On 2012-07-25 at 15:30:16 +0200, Felix Geyer wrote: > > I suspect that the issue is that the typical lack of entropy > > of virtual machines is the cause of the 60 minutes wait. > That test requires a lot of random data (5 MB).
Actually the build fails before said file is generated: the tests start by creating a number of gpg keys to test the gpg key creation code (and then actual use of said keys). The lock happens while gnupg is waiting for random data (which makes it hard to trick it into accepting not-so-random, nonlocking stuff). > It seems to be only used for testing if signing files work. > Do you have an idea why the data needs to be random > instead of just using a hardcoded pattern to generate > that file? > Relying on random data in unit tests is a bad idea anyway ... This data used to come in a file in the upstream tarball (generated via something like dd if=/dev/*random of=this_file); now it is getting generated at the beginning of doctests (after the unittest ones) to save space, saved and reused for future tests. I believe that if it was the problem it could be generated via some other method; the main cryterion is that it shouln't be trivial (so no small file, nor zero-filled file). -- Elena Grandi -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
