Tom Lane writes:
> That's a tad harsh, considering that the bug under discussion is
> psql's not pg_ctl's ;-)
Well, mumble. Why does pg_ctl start the postmaster in the background
without redirecting the input and output streams? Why does it say that
the postmaster started successfully before it actually started? Why does
the documentation say that -w waits for 60 seconds, when it's really only
a count-to-60 loop? Is starting the postmaster and checking for the
existence of the pid file in the next command really reliable? Why
doesn't /etc/init.d/postgresql use it? As long as those questions remain,
the only reliable way, in my mind, to start the postmaster is to start the
postmaster.
I would *like* to see pg_ctl become useful in terms of controlling log
file names or piping the postmaster output to a rotatelog script, but as
it stands it's just another layer of possible failure.
--
Peter Eisentraut peter_e@gmx.net http://yi.org/peter-e/