Hi, those tests sometimes fail because there is some randomness involved. They test the convergence of some algorithms, and sometimes the get stuck in local optima. The fix is to test the package during build with a fixed random seed, so that we only get "real" failing tests: if the fixed random seed no test is failing on the maintainer machine, no test should fail on the buildbots.
I already implemented the fix, and will try to find a sponsor to upload the fixed package in the next days... Thank you, Tiziano On Sat, Sep 22, 2012 at 8:20 PM, David Prévot <taf...@debian.org> wrote: > On Fri, Sep 21, 2012 at 07:00:04PM -0400, David Prévot wrote: >> Control: tags -1 unreproducible >> >> On Wed, Sep 12, 2012 at 03:16:42PM +0200, Lucas Nussbaum wrote: >> > Source: mdp >> > Version: 3.2+git78-g7db3c50-3 >> > Severity: serious >> > Tags: wheezy sid >> > User: debian...@lists.debian.org >> > Usertags: qa-ftbfs-20120912 qa-ftbfs >> > Justification: FTBFS in wheezy on amd64 >> >> Hi, >> >> I was able to reproduce the bug on Wheezy yesterday night, but not >> today. I couldn't find any relevant package that migrated in the mean >> time, so not closing this bug without confirmation or enlightenment. > > Attached build log and diff against the failed one. Please note that I > tried to rebuild it several time, and it failed twice so far, with > another failed test (FastICA, end of log attached) for the second > failure (didn't keep the log of the first one). Don't know how to > reproduce the issue “for sure”, but maybe the unreproducible tag should > be removed. > > Regards > > David > > -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
