https://bugs.kde.org/show_bug.cgi?id=524031

--- Comment #2 from Adam <[email protected]> ---
Apologies if I filed this against the wrong component. I'm new to reporting
bugs and may have pointed it at the most visible thing rather than the
actual cause.

New data. I captured the X11 SHAPE regions on the window in both states.

Broken (desktop icons unclickable; clicking the desktop kills interaction
with other windows on that output):

  window   0x1200008
  class    : xwaylandvideobridge / xwaylandvideobridge
  geometry : 3840x2132+0+28
  mapped   : viewable
  input    : XShapeGetRectangles(ShapeInput) -> count 0
  bounding : 1 rectangle, 3840x2132+0+0

Working (same session, bridge killed and restarted, desktop clickable):

  window   0x1200008
  class    : xwaylandvideobridge / xwaylandvideobridge
  geometry : 3840x2132+0+28
  mapped   : viewable
  input    : XShapeGetRectangles(ShapeInput) -> count 0
  bounding : 1 rectangle, 3840x2132+0+0

count 0 is an empty input region, not a failed query; the same call with
ShapeBounding on the same window returns a rectangle. The window id matches
in both because the restarted process reused the connection slot.

So the input region is empty and the window is click-through, exactly as you
said, and it reads identically whether or not the failure is present. The
X11-side shape is not what differs between the two states.

What does correlate is the process itself: killing it restores pointer input
on that output immediately, twice now. It had started 35 seconds after boot
and run for just over 24 hours without restarting, across several screen
lock and unlock cycles, so the window is not being recreated. Every
occurrence has followed a lock and unlock, never a fresh login.

That leaves a gap I can't close myself: the X11 input region is empty, but
the compositor still appears to route pointer events to the surface. If the
empty region isn't reaching the wl_surface, or isn't being acted on, that
would be an Xwayland or KWin issue rather than anything in this codebase.

Please reassign wherever it belongs. Happy to gather more data if there's
something specific worth capturing next time it occurs.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to