[FFmpeg-devel] [PATCH] dpxenc: also test 10 and 16 bit formats.

Reimar Döffinger Reimar.Doeffinger at gmx.de
Sat Apr 7 02:30:55 CEST 2012


On Sat, Apr 07, 2012 at 02:25:26AM +0200, Reimar Döffinger wrote:
> On Sat, Apr 07, 2012 at 01:58:23AM +0200, Michael Niedermayer wrote:
> > On Fri, Apr 06, 2012 at 05:57:05PM +0200, Reimar Döffinger wrote:
> > > Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
> > > ---
> > >  tests/lavf-regression.sh |    2 ++
> > >  tests/ref/lavf/dpx       |    6 ++++++
> > >  2 files changed, 8 insertions(+)
> > > 
> > > diff --git a/tests/lavf-regression.sh b/tests/lavf-regression.sh
> > > index f0fcac3..6174370 100755
> > > --- a/tests/lavf-regression.sh
> > > +++ b/tests/lavf-regression.sh
> > > @@ -221,6 +221,8 @@ fi
> > >  
> > >  if [ -n "$do_dpx" ] ; then
> > >  do_image_formats dpx
> > > +do_image_formats dpx "-pix_fmt rgb48"
> > > +do_image_formats dpx "-pix_fmt rgb48 -bits_per_raw_sample 10"
> > >  fi
> > >  
> > >  if [ -n "$do_xwd" ] ; then
> > > diff --git a/tests/ref/lavf/dpx b/tests/ref/lavf/dpx
> > > index a852ae0..2160809 100644
> > > --- a/tests/ref/lavf/dpx
> > > +++ b/tests/ref/lavf/dpx
> > > @@ -1,3 +1,9 @@
> > >  808ea110635774252439722a48329d61 *./tests/data/images/dpx/02.dpx
> > >  ./tests/data/images/dpx/%02d.dpx CRC=0x6da01946
> > >  305792 ./tests/data/images/dpx/02.dpx
> > > +5e1a777fa3f4094c9c4dd989cf9e8e8b *./tests/data/images/dpx/02.dpx
> > > +./tests/data/images/dpx/%02d.dpx CRC=0xe5b9c023
> > > +609920 ./tests/data/images/dpx/02.dpx
> > > +13dc41b1e1e36399a5e1f8b7e3344a81 *./tests/data/images/dpx/02.dpx
> > > +./tests/data/images/dpx/%02d.dpx CRC=0xf38d5830
> > > +407168 ./tests/data/images/dpx/02.dpx
> > 
> > fails on ppc, id guess we missed LE vs BE in the pixel format
> 
> Hm, I was quite convinced that DPX in independent from LE/BE,
> in which case the failure would rather indicate a encoder bug...

Hm, it actually swaps everything including the header depending
on whether the pixfmt is le or be.
I'll give it a try by forcing le instead, but not tonight
(or you or someone else can just change rgb48 to rgb48le).


More information about the ffmpeg-devel mailing list