[MPlayer-cvslog] r31560 - trunk/libmpcodecs/vd_ffmpeg.c

[Reinhard Tartler]

On Mon, Jun 28, 2010 at 17:25:33 (CEST), Reimar Döffinger wrote:

> On Mon, Jun 28, 2010 at 02:41:11PM +0200, Reinhard Tartler wrote:
>> On Mon, Jun 28, 2010 at 07:53:06 (CEST), Reimar Döffinger wrote:
>> 
>> > On Sun, Jun 27, 2010 at 04:36:49PM -0400, compn wrote:
>> >> On Sat, 26 Jun 2010 20:06:21 +0200 (CEST), reimar wrote:
>> >> >Author: reimar
>> >> >Date: Sat Jun 26 20:06:20 2010
>> >> >New Revision: 31560
>> >> >
>> >> >Log:
>> >> >Allow the FFmpeg VP8 decoder to work by disabling dr for it (thus we do not
>> >> >need edge emulation).
>> >> 
>> >> codecs.conf can be updated? ffvp8 is still marked 'buggy'.
>> >
>> > Might be changed to untested, but on x86 it is still quite a bit slower...
>> > Also, DR still is not possible, might make sense to keep a comment to that effect.
>> >
>> >> btw would this be needed as well?
>> >> 
>> >> Index: mp_taglists.c
>> >> ===================================================================
>> >> --- mp_taglists.c       (revision 31574)
>> >> +++ mp_taglists.c       (working copy)
>> >> @@ -112,6 +112,7 @@
>> >>      { CODEC_ID_TIERTEXSEQVIDEO,   MKTAG('T', 'S', 'E', 'Q')},
>> >>      { CODEC_ID_TXD,               MKTAG('T', 'X', 'D', 'V')},
>> >>      { CODEC_ID_VP6A,              MKTAG('V', 'P', '6', 'A')},
>> >> +    { CODEC_ID_VP8,               MKTAG('V', 'P', '8', '0')},
>> >
>> > So far not needed, but shouldn't hurt...
>> 
>> Without this patch, I have the following test matrix:
>> 
>> 
>> | avocodec compiled       | ./mplayer -demuxer lavf big_buck_bunny_480p.webm |
>> |-------------------------+--------------------------------------------------|
>> | static                  | OK                                               |
>> | shared w/ current trunk | OK                                               |
>> | shared w/ ffmpeg 0.6    | FAIL                                             |
>> 
>> 
>> This patch seems to be neccessary for the last case. I guess it's
>> because 0.6 doesn't contain the native vp8 decoder yet, but supports
>> vp8 only via libvpx.
>
> In this case I retract my ok.
> Someone forgot to backport the riff.c part of VP8 support, leaving
> VP8 in webm as the only combination working with FFmpeg, this is
> just broken.

indeed, I've just backported r23193 to ffmpeg 0.6, will be part of the
0.6.1 release. with that, vp8 playback does work.

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4