It was dropping to zero after the first frame because it hadn't been
incremented high enough. So the second frame would crash with:
```
#0 g_type_check_instance_cast
#1 META_DRM_BUFFER
#2 copy_shared_framebuffer_cpu
```
That's the CPU-copy path (fallback-fallback) that probably no one is using
but it does work after this fix. Exactly the same issue as was fixed
in `copy_shared_framebuffer_primary_gpu` by 36352f44f9.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2104>
A client can create a token without any seat, serial, or surface. In
this case, we'd still try to grab, which would run into some unforseen
code paths, potentially resulting in the following crash:
0) meta_wayland_tablet_seat_device_added (tablet_seat=0x55dff4271c90,
device=0x7f87b80655b0) at
../src/wayland/meta-wayland-tablet-seat.c:200
1) meta_wayland_tablet_seat_new (seat=0x0, manager=0x55dff3ec7b40) at
../src/wayland/meta-wayland-tablet-seat.c:283
2) meta_wayland_tablet_manager_ensure_seat (manager=manager@entry=0x55dff3ec7b40,
seat=seat@entry=0x0) at
../src/wayland/meta-wayland-tablet-manager.c:239
3) meta_wayland_tablet_manager_ensure_seat (seat=0x0, manager=0x55dff3ec7b40) at
../src/wayland/meta-wayland-touch.c:595
4) meta_wayland_seat_get_grab_info (seat=0x0, surface=0x55dff43ff5b0,
serial=0, require_pressed=0, x=0x0, y=0x0) at
../src/wayland/meta-wayland-seat.c:479
5) activation_activate (...) at
../src/wayland/meta-wayland-activation.c:261
Fix this by not trying to grab if not enough parameters was passed when
creating the token. Also add a test case that reproduces the above
crash.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2081>
When an activation times out, we'll be signalled two signals on the
startup sequence object: "timeout", and "complete".
Normally, the "complete" signal is emitted when a startup sequence is
completed succesfully by it being used for activation, and in this case,
the xdg_activation implementation should remove the sequence from the
startup notification machinery.
However, in the timeout case, we should not remove it, as the startup
notification machinery itself will deal with this. If we would, we'd end
up with use-after-free issues, as the sequence would be finalized when
removed the first time.
To avoid this, just clean up the Wayland side in the "timeout" signal
handler, leaving the "complete" signal handler early out if it was
already handled by it.
This avoids crashes like:
0) g_type_check_instance (type_instance=type_instance@entry=0xdd6740)
1) g_signal_handlers_disconnect_matched (instance=0xdd6740, ...)
2) meta_startup_notification_remove_sequence (sn=0x4cc890,
seq=0xdd6740) at
../src/core/startup-notification.c:544
3) startup_sequence_timeout (data=0x4cc890, ...) at
../src/core/startup-notification.c:504
4) g_timeout_dispatch (...) at ../glib/gmain.c:4933
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2081>
We set it via setenv(), and might not have the MetaX11Display at hand.
This fixes a crash when the stuck-client dialog (using zenity) appears
without any X1 client having appeared.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2081>
This test ensures that windows that were resized such that they extend
beyond the screen will be moved to be fully on the screen (if possible).
This has been working on X11 since forever, but on Wayland only since
the last commit.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2103>
The previous code was trying to detect client resizes by only
considering resizes without any pending configurations as client
resizes. There can however be pending configurations that do not involve
resizing, such as ones triggered by state changes. These may also stay
unacknowledged by the client until the next size change. This was
causing client resizes after showing the window (and therefore changing
its status to focused) to not be detected as client resize.
Fix this by checking whether the queue has any configuration with size
changes rather than just whether it is empty.
Fixes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2023
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2103>
Used to log multiple line entries. Just make continue prefix things, no
need to mess with maybe-prefixing; it'll just complicate using some less
custom logging functionality.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2094>
If some connectors disappeared, but the rest didn't change, we missed
actually removing the ones that disappeared, as we incorrectly assumed
nothing changed. Fix this by only assuming nothing changed if 1) we
didn't add any connector, and 2) we have the same amount of connectors
as before the hotplug event. The connector comparison checking makes
sure we report changes if anything of the still available connectors
changed.
Fixes: a8d11161b6
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2007
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2097>
When a docking station is disconnected, a few previously existing DRM
connectors may now be gone. When this happens, getting them via the
libdrm API results in NULL pointers returning, and we need to handle
this gracefully by making sure the connector state is properly updated.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2097>
The 'stop_after_next' will execeute one command, then not return to the main
loop until a 'continue' command is passed. Commands will still be
processed between 'stop_after_next' and 'continue'.
This is intended to be used to induce race conditions.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2066>
Currently the stored unconstrained_rect is only ever updated if there
was a move, resize or state change according to the move_resize_internal
implementation. For Wayland windows however resizes or state changes
are done in two steps, first the new configuration is sent to the client
and then once client acknowledges it, it is set on the mutter side in
another move_resize_internal call. Only the second call would result in
the unconstrained_rect being updated.
This started causing problems when unfullscreening windows was
immediately followed by a strut change. These strut changes started
happening in gnome-shell due to the visibility of the panel now being
considered for the struts and the presence of a fullscreen causing it to
be hidden until unfullscreen. In this situation first the unfullscreen
would resize the window to its pre-fullscreen size as expected, but then
the strut change triggers another window resize. This window resize is
based on the stored unconstrained_rect, which is still at the fullscreen
size because the unfullscreen resize only has sent its configuration,
but it has not been acknowledged yet. As a result the strut change
causes a resize to the fullscreen size which due to the constraints now
looks like a maximized window.
To fix this always update the unconstrained_rect when the requested size
has changed, but not when a previous request has been acknowledged
unless it is originating from the client itself.
If this included the move_resize_internal call from acknowledging the
size as well, it would be possible for this to be delayed long enough on
the client side to overwrite an intermediate resize originating from
mutter. And if this did not include resizes originating from the client,
clients would not be able to set an initial window size.
Fixes https://gitlab.gnome.org/GNOME/mutter/-/issues/1973
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2066>
meta_window_wayland_finish_move_resize() is called for both, finishing
a resize that has been requested through/by mutter and for resizes
directly done by the client. This introduces a CLIENT_RESIZE flag to
differentiate the former from the latter. Having this distinction is
required to know what the last requested size by either the client or
mutter is while ignoring older requests that might only have been
applied now.
This excludes client resizes when there are still pending
configurations, because the resize is known to be only temporary.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2066>
Because POSIX sh was, with hindsight, not a particularly well-designed
programming language, if we don't 'set -e', then we'll respond to failure
of a setup command such as cd by carrying on regardless.
Signed-off-by: Simon McVittie <smcv@debian.org>
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2009>
The assumption here seems to be that it's an overlay onto the
current environment which would make sense; but the implementation in
gnome-desktop-testing currently removes all other environment variables
(see GNOME/gnome-desktop-testing#1). This causes test failure when the
tests are run in Debian's autopkgtest framework, possibly because PATH
is cleared.
Signed-off-by: Simon McVittie <smcv@debian.org>
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2009>
Systems with AMD GPUs do not take advantage of Mutter's zero-copy path
when driving DisplayLink screens. This is due to a very slow CPU access
to the zero-copy texture. Instead they fall back on primary GPU doing a
copy of the texture for fast CPU access. This commit accelerates texture
copy by working through damage regions only.
Tests on a 4K screen with windowed applications show significant
reduction of GPU utilisation.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2033>
If one would end up with an actor attached to mapped actor, where the
attached actor doesn't itself have an up to date stage view list while
listening on the stage for updating, when clearing the stage views of
the list, anything that would query the stage views list at this time
would end up accessing freed memory.
This could happen if
1) An actor was added to a newly created container actor attached to
the stage
2) The actor got a timeline attached to it
3) The actor was moved to a container that already was mapped
4) A hotplug happened
After (1) both the container and actor would not have any stage views.
After (2) the timeline would listen on the stage for stage views
updates. After (3) the actor would still listen on the stage for stage
views updates. When (4) happened, the actor would be signalled when the
stage got its stage view cleared, at which point it would traverse up
its actor's tree finding an appropriate stage view to base its animation
on. The problem here would be that it'd query the already mapped
container and its yet-to-be-cleared stage view list, resulting in
use-after free, resulting in for example the following backtrace:
0) g_type_check_instance_cast ()
1) CLUTTER_STAGE_VIEW ()
2) clutter_actor_pick_frame_clock ()
3) clutter_actor_pick_frame_clock ()
4) update_frame_clock ()
5) on_frame_clock_actor_stage_views_changed ()
6) g_closure_invoke ()
7) signal_emit_unlocked_R ()
8) g_signal_emit_valist ()
9) g_signal_emit ()
10) clear_stage_views_cb ()
11) _clutter_actor_traverse_depth ()
12) _clutter_actor_traverse ()
13) clutter_actor_clear_stage_views_recursive ()
14) clutter_stage_clear_stage_views ()
...
Avoid this issue by making sure that we don't emit 'stage-views-changed'
signals while the actor tree is in an invalid state. While we now end up
traversing tree twice, it doesn't change the Big-O notation. It has not
been measured whether this has any noticible performance impact.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/1950
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2025>
Certains keys (such as ~ and |) are in the keyboard map behind the
second shift level. This means in order for them to be input, the
shift key needs to be held down by the user.
The GNOME Shell on-screen keyboard presents these keys separately on
a page of keys that has no shift key. Instead, it relies on mutter
to set a shift latch before the key event is emitted. A shift latch
is a virtual press of the shift key that automatically gets released
after the next key press (in our case the ~ or | key).
The problem is using a shift latch doesn't work very well in the face
of key repeat. The latch is automatically released after the first
press, and subsequent repeats of that press no longer have shift
latched to them.
This commit fixes the problem by using a shift lock instead of a shift
latch. A shift lock is never implicitly released, so it remains
in place for the duration of key repeat.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2045>
The `guess_candidates()` function scores each display that an input
device could be mapped to and then uses the `sort_by_score()` comparator
to find the best option. The function expects the list to be sorted from
best to worst, but the comparator currently sorts them in the opposite
order. This causes the function to end up returning the _worst_ match
rather than the the best. This commit reverses the sort order of the
comparator so that the best display can be returned as intended.
Closes: #1889
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1934>
This does two things to frown upon:
- Modifies ClutterEvent structs, while the effort is to have those
completely opaque, and readonly after creation from the input
thread side.
- Stores state in the ClutterInputDevice struct, event though those
are also considered static after creation, managed by the input
thread, etc.
Stop doing that. This makes all events just forwarded as-is in
the ClutterStage/clutter-main.c code.
Handling of click count sounds like material for a ClutterGestureAction
(or perhaps ClutterClickAction), all of both callers now do it in place
at the moment, while gestures lack a better state tracking and management.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2024>
