[FFmpeg-cvslog] r20202 - trunk/libavcodec/imgconvert.c

[Reimar Döffinger]

On Sun, Oct 11, 2009 at 09:08:14AM +0200, Reimar D?ffinger wrote:
> On Sun, Oct 11, 2009 at 09:05:26AM +0200, reimar wrote:
> > Author: reimar
> > Date: Sun Oct 11 09:05:26 2009
> > New Revision: 20202
> > 
> > Log:
> > 100l in avpicture_layout: width of chroma planes depends on format depth, too.
> > Fixes issue 1465
> 
> Hm, seems like there aren't any regression tests for the 16 bit raw
> formats at least, someone volunteering to add some?

Well, at least FATE seems to test it.
Please update the reference data for the v210 test, I verified with
Gimp's raw mode (that unfortunately does not support 16 bit per
component, but it is still good enough to see this) that the previous
output was wrong, the last 1/4 of the raw data was uninitialized/black,
and the right half of the U/V planes was missing, it is correct as it is
now (as far as it is possible to verify, considering gimp supports
neither raw planar YUV nor 16-bit-per-component raw - I guess there is
some ffmpeg commandline that could be used to verify more thoroughly,
but it's good enough like this for me).