>
> compiled under VC++
>
> common.h: changed unistd.h to stdlib.h in order to work under VC++
> sqrttab.h: changed float to double in order to supress 1000 warnings
> tables.c: idem
>
> musicin.c, l 145:
> info->copyright = 0; info->original = 1; info->error_protection = FALSE;
> l 237:
> case 'o': info->original = 0; break;
>
>
> Done this because of SCMS wich will be used in future hardware mp3 players
> like rio or nomad
>
> Gabriel Bouvigne - France
> [EMAIL PROTECTED]
> icq: 12138873
>
> MP3' Tech: www.mp3tech.org
>
Thanks, I added all those fixes to what will be lame3.03. I also
added some patches from Lars Ingebrigtsen to allow reading and
writing to stdin/stdout, and an option to turn off the status display.
Last night I found a new (but very rare) bug:
if the psy model encodes a sequence that looks like:
short_block end_block short_block
Since you need a start_block in front of every short_block,
and an end_block after every short_block, this is not possible.
So the encoder changes the end_block to a short_block as it should.
The problem is that it is now too late to go back and recompute the
noise threshold information (which is different depending on the block
type), so it just uses the thresholds from the previous short block.
The easy fix would be to always compute both short_block and long_block
thresholds, but I dont know how expensive this will be.
Mark
