Tim Peters schrieb:
[Andreas Krasa]
We are encountering some really strange problems with Zope 2.7.7 on our RedHat EL 4 Linux machines.
During the Zope 2.7.7 compilation works - however most of the time "make test" returns a random number of errors (somewhere between 20 and 30) ALL related to ZEO.
The funny thing is, we've managed to do a "make test" without any failures - however after doing a "make distclean" and compiling everything again "make test" produces the above mentioned errors (using *exactly* the same source code!).
I have absolutely no idea how this can happen - ANY hints are appreciated! Is this a known issue?
No. For example, it doesn't happen in the daily overnight testrunner reports.
What could it be related to?
ZEO <wink>? You'll have to give more info about which tests fail, and precisely how they fail. Because many of the ZEO tests create multiple processes, and try to assign sockets so that these processes can communicate, they're vulnerable to vagaries of OS process scheduling and socket use by other apps. For example, on a slow or overburdened (with other simultaneous work) machine, some ZEO tests can fail due to not getting enough cycles soon enough. The worst tests of that sort wait as long as a minute now for another process to "do something" they're waiting for before failing, but not even waiting a minute can _guarantee_ success.
Might be informative to run the tests on an otherwise-quiet machine.
Thank you Tim for the feedback! Our system is a Intel Xeon 3 GHz Dual-CPU with 2.5 GB RAM running RedHat Enterprise Linux 4 (SElinux disabled). As this is a test-machine it doesn't run any CPU-consuming tasks I can think of - the server load is usually somewhere between 0.00 and 0.10. But I'll check that nevertheless! Best regards Andreas