If the decoration is destroyed before the window is mapped, kwin can
respond with a configure event that has 0x0 size. New tests check that
problematic case.
BUG: 444962
Kwin announces a format with alpha when Dma-Bufs are available, even
when the texture doesn't contain any. This results in clients segfault
when trying to access the buffer assuming a maxsize wrt. the announced
format by ways of dimensions and bpp.
This patch moves the format masking to affect the Dma-Buf transport
only.
If eglSwapBuffers() fails, there won't be a buffer and so we need to
mark the frame as failed. Otherwise, the screen can be frozen.
eglSwapBuffers() can fail if some effect calls makeOpenGLContext()
between RenderBackend::beginFrame() and RenderBackend::endFrame(), which
is the case with the zoom effect. It can set wrong draw surface in
ZoomEffect::recreateTexture()
BUG: 445412
This simplifies focus related logic a bit. Instead of differentiating
between wayland and internal window focus, simply maintain window focus
that works both with regular wayland windows as well as the ones created
by kwin.
The thumbnail that represents the "desktop" in the switcher preview
is taken from the default wallpaper, and it will update automatically
as it changes in future versions of Plasma.
The current one is also added along the other thumbnails as a fallback.
BUG: 309401
FIXED-IN: 5.24
Currently, the test doesn't pass due to Xwayland encountering a protocol
error.
That protocol error occurs because the xdg-output is destroyed before
the wl_output.
This change moves the ownership of the xdg-output to wl-output so they
are destroyed together.
dmabuf-feedback allows the compositor to give the clients better feedback on what
formats and modifiers they should use, and for which device they should allocate.
This way they can reallocate for scanout whenever the compositor tells them to,
which makes direct scanout work for a lot more devices and applications.
KWin handles several types of pointing input devices, e.g. mice,
tablets, etc.
As is, enterEvent and leaveEvent are very ambiguous. This change
prepends "pointer" to those methods to make it explicit that they handle
pointer enter/leave events.
While it could be useful with tiled displays, the isFormatSupported and
supportedModifier functions can be called before prepareModeset, so where
m_formats is still empty. Additionally they're neither in a hot path nor
performance critical.
Whether or not we want to use explicit modifiers for our surfaces doesn't
matter for what format+modifiers drm planes support. This way direct scanout
works by default, without having to explicitly enable modifiers
While a stylus is in proximity we want to show its cursor. In this commit
it only gets shown on move because for a still unknown reason the position
is out of date before the first move event.
BUG: 443921
Ever since the effects were changed to static, each test of the
integration tests includes all the effects. The result of this is that
when doing a debug build each test is now 60MiB or more. With the amount
of tests, this results in ~8 GiB of diskspace used just for KWin's
binary output directory, which is rather excessive.
Since the tests all share a common framework library, we can change that
library to a shared library and that way avoid linking all the effects
into each test.
Most of this is shuffling around some link libraries in the integration
test CMakeLists, however, I needed to export the Xwayland class as it is
used by one of the tests but wasn't exported.
Tracking AbstractOutput properties in Toplevel is not extensible. Since
DecorationItem is the only one who needs Toplevel::screenScale(), make
it track the output device pixel ratio.
systemd takes care of setting and dropping master permissions when
sending PauseDevice and ResumeDevice signals.
When the ResumeDevice signal is received, the relevant drm device should
already have master permissions set up.
On the other hand, when the active property changes, there's still a
chance that systemd haven't granted drm master permissions to us.
For some reason, while touch doesn't have a middle button, the
taphandler for middle click close still triggered on tap. So make it
explicitly only support mouse/pen so we exclude touch events.
BUG: 445755
In case a modeset needs to be performed, the drm backend will test all
pipelines to ensure that new mode won't cause any bandwidth issues on
other outputs, etc.
To do that, it may delay presenting frames. If the new configuration
doesn't work, it needs to notify about failed frames.
However, the relevant code that notifies the RenderLoop about failed
atomic commits doesn't check if there's actually a pending modeset
present.
When switching between VTs, systemd can revoke master permissions from
kwin. To make things even more trickier, kwin can try to present a frame
in that short time span.
For many windows it will be invisible, so there's little point for
creating it ahead of time. Another reason for loading the close button
on demand is that Plasma components are typically heavy.
Currently, MouseArea.doubleClicked doesn't work in QtQuick effects.
Luckily, the overview effect doesn't use it, but it's still worth making
sure that it works as expected regardless of whether it's a kwin effect.
With this change, the OffscreenQuickView will keep track of the last
mouse press event. If the second button press occurs within the mouse
double click interval, the OffscreenQuickView will generate a
QEvent::MouseButtonDblClick event, similar to what QGuiApplication would
normally do when processing window system events.
The current code performed two functions:
- metadata was read async
- the event loop is spun between each effect loading
After we merged static and binary plugins this caused a problem for
things like blur and background contrast where loading an effect
influences what we advertise as supported. We don't want plasmashell to
have to reload all it's SVGs.
All KDE provided binary plugins are statically bundled so querying for
plugins is super duper fast anyway. It's not the same kbuildsycococa
path that it used to be.
The C++ plugins tend to be considerably faster in their constructor than
scripted counterparts, and if anything should be loaded lazily it can be
handled inside the effect.
Set the implicit size of the item to the client frameGeometry size.
This is also more convenient for tabbox usage than accessing it through the
client's property, had a proper notify signal and abstracts it from the
implementation.
It also can be mimicked by the preview ThumbnailItem used in the tabbox KCM,
which has no client property.
The TabBox preview uses a mock WindowThumbnailItem mimicking the
real item, but it was missing the `sourceSize` property, preventing
some effects to be previewed if they use it.
This change moves some pieces of the Overview effect in a new
QuickSceneEffect class that can be useful for other QtQuick based
effects.
The QuickSceneEffect class is meant to deal with rendering bits, input
handling, etc.
As the user of the QuickSceneEffect class, you need to override the
createView() method. The createView() method is invoked by the
QuickSceneEffect when it wants to create a QtQuick scene for the given
screen.
The design of the QuickSceneEffect is based on the Overview effect.
EffectQuickScene is not used strictly by effects, aurorae decorations
use it too to render window decorations.
This change renames the EffectQuickView/Scene to
OffscreenQuickView/Scene to clear up the naming scheme.
Add `tabletToolEvent()` to DecorationEventFilter class so it can handle
tablet events.
Add `tabletToolEvent()` to MoveResizeFilter class to avoid incorrect
titlebar highlight after a window is resized by using a stylus.
Add `tabletToolEvent()` to WindowActionInputFilter class so window focus
can be changed by clicking on the window content area when using a stylus.
Add `update()` to `TabletInputRedirection::tabletToolEvent()` to enable
switching window focus by using a stylus.
BUG: 432104
This prevents flickering behind plasma panels after changing compositing
settings.
If the contrast effect is disabled, the destructor of the ContrastEffect will
start the contrast manager remove timer. When it expires, the contrast manager
global will be removed and it will be finally destroyed after 5s (to work
around the wl_global race condition).
If compositing is toggled, the destructor of the ContrastEffect will start
the remove timer and the constructor of the ContrastEffect will stop the
timer, the contrast manager object will be unaffected.
The contrast effect has no options that could affect the shader, so the
reconfigure() method was removed.
This prevents flickering behind plasma panels after changing compositing
settings.
If the blur effect is disabled, the destructor of the BlurEffect will
start the blur manager remove timer. When it expires, the blur manager
global will be removed and it will be finally destroyed after 5s (to work
around the wl_global race condition).
If compositing is toggled, the destructor of the BlurEffect will start
the remove timer and the constructor of the BlurEffect will stop the
timer, the blur manager object will be unaffected.
The blur effect options don't affect the generated shader code, so the
shader status check was removed in reconfigure().
This enables kwin to get notified when a memory has been lost in the
system, allowing the driver to purge all the memory and have kwin create
new contexts.
This matches what we do on GLX.
The current code for creating the root shared context is very naive and
does not take into account aspects like robustness or versions.
It is not always possible to mix attributes, in particular robustness
meaning we are unable to opt into these features in future contexts.
This refactors everything to use a common path to make all contexts.
In the KPluginSelector we don't have the possibility to set the args for
each KCM. Because of that we use the metadata of the KPluginFactory as a
fallback to read the plugin keyword, which specifies the KCM we want to display.
This requires kcmutils 54b196a9bad88732debe0b49111af4755268f09f, which landed in the last release.
BUG: 445667
The person that added krunner search integration uses vim. It seemed
intuitive to that person that the Escape key should quit the search mode
instead of quitting the effect. But it seems like more intuitive thing
is to actually stop the effect.
BUG: 445708
If the opacity property is set to 0, the item will still receive mouse
events. One either needs to set the enabled or the visible property to
false to ensure that the WindowHeap receives no mouse events.
The reason for setting the opacity to 0 instead of the visible property
to false was that I thought later will nuke the paint node, but it seems
like that's not the case.
BUG: 445707
This is needed to ensure that mouse release events go to the view that
received mouse press events; otherwise some views can get stuck in a
state thinking that button xyz is pressed.
Currently, cursor shape set by QtQuick is not sync'ed, which can be seen
as the pointer not changing its shape to i-beam when hovering the search
field.
Currently, every time you launch the overview effect, QtQuick will go
out and parse QML files. With the overview effect gaining more features
and the code size getting bigger, it takes more time for the overview to
present the first frame after it got triggered.
With this change, the overview effect will keep the ScreenView
QQmlComponent object around that can be used to avoid reparsing qml code
every time the overview effect is launched.
It also ports the Overview effect from de-facto deprecated qml context
to initial properties, which yield slightly better startup times.
BUG: 445666
This allows effects to support touch input.
Unfortunately, Qt's TouchEvents require the full touch point state,
whereas KWin's internal touch handling only deals with one point at a
time. So we need to keep track of the touch state in EffectQuickView and
pass that on to Qt. Additionally, we need a QTouchDevice as lots of Qt
code depends on it.
On my Nvidia machine there was a massive lag exitingthe overview effect.
Hotspot showed it as being in QOpenGLVertexArrayObjectPrivate::destroy.
In this method we clean up some shared objects used in the context when
the context closes.
In order to do this we need the context to be current. If it is not
current Qt currently internally creates a temporary offscreen surface.
To fix this we need to have our context current during destruction,
which includes changing order so it is destroyed before the surface.
With this change, user will be able to press Tab and use Enter, Delete,
and F2 keys to activate, remove, and rename virtual desktops, respectively.
Co-authored: Fushan Wen <qydwhotmail@gmail.com>
If a window wants to be initially shown in fullscreen mode, it will
issue an xdg_toplevel.set_fullscreen request before the first surface
commit.
If a window wants to be shown in fullscreen mode and there hasn't been
any first surface commit, kwin will cache the request and apply
fullscreen mode when checking window rules in the initialize() function.
On the other hand, window rules are disabled for plasma surfaces. The
motivation behind that was to forbid user from messing with plasma's
surfaces (this change was suggested during redesign of xdg-shell
implementation).
As it turns out, there are cases where plasma may ask to show a window
in fullscreen mode, which also has a plasma surface installed, e.g.
fullscreen application dashboard.
In order to fix the dashboard, this change allows window rules to be
applied to xdg-toplevel windows that also have plasma surfaces installed.
As is, xdg-toplevel surfaces and plasma surfaces are very different in
nature. Adding more quirks to handle plasma surfaces in
XdgToplevelClient is not worth the effort and there are better
alternatives, e.g. layer-shell.
Based on user feedback, it will be great to have krunner functionality
integrated in the overview/present windows effect.
This change ports the overview effect to Milou for search results. With
it, one can search for existing windows or launch new applications.
BUG: 445207
kwin-strip-effect-metadata needs to run on the host. First, one needs to
build it, e.g.
cmake --build build-native --target kwin-strip-effect-metadata
then when cross-compiling, point cmake to the native build directory, e.g.
cmake -DKWIN_HOST_TOOLING=${path}/build-native
This improves plugin loading times. As is, the main issue is the number
of builtin effects and the fact that each has a lot of translated
strings, which combined adds up to noticeable loading times. KWin itself
will never read those translated strings, it only needs two pieces - the
plugin id and whether the plugin is enabled by default.
This change adds a little helper to strip unnecessary info from metadata
files.
Fixes a crash I have with dpms + suspend, which was caused by the udev
event for updating outputs being called before the output got enabled
again. When DrmGpu::updateOutputs got called it removed the crtc from
the inactive output and then disabled the output afterwards. Instead,
only remove crtcs if an output is really disabled.
This also allows to generalize the logic for lease outputs, and could
in the future allow for faster dpms on/off switching.
This unifies frame hooks for OpenGL and QPainter render backends. There
are a couple of reasons why it's a good idea - it provides one mental
framework to start painting a frame, the Compositor will be able to
start and submit frames. The last one is very cool because it gives the
Compositor more power over compositing.
Besides unifying frame hooks, this cleans up a bit the arg naming mess
in endFrame(). As is, "damage" and "damagedRegion" are very confusing
names. "damage" arg has been renamed to "renderedRegion," because that's
what it is. The renderedRegion arg specifies the region that has been
repainted by the Scene. It's different from the damagedRegion as that
one specifies the surface damage, i.e. the difference between the
current and the next frame, while the renderedRegion may include a
region that had to be repainted to repair the back buffer. The main
reason why we need renderedRegion is the X11 platform. On Wayland, it's
unused.
In the future, we will need to extend this api with output layers.
The lanczos filter checks the screen size before rendering in
LanczosFilter::updateOffscreenSurfaces(), so this is not needed.
This simplifies lifetime handling of the lanczos filter, e.g. we
don't need to bother with opengl context anymore, and makes the
scene use less Screens' features.
On Wayland, a window can have subsurfaces. The spec doesn't require the
main surface and its sub-surfaces to have the same scale factor.
Given that Toplevel::bufferScale() makes no sense with Wayland windows,
this change drops it to make code more reasonable and to prevent people
from using Toplevel::bufferScale().
With a persistent vbo, kwin will allocate one big enough buffer and
allocate memory out of it.
In order to prevent overwriting vertex buffer data that is currently
being accessed by the GPU, fences are inserted at the end of frame.
The signaled fences are destroyed after the buffer swap operation, which
seems a bit odd because the just inserted fence most likely won't be
signaled. Perhaps it's a historical artifact?
This change rearranges fence cleanup so it's performed right before
starting a new frame. With it, kwin will most likely re-use the
previously used memory chunk because there will be plenty of time for
the fence to become signaled.
Another motivation behind this change is to make refactoring SceneOpenGL
code easier. As is, m_backend->endFrame() is wrapped in
GLVertexBuffer::endOfFrame() and GLVertexBuffer::framePosted(). With
that, the Compositor can't call m_backend->endFrame(), which can be
desired for cleaning up render backend abstractions.
Currently, when screencasting a window, kwin may render a window into a
temporary offscreen texture, copy that offscreen texture to the dma-buf
render target, and discard the offscreen texture.
Allocating and deallocating offscreen textures is inefficient. Another
issue is that the screencast plugin uses Scene::Window::windowTexture().
It's a blocker for killing scene windows.
This change introduces a base ScreenCastSource type. It allows us to
move away from Scene::Window::windowTexture() and make the dma-buf code
path efficient with applications such as Firefox that utilize
sub-surfaces.
With the ScreenCastSource, kwin can also provide screen cast frames with
arbitrary device pixel ratio.
Like top level clients, apply plasmashell roles to popups as well (limiting them, don't allow dock or desktop roles in poups as they don't make sense)
This makes possible to recognize plasma tooltips as tooltips, treating them in a way closer to X, and makes morphingpopups work on wayland
After user edits the name of a desktop, the search field is no longer
focused. If the user starts typing text, one could expect that it will
be forwarded to the search field without requiring a click.
This change forwards unhandled key events to the search field to ensure
that searching is intuitive.
Since binary effects are installed in their own directory, checking
service type is redundant. Also, KPluginMetaData::serviceTypes() has
been deprecated.
Task: https://phabricator.kde.org/T14483
The ifdefs for have_gbm obfuscate the code unnecessarily - the drm backend
is not a great experience with qpainter, so in practice noone should ship
it without gbm anyways.
The Compositor contains nothing that can potentially get dirty and need
repainting.
As is, the advantages of this move aren't really noticeable, but it
makes sense with multiple scenes.
Backend parts are far from ideal, they can be improved later on as we
progress with the scene redesign.
The main idea behind the render backend is to decouple low level bits
from scenes. The end goal is to make the render backend provide render
targets where the scene can render.
Design-wise, such a split is more flexible than the current state, for
example we could start experimenting with using qtquick (assuming that
the legacy scene is properly encapsulated) or creating multiple scenes,
for example for each output layer, etc.
So far, the RenderBackend class only contains one getter, more stuff will
be moved from the Scene as it makes sense.
The proprietary NVidia driver now supports gbm, which vastly improves the
user experience. For older devices that will not get gbm support dropping
EglStreams will likely not have a big impact as it has several session breaking
issues anyways.
By removing the backend a lot of logic can be simplified, most notably multi-gpu.
