[FFmpeg-devel] [PATCH] matroskadec: Fix TTA duration calculation

[James Almer]

On 15/08/13 7:44 AM, Paul B Mahol wrote:
> On 8/15/13, Michael Niedermayer <michaelni at gmx.at> wrote:
>> On Thu, Aug 15, 2013 at 06:22:47AM -0300, James Almer wrote:
>>> Still not perfect due to rounding errors, but several samples should
>>> demux correctly now instead of none.
>>
>> please add a link to a sample to reproduce this bugfix in the commit
>> message
> 
> Samples are not needed at all, this can be easily reproduced.
> 
> And its incorrect to claim current one could not demux correctly any files.
> It just fails if last packet have number of samples smaller than all other
> packets before, which is 'almost' always but not always.

Yeah, bad choice of words from my part. It is about the last frame when its 
size is different than the rest.

> Question to OP: I wonder does other program(s) always decode last frame no
> matter what number of samples are in it?

No, when dealing with files where the last frame is different than the rest,
foobar2000 successfully decodes said frame in some, and fails in others.
With this patch ffmpeg does the same, but some files that fail with fb2k might 
work, and vice-versa.

Regards.