[Bug 1173727] New: Plasmashell crashes after relogin
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 Bug ID: 1173727 Summary: Plasmashell crashes after relogin Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: x86-64 OS: openSUSE Factory Status: NEW Severity: Major Priority: P5 - None Component: KDE Workspace (Plasma) Assignee: opensuse-kde-bugs@opensuse.org Reporter: simon.vogl@gmx.net QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- Created attachment 839354 --> http://bugzilla.opensuse.org/attachment.cgi?id=839354&action=edit Stack trace with debugsymbols Recently I noticed that plasmashell reports a crash in system tray after logging out and back in again a few times. At first plasma still seems to be working, but what's noticable is that all applications tend to start up A LOT slower. It never happens on first login, but after a few logins its easily reproductible even in a VirtualBox VM. I'm pretty sure the crash is not related to a graphics driver as it happens on both VirtualBox drivers and proprietary NVIDIA drivers. It may happen due to some patches applied to plasma5-desktop, KDE Frameworks or libqt5-base as I'm having great difficulties reproducting it in Arch Linux, however, it may also be an upstream bug Arch Linux has already patched, so no guarantees there. I've attatched a full stack trace with all debugsymbols installed, maybe that helps. Thanks for your efforts, DarkWav -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c1 Fabian Vogt <fabian@ritter-vogt.de> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |fabian@ritter-vogt.de, | |simon.vogl@gmx.net Flags| |needinfo?(simon.vogl@gmx.ne | |t) --- Comment #1 from Fabian Vogt <fabian@ritter-vogt.de> --- (In reply to Simon Vogl from comment #0)
Recently I noticed that plasmashell reports a crash in system tray after logging out and back in again a few times. At first plasma still seems to be working, but what's noticable is that all applications tend to start up A LOT slower. It never happens on first login, but after a few logins its easily reproductible even in a VirtualBox VM.
I can't reproduce the issue here at all, it might be related to the configuration, e.g. plasmoids or other settings. Can you try with a clean user account?
...
I've attatched a full stack trace with all debugsymbols installed, maybe that helps.
If I'm not mistaken, it shows that it tries to unload /usr/lib64/qt5/qml/QtQuick/Controls/libqtquickcontrolsplugin.so while plasmashell is still running in a separate thread - that looks very weird. Can you attach with gdb and try to get a deeper backtrace of all threads?
Thanks for your efforts, DarkWav
-- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c2 Simon Vogl <simon.vogl@gmx.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(simon.vogl@gmx.ne | |t) | --- Comment #2 from Simon Vogl <simon.vogl@gmx.net> --- Created attachment 839357 --> http://bugzilla.opensuse.org/attachment.cgi?id=839357&action=edit gdb run of plasmashell Thank you for your reply. I've reloggend until the issue happened, then killed plasmashell and restarted it inside GDB, then I ran "bt full". I've attatched the full GDB log. I hope that information is helpful. About reproducing: I can reproduce the issue with a both a clean user account and in a VirtualBox VM. It is not always easy, sometimes I have to relog 4-5 times to make the bug happen. However it might have to do with my CPU, ect. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c3 Fabian Vogt <fabian@ritter-vogt.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags| |needinfo?(simon.vogl@gmx.ne | |t) --- Comment #3 from Fabian Vogt <fabian@ritter-vogt.de> --- (In reply to Simon Vogl from comment #2)
Created attachment 839357 [details] gdb run of plasmashell
Thank you for your reply. I've reloggend until the issue happened, then killed plasmashell and restarted it inside GDB, then I ran "bt full". I've attatched the full GDB log. I hope that information is helpful.
Unfortunately not, you have to gdb -p $(pidof plasmashell) to attach to the crashed one and then run "thread apply all bt full".
About reproducing: I can reproduce the issue with a both a clean user account and in a VirtualBox VM.
It is not always easy, sometimes I have to relog 4-5 times to make the bug happen. However it might have to do with my CPU, ect.
-- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c4 Simon Vogl <simon.vogl@gmx.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(simon.vogl@gmx.ne | |t) | --- Comment #4 from Simon Vogl <simon.vogl@gmx.net> --- Created attachment 839358 --> http://bugzilla.opensuse.org/attachment.cgi?id=839358&action=edit gdb run with new instructions Thanks for the quick reply. I ran the instructions and attached the log. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c5 Fabian Vogt <fabian@ritter-vogt.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags| |needinfo?(simon.vogl@gmx.ne | |t) --- Comment #5 from Fabian Vogt <fabian@ritter-vogt.de> --- (In reply to Simon Vogl from comment #4)
Created attachment 839358 [details] gdb run with new instructions
Thanks for the quick reply. I ran the instructions and attached the log.
Unfortunately at that point the new plasmashell instance was already started to gdb attached to that one instead... Please run "pidof plasmashell" and attach to the lowest pid. Or ps -ef | grep drkonqi and you should see the pid as well. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c6 Simon Vogl <simon.vogl@gmx.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #839357|0 |1 is obsolete| | Attachment #839358|0 |1 is obsolete| | Flags|needinfo?(simon.vogl@gmx.ne | |t) | --- Comment #6 from Simon Vogl <simon.vogl@gmx.net> --- Created attachment 839359 --> http://bugzilla.opensuse.org/attachment.cgi?id=839359&action=edit gdb run with method 3 OK, I tried to attach GDB to the process with the lower PID. (2516 vs 2902, I chose 2516) I had to be quick though, after a few seconds the crashed plasmashell process would just disappear, but I managed to get a gdb stack trace in the end. Please pardon me for being bad at gdb debugging, I usually don't do that a lot ;) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c7 Fabian Vogt <fabian@ritter-vogt.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags| |needinfo?(simon.vogl@gmx.ne | |t) --- Comment #7 from Fabian Vogt <fabian@ritter-vogt.de> --- Ok, nothing new in that bt, unfortunately. I failed to reproduce the issue using the latest TW live cd, both with manual loggig out and in again and plasmashell --replace. My current guess is that something calls exit in thread 4, but unfortunately I don't know of a way to catch that after the fact. Can you reproduce this by just running "plasmashell --replace" a few times? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c8 Simon Vogl <simon.vogl@gmx.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(simon.vogl@gmx.ne | |t) | --- Comment #8 from Simon Vogl <simon.vogl@gmx.net> --- I actually needed to install TW to a disk in the VM to be able to reproduce the issue, the live CD cannot be used as far as I'm concerned. Additionally, this does not happen when logging in as root, but does happen when logging in as any normal user after a few logouts, even if freshly set up. It might also very well be a problem in some other component as running plasmashell --replace cannot trigger the bug after ~20 attempts whereas relogging usually triggers the bug first or 2nd try. I'm also suspecting this to be an upstream issue as I played around with Fedora 32 today, it worked great until I installed Plasma 5.19.2 from COPR. After that simmilar symptoms to what I experienced with TW started showing up. The symptoms are a minorly different, however, I belive this is because openSUSE uses dbus-daemon (without dbus-broker) while Fedora uses dbus-broker. ArchLinux somehow seems to get around this most of the time, however, it occasionally also affects Arch, its just a lot more rare there. I might test KDE neon later today, If it happens there its probably not openSUSE-related and it might be better to move this to KDE's bugtracker. Anyways thanks for your effort, I really appreciate it. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1173727 http://bugzilla.opensuse.org/show_bug.cgi?id=1173727#c9 Simon Vogl <simon.vogl@gmx.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |UPSTREAM --- Comment #9 from Simon Vogl <simon.vogl@gmx.net> --- OK, so I have some very interesting news. 1. The issue somehow does not exist on Arch Linux or KDE neon. 2. The issue does in fact exist on Fedora when using Plasma 5.19.2 from copr, meaning this glitch is in no way exclusive to openSUSE. 3. The issue seems to be related to plasmashell not shutting down correctly rather then having problems with startup. The issue can be fixed by doing the following: Make a shell script at ~/.config/plasma-workspace/shutdown/ that contains #!/bin/bash killall plasmashell make the file executable. After this, the issue completely disappears on both openSUSE and Fedora and the system works as expected. I think this is something the KDE developers should take a look at and it doesn't make much sense to discuss issues that also affect Fedora in this bugtracker. When I have time I will create a bug report in the KDE Bugzilla, but for now we can see this issues as resolved as it isn't openSUSE-specific and a workaround exists. Again, thank you very much for your support, I hope KDE will be able to resolve the issue ;) //CLOSED (RESOLVED UPSTREAM) -- You are receiving this mail because: You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠suse.com