[FFmpeg-devel] donation for snow

[Jason Garrett-Glaser]

> And like always the problem is the same, few are cleaning code up,
> fixing the bug that annoys one or the latest feature like the new lossless
> mode in h264 is just more sexy than (the actually easier) spliting and
> documenting of code.

On that topic, if its so sexy, any chance it can get committed, even
if it requires a duplicate codepath to avoid speed loss? ;)

(By the way, on that topic, I talked with the CoreCodec guys, and they
implemented it the same way my original patch does, which indeed
works: that is, putting the residuals into 8-bit pixel values and then
doing overflow-sums onto them.  This works as long as you assume the
input isn't supposed to overflow and clip, which the spec says should
be the case for compliant streams.)

Dark Shikari