This one's simple enough to reproduce (see SQL script below), but, there are some comments in src/backend/catalog/namespace.c that seem questionable and incorrect:
The proposed patch reverts a change deliberately applied in namespace.c
rev 1.15 (4/29/02). I think you need to go back and consult the schema
privilege discussions that occurred just before that; I'm much too tired
to do so myself right at the moment ...
I can see that it was done in rev 1.15, but I haven't seen any discussion that suggests that it was deliberate beyond what's in the comment... but that's lacking rationale, IMHO. The thread that I think you're referring to begins here:
http://archives.postgresql.org/pgsql-hackers/2002-04/msg01035.php
But here's pretty much the only relevant thread:
http://archives.postgresql.org/pgsql-hackers/2002-04/msg01191.php
But it doesn't have a conclusion, synopsis, or any agreement that comes close to, "here's why we check the perms for the session user and not the current user." Having the permissions for CREATE TEMP TABLE check on the session user defeats the purpose of having functions run as SECURITY DEFINER.
Without any rationale as to why CREATE TEMP TABLEs checks the session user in the archives, could we open this up for discussion again? To me, it seems to fly directly in the face of a function running as SECURITY DEFINER. At the moment, this behavior cripples the usefulness of having a TEMP table be used as a trusted cache for data.
-sc
-- Sean Chittenden
---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives?
http://archives.postgresql.org
