[MPlayer-cvslog] r19292 - trunk/libmpcodecs/ad_hwmpa.c

[Nico Sabbi]

Uoti Urpala wrote:

>On Tue, 2006-08-01 at 22:54 +0200, Nico Sabbi wrote:
>  
>
>>confirmed: advised by me and working for him with -ao ivtv  :)
>>Additionally some time ago Uoti  explained me that letting decode_audio()
>>decode only one frame (as do all other ad_*) per call is perfectly 
>>acceptable,
>>because the function will be simlpy called more often
>>    
>>
>
>Yes the new behavior of the decoder is acceptable in itself. However the
>aos that always return 0 from get_delay() are severely broken, and
>relying on other parts to work around that is not ok.
>  
>

agree; when I understand how to fix ao_mpegpes correctly I'll do it

>After looking at the code it seems this change shouldn't even affect the
>data going to the aos, so I think there's also some other bug affecting
>the sync. Maybe demuxing/decoding audio later rather than earlier
>affects something it shouldn't, or maybe the values used to calculate
>duration of buffered audio are set incorrectly.
>  
>

tha major flaw in the ad layer is, IMO, the inability to tell between 
real decoders
and pass-through ones. dec_audio.c  really expects to be fed with 
decoded audio
(look at ad_sample.c and ad_hwac3.c)