The logic for handling the active window when it is a transient, added in
https://invent.kde.org/plasma/plasma-workspace/-/merge_requests/509, had
a bug (fixed in 67003f740888fa9d1bcee60cd28ce9a862d2452a).
This reworks the implementation and actually makes the original feature work
(a signal also wasn't properly handled).
BUG: 438222
FIXED-IN: 5.22.1
If there's no texture factory function, return the old node immediately.
Also, remove a guard against the code that calculates centered geometry
of the texture node. We cannot hit that case as we check whether the
factory function has returned a valid texture before that.
Needed to make panel Adaptive Transparency work with "Show window thumbnails: always"
Needed to make Minimize All work with "Show window thumbnails: always"
An iconified task/window is in reality hidden. For example, when
`Show Desktop` action is performed, tasks are 'Minimized'. But when
tasks are minimized, they are 'Hidden'.
This property is essential to detect check if a window is truly
iconified or not when the hidden window previews is set to 'Always'.
We recently changed it so kwin defaults to memptr because some clients
couldn't deal with dmabuf. We still want dmabuf to be used as it's less
stressful to the system and we support it just fine.
Tell QQuickWindow when there is an alpha channel, otherwise it's
considered opaque and Qt/mesa renders it disregarding what's underneath.
Sometimes windows will have translucent areas and we want these to be
rendered cleanly on top of the scene. Alternatively it shows weird
noise.
A window can be moved to another activity externally, for example using
the user actions menu in kwin. The window model should notify the task
manager when such an event occurs.
Window activation is tracked to implement a new front-
end feature to activate the most recently active window (or fall
through to stacking order otherwise) subsequently.
CCBUG:370258
The KWin setting for whether or not to wrap around was not respected, so
wrap was always true even if the user has turned it off. This commit
fixes that.
BUG: 425787
The test depends on the applications being installed and the path being
correctly set. To make it more robust the path is set inside of the test
and the files the test used are shipped with it.
This isn't always static, espcially in GTK apps where the xproperty is
set later by gmenuproxyodel.
The Appmenu applet already watches for this role changing, but it wasn't
hooked up.
The wayland implementation does not have the same bug.
BUG: 422786
We do this sometimes, but not consistenly.
The benefit is twofold.
First it improves the diff when adding new values since no existing line needs to be touched.
Second it prevents clang-format from collapsing the definition into a single line, which is undesired for large enums.
The LauncherTasksModel tried to avoid a model reset by making certain
assumptions to save a few cycles. Ultimately this was misguided, in-
correctly not evicting the internal app data cache as well as causing
a lot more work in the view trying to compute layout deltas and some-
times causing a few frames of ugly animations.
There likely _is_ some optimization potential here, but it needs to
be done smarter later on.
When computing the delta between the visible and the last saved launcher
order state, following the reordering of launchers, the code was incorrectly
comparing resolved and unresolved (i.e., for config storage) launcher URLs.
This lead to an incorrect result and therefore pinned tasks jumping around
in the Task Manager applet.
BUG:426880
Nicolas Fella
Aleix Pol
Henri Chain
Vlad Zahorodnii
Bharadwaj Raju
Abhijeet Viswa
David Redondo
Alexander Lohnau
Kevin Ottens
Usarin Heininga
Nate Graham
Niccolò Venerandi
David Edmundson
Mikel Johnson
Laurent Montel
Eike Hein
Jonathan Riddell
Bhushan Shah