"JJAR" <jjarcore@eresmas.com> writes:
> [ parallel regression tests hang up in second parallel group ]
> I come to this point and nothing happens. The CPU doesn't work at all ( only
> 5% CPU power consumption ) and the hard disk is quiet so the process is
> dead.
Hmm. You might be looking at a shell bug. There are shells out there
that lose track of how many children they have when asked to spawn more
than a dozen or so parallel children. For example, I see extremely
similar behavior when I try to run 'make check' using the standard
/bin/sh shell on HPUX 10.20: it locks up partway through the second set
of parallel tests. I have to select SHELL=/bin/ksh instead to get
the tests to complete. I'm not sure it's quite the same issue, because
HPUX's sh goes to 100% CPU usage after getting confused in this way,
but it still seems like a likely explanation.
Can you run the non-parallel test (make installcheck) successfully?
Also, try splitting the parallel groups in the parallel_schedule file
into smaller numbers of tests.
regards, tom lane
