elliefm
commented
7 years ago 1) test_replication_mailbox_new_enough(Cassandane::Cyrus::Replication)
Perl exception: child process (binary tar pid 19302) exited with code 2
at Cassandane/Instance.pm line 1427.
Cassandane::Instance::__ANON__(HASH(0x637ae88), 2) called at Cassandane/Instance.pm line 1609I wonder if this one is due to the cassandane directory running out of disk space? The exit 2 from tar is supposedly for unrecoverable fatal errors. Cassandane runs churn up quite a bit of disk space (especially if not cleaned up between runs). You might try setting cleanup = yes in the cassandane.ini, or running testrunner.pl with the --cleanup argument, and see if this goes away?
Haven't looked into the test_references_chain failure yet, I'm not even sure what this suite tests so I have no gut feeling about what it might be
jasontibbitts
I've found that at least test_references_chain(Cassandane::Cyrus::Thread) and test_replication_mailbox_new_enough(Cassandane::Cyrus::Replication) will sometimes pass and sometimes fail. They don't pass or fail together. As far as I can tell there are no dependency changes between the various builds.
I am calling testrunner.pl with various values for the '-j' option depending on the number of available CPUs, so I suspect there's a race somewhere, but I haven't been able to reproduce any of this on my personal build machine (12 threads, so '-j 12'). I've seen it on a builder with 6 threads and another with 48. I have a machine here I can use that has 56 threads that I can use for testing but I need to do some setup on it first.
Here is the full output from those tests. These come from different runs.