Closed ericjmcalvin closed 3 years ago
I'm pretty sure it's a Qt bug. I found a few similar reports on bugreports.qt.io. Most likely Qt doesn't handle some system errors correctly and eventually crash. For example in cases of some incompatibility with video driver.
So if you use non-intel video, try to switch to intel.
I'm pretty sure it's a Qt bug. I found a few similar reports on bugreports.qt.io. Most likely Qt doesn't handle some system errors correctly and eventually crash. For example in cases of some incompatibility with video driver.
So if you use non-intel video, try to switch to intel.
Well, it is a bit more complicated than that. We are using vncviewer to connect to a virtual headless session (running Xvnc (TigerVNC)). The window manager in the VNC session is FVWM. The underlying hardware on the machine running Xvnc is Nvidia with G05 (435) drivers.
From there, we SSH to another machine (with X Forwarding enabled) to run psi+ (that remote machine also has nvidia drivers (G05/440).
Well I also had similar crashes with nvidia. all them were related to nvidia driver upgrade without consequent reboot. After the reboot everything worked properly. But I usually had a crash on start, so it was not something random.
If you are somehow able to reproduce it (like exact steps to make it crash) it would be super helpful. Currently from the stack trace above I see there is no explicit issue with Psi. But I remember Qt changed something in multithreading handling in webengine. If this is an issue, like Qt-5.12 still has some unsafe code, then Qt upgrade would help as well as using Psi version compiled without webengine.
@ericjmcalvin Could you check if this crash reproducible in our AppImage builds of Psi+?
@ericjmcalvin Could you check if this crash reproducible in our AppImage builds of Psi+?
Apologies for the delayed reply. I am unable to use AppImages in this particular environment (and I can't reproduce it successfully outside of it). However, other users did experience the issue (this time psi-plus would immediately core dump and the backtrace showed nvidia_glcore.so as the culprit. Also, running glxgears on the same system caused the same core dump backtrace. So this may be due to iGLX perhaps (as this is displayed over SSH + XForwarding). So, after upgrading to the latest Nvidia drivers (460) the issue so far seems to be resolved. I don't think we need to keep this issue opened as it is obviously either a QT library or Nvidia library bug and not psi-plus. Thank you for your help!
OS: openSUSE 15.2 x86_64 psi+ version: psi+-1.4.1159+0-lp152.1.1.x86_64
Randomly when in a chat room, psi+ will segfault. I managed to get a core file and loaded it into GDB. Here is the gdb output: