Red Hat Bugzilla – Bug 478559
assertion pthread_setspecific(t->key, userdata) == 0'
Last modified: 2009-09-10 14:24:14 EDT
Description of problem:
every once in awhile, gimp will just crash on me with the following
E: thread-posix.c: Assertion 'pthread_setspecific(t->key, userdata) == 0' failed at pulsecore/thread-posix.c:200, function pa_tls_set(). Aborting.
gimp: terminated: Aborted
(script-fu:11918): LibGimpBase-WARNING **: script-fu: gimp_wire_read(): error
It seems to be an abort from the pulseaudio layer (although what pulseaudio has to do with the gimp I have no idea).
Unfortunately, I can't make this crash happen at will, but it's happened enough that it's very annoying.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
No reproducer. I use gimp and once in awhile and gimp will crash with the above assertion.
I don't use pulseaudio in my setup. basically, i remove the pulseaudio alsa plugin package.
Created attachment 328047 [details]
backtrace of assertion
I caught the crash and here's the backtrace. In this case, I had two files open and I closed one then crash.
ok, I did some looking around in libcanberra and set this in my environ:
$ printenv |grep CAN
I get gnome theme sounds now, and I'm betting gimp->libcanberra doesn't call into pulse anymore (which is what i want).
At this point I don't know which component has the bug (gimp/libcanberra/pulseaudio). I leave this to the experts.
Argh, just found this while cleaning up Bugzilla, sorry for the late reaction. Looking at the backtrace, I have a strong hunch that the problem was in libcanberra itself (don't know if there are updates already which would mitigate this issue), changing component.
Hmpf, smells as if something ran out of TLS handles. That really should not happen. Something might be leaking TLS instances.
Could you please retry with something newer than F10? F11? Rawhide?
This machine is at F11+updates now. I'll remove the canberra->alsa shim (I've been running stable that way ever since) and see what happens for the next few days. I'll report back.
So I haven't run into the assertion at all now, so this probably got fixed at some point. Can be closed I think.
OK, closing. Feel free to reopen if this resurfaces.