On Sun, Feb 19, 2012, at 05:03 PM, Jenkins wrote: > See <http://ci.cyrusimap.org/job/cyrus-imapd-master/402/> > > ------------------------------------------ > [...truncated 1246 lines...] > for file in ./imtest.1 ./pop3test.1 ./nntptest.1 ./lmtptest.1 > ./smtptest.1 ./sivtest.1 ./mupdatetest.1 ./installsieve.1 ./sieveshell.1; > \
There are at least four separate problems happening here, all at once :( 1) the Master.service_ipv6 test is failing because the older netstat program on ci.cyrusimap.org reports connect TCP/IPv6 sockets slightly differently than Cassandane was expecting. I fixed that here http://git.cyrusimap.org/cassandane/commit/?id=a76ec6226f0bad1bd81238a8b80360d58c894080 2) The Master.maxforkrate test was accidentally triggering an old and obscure race condition bug in the master process. I made the test not trigger that bug anymore http://git.cyrusimap.org/cassandane/commit/?id=10999db507143bdc668d49f59349f23ba61fc8e1 and added a new test just for that bug http://git.cyrusimap.org/cassandane/commit/?id=ad9bf14faddef02f070f5b82fb8f408e37826b1c and fixed the bug http://git.cyrusimap.org/cyrus-imapd/commit/?id=d288a9ad66636ab406e537b1fb57f05f016e1f38 3) The Master.maxforkrate test has detected that the maxforkrate parameter is not being enforced correctly, which is strange because that code did in fact work fine in that test when last modified. I suspect some kind of environmental problem is breaking the fork rate calculation, will investigate further. 4) Cassandane sometimes leaves master and lemming processes lying around. I haven't been able to reproduce that problem, although I have "solved" it several times before. Those leaked processes are never cleaned up and hog the TCP ports that Cassandane expects to be able to use, causing subsequent Cassandane runs to fail spuriously. I'm not entirely sure of the best way to address this, but I'm thinking of something like a sledgehammer which kills all processes running as the "cyrus" userid. So sadly it will be another day or so before the build gets back to stable. -- Greg.
