[FFmpeg-devel] [PATCH] avcodec/keyframe_bsf: add keyframe bitstream filter

[Michael Niedermayer]

On Sun, Jun 08, 2014 at 09:16:25PM +0200, Reimar Döffinger wrote:
> On Sun, Jun 08, 2014 at 12:40:03PM +0200, Michael Niedermayer wrote:
> > On Sun, Jun 08, 2014 at 11:29:18AM +0200, Reimar Döffinger wrote:
> > > On Tue, Jun 03, 2014 at 11:05:39PM +0200, Michael Niedermayer wrote:
> > > > On Tue, Jun 03, 2014 at 08:43:03PM +0100, Derek Buitenhuis wrote:
> > > > > On 6/3/2014 8:20 PM, Aman Gupta wrote:
> > > > > > Signed-off-by: Aman Gupta <ffmpeg at tmm1.net>
> > > > > > ---
> > > > > >  libavcodec/Makefile       |  1 +
> > > > > >  libavcodec/allcodecs.c    |  1 +
> > > > > >  libavcodec/keyframe_bsf.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++
> > > > > >  3 files changed, 50 insertions(+)
> > > > > >  create mode 100644 libavcodec/keyframe_bsf.c
> > > > > 
> > > > > Two things:
> > > > > 
> > > > 
> > > > > 1) Won't this create some pretty broken files with some codecs?
> > > > 
> > > > which codec/case where you thinking of here ?
> > > 
> > > I don't know about broken files, but it should be
> > > quite useless for audio and questionable for subtitles at least.
> > > Which should be mentioned in the documentation.
> > > Not sure what will happen for H.264 with refresh interval (sorry,
> > > the proper name escaped my memory)?
> > 
> > > I guess it will create empty files?
> > 
> > either that, or it would create files with useless frames.
> > 
> > 
> > > In addition, for MPEG-2 etc, won't the timestamps possibly
> > > be quite messed up?
> > 
> > I would hope that wouldnt happen, if it does it should be investigated
> 
> Ok, I was just expecting that for the filtered frame some of the code
> to detect e.g. has_b_frames might not work and thus pts/dts at least
> won't quite match exactly.
> Even if it might not be significant, it's just a suspicion that it
> might show some non-obvious behaviour.

hmm, yes, thats a plausible problem
i think theres not too much sense in trying to fix it though before
we have a affected file / testcase

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Let us carefully observe those good qualities wherein our enemies excel us
and endeavor to excel them, by avoiding what is faulty, and imitating what
is excellent in them. -- Plutarch
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140608/97f7f35a/attachment.asc>