Re: Bug #899: pg_dumpall produce erroneous files if the databbase

=?iso-8859-1?q?Laurent=20FAILLIE?= <l_faillie(at)yahoo(dot)com> writes:
> ... yes but it is not the probleme : as no user is
> attached w/ procedural languages (it's clearly
> commented in the source code of pg_dump), it should
> generate ACL of then in a scope of a normal user, and
> not as super user. Again, I solve this issues in my
> patches.

I don't like the parts of this that are concerned with rearranging dump
order; it's just adding complexity that won't do much except confuse
people working on the code. The one and only true solution for this
class of problems is to make pg_dump pay attention to pg_depend
dependencies. Now that the raw data is available, I think we're past
the point where we should keep adding heuristic kluges that try to
substitute for dependency analysis. I fear they're as likely to break
corner cases as fix 'em :-(

As for the ACL changes, I don't see the point of 'em any more than Bruce
does. Explain again please?

regards, tom lane