Christoph Haller <ch@rodos.fzk.de> writes:
> Furthermore, the regression test fails badly on
> parallel group (20 tests): point lseg
> running forever. The process list shows
> ps -fu rodos
> UID PID PPID C STIME TTY TIME COMMAND
> rodos 14786 6955 228 10:26:43 pts/0 3:08 /bin/sh ./pg_regress
> --temp-install --top-builddir=../../..
> rodos 6796 18757 0 10:25:18 pts/0 0:00 gmake check
> rodos 14892 14786 0 10:26:46 pts/0 0:00 <defunct>
In other words, the shell is failing to notice that some of its child
processes have exited. This is a shell bug; we've seen it before on
some platforms (obsolete HPUX versions, in my experience). You can
probably work around it by reducing the maximum concurrency setting
in the regression script. See the regression test discussion in the
administrator's guide.
regards, tom lane