[FFmpeg-devel] [PATCH 1/8] libavcodec/jpeg2000_parser: Speed up long skips

[Tomas Härdin]

tor 2022-06-02 klockan 21:19 +0200 skrev Michael Niedermayer:
> On Thu, Jun 02, 2022 at 11:54:39AM +0200, Tomas Härdin wrote:
> > ons 2022-06-01 klockan 18:23 +0200 skrev Michael Niedermayer:
> > > On Wed, Jun 01, 2022 at 06:21:19PM +0200, Michael Niedermayer
> > > wrote:
> > > > On Tue, May 31, 2022 at 11:58:39AM +0200, Tomas Härdin wrote:
> > > > > 
> > > > 
> > > > >  jpeg2000_parser.c |   11 +++++++++++
> > > > >  1 file changed, 11 insertions(+)
> > > > > 634546fb5a0eb281eea87ad7471c503f5bc9e8ab  0001-libavcodec-
> > > > > jpeg2000_parser-Speed-up-long-skips.patch
> > > > > From fedd7f9ae2c691a25c37be935d7547be61d46017 Mon Sep 17
> > > > > 00:00:00
> > > > > 2001
> > > > > From: =?UTF-8?q?Tomas=20H=C3=A4rdin?= <git at haerdin.se>
> > > > > Date: Fri, 20 May 2022 11:38:25 +0200
> > > > > Subject: [PATCH 1/8] libavcodec/jpeg2000_parser: Speed up
> > > > > long
> > > > > skips
> > > > > 
> > > > > ---
> > > > >  libavcodec/jpeg2000_parser.c | 11 +++++++++++
> > > > >  1 file changed, 11 insertions(+)
> > > > 
> > > > breaks
> > > > j2kref/codestreams_profile1/p1_04.j2k
> > > 
> > > [jpeg2000 @ 0x7fb0b8002600] Psot 66195 too big
> > > [jpeg2000 @ 0x7fb0b8002600] error during processing marker
> > > segment
> > > ff90
> > > Input #0, j2k_pipe, from 'j2kref/codestreams_profile1/p1_04.j2k':
> > >   Duration: N/A, bitrate: N/A
> > >   Stream #0:0: Video: jpeg2000 (JPEG 2000 codestream restriction
> > > 1),
> > > gray16le(12 bpc), 1024x1024, 25 fps, 25 tbr, 25 tbn
> > > [jpeg2000 @ 0x7fb0b8003240] unsupported marker 0x97C8 at pos
> > > 0x6F0
> > > [jpeg2000 @ 0x7fb0b8003240] Missing EOC Marker.
> > > [jpeg2000 @ 0x7fb0b8032280] unsupported marker 0x97C8 at pos
> > > 0x6F0
> > > [jpeg2000 @ 0x7fb0b8032280] Missing EOC Marker.
> > > [jpeg2000 @ 0x7fb0b8006280] Psot 66195 too big
> > > [jpeg2000 @ 0x7fb0b8006280] error during processing marker
> > > segment
> > > ff90
> > 
> > Took a while to figure out, but this is due to buf_size - i - 9
> > being
> > changed to unsigned because of the uint32_t. Try attached patch.
> > The
> > rest of the set should work with it.
> 
> works fine here

Great

> > Can we roll these tests into FATE?
> 
> feel free to do so. 
> ill test if someone posts a patch

How are you testing them? Just decoding and checking that the output is
the same as some reference output? Just looking for non-zero exit code?

/Tomas