Thread: pg_ctl bug

pg_ctl bug

From
oberpwd@anubis.network.com (Wade D. Oberpriller)
Date:
Hello,

I've noticed a bug if pg_ctl. If you do the following:

pg_ctl -w -D $PGSQL/data -o '-p 10000' start;

The pg_ctl script will loop doing a 'psql -l' to connect to the postmaster
process on port 5432 (but postmaster is using 10000 in this case). The
pg_ctl script should take the port number as its own option (-p port) and then
pass it to the postmaster (or it should parse the options sent to postmaster).
In any case it, the -w option should work whether or not postmaster is
running on 5432 or 10000.

Wade Oberpriller


Re: pg_ctl bug

From
Tatsuo Ishii
Date:
> Hello,
>
> I've noticed a bug if pg_ctl. If you do the following:
>
> pg_ctl -w -D $PGSQL/data -o '-p 10000' start;
>
> The pg_ctl script will loop doing a 'psql -l' to connect to the postmaster
> process on port 5432 (but postmaster is using 10000 in this case). The
> pg_ctl script should take the port number as its own option (-p port) and then
> pass it to the postmaster (or it should parse the options sent to postmaster).
> In any case it, the -w option should work whether or not postmaster is
> running on 5432 or 10000.

Moreover, using 7.1's postgresql.conf, we could start postmaster on an
unusual port number without an explicit option to pg_ctl. Maybe we
should throw away 'pg_ctl start' and add the same functionarity to
postmaster itself?
--
Tatsuo Ishii