On Sat, Jun 20, 2015 at 04:25:55PM +0200, Michael Niedermayer wrote: > On Sat, Jun 20, 2015 at 01:19:29PM +0200, Sebastien Zwickert wrote: > > The pixel buffer base address is never unlocked this causes > > a bug with some pixel format types that are produced natively > > by the hardware decoder: the first buffer was always used. > > Unlock the pixel buffer base address fixes the issue. > > > > --- > > ffmpeg_vda.c | 2 ++ > > 1 file changed, 2 insertions(+) > > LGTM > > thanks >
FYI this is supposed to fix normal cases like: wget http://lucy.pkh.me/samples/natsuno.mp4 ffmpeg -hwaccel vda -i natsuno.mp4 -t 15 out.avi For some reason it's working if requesting kCVPixelFormatType_420YpCbCr8Planar (yuv420p) but not kCVPixelFormatType_422YpCbCr8: (uyvy422, default with ff) or kCVPixelFormatType_32BGRA (rgb32) I can't confirm it fixes the issue because I haven't the hw to test right now, but I believe it does :) Thanks for the patch. On a side note, it would be nice to allow a CLI option to request a special pixel format to the decoder (don't we have that stuff already?). -- Clément B.
pgpCR_WGbE2Ba.pgp
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
