[MPlayer-dev-eng] WAV File Demuxing

[Alex Beregszaszi]

> > Yes, it fixes it, thank you ;)
> > Also why does it handles data with junk so bad ? (adpcm decoder)
> 
> 	You can draw comparisons between the ADPCM decoder and many other
> things, such as life: What you get out of it depends on what you put into
> it...:) The ADPCM decoder(s), like all of the audio and video decoders,
> expect data to arrive in a certain format. The first 7-14 bytes (depending
> on stereo mode) of MS ADPCM data is supposed to have a bunch of numbers
> used in decoding the next 1000-2000 PCM samples. If you were to feed MP3
> data to an MP3 decoder starting with the wrong byte in the stream, I'm
> pretty sure the resultant audio wouldn't sound very pleasing either. The
> raw PCM data from WAV files wasn't playing correctly either, but because
> each 8-16 bits decodes to an individual sample independently, the problem
> wasn't as audible.

Also by MP3 it isn't a problem, you must only search for a valid mp3 header.
Are there no ways by ADPCM to check for a 'header'? 

> 
> 	A big issue with predictive algorithms like this is that they're
> not very fault-tolerant. If there's an error in the stream, everything
> after the error comes out wrong. This may be why IMA ADPCM has such small
> block sizes. Of course, if the preamble is bad to begin with (as was the
> case here), none of the sound will be right.
> 
> 	However, I have added some new sanity checks to the ADPCM decoders
> and will commit those sometime, so hopefully it becomes more visible when
> something goes wrong and the audio sounds bad.
> 
> > It always gets locked when it finished to play an WAV.
> 
> 	Does anyone know why yet? Is there a movement afoot to fix the
> issue?

Dunno.

Also i've got a second problem.
I've fixed (demuxer gets filetype from networking) MP3 streaming (HTTP/ICY),
it plays fine, but it stops at 1.9s (i think the buffer gets full (20kb buffer,
it said))

-
alex