[Ffmpeg-devel] [BUG] seeking crash in x264 with b_pyramid

[Corey Hickey]

Erik Slagter wrote:
>>It usually takes about one
>>minute of running the command I provided before any problems occur; just
>>opening the file and seeking a few times is almost certainly not going
>>to be sufficient.
> 
> Does this also happen with material not encoded using b_pyramid?

No, and now that Loren has fixed the bug, it doesn't happen with
b_pyramid either.

> I also have issues, but not related to b_pyramid. For instance, mplayer
> (very latest CVS) plays raw h264 files double speed,

$ avi2raw pyramid.avi pyramid.264

$ mplayer pyramid.264

Yes, that does seem to be the case. On the other hand:

$ mencoder pyramid.avi -ss 10 -ovc copy -o pyramid-ss.avi

$ avi2raw pyramid-ss.avi pyramid-ss.264

$ mplayer pyramid-ss.264

That seems to be ok. I suspect that avi2raw is calculating fps from the
pts difference between the first two video frames. The second video
frame in pyramid.avi is empty, so avi2raw ignores it and ends up
calculating fps from the pts difference between the first and third
video frames.

Using 'mencoder -ss 10 -ovc copy' bypasses the empty frames at the
beginning.

I was going to recommend using the harddup filter for mencoder, but that
doesn't seem to work either. Indeed, avi2raw still says "warning: 1 zero
length frames ignored". That may be an mencoder bug; I don't know.

> and using some
> combinations of options to x264, I get awful artifacts now and then. I
> am trying to pinpoint the exact combination now.

Hmm; I'm curious what you find out.

-Corey