On Tue, Dec 29, 2015 at 01:33:21AM +0100, Axel Beckert wrote: > Emilio Pozuelo Monfort wrote: > > Control: severity -1 important > > > > Downgrading, as the package has been built on every release architecture.
> And as far as I can see this issue is not in the test suite but in the > library itself as it seems to produce different results by pure chance > where they should be deterministic. I.e. the test suite found a real > bug from my point of view. > > If it wouldn't have wait for perl itself to migrate, I'd set the > severity to serious again to prevent its migration to testing (1.21 is > not in testing yet). The alternative is to remove it from testing. The changes between 1.20 and 1.21 are minimal, those can't be the reason for the failure. Probably either changes in perl or libc broke it. I expect 1.20 fails in the same way at least on sid, and probably in testing too. I've made some progress on it fwiw, and it seems to be something with two print statements in different modules getting the same opcode address on the perl side, therefore hashing down to a same hash key and getting counted as one. Will update the upstream bug when I've got.to the bottom of it. -- Niko Tyni [email protected]
