Kurt Roeckx wrote: > On Wed, Mar 04, 2009 at 11:42:37AM +1100, Erik de Castro Lopo wrote: > > Samuel Mimram wrote: > > > > > The thing is that the build failed on every arch (except i386), see : > > > > > > https://buildd.debian.org/build.php?pkg=libsndfile > > > > I've just confirmed this on Debian Testing on Amd64. > > Note that there are 3 errors in the buildd logs: > - alpha shows: > test_wrapper.sh: line 33: 31486 Floating point exception./floating_point_test > - ia64, mips, mipsel, hppa, armel show one of: > Line 231: Actual SNR ( 58.7) > target SNR (-31.0). > Line 231: Actual SNR ( 57.4) > target SNR (-31.0). > Line 279: Actual SNR ( 59.3) > target SNR (-29.0). > Line 231: Actual SNR ( 0.0) > target SNR (-31.0).
Those are a separate issue that I missed before. I'll have a look. > - amd64, s390 and powerpc show: > Error : max_abs 1.020731 should be < 1.02. This is the one my patch fixed. > I have no idea what this test is testing exactly. As the upstream author of libsndfile I can tell you that for what that is testing, the difference between 1.02 and 1.021 is irrelevant. Its a product of floating point rounding inside libvorbis. I did all my development and testing with libvorbis from SVN which due to some improvements, gave results < 1.02. The libvorbis in debian is older and produces slightly worse results for that test. > But > either the test is wrong, or the result is wrong. The test is right for libvorbis-1.2.1 (still unreleased) and later but wrong for libvorbis < 1.2.1. Erik -- ----------------------------------------------------------------- Erik de Castro Lopo ----------------------------------------------------------------- "life is too long to be an expert at harmful things, including such evilness as C++ and perl." -- Erik Naggum -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
