Bug#561553: ffmpeg: segmentation fault when creating mjpeg file from jpg sequence

[B. Bogart]

Disabling mmx does solve the segfault.

The cause of the problem is jpgs of different classes in my sequence. I 
ended up with some pseduoclass images in the sequence of largely 
directclass jpgs, which kill ffmpeg. With mmx enabled this results in a 
segfault, with mmx disabled ffmpeg just quits (return 1).

-v 10 does not give any hints about the quit, an error message would be 
useful when a pseudoclass jpeg in found in a sequence and exiting with 
an error message.

Thanks,
B. Bogart


Reinhard Tartler wrote:
> "B. Bogart" <ben at ekran.org> writes:
> 
>> Reinhard Tartler wrote:
>>> this is a good indication that this bug is also present in current
>>> trunk, i.e., this bug needs to be confirmed and upstreamed.
>> Compiling ffmpeg fresh for current svn also has the same problem.
> 
> okay, then this bug definitly upstreaming.
> 
> see http://ffmpeg.org/bugreports.html for details, much appreciated if
> you could help us out here.
> 
>>> Just a try: please disable your /usr/lib/i686/cmov/ directory (e.g.,
>>> by renaming). If this causes the symptom to disappear, then the
>>> defect is in the hand written assembler in i386/dsputil_mmx.c
>> renaming the directory has not changed the result:
>>
>> Program received signal SIGSEGV, Segmentation fault.
>> [Switching to Thread 0xf73776c0 (LWP 19789)]
>> 0xf79c77b0 in get_pixels_mmx (block=0x85fdbf0, pixels=0x0, line_size=336)
>>     at
>> /build/siretart-ffmpeg-debian_0.svn20080206-18-i386-ZyxKlX/ffmpeg-debian-0.svn20080206-18/libavcodec/i386/dsputil_mmx.c:225
>> 225
> 
> Hm, do we really enable mmx in the non-optimized build? I wonder if that
> is a clever idea, but this would be a seperate bug anyways.
> 
> Since you already have confirmed this bug in current trunk, would you
> mind to check if the symptom disappears if you run ./configure without
> mmx, e.g., add the configure switch --disable-mmx?
>