The configure script for libmad chooses fixed-precision math type of "default" on my amd64 machine. According to the release notes, this is sub-optimal. The ebuild should probably set the fpm model to 'intel' on amd64 machines. If someone knows better, that it should maybe be 64bit, please go with that.
Created attachment 92592 [details] libmad ebuild that sets fpm to intel for amd64 machines ebuild that sets the fpm model. It builds fine on my ~amd64 box, and the library is able to decode mp3's.
Created attachment 94716 [details] libmad ebuild that sets fpm to intel for amd64 and x86 machines ebuild should use --enable-fpm=intel on x86
Arch teams please advise (possibly looking at the code if it works).
Okay, I've added the various fpmath choices for x86, amd64 (using 64-bit that performs better than intel) and ppc in r2. If trouble is found, just tell me.