On 05/09/13 18:51, Julien Cristau wrote: > I couldn't reproduce this on falla (ran make check a few times). Can a > kfreebsd porter take a look?
It built successfully for me locally in a wheezy chroot and twice in a sid chroot. It seems test-poll tries to listen on 127.0.0.1:12345 for TCP connections. So that can fail for a couple of reasons: 1. something else listening on that port already 2. connection in the past few minutes is in TIME_WAIT state - for some reason it still fails even though SO_REUSEADDR is requested here (and I think this is important - needs followup). There seems to be a race within the test program that can trigger this problem itself sometimes. Both of those produce exit status 77 and the test is 'skipped', leading to a successful build. If you are 'lucky', the test program gets a bit further, but (only sometimes) hits a different problem in test_socket_pair() "expecting POLLHUP after shutdown" returning status 2, which is treated as an actual failure. The test sometimes runs fully and passes - especially when running under ktrace. So I think there is a race for the connection over the loopback interface to be fully closed, before poll1_wait() is called. BTW it would be nice if failures in the testsuite would output the contents of libmissing/tests/test-suite.log into the build log. Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-bsd-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/522938ef.1010...@pyro.eu.org
