On Tue, Dec 27, 2011 at 06:16:33PM -0800, Ronald S. Bultje wrote: > Hi, > > On Tue, Dec 27, 2011 at 5:43 PM, Derek Buitenhuis > <[email protected]> wrote: > > From: Reimar Döffinger <[email protected]> > > > > Signed-off-by: Reimar Döffinger <[email protected]> > > --- > > tests/codec-regression.sh | 5 +++++ > > tests/ref/vsynth1/v210 | 4 ++++ > > tests/ref/vsynth2/v210 | 4 ++++ > > 3 files changed, 13 insertions(+), 0 deletions(-) > > create mode 100644 tests/ref/vsynth1/v210 > > create mode 100644 tests/ref/vsynth2/v210 > > > > diff --git a/tests/codec-regression.sh b/tests/codec-regression.sh > > index 57e5f07..097931d 100755 > > --- a/tests/codec-regression.sh > > +++ b/tests/codec-regression.sh > > @@ -278,6 +278,11 @@ do_video_encoding rgb.avi "-an -vcodec rawvideo > > -pix_fmt bgr24" > > do_video_decoding "" "-pix_fmt yuv420p" > > fi > > > > +if [ -n "$do_v210" ] ; then > > +do_video_encoding v210.avi "-an -vcodec v210" > > +do_video_decoding "" "-pix_fmt yuv420p" > > +fi > > Isn't v210 10-bit, so shouldn't we use 10-bit output (-pix_fmt > yuv420p10le, see h264 tests also)?
We have no H.264 encoder, thus we have no codec regression tests. Codec regression tests compare input to decoded output. As long as the input is always 8 bit, the decoded output needs to be 8 bit as well, regardless of the encoder format. As a long term solution, we should probably have a 16-bit input reference and use that for any > 8 bit formats, but that will be a larger change. _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
