You are not logged in.
Today I migrated completely from GNOME to the KDE/Plasma Environment and ran into a small problem. Whenever I want to shutdown or reboot the system via the application launcher akonadiserver crashes. In the few seconds between the crash report an the actual shutdown I was able to save the crash report (excerpt, im not sure which part is the important one):
Application: Akonadi Server (akonadiserver), signal: Segmentation fault
Using host libthread_db library "/usr/lib/libthread_db.so.1".
[Current thread is 1 (Thread 0x7efe4c8377c0 (LWP 983))]
I apart from a test configuration in KMail (which is now removed again) I didn't use any of the akonadi-related applications. The crash doesn't occur when I'm shutting down the system in the command line.
akonadi version is 16.12.3
What should do next / what kind of addtional information is needed?
Regards,
thurse
Last edited by thurse93 (2017-03-23 20:51:25)
Offline
This almost always happens for me as well. I've gotten in the habit of just running 'akonadictl stop' in a terminal prior to shutting down to avoid it.
Offline
Ahh... well yeah, that does the trick. I put your command in a little bash script, placed it in .config/plasma-workspace/shutdown/ and it works flawlessly.
But I'm starting to have doubts whether one can find the real source of the issue. There seem to be a lot of problems with this akonadi thing on the internet, just not my particular crash report.
Last edited by thurse93 (2017-03-24 07:10:57)
Offline
Mh, it seems like it didn't really work. Akonadi doesn't crash within approximately 5 minutes after the system has started. After that (or if i force some "akonadi-related stuff", e.g. change some calendar settings in the digital clock widget) akonadiserver does crash in a segfault every time, even with akonadictl stop written out in a terminal.
Offline
No shit? Akonadi crashes?? ... ;-P
You should get a crash dialog w/ a backtrace if you install drkonqi - otherwise either attach gdb to akonadi or inspect the backtrace using coredumpctl.
Offline
I have to say as a long time GNOME user I'm a bit irritated that a backend service crashes this often and (apparently) out of so many reasons...
Anyways, after building a debug-version of akonadi from the abs and installing qt-base-debug from the qt-debug-repo the backtrace crash report of drkonqi seems to reveal some more details... which I still do not really understand: http://pastebin.com/zeS7Sbej
Do I have to install even more stuff? Or can someone already point out the important parts and guide me from there?
Offline
File a bug at bugs.kde.org (from eg. within the dialog)
There's no mention of akonadi specific symbols, but the entire thing looks like a race where some mutex times out and ultimately tries to stop a QTimer that dereferences to nullptr. (ie. the troublemaker is actually thread #14) - the dbus call in thread #2 alone should be worrysome (iirc, multithreaded dbus isn't supported by Qt)
Afaict, people simply don't use akonadi - it's been a complete PITA from day #1 (and before that, kmail must be considered to have been one of the best available MUAs - if you rule out mutt for being mutt ;-)
Offline
Are you using BFS/MuQQS or similiar? It's bug in this scheduler, I think.
Offline
Are you using BFS/MuQQS or similiar? It's bug in this scheduler, I think.
Yes, atm I'm using the linux-ck-ivybridge kernel from graysky's repo-ck and indeed there are no crashes when booting vanilla. So akonadi _could_ be completely out of this???
Last edited by thurse93 (2017-03-26 19:55:02)
Offline
there's either a bug in the scheduler or it's just more prone to expose potential races - threading is random by definition and should not depend on certain scheduler (cpu/core count/...) aspects.
Offline
Ok, I filed the bug report: https://bugs.kde.org/show_bug.cgi?id=378135
Offline
pb wrote:Are you using BFS/MuQQS or similiar? It's bug in this scheduler, I think.
Yes, atm I'm using the linux-ck-ivybridge kernel from graysky's repo-ck and indeed there are no crashes when booting vanilla. So akonadi _could_ be completely out of this???
All akonadi crashes I've noticed lately is BFS related. It appears only on MuQSS or VRQ (see: http://cchalpha.blogspot.com/search/label/VRQ) for - if I remember - kernel 4.9. It doesn't matter that kernel is from graysky's repo or I've build it myself. Akonadi crashes only on kernel with MuQSS or VRQ patch. So this bug should be reported to Con Kolivas or/and Alfred Chen not to KDE.
BTW: I'm AMD APU user.
Last edited by pb (2017-04-02 15:37:47)
Offline