Re: lastval()

Neil Conway wrote:
> If you're busy, I can clean this up and apply it.

Attached is a revised patch. Per subsequent discussion, I stuck with
your approach of keeping a pointer to the sequence object, rather than
just the last int64 produced by nextval(). That means we emit an error on:

CREATE SEQUENCE seq;
SELECT nextval('seq');
DROP SEQUENCE seq;
SELECT lastval();

It also means that setval() _does_ affect lastval(), and that we do
permission checks properly. Barring any objections I'll apply this later
tonight or tomorrow.

BTW, I noticed that the "permission denied" messages throughout the
source don't quote the name of the identifier for which permission has
been denied. This violates the error code conventions: "Use quotes
always to delimit file names, user-supplied identifiers, and other
variables that might contain words." Is there a reason for this?

-Neil