| Summary: | KWin crashes when switching from Breeze to Plastik | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | Erin Yuki Schlarb <erin-kde> |
| Component: | general | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED UPSTREAM | ||
| Severity: | crash | Keywords: | drkonqi |
| Priority: | NOR | Flags: | thomas.luebking:
nouveau+
|
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Platform: | unspecified | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
| Attachments: | qdbus org.kde.KWin /KWin supportInformation | ||
|
Description
Erin Yuki Schlarb
2015-02-07 19:37:40 UTC
ShaderBinder binder(ShaderManager::ColorShader); binder.shader()->setUniform(GLShader::Offset, QVector2D(0, 0)); Means that scene-color-*.glsl did not compile. Since this only happens under "random" (technically there should not be much difference between Plastik and Aurorae) conditions, this is very likely due to nouveau. - Can you try the nvidia blob? - Please attach the output of "qdbus org.kde.KWin /KWin supportInformation" Created attachment 90968 [details]
qdbus org.kde.KWin /KWin supportInformation
> glCoreProfile: true Try OpenGL 2.0 > OpenGL platform interface: GLX and/or EGL > OpenGL renderer string: Gallium 0.4 on NVA5 > GPU class: G80/G90 NVA5 would be GT218, but both are Tesla chips - they're supported by the current nvidia blob if you want to try. So your conclusion is that this is a driver bug and can't be caused by a bug in KWin? Those settings you mentioned didn't change anything; also the binary driver (aside from being faster) sucks in pretty much every other regard that I didn't even bother getting it to work after last distro hop (after finding that nouveau had improved driver stability considerably). If you believe that this might not be driver bug please tell me. I don't want to go through that hassle needlessly. Thank you for responding so quickly! (In reply to Alexander Schlarb from comment #4) > So your conclusion is that this is a driver bug "likely" > and can't be caused by a bug in KWin? If the shader was incompliant, I'd expect compilation to fail everytime. If compilation would fail for other reasons, or not be triggered (dangeling singleton instance pointer), I'd expect the from/to deco to not have much impact, but have it happen on eg. every deco change (if not any reconfig) I'm certainly not ruling out that the GPU is maybe left in a invalid state due to a previous error caused by either KWin or QtQuick (aurorae/plastik) You may want to export LIBGL_DEBUG=verbose export MESA_DEBUG=1 kwin_x11 --replace & And see whether any errors are noted before the crash happens (be prepared that this will print a lot of text - exporting GALLIUM_LOG_FILE might be reasonable) Well, I guess I missed your response there Thomas, sorry about that… ^_^ Anyway: I cannot reproduce this issue anymore and everything works as expected™. Since Nouveau have done a lot of refactoring and stability fixes upstream, I fairly sure that that has fixed this issue as well. :-) Thanks for your help! |