https://bugs.kde.org/show_bug.cgi?id=524031
Bug ID: 524031
Summary: Window becomes fullscreen, invisible (opacity 0) and
input-accepting, silently blocking all pointer input
on one output
Classification: Applications
Product: xwaylandvideobridge
Version First 0.5.0
Reported In:
Platform: Fedora RPMs
OS: Linux
Status: REPORTED
Severity: major
Priority: NOR
Component: general
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Target Milestone: ---
The xwaylandvideobridge window entered a state where it covered an entire
output at 3840x2160 with opacity 0 while still accepting pointer input.
Because the window is also flagged skipTaskbar, skipPager and skipSwitcher,
there is no user-facing way to see it, reach it, or close it. It sits in
NormalLayer, so it is permanently stacked above the desktop and competes
with ordinary windows for stacking position.
The user-visible result is that desktop icons on the affected output cannot
be clicked at all, and ordinary windows on that output stop responding to
clicks whenever the desktop is clicked.
This has happened approximately five times. Every occurrence has followed a
screen lock and unlock cycle. Prior to identifying the cause, recovery was
by rebooting.
STEPS TO REPRODUCE
1. Plasma Wayland session with two outputs (here: 3840x2160 HDMI primary at
0,0 and 1920x1080 DisplayPort secondary at 3840,977).
2. Lock the screen, leave it locked, then unlock. Across roughly five
occurrences the failure has always been present after a lock and unlock
cycle and has never appeared without one. The precise conditions within
that cycle are not isolated.
3. Click the desktop wallpaper on the affected output.
4. Attempt to click into any ordinary window on that same output.
OBSERVED RESULT
Desktop icons on the affected output do not respond to clicks at any point.
After clicking the wallpaper, ordinary windows on that output stop receiving
pointer input entirely. The cursor shape does not update over them, for
example no I-beam appears over a Konsole text area, confirming the surface
receives no pointer enter or motion events at all.
Activating a window from the Task Manager restores interaction with it, until
the desktop is clicked again, at which point it goes dead once more. This is
repeatable on demand.
Clients on the other output are unaffected throughout. The Plasma panel
remains usable because layer-shell surfaces stack above normal windows. The
KWin debug console remains usable because it is an internal window rather
than a client.
Restarting plasma-plasmashell.service does not help.
The KWin debug console showed the following on the xwaylandvideobridge
window:
caption Wayland to X Recording bridge - Xwayland Video Bridge
resourceClass xwaylandvideobridge
desktopFileName org.kde.xwaylandvideobridge
size 3840x2160
pos 0,0
frameGeometry 0,0 3840x2160
output HDMI-A-1
opacity 0
wantsInput true
layer NormalLayer
managed true
maximizeMode MaximizeFull
hidden false
minimized false
skipTaskbar true
skipPager true
skipSwitcher true
stackingOrder 4
Killing the process immediately restored normal input on that output.
EXPECTED RESULT
A window should not be simultaneously invisible and input-accepting. If the
bridge window is not meant to be seen or interacted with, it should not
accept pointer input. If it does accept input, it should be visible or
otherwise reachable.
The combination of opacity 0, wantsInput true, and skipTaskbar plus
skipPager plus skipSwitcher closes every channel a user has for noticing or
reaching the window while leaving input active. That is what makes this
failure effectively undiagnosable without the KWin debug console, which most
users will not know about. A visible window blocking clicks would be a
trivial diagnosis by comparison.
ADDITIONAL INFORMATION
The fullscreen geometry may or may not be a separate defect. If it reflects
a fullscreen capture source then the size may be legitimate, in which case
the bug is that the window remained mapped and input-accepting outside an
active share. Either way the opacity and input combination stands on its own.
Screen sharing is used rarely on this system and no share was active at the
time the state was observed, nor in the sessions preceding any of the
occurrences.
Displays are powered off physically at the switch rather than left to
power-saving blanking, so outputs are removed and re-added rather than merely
blanked. This may or may not interact with the lock and unlock cycle.
SOFTWARE/OS VERSIONS
Operating System: Fedora Linux 44
KDE Plasma Version: 6.7.4
KDE Frameworks Version: 6.28.0
Qt Version: 6.11.1
Kernel Version: 7.1.5-200.fc44.x86_64
Graphics Platform: Wayland
xwaylandvideobridge: 0.5.0-1.fc44
Graphics Processor: NVIDIA GeForce RTX 3060
NVIDIA driver: 610.43.03
--
You are receiving this mail because:
You are watching all bug changes.
