[FFmpeg-devel] libavfilter: single ffplay.c diff?

Vitor Sessak vitor1001
Sun Oct 12 22:21:18 CEST 2008


Stefano Sabatini escreveu:
> On date Thursday 2008-10-09 22:32:47 +0200, Vitor Sessak encoded:
>> Stefano Sabatini escreveu:
>>> On date Wednesday 2008-10-08 20:20:10 +0200, Vitor Sessak encoded:
>>>> Stefano Sabatini escreveu:
>>>>> On date Thursday 2008-10-02 23:21:47 +0200, Vitor Sessak encoded:
>>>>>> Stefano Sabatini wrote:
>>>>>>> On date Wednesday 2008-10-01 07:57:09 +0200, Vitor Sessak encoded:
>>>>>>>> Stefano Sabatini wrote:
> [...]
>>>>>>> I'll hope to be able to work at the vsrc_movie.c ffplay.c refactoring
>>>>>>> now...
>>>>>> If you want me to test, cleanup and commit the patch I posted, I can do 
>>>>>> that real soon, just let me know. If want to do it yourself, do not 
>>>>>> hesitate in asking me if you have any doubt.
>>>>> Please do it if you can, I won't have much time this weekend so I'll
>>>>> try to use it for the other patches.
>>>> Done.
>>> Thanks.
>>>
>>> But now I see that ffplay is not anymore updated against latest SVN
>>> and the chechout.sh script fails.
>>>
>>> Attached patch update both ffplay diff and checkout.sh.
>> Thanks for the patch, but you can just commit such obviously correct 
>> patches without asking.
> 
> OK, committed.
> 
> BTW, Vitor, do you have an opinion regarding this:
> http://thread.gmane.org/20081005222842.GA31799 at geppetto
> ?
> 
> Is that to be considered a libavfilter bug, could we find a workaround
> for it or do you consider that that flag implementation could be
> useful?

My opinion is that it is a bug in the libavfilter patch against 
ffplay.c. If you use vsr_buffer.c, the filter flag the frame as 
"AV_PERM_WRITE | AV_PERM_PRESERVE | AV_PERM_REUSE2". That means that the 
filter framework don't expect this frame to be available later nor it 
can modify it (that is important for codecs that will use the frame 
later for ME). So the filter framework will copy this frame if it will 
need it anytime after the first request_frame() call, and there should 
be no problem if it is free'd...

-Vitor






More information about the ffmpeg-devel mailing list