From: Oliver Elphick [mailto:olly@lfix.co.uk]
>
> I can't see any bug in the initdb script. It looks straightforward.
>
> Did the user perhaps delete and recreate the postgres user at some stage?
> Then the hard-coded sysid in the installation would orphan user-created
> functions when the database was upgraded.
No, that never happened. Also, there are no user-created functions
(or types) in the database to begin with, just tables, indexes, and
sequences.
This is a brand new installation; I had 6.5 (potato's version)
installed previously, but I did a pg_dumpall and purged the old
install before installing 7.1.1-3 and importing the dumped data
using psql.
If you'd like to see any other queries on my template1 database,
to see if you can figure out where the line is drawn between user
ids 103 and 31, I'd be happy to oblige.
One thing to note, I did have to manually execute the adduser call
from the install script, before I could succesfully install 7.1.1-3,
due to the shell=/bin/false bug I reported last week. Could that be
part of the problem?
Should we move this thread to BTS?
- Marc