Originally Posted by eamiller
In regards to the glitches, I should clarify a little. With Xv, the mpeg glitches (if they show) are limited to a very small region of the image and the frequency is very infrequent (though the small area of effect probably hides the glitches better).
With VDPAU, the glitches seem to affect a large portion of the frame, sometimes the whole frame. Not only does this cause the affected part of the frame to become garbled, but it will also shift sideways at times.
I've attached 2 images showing approximately the same part of a recorded HD file. In this particular case, the VDPAU artifacts lasted several seconds in approximately the same place (as is plainly visible), in this case Xv showed no visible artifacts. There was a scene right at the start of the video that both VDPAU and Xv showed glitches, and the Xv was far less noticible (I wasn't able to get a freeze frame on it as it happened very quickly).
Note that what you describe has not much to do with xv vs VDPAU, but more with software vs hardware decoding.
Both FFmpeg's MPEG-2 video decoder and libmpeg2 contain rather sophisticated error-concealment algorithms that seem to be difficult to find in hardware decoders (lets face it: We are all happy that VDPAU doesn't freeze the whole PC when decoding a buggy stream).
That means if you expect corrupted media, you should use software decoding: You can still use VDPAU's rendering capabilities (including de-interlacing) and benefit from advanced error-concealment.
(If, otoh, you would be able to watch hardware-decoded video on xv - this possibility is intended by VDPAU's API, there is just no patch at least for MPlayer available - you would see the same/very similar corruption as you see now with hardware decoding and VDPAU rendering.)