Instead of best-guessing, at BGR (which in retrospect was a bad guess),
offer whatever resembles most the internal representation. This way the
frame gets to be least treated as it goes into the client.
...before releasing the connection!
this was randomly causing crashes because of use-after-disconnect
problems. since we would tear down the connection in ~X11WindowedBackend
but outputs wouldn't get cleaned up until QObject children cleanup of
the X11WindowedBackend object (or more precisely the OutputBase). this
would then result in ~X11WindowedOutput accessing a backend connection
that had already been closed
to help debug this type of problem moving forward let's also reset the
connection to nullptr during destruction. it's kinda pointless but it
makes it easier to spot use-after-disconnect
BUG: 466183
This reverts commit 4b1ef33c1e.
It introduced a crash. The large icons task switcher can create and
destroy the shadow in the middle of painting. When a shadow is
destroyed, kwin can make opengl context current. It will reset current
QOpenGLContext, and QtQuick can crash.
BUG: 466279
The input events can be processed when the workspace is not available,
e.g. during startup or shutdown, so add a corresponding guard.
As a long term plan, we need to decouple Workspace from low-level input
backend parts, but it will be a too invasive change for now.
BUG: 449317
BUG: 465456
Checking if the pointer needs to be confined before doing `std::clamp` on
the position is useless and causes problems. With this change, the pointer
will always be confined to exactly the screen, without any fractional offsets.
BUG: 461911
Currently the Shadow class is scene specific, which adds coupling
between Window and scene bits.
This change introduces ShadowTextureProvider that contains scene
specific stuff so the Shadow acts like a data source and is not coupled
to concrete scene.
It's gone in KF6
It's only used to obtain the install location of the package and resolve files relative to that
Use QStandardPaths to locate those files directly
Aurorae uses a QQuickWindow that receives an implicit scale through the
QPA. Unfortunately there's a known quirk with device pixel ratio not
being followed when offscreen surfaces are never shown and an
independent render control is used.
Without this QQuickRenderControl::grab ends up using a buffer the size
of the window * dpr, but with the contents rendered at 1x in the corner
along with incorrect device pixel ratio information.
This patch copies a used solution from EffectQuickView. This can be
resolved more neatly with newer Qt.
BUG: 452730
BUG: 465790
Since normal geometry is now rounded, we need to make sure to round
anything that should match that geometry, otherwise we risk things
peeking out of the normal geometry.
BUG: 464526
Before this commit, the device pixel ratio of the cursor was set to ignore
the Xwayland scale override, while the size of the cursor surface was scaled.
This change removes the scaling of the cursor surface size, to make the cursor
be no longer cut off in rendering.
BUG: 466094
Converting to wl_fixed is lossy. Especially for non-round values like 1.15.
This causes various visual glitches and off sizes.
Given we're using units of 120ths for the fractional scale protocol it makes sense
to fix the passed values for anything else over the wayland protocol.
BUG: 465850
Testing done:
ran a 1920 screen at 1.15 scale, got a logical size of 1670 which matches what it should be
The only one that does it differently is the DRM backend and it's just
an extension that takes GBM into account, otherwise it's effectively
copy-pasted code.
We need to register data types such as QVector<KWin::VirtualDesktop *>
both for QML and JavaScript scripts.
If you have no JavaScript scripts enabled, those container types won't
be registered and QML will fail to read properties such as
Window.desktops, which can result in already visible windows animating
opacity from 0 to 1 when opening overview effect.
To fix that, register all accessible data types in scripting apis in
Scripting::init().
