On Tue, 19 Sep 2000, Ross J. Reedstrom wrote:
> I'd guess that Andreas must have looked at the template1 database with
> pgaccess.
Yes, this was the reason.
> That would create the pga_* tables in there, which would then
> propagate to all newly crreated dbs. Pgaccess makes these files behind
> the users back. It might be reasonable for it _not_ to create them in
> template1, if possible, without prompting the user, at least. I'll ask
> Constantin about it.
That's a nice idea!
This reminds me to a further issue on this topic:
By accident I filed a dump not to the intended database, but to
template1. This is not hard to do because I wrote a script like
#!/bin/sh
MYDB=<some_function>
cat dumpfile | psql $MYDB
unfortunately I hadn't checked whether $MYDB could be "" :-(.
So I filled my template1 database with a lot of rubish.
Nice exercise to remove this rubish which introduced me a little bit
deeper into PostgreSQL internal tables :). Hope that I got rid off
all this stuff.
So the idea is to make it a little bit harder to put something into
template1 or, alternatively serve a method which helps out such kind
of situation.
.... just an idea ...
> The workaround is to go into template1 with psql and drop the pga_*
> tables, then never use pgaccess to look in there.
Or just to call my script which removes all tables and sequences
which are not created by user postgres :).
Kind regards
Andreas.