[FFmpeg-devel] libavfilter-soc update and variable frame size/pixel format

Vitor Sessak vitor1001
Wed Jul 1 20:25:27 CEST 2009

Baptiste Coudurier wrote:
> Hi Michael,
> Michael Niedermayer wrote:
>> On Mon, Jun 29, 2009 at 09:57:04PM +0200, Stefano Sabatini wrote:
>>> On date Monday 2009-06-29 21:01:15 +0200, Michael Niedermayer encoded:
>>>> On Mon, Jun 29, 2009 at 08:54:38PM +0200, Stefano Sabatini wrote:
>>>>> On date Monday 2009-06-29 01:26:11 +0200, Michael Niedermayer encoded:
>>>>>> On Mon, Jun 29, 2009 at 01:05:19AM +0200, Stefano Sabatini wrote:
>>>>>>> Hi, since r19226/19279 ffmpeg supports variable frame size/pixel
>>>>>>> formats.
>>>>>>> I managed to update libavfilter-soc, and my current solution is to
>>>>>>> simply don't support it if libavfilter is enabled.
>>>>>>> That's because those are ones of the problem lavfi should natively
>>>>>>> manage, rather than leave the problem to the application, also I don't
>>>>>>> want to spend time on this since that will be removed anyway once it
>>>>>>> will be supported by the lavfi framework.
>>>>>>> See updating patch attached, OK to apply?
>>>>>> i dont review diffs of diffs :)
>>>>> In attachment the diff. Note that I'm just keeping the old behaviour
>>>>> when CONFIG_AVFILTER is enabled.
>>>> i also dont review that
>>>> we have A ---diff1--->B
>>>> and
>>>> we have A ---diff1--diff2->B
>>>> i review diff2, i wont review a combined diff1+diff2 nor a diff between
>>>> diff1 and diff1+2
>>>> not sure why you post the change like this ...
>>> The diff of diff still looks to me the more readable alternative,
>>> considering that the difference between the two patched versions
>>> (19066 and 19289) contains many diffs related to the ffmpeg
>>> development.
>>> Also to get such a patch is quite painful, and that was the reason for
>>> which I avoided it the first time, to get it I had to go through:
>>> svn export svn://svn.ffmpeg.org/ffmpeg/trunk/ffmpeg.c ffmpeg.c.r19XXX -r 19XXX
>>> patch -p0 < ../diffs/02_ffmpeg_filters.diff.r19XXX
>>> svn export svn://svn.ffmpeg.org/ffmpeg/trunk/ffmpeg.c ffmpeg.c.r19YYY -r 19YYY
>>> patch -p0 < ../diffs/02_ffmpeg_filters.diff.r19YYY
>>> diff -u ffmpeg.c.r19XXX  ffmpeg.c.r19YYY > patch
>>> Note that the new version passes regression tests as before (that is all but the
>>> old mxdf10 one), so I believe it should quite safe to apply it.
>>> Regards.
>>> -- 
>>> FFmpeg = Forgiving & Faboulous Merciful Patchable Evangelical Gangster
>>>  ffmpeg.c.r19289 |  161 ++++++++++++++++++++++++++++++++++----------------------
>>>  1 file changed, 99 insertions(+), 62 deletions(-)
>>> 2763717cd64dbdcf3590f734da6a1a4215fbca3d  patched-ffmpeg-10966-19289.patch
>> patch rejected, mixes unrelated changes
>> note, before you waste more of our both time, acceptable patches are either
>> 1. mechanical updates that is application of changes from ffmpeg svn that
>>    are not in soc yet.
>> 2. clean patches passing the full set of rules about clean patches 
>> In that sense i like to again mention, that lavfi should be moved into ffmpeg
>> svn quickly because it does become more painfull with time (now a merge
>> _requires_ support for changing width/height as well, a short while ago it
>> didnt but now it would be a feature regression)
> Can you or someone else please state what needs to be done to finallly
> move lavfi into ffmpeg svn ? Maybe it would be possible to help.

IMHO what is missing is:

1- get the Stefano's padding filter accepted by Michael (see thread 
"libavfilter-soc: direct rendering"). This patch is fine by me, but it 
introduces a useless memcpy() in ffmpeg.c
2- get vf_scale.c and vf_crop.c accepted
3- get the ffmpeg.c and ffplay.c patches accepted (requires also getting 
vsrc_buffer.c accepted)
4- making libavfilter work with variable frame dimensions/colorspace

Maybe Stefano and Michael want also to comment too?


More information about the ffmpeg-devel mailing list