Joachim: > About your patch: I was hoping that upstream can bless it quickly, but > given that this blocks quite a few patches, and that there is a > presumably good test suite in the package, I think I can simply apply > it and see what happens. > > Will do that right away.
Most of the results are in: https://buildd.debian.org/status/package.php?p=haskell-cryptonite&suite=experimental It's failing in a consistent way on the big-endian architectures: powerpc, s390x, ppc64, and we'll probably see the same on mips in due course. If anyone wants to look at the problem, it's easy to find the relevant part of the code (look at the test names in the log), and there's not a lot of it. Presumably there's a place in there where little-endian is assumed. There are different failures on armhf and kfreebsd-amd64. The failures are somewhat similar, even though the architectures have little in common, and 0.6-1 was built successfully on kfreebsd-amd64. Also the error message says "Use ... to reproduce", suggesting that the test was random. So it looks like a genuine, architecture-independent bug that is only sometimes, randomly exposed by the test suite. To test that theory I'm going to ask for a give-back on those two architectures. Edmund
