For example, if we
only receive an ABS_X event, we transform the already transformed
device->abs.y again in transform_absolute() when applying the calibration.
Now that we use AC_SYS_LARGEFILE, we need to pull in config.h at least
whereever we use mmap(). Fixes at least the test-suite and simple-shm
on 32 bit systems.
window->x/y is the coordinate of the top-level surface (whether that's
the frame window or an override-redirect window) and the wayland surface
should be placed there, without the t->margin offset.
The coordinate transformation was broken (worked for first output where
output->x/y was 0,0, broke on all other outputs). We can just use
surface->geometry.x/y directly. We can't use the full transformation,
the best we can do is to move the X window to the geometry.x/y location.
Get rid of the static old_sx/sy hack as well.
We only get configure notify for toplevel (frame or override-redirect window)
and those are the cases where we want to update window->x/y. The way the
code worked, we'd exit immeidately in those cases and window->x/y would
not be updated.
This patch fixes a bug found by Marek Romanowic: the RDP peer output must
be enabled by default, or we have to unfocus/focus the RDP client window to
have disable/enable output messages sent (and finally receive updates).
config.h includes were missing in a few files, including input.c, the
lack of which caused the X11 backend to segfault instantly due to not
having an xkbcommon context.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
This allows a surface to live on after its resource has been
destroyed. The ref-count can be increased in a resource destroy signal
listener, to keep the surface around for a destroy animation, for example.
We don't handle them in any way now and having your steering wheel move
the cursor isn't useful. Applications can still open evdev devices and
access them directly like they already do.
This patch adds 3 new options to weston.ini to allow
the user to change default constant_accel_factor,
min_accel_factor and max_accel_factor. If no options
are set, it falls back using defaults as it did before.
v2: create weston_config_section_get_double and use it
instead of manualy converting string to double.
v3: add default values in weston_config_get_double
instead of using conditionals.
v4: don't pass diagonal as pointer.
Other clients of an evdev device need to have the events they receive
be separated, in moment in time, from other events by an EV_SYN/
SYN_REPORT. This is the responsibility of the client who writes events
into the stream.
This removes the use of wl_client_get_display() where the client is
derived from the focussed resource. This starts the removal of the
assumption of a single resource on a client that would be notified about
events on the focussed surface.
device->mt.slot is uninitialized when we're not receiving the
evdev slot events. Always use ID 0 as we do when we generate the
touch down and motion events.
The current code works if pw->pw_shell is bash because:
"If the shell is started with the effective user (group) id not equal to
the real user (group) id, and the -p option is not supplied, these actions
are taken and the effective user id is set to the real user id."
Thus, for bash, weston's EUID == UID.
For zsh, the -p option "is enabled automatically on startup if the effective
user (group) ID is not equal to the real user (group) ID."
Thus, weston's EUID = 0, and if pw_shell is zsh, /run/user/$UID/wayland-0 is
created with euid root and not writeable by the user, causing all clients to
fail.
Fix this by always dropping privileges to the user.
Regression introduced in 636156d.
For touchpads, device->dispatch is set up when exiting
evdev_handle_device() and a potential source for a memleak.
This can't actually happen at the moment, as evdev_handle_device() won't
fail for touchpads after setting up the dispatch but prevent this from
happening in the future.
If touches are already present on the device, absinfo has the currently
active touch slot. There's a race condition where the slot may change before
we enable the fd and we thus miss out on the ABS_MT_SLOT event. It's still
slightly more correct than assuming whatever comes next is slot 0.
For Protocol B devices, mtdev merely routes the events and is not needed.
For Protocol A devices, mtdev is needed, so fail for those devices now if we
mtdev fails.
mtdev as currently used in weston is a noop. mtdev's purpose is to convert
Protocol A devices (without ABS_MT_SLOT) to Protocol B devices (slots).
For Protocol B devices mtdev merely routes the events, so checking for
slots and then using mtdev based on that adds no functionality.
Check for ABS_MT_POSITION_X/Y instead and use that to categorise a device
as MT device. mtdev will provide us with a slotted protocol for all devices.
https://bugs.freedesktop.org/show_bug.cgi?id=54428
If weston is compiled against a gl2ext.h that doesn't have the subimage
extension, but then run against a gles2 library that does provide it,
we end up disabling the glTexImage2D falback without having the subimage
code paths compiled in.
Earlier versions of gl2ext.h defined the GL_EXT_unpack_subimage tokens
without the _EXT suffix. That's been fixed and we're using the _EXT
tokens now, but just in case we're using an older, broken header, define
the _EXT tokens manually.
It is defined by the mesa #include, which is just a copy of the official
Khronos header. It's just defined in a different section than the
extension tokens. In the mean time, the extension tokens were renamed
to add a _EXT suffix (eg GL_UNPACK_ROW_LENGTH -> GL_UNPACK_ROW_LENGTH_EXT)
and we silently failed to used the subimage extension.
fbdev_output_destroy will free the memory passed into in and since we
want to pass the device name into fbdev_output_create we need to save
this to an intermediate value
Prior to freeing the memory in which the link node for the signal is
emedded we should remove the link node from the list to prevent the list
from being corrupted.
https://bugs.freedesktop.org/show_bug.cgi?id=67231
Prior to freeing the memory in which the link node for the signal is
emedded we should remove the link node from the list to prevent the list
from being corrupted.
https://bugs.freedesktop.org/show_bug.cgi?id=67231
Rather than using the dimensions in the mode we can use the recently
added output width and height members which are updated to reflect any
output rotation.
https://bugs.freedesktop.org/show_bug.cgi?id=66798
Don't NULL the resource pointer before calling weston_surface_destroy().
We use to have more of a distinction between compositor created surfaces
and client surfaces, and weston_surface_destroy couldn't be used for
client surfaces. Now it all goes through weston_surface_destroy() and
we can remove the assert and the NULL-ing of resource, which caused the
marshalling warning.
The list of surfaces used by weston_compositor_pick_surface() is
maintained in list of surfaces stored on the compositor. This list is
generated from the surfaces across all the layers using
weston_compositor_build_surface_list.
When destroying a surface the surface is "unmapped" with
weston_surface_unmap which removes it from the layer list. However since
the compositor surface list was only being rebuilt when the output was
repainted a call to weston_compositor_pick_surface before the next
output repaint would use an outdated surface list containing surfaces
that have been partially destroyed.
https://bugs.freedesktop.org/show_bug.cgi?id=65986https://bugs.freedesktop.org/show_bug.cgi?id=66173https://bugs.freedesktop.org/show_bug.cgi?id=66198
We can get a destroy notify for the frame window after we've removed it
from the hash table. This turns into a NULL pointer deref when we look up
the window and try to use it for debugging printout.
Fixes the failing xwayland test case.
Calling weston_pointer_start_grab can lead to a code path (in this case
when the shell surface is unresponsive) that can try and remove the
popup grab to setup a shell grab.
Ending the popup grab requires removing the surface from the grab's
surfaces list - however the grab had not yet been fully setup so the
grabbed surface was not yet in this list.
With this change we ensure we add the surface to the list before setting
up the pointer grab.
https://bugs.freedesktop.org/show_bug.cgi?id=66167
Make sure the center point of a rotation is not rounded to an
integer. It makes the calculation consistent with others in the
shell. It also ensures surfaces rotated 180 degrees are at the
exact same place.
This commit sets the version numbers for all added/created objects. The
wl_compositor.create_surface implementation was altered to create a surface
with the same version as the underlying wl_compositor. Since no other
"child interfaces" have version greater than 1, they were all hard-coded to
version 1.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Window contents cannot be assumed to be fully opaque for windows drawn with
a RGBA visual. The optimization of setting a full opaque region is limited to
windows with a color depth != 32.
In embedded environments, devices that appear as evdev "keyboards" often
have no resemblence to PC-style keyboards. It is not uncommon for such
environments to have no concept of modifier keys and no need for XKB key
mapping; in these cases libxkbcommon initialization becomes unnecessary
startup overhead. On some SOC platforms, xkb keymap compilation can
account for as much as 1/3 - 1/2 of the total compositor startup time.
This patch introduces a 'use_xkbcommon' flag in the core compositor
structure that indicates whether the compositor is running in "raw
keyboard" mode. In raw keyboard mode, the compositor bypasses all
libxkbcommon initialization and processing. 'key' events containing the
integer keycode will continue to be delivered via the wl_keyboard
interface, but no 'keymap' event will be sent to clients. No modifier
handling or keysym mapping is performed in this mode.
Note that upstream sample apps (e.g., weston-terminal or the
desktop-shell client) will not recognize raw keycodes and will not react
to keypresses when the compositor is operating in raw keyboard mode.
This is expected behavior; key events are still being sent to the
client, the client (and/or its toolkit) just isn't written to handle
keypresses without doing xkb keysym mapping. Applications written
specifically for such embedded environments would be handling keypresses
via the raw keycode delivered as part of the 'key' event rather than
using xkb keysym mapping.
Whether to use xkbcommon is a global option that applies to all
compositor keyboard devices on the system; it is an all-or-nothing flag.
This patch simply adds conditional checks on whether xkbcommon is to be
used or not.
v3 don't send zero as the file descriptor - instead send the result of
opening /dev/null
v2 by Rob Bradford <rob@linux.intel.com>: the original version of the
patch used a "raw_keycodes" flag instead of the "use_xkbcommon" used in
this patch.
v1: Reviewed-by: Singh, Satyeshwar <satyeshwar.singh@intel.com>
v1: Reviewed-by: Bob Paauwe <bob.j.paauwe@intel.com>
This change tweaks weston_pointer_clamp to take into consideration if a
seat is constrained to a particular output by only considering the
pointer position valid if it is within the output we a constrained to.
This function is also used for the initial warping of the pointer when a
constraint is first established.
The other two changes are the application of the constraint when either
a new device added or a new output created and therefore outputs and
input devices can be brought up in either order.
v2: the code in create_output_for_connector has been spun off into a
new function setup_output_seat_constraint (Ander). The inappropriate
warping behaviour has been resolved by using weston_pointer_clamp
(Pekka).
This refactors the code out from clip_pointer_motion into a function of
its own which can then be used elsewhere to clamp the pointer
coordinates to the range of the outputs.
This change also makes the caller of clip_pointer_motion use this new
function.
This change spills the code for looking up a seat by name and then
potentially creating it if it doesn't exist into a new function called
udev_seat_get_named.
This change allows us to reuse this code when looking up the seat
when parsing seat constraints per output.
The fix to not call glTexImage2D() on every attach does not properly
set the texture damage region appropriately when the surface has a
buffer transform with 90 or 270 degrees rotation, since it would simply
multiply the buffer dimensions by the buffer scale, but in this case
width and height are inverted.
A possible fix for this would be to add the properly transformed region
to the texture damage region. However, there is a conversion back to
buffer coordinates when doing the actual upload and the entire buffer
needs to be uploaded anyway. So we just set a flag signalling that and
handle that special case in gl_renderer_flush_damage().
Checking for gs->num_images for determining the previous buffer type
when attaching is not reliable. The number of images is never cleared
in the SHM path, so after a switch from an EGL buffer to SHM, every
following attach of an SHM buffer will happen with gs->num_images > 0,
and the code will assume the previous buffer was an EGL one.
Fix this by adding a buffer_type field to gl_surface_state.
We were assigning drag from the resource user data, which was wrong
(resource data is the weston_seat) and confusing since drag is later
assigned newly malloc()ed memory.
We used to refcount the data source, but switched to using a destroy signal
instead. When we switched we forgot to free the source insted of
unreffing it.
With the change to move free()ing of the wl_resource into wayland-server, we now have
a few cases where we double free the resource in the destructor. This patch
removes those.
This commit adds a weston_buffer structure to replace wl_buffer. This way
we can hold onto buffers by just their resource. In order to do this, the
every renderer.attach function has to fill in the weston_buffer.width and
weston_buffer.height fields.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Because of its links to selection.c and xwayland, a destroy_signal field
was also added to wl_data_source. Before selection.c and xwayland were
manually initializing the resource.destroy_signal field so that it could be
used without a valid resource.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Instead, forward signal to weston and wait for weston to clean up nicely.
Weston relies on weston-launch being around to shut down correctly,
so don't exit until we get the SIGCHLD from weston. This make
killall weston-launch work properly.
https://bugs.freedesktop.org/show_bug.cgi?id=62910
When the spring goes outside the envelope, we have a few options for
bringing it back: either just let it overshoot, bounce off the limit or
just clamp it. Instead of controlling that with #ifdef, let's make it
a part of the spring state.
The weston_spring is a very flexible and powerful mecanhism for driving
animations. However, it can be a little difficult to tame, but this
little helper can plot the response of the spring to a set of initial
parameters and makes it easy to tune and tweak the spring behavior.
This commit converts shell_surface.resource to a pointers and updates
shell.c to use wl_resource_get accessors for shell_surface, desktop_shell,
screensaver, and workspace_manager related resources.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
xeyes works as expected now. subwindows are popped also as expected. This
patch should fix the following:
https://bugs.freedesktop.org/show_bug.cgi?id=59983
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
When printing debug information about atoms, the XWM would crash if the X
server failed to respond to a request about atom names. In practice this
occurred when the server itself crashed, e.g. when starting mplayer with the
"xv" vo.
This is the first in what will be a series of weston patches to convert
instances of wl_resource to pointers so we can make wl_resource opaque.
This patch handles weston_surface and should be the most invasive of the
entire series. I am sending this one out ahead of the rest for review.
Specifically, my machine is not set up to build XWayland so I have no
ability to test it fully. Could someone please test with XWayland and let
me know if this causes problems?
Because a surface may be created from XWayland, the resource may not always
exist. Therefore, a destroy signal was added to weston_surface and
everything used to listen to surface->resource.destroy_signal now listens
to surface->destroy_signal.
v4:
Incorporated krh and anderco's comments. Now adding newly allocated
buffer's dimensions to texture_damage
v3:
* Removed unnecessary parentheses
* Added check for switching from EGL image to SHM buffer
* Moved shader assignment out of IF condition
v2:
Fixed the wrong comparison
v1:
Depending on specific DRI driver implementation, glTexImage2D() with data
set to NULL may or may not re-allocate the texture buffer each time it is
called. Unintended consequences happen if later glTexSubImage2D() is called
to only update a sub-region of the texture buffer.
I've explored moving glTexImage2D() from gl_renderer_attach() and simply
mark the texture dirty, but the current implemention seems cleaner because
I won't have to worry about calling ensure_textures() and re-assigning
gs->shader unnecessarily.
AC_USE_SYSTEM_EXTENSIONS enables _XOPEN_SOURCE, _GNU_SOURCE and similar
macros to expose the largest extent of functionality supported by the
underlying system. This is required since these macros are often
limiting rather than merely additive, e.g. _XOPEN_SOURCE will actually
on some systems hide declarations which are not part of the X/Open spec.
Since this goes into config.h rather than the command line, ensure all
source is consistently including config.h before anything else,
including system libraries. This doesn't need to be guarded by a
HAVE_CONFIG_H ifdef, which was only ever a hangover from the X.Org
modular transition.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
[pq: rebased and converted more files]
By labelling devices with ENV{WL_SEAT} in udev rules the devices will be
pulled into multiple weston seats.
As a result you can get multiple independent seats under the DRM and
fbdev backends.
And as a result of this stop iterating through the compositor seat list
(of one item) and instead access the udev_input structure directly.
This enables a refactoring to pull out the weston_seat into a separate
structure permitting multiple seats.
The kernel is supposed to set this when drmModeSetCrtc() is called but
at least the i915 driver wouldn't do that in all cases. A fix for this
should be released with kernel 3.10, but we work around the issue in
older kernels by always forcing DPMS to ON when doing a mode set.
https://bugs.freedesktop.org/show_bug.cgi?id=64873
exit() calls atexit() handlers and C++ destructors (e.g. a C++
weston module) which could destroy state that the main process
depends on (e.g. ioctl's, tmpfiles, sockets, etc...). If an exec
fails, call _exit() instead of exit().
v2: prefer _exit over _Exit
Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
The old code had an off-by-one error on the y coordinate
where it says height - (cur_y - y). And it does the vflipping of
the *destination* buffer, whereas what is really needed is to
vflip the whole source buffer. This only affects when you read
out part of the image, such as when using the screen recoder.
Also, instead of doing the flipping manually we just let pixman
handle it.
When a window is fullscreened with DRIVER method and we succeeded
in changing mode we need to actually move the surface to the
origin of the output, or it won't be used for scanout.
