Summary:
Currently, testDontCrashReinitializeCompositor is unstable because it
doesn't take into account that effects can be loaded asynchronously.
In general, we don't need all default effects in this test, so let's
disable them.
Test Plan: This should fix ASan issues.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17386
Summary:
If updating a shadow means that there should be no shadow, shadow ends
up deleting itself midway through the method.
It's cleaner and safer to leave that to the caller.
This new change matches the existing documentation for
Shadow::updateShadow which states
> In case the Property has been withdrawn the method returns @c false.
> In that case the owner should delete the Shadow.
Test Plan:
Created an ASAN of kwin
Relevant testMaximised no longer fails
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17380
Summary:
213239a0ea tried to address the case when
a wayland client gets shadow after it was mapped, but because of poor
testing from my side, another bug was introduced. If a decoration tooltip
or the user actions popup is shown, then in some cases it can be blank.
Usually, SurfaceInterface::shadowChanged proceeds SurfaceInterface::sizeChanged,
so when the shadow is installed, window quads cache is rebuilt. But
because shell client already knows the geometry of the internal client,
goemetryShapeChanged is not emitted, thus the cache is not updated.
It would be better just to invalidate the cache when the shadow is
installed, uninstalled, or updated. This reduces the number of
unnecessary invocations of Scene::Window::buildQuads and also moves
handling of the window quads cache away from the Shadow class.
BUG: 399490
FIXED-IN: 5.15.0
Test Plan: Decoration tooltips are no longer blank.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17215
Summary:
Currently, --about doesn't have any effect. In addition to that, the about
data contains outdated information.
This change hopefully addresses all previously mentioned issues.
Test Plan: Ran kwin_wayland --author.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17319
Summary:
Use the new functionality in KWayland to support drag and drop via touch
screens.
Either a drag and drop session with pointer or touch is possible, but not
both at the same time. Pointer/touch gets deactivated if a touch/pointer
drag and drop session is active.
Test Plan: Manually.
Reviewers: #kwin, davidedmundson
Subscribers: davidedmundson, alexde, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15466
Summary:
This patch aims at improving the Toplevel, internal window and decoration
focus tracking.
In detail the goals are:
* Clean tracking of beneath and focus Toplevel as well as decoration and
internal windows. Splitting this up in well defined sub routines.
* Minimal find Toplevel operations on window stack.
* Reduce code duplication in pointer and touch child classes.
* Reuse tracking in drag operations.
* Allow direct usage of Wayland input interfaces for decoration and internal
windows in the future.
* Update touch focus on external events like VD switches correctly.
Test Plan: Manually and existing autotests.
Reviewers: #kwin
Subscribers: kwin, zzag
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15595
Summary:
Some members were declared protected. Better style is to have them private
with public or protected getters and setters.
This also removes the unnecessary m_input variable.
Test Plan: Builds and runs.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15519
Summary:
We lost information when using QMouseEvent::globalPos, since
it is integer. Use instead InputRedirection::globalPointer,
which is updated before the filters are processed and is
float.
Test Plan: Manually.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: anthonyfieroni, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15502
Summary:
Compositing is suspended/finished in a very hard way fashion, effect
windows are destroyed without notifying effects about it.
AnimationEffect tries gracefully release deleted windows, but because
in some cases(like when suspending compositing) a deleted window can
be already destroyed, a segmentation fault can happen.
This change adjusts the order in which effect windows and effects are
destroyed, so AnimationEffect (and other effects) cannot access dangling
pointers.
BUG: 400788
FIXED-IN: 5.15.0
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17311
Summary:
When suspending compositing, SceneOpenGLShadow cannot cleanup cached
decoration shadow textures because the effects handler is already gone.
This sometimes can result in a crash when running kwin_x11 --replace
(we're hitting an assert statement).
To fix that, let's use the scene instead of the effects handler for
making the OpenGL context current.
Test Plan:
No longer hit the assert statement:
ASSERT: "m_cache.isEmpty()" in file /home/vlad/Workspace/KDE/src/kde/workspace/kwin/plugins/scenes/opengl/scene_opengl.cpp, line 2025
Application::crashHandler() called with signal 6; recent crashes: 1
QCoreApplication::applicationFilePath: Please instantiate the QApplication object first
KCrash: crashing... crashRecursionCounter = 2
KCrash: Application Name = kwin_x11 path = /home/vlad/Workspace/KDE/usr/bin pid = 5407
KCrash: Arguments: /home/vlad/Workspace/KDE/usr/bin/kwin_x11 --replace
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17306
Summary:
WaylandServer provides the initialization flag to disable lock screen
integration. This couldn't be used yet as there is no command line flag
for it. This change adds a new command line argument and hooks up the
functionality.
The no lock screen integration is useful for the embedded use case where
KWin is just used to render one application.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17279
Summary:
EffectsHandlerImpl::unloadAllEffects can be very useful when writing
tests, e.g.:
auto effectsImpl = qobect_cast<EffectsHandlerImpl *>(effects);
QVERIFY(effectsImpl);
effectsImpl->unloadAllEffects();
QVERIFY(effectsImpl->loadEffect(QStringLiteral("kwin5_effect_foobar")));
but because unloadAllEffects doesn't clean up effect_order, the old
effects can be re-added back into loaded_effects when a new effect is
loaded. Such behavior can result in a segfault.
Test Plan: Existing tests pass.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17283
Summary:
For supporting Wayland windows in the kwin_rules_dialog we need a way to
pass a window id for Wayland windows to the dialog. This id needs to be
sent to the dbus interface to query window information just like the
interactive query. For Wayland windows we don't really have a window id
and it would require to also pass the windowing system to
kwin_rules_dialog and back through the dbus interface.
To not complicate things this change introduces a windowing system
independent id based on UUID. This could in future also be used
internally for areas where it's window id based and used in both
windowing systems.
Test Plan: Adjusted test cases to verify the uuid is generated and passed to Deleted
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16986
Summary:
Main reason for using QPainter was the fact that build.kde.org did not
support OpenGL back when the test got introduced. As we have vgem
support nowadays we can switch back to OpenGL.
Test Plan:
100% tests passed, 0 tests failed out of 130,
but the test is flaky with and without this change
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16940
Summary:
When running ctest in my session a few OpenGL/waylandonly tests crashed
on tear down. This does neither happen on build.kde.org nor when running
ctest in a tty. Comparing the env variables of tty and session pointed
to the session variables. Unsetting those makes the test not crash. This
makes sense as e.g. plasma-integration no longer gets loaded.
As our test suite is intended to test KWin and not plasma-integration or
gnome integration we should have a clean and reproducable environment,
so the variables are unset.
Test Plan: 100% tests passed, 0 tests failed out of 130
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16939
Summary:
When running the complete test suite the ScriptedEffectsTest::testShortcuts
registers the shortcut meta+shift+y. But GlobalShortcutsTest::testX11ClientShortcut
also tries to use this shortcut for a window shortcut which fails if it
is already registered. So when running the complete test suite it
depends on the order of execution whether the GlobalShortcutsTest passes
or not.
Test should be clean - also the globalshortcuts. As the config is read
from test directory anyway, we can delete the file prior to init of
kglobalaccel.
Test Plan: Run ctest, test did not fail anymore
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16932
Summary:
One of the things I dislike most about our virtual keyboard integration
is it's size. It's at least on my system covering too much vertical
space. Unfortuantely the keyboard API does not offer a way to control
the vertical space - one can only control the width. Thus it's also not
possible to just provide the keyboard in the optimal size.
This change tries to address the problem by adding a resize button. When
pressed one can dynamically decrease and increase the width of the
keyboard and thus also the height.
The button is added to the top/right corner. Ideally it would be added
to the symbol button row, but either I couldn't find the appropriate API
hooks or it's not possible.
Test Plan: Started KWin with the change and tested with touch input
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16485
Summary:
The idea of opening the compositing kcm was to show a warning. But that
broke quite some time ago without noticing. We had two ways:
* pass through --args command line argument
* use dbus call to already open kcm
Neither of the two ways is working. The kwincompositing doesn't parse
the arguments and the dbus interface doesn't exist any more.
Following the advice to remove functionality nobody noticed that it is
broken, this is removed with this change. This probably broke with
introducing the new KCM which happened IIRC for Plasma 5.0.
BUG: 393845
FIXED-IN: 5.14.0
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13235
Summary:
It crashes occasionally as the connection thread is still processing events
when the app is tearing down. These changes bring it in line with the
other tests using WAYLANDTEST_MAIN.
Test Plan:
Relevant test now passes.
Ran repeatedl without a crash
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17254
Summary:
EffectWindow proxies its properties from the client/deleted's
properties.
QObject::property(char*) is a slow string search. It's a loop
of string comparisons not a hash lookup!
QML's use of properties is different, there's a property cache.
It's fetched multiple times for every window in every paint of some
effects (such as blur). Hotspot shows this as a significant amount of
the render pass (X11) with nothing in kwin animating.
This patch replaces the macro that does
parent()->property("propertyName")
with a macro calling the relevant function directly without metaobjects.
This also improves type safety for future changes.
Test Plan:
Existing unit tests
Ran it for a bit
Reviewers: #kwin, graesslin
Subscribers: graesslin, zzag, broulik, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16602
Summary:
Some applications are not able to inhibit the idle behavior because
each of them creates an inhibitor object before the corresponding
ShellClient object becomes ready for painting.
BUG: 401499
FIXED-IN: 5.15.0
Test Plan: idle-inhibit client (from wlroots/examples) works.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17227
Summary:
In order to be compatible with the monolithic Fade effect, this effect
animates special windows (e.g. notifications, etc) as well, but some
of those special windows shouldn't be animated by this effect (e.g.
KDE Plasma splash screen).
Test Plan:
Started new Plasma on Wayland session, the KDE Plasma splash screen
smoothly faded out without flickering/blinking, etc.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17209
Summary:
I forgot to add a kconf update that moves existing users of the builtin
Dim Screen effect to the scripted version of the effect, sorry.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16957
Summary:
Currently, there is a guarantee that a client, which is about to be removed,
is no longer in the stacking order(both in constrained and unconstrained)
when Workspace::removeClient is called. However, because the client gets
removed from m_allClients after clientRemoved is emitted, it can be
re-inserted back into the stacking order.
In general, the pattern is to do some work and then notify others about
what you've done by emitting a signal. In the case of Workspace::removeClient,
we emit clientRemoved way before the client actually gets removed.
CCBUG: 392412
CCBUG: 400854
Test Plan:
* Enable the following script:
```lang=js
workspace.clientAdded.connect(function (client) {
if (client.skipTaskbar || client.modal || client.transient) {
return;
}
workspace.desktops = workspace.desktops + 1;
workspace.currentDesktop = workspace.desktops;
client.desktop = workspace.currentDesktop;
});
workspace.clientRemoved.connect(function (client) {
if (client.skipTaskbar || client.modal || client.transient) {
return;
}
workspace.desktops = workspace.desktops - 1;
});
```
* Open an app, close the app.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17069
Summary:
The window type can be adjusted by rules, so we probably would like to have
effective window type instead of direct.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17087
Summary:
Apparently this test is flaky, based on the output I recieved we had
processed the un-fullscreen but not the un-maximise.
It seems possible the wayland thread could process when there's only one
request in the queue.
Test Plan:
Couldn't reproduce the original issue, it's a blind patch but the
current tests still pass.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16955
Summary:
The primary reason for rewriting this effect was to clean up code and
fix spawning of multiple animations for a single window when user cycles
through virtual desktops very quickly.
Visually, the rewritten version doesn't deviate from the old version.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16454
If a full screen effect has been activated, we need to brighten main
windows because the full screen effect can dim windows on its own
(one such effect is the Present Windows, for example).
If the full screen effect is no longer active, we need to dim main
windows back.
Summary:
The ported effect looks quite similar to the C++ version except one
thing: it works correctly when user activates/deactivates a full
screen effect, for example the Desktop Cube effect.
Other than that, there are no behavioral or visual differences.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16452
Summary:
If a window is unminimized and the minimize animation is still active,
don't cancel the latter and start a new animation. Instead, try to
reverse already active animation and if that attempt failed, start a new
animation. This results in more smoother animations.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16451
Summary:
Effects that prefer to manipulate direction of animations sometimes need
to create animations in some particular state so later on they can be
played backward (swapping from and to is not enough and it would be wrong).
The proposed complete function lets such effects to fast-forward animations to
to the target position so they can be played backwards later on.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16450
Summary:
Consider current implementation of the Squash effect: if a window was
minimized, an animation will be started; if the window is unminimized
and the animation is still active (that can happen when user clicks on
app's icon really fast), the animation will be stopped and a new one will
be created. Such behavior can lead to rapid jumps in the observed
"animation".
A better approach would be first try to **reverse** the already active
animation, and if that attempt wasn't successful, start a new animation.
This patch introduces a new function to the scripted effects API that
lets JavaScript effects to control direction of animations. The
prototype of the function looks as follows:
redirect(<animation id(s)>, <direction>, [<termination policy>])
the first argument is an animation id or a list of animation ids, the
second argument specifies the new direction of the animation or
animations if a list of ids was passed as the first argument. The
third argument specifies whether the animation(s) should be terminated
when it(they) reaches the source position, currently it's relevant only
for animations that are created with set() function. The termination
policy argument is optional, by default it's Effect.TerminateAtSource.
We can use this function to fix issues with rapid jumps in the Squash
effect. Also, redirect() lets us to write effects for simple animations
in slightly different style: first, we have to start the main animation
(e.g. for the Dialog Parent effect, it would be dimming of main windows)
and then change direction of the animation depending on external events,
e.g. when the Desktop Cube effect is activated.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, abetts, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D16449
