Hello list,

since updating both gstreamer and alsa this afternoon, I'm having
regular segfaults in quodlibet. However, since ql has not been updated I
have no idea against which package I should file a bug.

So, I set out to catch a core dump, and I'm trying to make sense of the
backtrace but apparently I'm still missing some -dbg files. Here is
what I have so far:

aschuring@neminis:~$ gdb `which python` core
[..]
[New Thread 3134]
[New Thread 3135]
[New Thread 3143]
[New Thread 3144]
[New Thread 3147]
[New Thread 3113]
[New Thread 3114]
[New Thread 3119]
[New Thread 3133]
warning: Can't read pathname for load map: Input/output error.
[Reading symbols..]
Core was generated by `python /usr/bin/quodlibet --start-playing'.
Program terminated with signal 11, Segmentation fault.
#0 0xb0158b4a in ?? ()
(gdb) thread 1
[Switching to thread 1 (Thread 3134)]#0 0xb0158b4a in ?? ()
(gdb) bt
#0 0xb0158b4a in ?? ()
#1 0x00000900 in ?? ()
#2 0x00000000 in ?? ()
(gdb) thread 2
[Switching to thread 2 (Thread 3135)]#0 0xb7828424 in __kernel_vsyscall ()
(gdb) bt
#0 0xb7828424 in __kernel_vsyscall ()
#1 0xb758a696 in __poll (fds=0xb760dff4, nfds=1, timeout=40)
at ../sysdeps/unix/sysv/linux/poll.c:87
#2 0xb5526c96 in ?? () from /usr/lib/libasound.so.2
#3 0xb5526e4b in snd_pcm_wait () from /usr/lib/libasound.so.2
#4 0xb0067e92 in gst_alsasink_write(asink=0xd634f88, data=0xd78b3b8,
length=7520) at gstalsasink.c:865
#5 0xb0149d47 in audioringbuffer_thread_func (buf=0xafb211f0) at
gstaudiosink.c:244
#6 0xb715bb6f in ?? () from /lib/libglib-2.0.so.0
#7 0xb77fc955 in start_thread (arg=0xafaffb70) at pthread_create.c:300
#8 0xb7597e7e in clone ()
at ../sysdeps/unix/sysv/linux/i386/clone.S:130


As you can see, the segfault occurs in a thread without symbols. Will
it make a difference if I install each and every -dbg package indicated
by gdb (that's about a zillion-and-one packages), or am I facing a
corrupted stack?

I've now reverted to alsa-base from Squeeze, but that shows the same
problem. Is there another way that I can get more info from a core
dump, or should I just stick to downgrading/upgrading packages until
I've found the cause?


Thanks,
Arno


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110429221204.54f082f4@neminis.loos.site">http://lists.debian.org/20110429221204.54f082f4@neminis.loos.site

On Fri, 29 Apr 2011 22:12:04 +0200, Arno Schuring wrote:

> since updating both gstreamer and alsa this afternoon, I'm having
> regular segfaults in quodlibet. However, since ql has not been updated I
> have no idea against which package I should file a bug.
>
> So, I set out to catch a core dump, and I'm trying to make sense of the
> backtrace but apparently I'm still missing some -dbg files. Here is what
> I have so far:

(...)

> warning: Can't read pathname for load map: Input/output error. [Reading
> symbols..]
> Core was generated by `python /usr/bin/quodlibet --start-playing'.
^^^^^^^^^^^^^^^

Have you tried to launch the app with no options or just crashes in the
same manner? :-?

> As you can see, the segfault occurs in a thread without symbols. Will it
> make a difference if I install each and every -dbg package indicated by
> gdb (that's about a zillion-and-one packages), or am I facing a
> corrupted stack?

I agree that installing "-dbg" packages for gstreamer and gst-python can
be an insane task -unless you have a testing machine where you can do and
undo at your wish- :-). I don't know the package but maybe it is possible
to recompile "quodlibet" with the symbols on it... just a wild-guess.

> I've now reverted to alsa-base from Squeeze, but that shows the same
> problem. Is there another way that I can get more info from a core dump,
> or should I just stick to downgrading/upgrading packages until I've
> found the cause?

Why not just opening a bug report on BTS against "quodlibet" and wait
until someone tells you the next step? Should devels/packagers need the
symbols then also tell you the exact required extra packages to get them.

Greetings,

--
Camaleón


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: pan.2011.05.01.17.46.18@gmail.com">http://lists.debian.org/pan.2011.05.01.17.46.18@gmail.com

Hi again,

Thanks for the response.

Camaleón (noelamac@gmail.com on 2011-05-01 17:46 +0000):
> On Fri, 29 Apr 2011 22:12:04 +0200, Arno Schuring wrote:
>
> > since updating both gstreamer and alsa this afternoon, I'm having
> > regular segfaults in quodlibet. However, since ql has not been
> > updated I have no idea against which package I should file a bug.
> >
[..]
> > warning: Can't read pathname for load map: Input/output error.
> > [Reading symbols..]
> > Core was generated by `python /usr/bin/quodlibet --start-playing'.
> ^^^^^^^^^^^^^^^
>
> Have you tried to launch the app with no options or just crashes in
> the same manner? :-?
In fact, the app would not crash while playing but at the end of the
song (and not even on all songs). After fruitlessly digging around for
awhile, I was able to reproduce this with totem and rhythmbox as well.
Even then, it took me about four hours to find the offending package
(liborc).

Long story short, this has been resolved. I have not opened a bug
because a newer version (from incoming.debian.org) already had a fix.
For those interested, I wrote down my experiences here:
http://lists.alioth.debian.org/pipermail/pkg-gstreamer-maintainers/2011-April/008293.html


Regards,
Arno


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110502000404.18ff7992@neminis.loos.site">http://lists.debian.org/20110502000404.18ff7992@neminis.loos.site