GNOME Bugzilla – Bug 590252
[flacdec] Totem segfaults while seeking flac
Last modified: 2009-07-30 16:11:59 UTC
After bug 578612 was fixed I still get occasional segfaults when seeking in flac files. The problem does not seem to be the same as bug 589849 as it does not occur at the end of a song.
Created attachment 139554 [details] GST_DEBUG=*sink:5,flac*:5,*audio*:5 log
Since it's a crash, a stacktrace made with (gdb) thread apply all bt would also be nice.
[New Thread 0xb5f0bb90 (LWP 17108)] [New Thread 0xb4fffb90 (LWP 17109)] [New Thread 0xb47afb90 (LWP 17110)] [Thread 0xb47afb90 (LWP 17110) exited] [Thread 0xb4fffb90 (LWP 17109) exited] [New Thread 0xb4fffb90 (LWP 17111)] [Thread 0xb5f0bb90 (LWP 17108) exited] [New Thread 0xb5f0bb90 (LWP 17112)] [Thread 0xb4fffb90 (LWP 17111) exited] [New Thread 0xb4fffb90 (LWP 17113)] [New Thread 0xb47afb90 (LWP 17114)] [New Thread 0xafee8b90 (LWP 17115)] [New Thread 0xaf6e7b90 (LWP 17116)] [New Thread 0xaeee6b90 (LWP 17117)] [Thread 0xafee8b90 (LWP 17115) exited] (totem:17102): GStreamer-CRITICAL **: gst_caps_is_fixed: assertion `GST_IS_CAPS (caps)' failed (totem:17102): GStreamer-CRITICAL **: gst_caps_is_any: assertion `GST_IS_CAPS (caps)' failed (totem:17102): GStreamer-CRITICAL **: gst_caps_is_empty: assertion `GST_IS_CAPS (caps)' failed (totem:17102): GStreamer-CRITICAL **: gst_caps_is_any: assertion `GST_IS_CAPS (caps)' failed (totem:17102): GStreamer-CRITICAL **: gst_caps_is_empty: assertion `GST_IS_CAPS (caps)' failed Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0xb4fffb90 (LWP 17113)] 0xb7e931a4 in gst_caps_subtract (minuend=0x8fc9880, subtrahend=0x8fc9640) at gstcaps.c:1382 1382 sublen = subtrahend->structs->len; (gdb) thread apply all bt
+ Trace 216716
Thread 1 (Thread 0xb6d5a750 (LWP 17102))
This looks very much like bug #575349, which I also got with flacdec, but that was possibly because I was only testing with flac, the original reporter was using mad as far as I can tell.
If you think it's the same, feel free to mark as dupe, I will follow the other bug. I cannot reproduce this with my vorbis files but I just ripped to mp3 and I see the same problem there as well.
> If you think it's the same, feel free to mark as dupe, I will follow the other > bug. I cannot reproduce this with my vorbis files but I just ripped to mp3 and > I see the same problem there as well. Ok, that means it's not a bug in flacdec then and most likely a duplicate. Unfortunately I have absolutely no idea where the issue might be. Sounds like a refcount problem somewhere, but where (and why does it not happen more often)? It would be interesting to try a pipeline that doen't contain any basetransform-based elements (audioconvert/audioresample) to narrow it down a bit. Something like filesrc ! flacdec ! pulsesink or so. *** This bug has been marked as a duplicate of 575349 ***