[MPlayer-dev-eng] [PATCH] Real demuer audio cleanup

[Roberto Togni]

On Tue, 13 Dec 2005 18:57:37 +0800
陆然 <hephooey at fastmail.fm> wrote:

[...]
> I've found a strange behavior of the new code, since I've just tested this 
> with one file, I don't know it's an exception, or a general problem. the file 
> have a ra10 cook audio, playing with binary codec. 
> 
> First thing I notice is that the vibration of 'A-V' is much bigger than the 
> old code, +/- 1s compare to +/- 0.002s. 
> 
> The bigger problem is the movie will eventually go out of sync, the output is 
> perfect normal, 'A-V' is around 0 with some oscillation. But actually the 
> audio is slower than video, I notice this after about 10 minute after the 
> movie started, the audio at that time is about 1s after the video. The last 
> number of the output of the bitrate is different, don't know if that is 
> related.
Can you try to seek and check if the desync disappears? If yes, does it
starts drifting again after that?

Can you also try to force the old bitrate (12905) in i_bps and check if
the sync is ok in that way? In ad_realaud.c on line 364 replace
"sh->i_bps=sh->wf->nAvgBytesPerSec;" with "sh->i_bps=12059;" If it
fixes it, it's probably a rounding problem when converting from bits to
bytes (it truncates now, it should probably round).

> 
> One more thing, with the new code I get a crash when mplayer try to decode the 
> audio with ffmpeg codec, the old code just gave a error message and go on the 
> the binary codec.
> 
The older code did not have the cook decoder enabled (unless you added
it yourself); anyway i can confirm the crash with the sample you
provided, both with MPlayer and ffplay.

Ciao,
 Roberto