I can indeed confirm that this is a compiler optimization bug, recompiling
mpeglib with -O0 fixes the problem. Still investigating exactly which
optimization is causing the bug so a report can be filed against g++.
Just in case anyone's interested, the quickly-hacked -O0 deb I'm using is
here:
I captured the output of the same 30 seconds of audio from Noatun/arts/mpeglib
and also from mpg321/libmad, then encoded them with FLAC. Available here:
http://www.marm.org.uk/mpeglib-decode-bug-mad-output.flac
http://www.marm.org.uk/mpeglib-decode-bug-mpeglib-output.flac
--
marm
Package: mpeglib
Version: 4:3.2.2-1
Severity: important
Tags: sid
Since upgrading to mpeglib 3.2.2-1, decoding of mp3s using mpeglib sounds
dreadful, full of strange MPEG artifacts. Version 3.2.1-1 works perfectly,
as do players built around other MP3 decoders (e.g. XMMS, mpg123, mpg321).
Given
3 matches
Mail list logo