https://bugs.kde.org/show_bug.cgi?id=363224
--- Comment #41 from Martin Gräßlin <mgraess...@kde.org> --- I'm sharing my theory on what we see from the backtraces we have: The crash happens when a window is exposed to the window effect system. This is the part that compositing is required. The code in question delays a call to the end of current event cycle. And it looks like the window gets destroyed while the call is being delayed. This would mean that the crash happens when a window gets opened and closed in a very short time frame. In fact the window must be closed when KWin gets the event that the window is created. The destroy event must already be queued. It's the only explanation I have for the window getting destroyed during the delay. Now I tried to verify this assumption with test applications which simulates this condition. But I failed. Thus my theory: it only happens on older systems or on systems under stress at that time. The quite often mentioned VMware and virtual box could verify the stress situation: resources are passed to the VM and taken away from the main system. What I need: Verification that my theory is correct. Please run xev and trigger the crash. It should show us whether there was a short lived window. In addition I will need someone who can test a patch if my theory is correct. Without being able to reproduce I just cannot write and test a patch. I can write it, but need someone to test and verify. -- You are receiving this mail because: You are watching all bug changes.
