[FFmpeg-devel] [ANNOUNCE] ffmpeg-mt merged

[Thomas Worth]

On Mon, Mar 21, 2011 at 2:23 PM, Vitor Sessak <vitor1001 at gmail.com> wrote:
> On 03/21/2011 09:42 PM, Thomas Worth wrote:
>>
>> On Mon, Mar 21, 2011 at 12:56 PM, Michael Niedermayer<michaelni at gmx.at>
>>  wrote:
>>>
>>> Hi
>>>
>>> I had between  work, sleeping, washing clothes, cleaning up my appartment
>>> rebuilding the server infrastructure, eating and a few other things a few
>>> hours
>>> time today so i
>>>
>>> * merged ffmpeg-mt in
>>> * fixed it to pass fate
>>>
>>> Please help testing and reporting bugs to our tracker. And dont just stop
>>> there
>>> if you can fix a bug thats even better! :)
>>> And someone should write a news entry ...
>>
>> This is great news, however the seeking problem with H.264 persists
>> and now there is no alternative (it was not a problem with normal
>> FFmpeg).
>>
>> Unfortunately I don't know enough about the guts of avcodec to fix the
>> problem, however I do believe the problem is related to
>> ff_thread_decode_frame(). So, h264.c or pthread.c, perhaps? Sorry I
>> can't be more helpful!
>
> The best way would be to reproduce the problem using either ffplay or the
> ffmpeg binary.

Good idea, I'll try that. Does ffplay allow seeking on a per-frame
basis? To reproduce this one would need to set AVSEEK_FLAG_ANY to
return non-I frames (I frames seem to work OK during seek, it's only
non-I frames that get messed up).