Created bug on bugs.kde.org: https://bugs.kde.org/show_bug.cgi?id=446779
First suspect as always: nvidia-drivers. Did you try other versions? Did it happen with previous driver version?
Only detected this behavior with latest nvidia-drivers, but since there was also an update for kwin, I'm in doubt of the origin of this issue.
This happens randomly and normally when I login immediately after the boot. Haven't checked yet the downgrade to previous driver version. I can't go back to stable version because I need to use nvidia-cuda-toolkit that depends on nvidia-drivers latest version.
Doubtful in a .4 release, plus no such issues reported by users other than those with Nvidia inside, but confirmed by other Nvidia users upstream.
Quote from upstream bug: > I can confirm this issue with a GTX 1070. This only happens with the latest > "new feature branch" nvidia driver (495.44). With the current "production > branch" version (470.86), it never fails.
(In reply to Samuel Bernardo from comment #3) > This happens randomly and normally when I login immediately after the boot. > Haven't checked yet the downgrade to previous driver version. I can't go > back to stable version because I need to use nvidia-cuda-toolkit that > depends on nvidia-drivers latest version. Latest nvidia-cuda-toolkit no longer depends on 495.xx, if affected you should be able to use stable 470.xx just fine.
When this happens, could you please find in kwinrc the value given to OpenGLIsUnsafe. My knowledge is certainly outdated but as I can remember, the setting of OpenGLIsUnsafe to false follows a decision taken after a given timeout. Could then well be that, with latest NVIDIA drivers, the timeout is set a bit too low.
Feel free to re-add kde proj whenever there is something we can actually do.
(In reply to Eric F. GARIOUD from comment #7) > When this happens, could you please find in kwinrc the value given to > OpenGLIsUnsafe. > My knowledge is certainly outdated but as I can remember, the setting of > OpenGLIsUnsafe to false follows a decision taken after a given timeout. > Could then well be that, with latest NVIDIA drivers, the timeout is set a > bit too low. Very interesting, will try that next time. But since I upgraded nvidia-drivers to 495.46-r10 the crash didn't happen again, at least until now.
Is this still happening with nvidia-drivers-510.39.01? Not that I think there's anything I can do about this either.
(In reply to Samuel Bernardo from comment #9) > But since I upgraded nvidia-drivers to 495.46-r10 the crash didn't happen > again, at least until now. Well, then again given this I guess it's already solved? Personally wasn't able to reproduce.