[issue2141] avcodec crash in amarok

[Carl Eugen Hoyos]

Carl Eugen Hoyos ceho...@rainbow.studorg.tuwien.ac.at added the comment:

If there really is an issue with float_to_int16_interleave_sse2(), I believe it
should also be reproducible with ffmpeg -i wmafile -f null -

Note that if you are able to reproduce the issue, you will have to verify that
it is also reproducible with latest svn, since other versions are not supported
on this bug tracker.

--
status: new - open
substatus: new - needs_more_info


FFmpeg issue tracker iss...@roundup.ffmpeg.org
https://roundup.ffmpeg.org/issue2141

[issue2141] avcodec crash in amarok

[Mans Rullgard]

Mans Rullgard m...@mansr.com added the comment:

Ronald S. Bultje iss...@roundup.ffmpeg.org writes:

 #6  0x014d789d in float_to_int16_interleave_sse2 (dst=0xd788b28,
 src=0x960470a8, len=2048, channels=2) at
 /usr/src/debug/ffmpeg-0.6/libavcodec/x86/dsputil_mmx.c:2475
 #7  0x01450cb4 in wma_decode_frame (s=0xd7c79b0, samples=0xd788b28) at
 /usr/src/debug/ffmpeg-0.6/libavcodec/wmadec.c:810
 #8  0x014522e1 in wma_decode_superframe (avctx=0xb9da2e0, 
 data=0xd788b28,
 data_size=0x960471fc, avpkt=0x96047134) at

 Probably true, although Amarok is responsible for aligning samples (not 
 16-byte aligned here), we should align output[] (stack-allocated, and 
 thus not 16-byte aligned). See line 807 in wmadec.c.

That is just an array of pointers to each channel.  It does not need
to be aligned, the pointers within it must be, and they are.  The bug
is firmly in Amarok territory.


FFmpeg issue tracker iss...@roundup.ffmpeg.org
https://roundup.ffmpeg.org/issue2141

[issue2141] avcodec crash in amarok

[Ronald S. Bultje]

Ronald S. Bultje rsbul...@gmail.com added the comment:

(Actually, in this case it probably means the bug is in Xine, not Amarok, 
looking at the backtrace in the original bugreport.)


FFmpeg issue tracker iss...@roundup.ffmpeg.org
https://roundup.ffmpeg.org/issue2141