Bug#648843: ffmpeg: segmentation fault when transcoding mkv to mpg, Bug#648843: ffmpeg: segmentation fault when transcoding mkv to mpg

[Reinhard Tartler]

tags 648843 upstream
stop

Hi Eric,


On Di, Nov 15, 2011 at 16:30:05 (CET), Debian BTS wrote:


[...]

>
>     $ ffmpeg -i 05\ The\ Rebel\ Flesh.mkv -target ntsc-dvd /tmp/5.mpgffmpeg version 0.7.2-4:0.7.2-1+b1, Copyright (c) 2000-2011 the Libav developers
>       built on Oct 27 2011 17:43:14 with gcc 4.6.2

[...]

>     [dvd @ 0x245f0e0] buffer underflow i=1 bufi=1017 size=1792ate=4010.4kbits/s dup=4777 drop=0    
> 	Last message repeated 3 times051252kB time=2596.59 bitrate=3316.6kbits/s dup=12904 drop=0    
>     Input stream #0.1 frame changed from rate:48000 fmt:flt ch:6 to rate:48000 fmt:flt ch:2
>     Warning, using s16 intermediate sample format for resampling
>     Segmentation fault

Can you please install the package libav-dbg and attach a stacktrace for
this crash?

> I had first tried the same command with "-threads 6", and it
> printed the following before crashing:
>
>     ffmpeg: malloc.c:4631: _int_malloc: Assertion `(unsigned long)(size) >= (unsigned long)(nb)' failed.
>     Aborted
>
> Then I tried the one above, in case it was a multi-threading issue,
> but that doesn't appear to be the case.

Well, if it behaves differently depending on the number of threads, then
I'd consider it as an upstream bug. However, this may very well be a
different issue.

> The input .mkv file is 1.1GB; I can upload it somewhere upon request.

Please try to reproduce the issue with a shortened version of the
file. See also http://libav.org/bugreports.html, section
"Submitting Sample Media" for advice on that topic.


[...]

On Di, Nov 15, 2011 at 16:51:25 (CET), Eric Cooper wrote:

> Here's some additional information.  The resulting .mpg files all seem to
> play OK, and do not seem to be truncated.
>
> I tried transcoding another .mkv file to mpg, and a similar failure occurred.
> But ffmpeg produced much more debugging info this time:
>

[...]

> Input stream #0.1 frame changed from rate:48000 fmt:s16 ch:6 to rate:48000 fmt:flt ch:6
> [dvd @ 0x11edbc0] buffer underflow i=1 bufi=1638 size=1792te=4084.1kbits/s dup=296 drop=0    
>     Last message repeated 10 times46158kB time=2705.30 bitrate=3470.7kbits/s dup=13445 drop=0    
> Input stream #0.1 frame changed from rate:48000 fmt:flt ch:6 to rate:48000 fmt:flt ch:2
> Warning, using s16 intermediate sample format for resampling
> *** glibc detected *** ffmpeg: double free or corruption (out): 0x0000000003d30060 ***

[...]

This is a different symptom, which again may or may not mean a different
bug. Again, I'd like to see a backtrace and ideally a valgrind trace for
this testcase.

See the following pages for additional information on Backtraces and Valgrind:

http://wiki.debian.org/HowToGetABacktrace
https://wiki.ubuntu.com/Backtrace
https://wiki.ubuntu.com/Valgrind

Cheers,
Reinhard

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4