Summary: | kwin_x11 crashes if Minecraft 1.6.4 (Modded) starts and finishing loading | ||
---|---|---|---|
Product: | [Plasma] kwin | Reporter: | Linus <wendefelix1998> |
Component: | decorations | Assignee: | KWin default assignee <kwin-bugs-null> |
Status: | RESOLVED WAITINGFORINFO | ||
Severity: | crash | ||
Priority: | NOR | ||
Version First Reported In: | 5.4.2 | ||
Target Milestone: | --- | ||
Platform: | Ubuntu | ||
OS: | Linux | ||
URL: | https://youtu.be/jHEEnC79HZc | ||
Latest Commit: | Version Fixed In: | ||
Sentry Crash Report: | |||
Attachments: |
Kwin-Crashlog
Kwin-Crashlog (with 'grep -i buffer') |
Description
Linus
2016-02-09 15:53:28 UTC
Sounds terribly like bug #344326 Can you run "xev -root -event randr", start minecraft and see whether the former produces any output? Suspending the compositor before running minecraft (SHIFT+Alt+F12), resp. applying a window rule to have mincraft block the compositor (last tab) should bypass the problem (as toggling the compositor twice should simply clear the stage after the issue occurred) I tried it. "$ xev -root -event randr" did not gave me any output while crashing. Blocking the Compositor with SHIFT + ALT + F12 seems to work best. As I tried it with the Window-Rule and start Minecraft the Compositor got blocked but then my middle panel (Windowlist) freezed and kwin shuts down (windowdecoration was gone). After I restarted both (kwin_x11 and plasmashell) I noticed that the Hotkey ALT + F2 to enter commands did not work (sure only till the next reboot) but my Hotkey to pop up my Application-Launcher (SUPER + Y). To sum it up: - Not a cause of the "Root-Window" ("$ xev -root -event randr" with no output) - Disabling Compositor works but caused a big crash as I set it as windowrule. I found out that if kwin_x11 crashes (shattered screen) I can use the Hotkey to disable/enable the compositor twice instead of restarting kwin_x11. Then this image is restored. This means that it definitly is a bug of the compositor. Maybe a fullscreen or OpenGL problem? The "crash" you report (ie. the distorted image) sounds terribly like bug #344326 while the segfaults of comment #2 sound terribly like bug #341497, what is why I believe minecraft alters the screen resolution (despite you didn't log any event) run QT_LOGGING_RULES="kwin_core.debug=true" kwin_x11 --replace & and watch out for ""Buffer detailed info:" messages (maybe grep -i for "buffer", If the message changed, or workaround won't work anymore) Created attachment 97174 [details]
Kwin-Crashlog
Created attachment 97175 [details]
Kwin-Crashlog (with 'grep -i buffer')
I run it with the command "QT_LOGGING_RULES="kwin_core.debug=true" kwin_x11 --replace 2> kwin.log" I added the attachments (Kwin-Crashlog) and (Kwin-Crashlog (with 'grep -i buffer')) Seems as if the VBOs get transferred forth and back when starting minecraft in return of a randr event (the refresh rate output) what certainly means we're in the territory of bug #344326. KWin also reloads VBOs, so the question is what fails next. You then seem to apply a kwin rule to block compositing for minecraft => is this a resolution or maybe rather the trigger of the problem? I think yes, because I had a similar bug a long time ago with Plasma where Kwin mostly crashing everytime, I adding/removing/activating/deactivating a monitor. Currently I have two Monitors: DVI (BenQ): Primary Monitor [1920x1080] VGA (Samsung TFT): Secondary Monitor [1024x768] HDMI (Samung SmartTV): Gaming Monitor (uses when play Steam-Games with Controller) [1920x1080; Mirror of BenQ] Screenshot of KDE's Monitor-Settings: http://i.epvpimg.com/pkvcd.png Everytime I starting my SmartTV (if not at current boot), kwin crashes. My Laptop which I using in School and which is running ArchLinux with latest Updates, crashes often when I unplug a Monitor. But then kwin restarts and shows me a Crash-Dialog. It seems that kwin has mayor problems with multiple Monitors or screen resolutions. I think that it isn't a fault of my Nvidia GTX 750 Ti because my Laptop only has a Intel i3 with HD Graphics 2000 which should work fine on Linux. crashes on randr events are a common bug in qt5, see bug #341497 on details - that doesn't indicate anything and is not related to the VBO transfer problem (which is probably the cause for the screen distortion and *may* crash kwin as well, but on different stack positions) So again: you added the blocking rule to mitigate the distortion? |