[MPlayer-dev-eng] branching off rc4

[Reimar Döffinger]

On Sat, Jun 26, 2010 at 06:19:07PM +0200, Reinhard Tartler wrote:
> On Fri, Jun 25, 2010 at 22:10:47 (CEST), Carl Eugen Hoyos wrote:
> 
> > Reinhard Tartler <siretart <at> tauware.de> writes:
> >
> >> VP8 seems to be currently broken in trunk
> >
> > Could you elaborate (or fix)?
> 
> I fear not. Here is the relevant excerpt from irc:
> 
> 20:03 <siretart> KotH: however, a sample VP8/WebM file fails to playback with mplayer/trunk, but works fine with the system ffplay
> 20:04 <KotH> hmm.. right vp8 still doesnt fully work
> 20:04 [KotH has not been following the discussion around vp8]
> 20:04 <KotH> anyone know whats missing there?
> 20:04 <dalias> seems ffmpeg does not support edge emulation for vp8
> 20:04 <dalias> and mplayer's lavc wrapper requires edge emulation
> 20:05 <KotH> dalias: edge emulation?
> 20:05 <dalias> mpeg style codecs make use of region outside the edges of the frame for motion vectors
> 20:06 <dalias> this phantom region is usually just a copy of the adjacent in-frame pixel
> 20:06 <dalias> edge emulation is a trick whereby, instead of actually creating this outside-frame crap for the mc functions..
> 20:07 <dalias> ..you have special mc functions which duplicate the effect
> 20:07 <KotH> and why isnt that internal to lavc?
> 20:07 <dalias> because without edge emulation, stride must be (significantly, at least by 1 block or so) > width
> 20:08 <dalias> and mplayer seems not to support the codec choosing its stride
> 20:08 <dalias> not for lavc at least
> 20:08 <KotH> so it's basically an mplayer bug not working nicely together with lavc?
> 20:08 <dalias> yeah
> 20:09 <KotH> could you send that as mail reply to the "branching off rc4" on -dev-eng?

As said: the stride is the smaller problem (we already have code to handle that issue btw.),
the problem is the x and y offset of the image data.