Howdy Patrick,
> Would you recommend me to use the dist10 decoder as a basis for a
> further implementations ?
Pretty much everyone does, so come on in, the water's fine... (You don't
really have much choice - there are bugs in the spec which can only be fixed
(readily) by looking at the dist10 source. Even if you were to code from
scratch, you'd still need both the spec and dist10 (or another working
decoder) for reference.)
> I heard about some bugs in relationship with
> that decoder. How serious are they ?
Depending on your platform, you should be able to get a working executable
within a couple of days. Of course, really cleaning it up, optimizing it,
etc. is likely to take more like two months.
Beyond that, most of the major bugs are well documented in the version logs
for LAME. I say 'major', but the only really major one that comes to mind
is the flipping of the count1 quads in the huffman block of the dist10
_encoder_ - minor bugs abound, but rarely have a big impact on the audio
quality. Assuming you aren't foolishly trying to work with mixed blocks or
something weird like that, the code is really (functionally) pretty clean,
if massively inefficient and often obfuscated.
Hope that helps,
Alex
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
