Commit Graph

28818 Commits

Author SHA1 Message Date
Carlos Garnacho
58bcd30ee6 clutter: Always unregister point on GRAB_NOTIFY leave event
The ClutterGestureAction base code would correctly try to cancel a
gesture if it would receive GRAB_NOTIFY leave events (that would indicate
other portions of the actor tree stole input away from the gesture actor),
but it would mistakenly do so only if the gesture was already initiated,
possibly leaving stale point information if the gesture collected input
but didn't initiate yet.

This could be indirectly seen clicking with the mouse on OSK keys with
no motions in between, clicks would accumulate on the swipeTracker
gestures until the trigger point, so the third click could drag the
workspaces.

We do always want to unregister the related device/sequence here, do that
while still cancelling any already initiated gesture.

Fixes: https://gitlab.gnome.org/GNOME/mutter/-/issues/1907
Fixes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4987
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2334>
2022-03-28 17:11:02 +00:00
Jonas Ådahl
5c5e30638d ci: Run KVM tests
This commit makes the KVM tests run in special VM runners tagged with
the 'kvm' tag. In order to avoid building the kernel image used for
running the tests each pipeline, it's built as part of the CI image
building.

For now, KVM tests are only run on the x86_64 architecture. The reasons
for this are two that the kernel image building script doesn't yet handle
any other architecture than x86_64 due to differences in how the image
is built and handled, as well as the fact that there only exists a kvm
tagged runner for x86_64.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Bartłomiej Piotrowski
7a7fb863a2 tests/virtme-run: Expose two CPUs
We're multi threaded, so make things more like reality by exposing more
CPUs using the '-smp 2' qemu option.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Bartłomiej Piotrowski
7435bfd0e5 tests/virtme-run: Disable pdcm
There are issues when running a nested KVM with Fedora 35 when pdcm is
enabled. Work around this for now by disabling it.

Related: https://bugzilla.redhat.com/show_bug.cgi?id=2022075
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Bartłomiej Piotrowski
4ce233b89e tests/virtme-run: Create fake stdin
When running in CI, the stdin may be /dev/null, which causes issues
for qemu. Avoid this issue by creating our own fake stdin.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Jonas Ådahl
31bc1c80ab tests/kvm: Forward the test environment variables to the VM
This will help with finding the right libdefault.so and correctly setup
the GLib test framework.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Jonas Ådahl
3650cb941f ci: Break up some long meson lines
Placing one configuration option per line makes it easier to see what is
enabled or not.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Jonas Ådahl
bc2d5861b0 build: Include KVM test enablement in summary
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Jonas Ådahl
189068d465 tests/build: Set up test environment using a dict()
This will allow us to reuse the keys and values more easily, as later
commits will rely on being able to iterate over the keys and values to
construct explict env strings for passing into special test cases.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2152>
2022-03-28 16:17:21 +00:00
Christian Hergert
858f89084e winsys: Check for KHR_swap_buffers_with_damage
Previously, we would only check for EXT_swap_buffers_with_damage which
generally will find an implementation. However, some EGL implementations
do not appear to support that naming of the extension, preferring to
only advertise KHR_swap_buffers_with_damage.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2316>
2022-03-28 15:23:28 +00:00
Nathan Follens
3c9622bea9 Update Dutch translation 2022-03-25 11:48:42 +00:00
Robert Mader
f9857cb8bd wayland/surface: Disable WL_SURFACE_ERROR_INVALID_SIZE check again
The check was introduces in ab17c98c94
but turned out to trigger client crashes in common cases.
Turn it into a warning for now until clients are fixed.

Closes https://gitlab.gnome.org/GNOME/mutter/-/issues/2083
Closes https://gitlab.gnome.org/GNOME/mutter/-/issues/2163

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2335>
2022-03-24 15:15:08 +00:00
Jonas Dreßler
0280b0aaa5 events: Use the event target actor to determine window for event
We use get_window_for_event() to check whether an event happened on top
of a window or on top of shell UI to decide whether to bypass delivering
the event to Clutter. In case of crossing events though, we can't just
use the device actor to determine whether to forward the event to
Clutter or not: We do want to forward CLUTTER_LEAVE events which
happened on top of shell UI. In that case the device actor is already a
window actor (the pointer already is on top of a window), but the shell
still needs to get the LEAVE crossing event.

Since the event source actor got removed from the detail of
ClutterEvent, the context we're looking for (which actor did the pointer
leave) is now the target actor that the event gets emitted to. Since the
last commit, we also made event filters aware of this context by passing
the target actor to them, so use this context now to determine whether
we're on top of a window or not.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2321>
2022-03-21 23:08:40 +00:00
Jonas Dreßler
2aad56b949 clutter: Pass target actor of events to event filter functions
We'll need the additional context of which actor the event will be
emitted to in mutters event filter (see next commit), so pass that
target actor to the event filters that are installed.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2321>
2022-03-21 23:08:40 +00:00
Sebastian Keller
703d8a77f1 clutter: Mark clutter_stage_grab() return value as transfer full
The caller is expected to unref the returned grab and all C code already
does this, but grabs requested from js were resulting in the memory
getting leaked due to it being annotated as transfer none.

Fixes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2189
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2345>
2022-03-21 21:25:28 +01:00
Марко Костић
8edd2c7d06 Update Serbian translation 2022-03-19 08:03:20 +00:00
Milo Casagrande
7ead31aac9 Update Italian translation 2022-03-17 08:35:28 +00:00
Charles Monzat
882af01d9d Update French translation 2022-03-15 15:43:47 +00:00
Balázs Úr
6b3ab88fcd Update Hungarian translation 2022-03-13 01:16:52 +00:00
Florian Müllner
9249aba72a Bump version to 42.0
Update NEWS.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2337>
2022-03-12 16:57:23 +01:00
Alexander Shopov
73780ea767 Update Bulgarian translation 2022-03-11 07:03:09 +00:00
Fabio Tomat
6ad6e92e74 Update Friulian translation 2022-03-08 06:15:47 +00:00
Florian Müllner
6effaae7a7 Bump version to 42.rc
Update NEWS.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2326>
2022-03-07 15:43:16 +01:00
Jonas Ådahl
11ae4c2084 ci: Put wild cards in single quotes
To not expand coverage-* into actual coverage files, put the argument
in quotes, so that all files are added.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2324>
2022-03-07 09:54:43 +00:00
Ivan Molodetskikh
690b8806dd backends/native: Change uint32_t modifiers to uint64_t
They were erroneously cast to uint32_t. In particular this prevented the
invalid modifier from working as it is equal to a 64-bit -1.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2322>
2022-03-06 16:56:27 +03:00
Piotr Drąg
a0082c3148 Update Polish translation 2022-03-06 12:44:13 +00:00
Daniel Șerbănescu
9e3aba08ce Update Romanian translation 2022-03-06 11:34:14 +00:00
Gwan-gyeong Mun
756b9113af Update Korean translation 2022-03-05 18:04:30 +00:00
Robert Mader
7f93004809 compositor/native: Check that the surface can be scanned out untransformed
Before scanning out the surface of a native client we have
to check the following attributes that influence the
relationship between buffer and the defined result on screen:
 - buffer scale
 - buffer transform
 - viewport

In the future we can loose these checks again in cases where the
display hardware supports the required operations (scaling, cropping
and rotating).

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2276>
2022-03-05 14:14:37 +00:00
Robert Mader
6330acfaa2 compositor: Some header cleanups
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2276>
2022-03-05 14:14:37 +00:00
Sebastian Keller
251bb45ec0 window/wayland: Always include a size when sending a configuration
Prior to 67033b0a mutter was accidentally including sizes for
configurations that were just focus state changes. This was not leading
to any known problems on the client side, but it was causing issues in
mutter itself when detecting whether a resize originated from the client
or the server.

Not including sizes in focus change configurations anymore however
revealed a bug in gtk. It was storing the window size when in a fixed
size mode (tiled/maximized/fullscreen), but not on any other server side
resizes. It was then restoring this stored size whenever there was a new
configuration without a size while in floating mode, i.e. the focus
change configurations generated by mutter after 67033b0a.

This change now addresses the issue 67033b0a was fixing in a way that
restores the previous behavior of always including the size whenever
sending a configuration.

Fixes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2091
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2238>
2022-03-05 13:18:58 +00:00
Jonas Ådahl
a3d689e711 meson: Bump meson version requirement
We use features from 0.55.0 in the KVM test paths; make that requirement
explicit.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2320>
2022-03-05 13:08:17 +01:00
Jonas Dreßler
0fa64f15e3 clutter/base-types: Remove weird random unicode char
Have I discovered the hidden treasure that the evil wizard has buried
when he ruled the land of Clutter through its golden years?

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2298>
2022-03-05 09:51:44 +00:00
Jonas Ådahl
cccf133f55 ci: Increase warnlevel from 1 to 2
This will hopefully catch more things than -Wall.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
eeb96e95a5 meson: Stop passing -Wall
Passing -Wall results in meson logging a warning, suggesting to use
different warning levels instead. Lets do what meson say.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
6916402456 meson: Set the correct required version
We use features from 0.54.0 so make that requirement explicit.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
46f81afb3b kms/impl-device/atomic: Fix function type signature
The property enum type was wrong.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
590e32dc62 wayland/tablet-pad: Fix fallthrough warning
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
339c315ed9 input-settings/native: Fix switch fallthrough warning
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
43870a5f55 input-settins/x11: Fix switch fallthrough warning
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
79319a3a91 barrier/native: Fix warning about implicit enum conversion
Also add static asserts to make sure the values match so casting can be
done.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
ab21bedcfc clutter/gesture-action: Fix a fallthrough warning
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
863a519f00 clutter/desaturate-effect: Use fabs() with passed double
Fixes a warning about passing a double to fabsf().

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
ad76c60ce3 clutter/click-action: Fix a couple of fall-through warnings
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
63cc69b342 cogl/tests: Fix warning about passing a double to fabs
The warning was in a home baked g_assert_float_with_epsilon(). Change to
use the glib one directly.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Jonas Ådahl
c9125aba73 cogl/gl: Fix switch passthrough warnings
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2319>
2022-03-04 23:12:24 +00:00
Carlos Garnacho
40627f88cf ci: Run coverage in all tests, and merge results
Instead of having different coverage outputs for different architectures
and compilation flags, have each of those tests run coverage in order
to generate a JSON report, and have that merged at a final common job.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2315>
2022-03-04 22:47:04 +00:00
Sebastian Wick
a76542e191 meson: Let the linker create RPATH instead of RUNPATH tag
Running meson test from the build directory in a jhbuild environment
picks up libraries which have been installed previously because
LD_LIBRARY_PATH takes precendence over RUNPATH. Make the linker generate
the RPATH tag again.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2307>
2022-03-04 22:16:12 +00:00
Jonas Ådahl
1914f533b8 tests/screen-cast: Add input injection tests to screen cast client
This tests that cursor metadata are handled properly, and that the
display server handles input events correctly, including ones that are
sent too early.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2270>
2022-03-04 18:13:45 +00:00
Jonas Ådahl
1e50ca0057 remote-desktop/virtual: Drop absolute events if not active
If the remote desktop service emits absolute input events (e.g. absolute
pointer events) before the stream has started streaming, we don't have a
virtual monitor, as the size has not been negotiated yet. When this
happens, just drop the event. Remote desktop services should probably
make sure not to send events before the streaming has started, but them
doing so anyway shouldn't trigger a crash, which would be the case
otherwise.

Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2270>
2022-03-04 18:13:45 +00:00