[FFmpeg-devel] [PATCH] Google WebP support

[Pascal Massimino]

Ronald,

On Thu, Oct 28, 2010 at 5:14 AM, Ronald S. Bultje <rsbultje at gmail.com>wrote:

> Hi,
>
> On Thu, Oct 28, 2010 at 1:17 AM, Pascal Massimino
> <pascal.massimino at gmail.com> wrote:
> > On Tue, Oct 19, 2010 at 10:34 AM, Pascal Massimino <
> > pascal.massimino at gmail.com> wrote:
> >
> >> Aurelien,
> >>
> >> On Tue, Oct 19, 2010 at 4:30 AM, Aurelien Jacobs <aurel at gnuage.org>
> wrote:
> >>
> >>> On Mon, Oct 18, 2010 at 07:39:40PM -0700, Pascal Massimino wrote:
> >>> > [...]
> >>> >
> >>> > About the other webpdec.c/webpenc.c: metadata now seem to no longer
> >>> works.
> >>> > AVMetadataConv have been deprecated just after i finished addiung
> them.
> >>> > Should i come back to manually mapping tag/metadata keys? (pretty
> much
> >>> > like AVMetadataConv was doing?)
> >>>
> >>> No. It is not deprecated. It has been moved to internal API. So now,
> >>> your demuxer has to call ff_metadata_conv() by itself, instead of
> >>> exporting the conversion table in AVInputFormat. That's the only change
> >>> required.
> >>>
> >>
> >> actually, i don't need to change stream->metadata to use native tag
> (since
> >> i'd need to convert back to generic down the line afterward. And it's
> just
> >> about
> >> putting the tags on the wire so i shouldn't change the object's generic
> tag
> >> i'd say).
> >> $attached?
> >>
> >>
> > ping? does someone want the webp{enc,dec}.c patch? [ ] yes   [ ] no
>
> I personally thought that your later img2.c version was much cleaner,
> regardless of the few extra ifs. It's smaller and works for -i
> bla%02d.webp input strings. What was the reason you didn't continue
> that one?
>

well, if i submit the change in img2.c, it's take precedence when handling
.webp image and then i'll lose the ability to handle metadata (and it's also
an annoying code duplication -or lack of, regarding metadata!- between
img2.c and webp{enc,dec}.c). Oh, and there's the force-keyframe problem
too.
So, if i change img2.c as is now, i'll have to use -f webp to fallback to
the
real expected behaviour while these img2 problem are being fixed.
Speaking of fixing img2 behaviour, i'd say that it should be changed to
call muxer/demuxer instead of simply codec, since the muxer is the one
packaging the whole webp packets. But then it's a whole lot more work
and i'm not even sure how that is possible to start with...

skal


> Ronald
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>