[FFmpeg-devel] [PATCH] Bink file demuxer and audio decoder

[Daniel Verkamp]

On Wed, Jan 27, 2010 at 5:14 AM,  <pross at xvid.org> wrote:
[...]
> Should the decoder be called binkaudio{1,2} or binkaudio{rdft,dct} ?
> the 1 and 2 numbers were arbitrarily assigned by me.

I would prefer binkaudio_{rdft,dct}, but not strongly.

Also, after some review of the demuxer, I think we could split audio
frames better; currently, we split them in the decoder.  The specific
example here is the first chunk of most files - it has enough audio
for several video frames (don't remember the exact number), but if we
split it in the demuxer, the split frames would still decode fine, and
we could eliminate the frame loop in the decoder.  This would also
avoid problems like overflowing the available audio decoding buffer
provided by ffmpeg...  The Smacker demuxer would also have to be
updated then.  If this sounds like the right way to go, I can provide
a patch.

Also, there is some memory corruption from reading beyond the end of a
buffer (spotted with Valgrind) in the audio decoder with DCT-format
audio; I'll investigate further, but it is crashing on the
float_to_int16_interleave() call on stereo files.  I assume there is a
problem with the sizes passed to this function, but I haven't verified
yet.

Thanks,
-- Daniel Verkamp