Missing data with vp8dec/avdec_vp8 and filesink

[Jeppe Ledet-Pedersen]

On 2014-08-26 10:41, Tim Müller wrote:
> On Tue, 2014-08-26 at 08:38 +0200, Jeppe Ledet-Pedersen wrote:
>
> Hi Jeppe,

Hi Tim,

>> We recently switched to the avdec_vp8 decoder because it seems to run
>> better on our test devices. However, with the libav decoder, the
>> output-local file is always missing the last data, so the test fails:
>>
>> test % ls -l output*
>> -rw-r--r-- 1 jlp jlp 42739200 Aug 25 17:02 output-local
>> -rw-r--r-- 1 jlp jlp 44620800 Aug 25 17:01 output-remote
>> test % md5sum output*
>>
>> e947e40881ff9ea47b3c341c69bab0aa  output-local
>> 6c07a9ef488d5ca6150e762ec5c8c6d0  output-remote
>>
>> The remote transfer outputs the same data as before. Shouldn't the
>> pipeline output all the data on EOS or isn't this guaranteed? Am I
>> missing something obvious?
>
> It looks like a bug. Might be fixed already in git master and 1.4
> branch:
>
> https://bugzilla.gnome.org/show_bug.cgi?id=734608
>
> so the fix should be in the upcoming 1.4.1 bug-fix release.

Thank you for the fast reply.

I just tried with the 1.4 branch from git and got the same result - the 
output from avdec_vp8 is still smaller than with vp8dec. I have also 
tried to cherry-pick this commit (which hasn't hit the 1.4 branch yet) 
with the same result:

http://cgit.freedesktop.org/gstreamer/gst-libav/commit/?id=8c00a359d0f012014c2337bd45f03e4a549a00bd

If I try to build gst-libav master, I get an "FF_IDCT_XVID undeclared" 
error in gstavcfg.c:179, so I can't test that.

With gst-launch-0.10 (version 0.10.36) and ffdec_vp8, I get the expected 
file size, so this does look like a regression bug.

Can you reproduce the bug on master?

Thanks,

-Jeppe