Bryn Llewellyn <[email protected]> writes:
> Thanks, all, for your replies. I'd assumed that the arguments of "set
> search_path" had to be SQL names. so I tried "". But that caused an error. I
> didn't try the ordinary empty string because I'd assumed that, as an illegal
> SQL name, it would be illegal in "set search_path". Hmm...
search_path's value is not a SQL name. It's a list of SQL names
wrapped in a string ... and the list can be empty.
A bit off topic: I'm not sure how you came to the conclusion that
superusers can't write into pg_catalog. They can. But I don't see
much point in being paranoid about whether the contents of pg_catalog
are trustworthy. If an adversary has already obtained superuser
privileges, he needn't bother with anything as complicated as
trojan-horsing something you might call later.
regards, tom lane
