[FFmpeg-devel] [PATCH] avformat/img2dec: skip DHT segment in jpeg_probe()

[Matthieu Bouron]

On Wed, Apr 01, 2020 at 04:29:16PM +0200, Carl Eugen Hoyos wrote:
> Am Mi., 1. Apr. 2020 um 16:24 Uhr schrieb Matthieu Bouron
> <matthieu.bouron at gmail.com>:
> >
> > On Wed, Apr 01, 2020 at 04:09:08PM +0200, Carl Eugen Hoyos wrote:
> > > Am Mi., 1. Apr. 2020 um 15:48 Uhr schrieb Matthieu Bouron
> > > <matthieu.bouron at gmail.com>:
> > > >
> > > > On Wed, Apr 01, 2020 at 03:38:45PM +0200, Carl Eugen Hoyos wrote:
> > > > > Am Mi., 1. Apr. 2020 um 10:57 Uhr schrieb Matthieu Bouron
> > > > > <matthieu.bouron at gmail.com>:
> > > > > >
> > > > > > ---
> > > > > >  libavformat/img2dec.c | 1 +
> > > > > >  1 file changed, 1 insertion(+)
> > > > > >
> > > > > > diff --git a/libavformat/img2dec.c b/libavformat/img2dec.c
> > > > > > index 40f3e3d499e..93cd51c1932 100644
> > > > > > --- a/libavformat/img2dec.c
> > > > > > +++ b/libavformat/img2dec.c
> > > > > > @@ -775,6 +775,7 @@ static int jpeg_probe(const AVProbeData *p)
> > > > > >                  return 0;
> > > > > >              state = EOI;
> > > > > >              break;
> > > > > > +        case DHT:
> > > > > >          case DQT:
> > > > > >          case APP0:
> > > > > >          case APP1:
> > > > >
> > > > > Please provide a sample file.
> > > >
> > > > This patch does not fix a particular issue. IMHO, we should skip the DHT
> > > > segment just like the other segments (and avoids parsing the data it
> > > > contains) unless there is a particular reason for it ?
> > >
> > > I am (somewhat strongly) against patches that do not fix issues but
> > > Michael will hopefully comment.
> > >
> > > Or are you trying to improve probing speed?
> >
> > I am trying to fix an issues with JPEG files containing MPF metadata
> > appended at the end.
> > Theses files makes the jpeg_probe function fails (returns 0) because the
> > MPF metadata contains JPEG images (so jpeg_probe() finds a SOI marker
> > after EOI). I will send a separate patch for that and provide a
> > sample.
> 
> Are these valid jpeg files?

They are valid JPEGs with MPF metadata
(https://exiftool.org/TagNames/MPF.html).

See
https://ffmpeg.org/pipermail/ffmpeg-devel/2020-April/259489.html


> 
> > This patch should improve the probing speed by little though.
> 
> Then please measure the speedup and mention it in the
> commit message.

I tried to measure the speed gain using perf record ffmpeg -i [...]
without success (i can't get a meaningful difference between the different
runs which makes sense as the DHT segment size from the file I test against is
419 bytes).

> 
> Thank you, Carl Eugen
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".

-- 
Matthieu B.