Re: async notification patch for dblink

Joe Conway <mail@joeconway.com> writes:
> Sure, I guess I ought to use the latest-and-greatest. Any other comments
> before I commit?

That "be_pid/be_pid" hack in the regression test is pretty ugly, and
doesn't test anything very useful anyway seeing that it's integer
division.  Could you do something likebe_pid = pg_backend_pid() AS is_self_notify
instead, to verify that it's a self-notify?  (This is not quite right
because you'd need to execute pg_backend_pid() at the remote end, but
I'm not awake enough to remember the dblink syntax for that.  Maybe
it's too complex to be worth it, but I think demonstrating how to
check for self-notify would be a useful bit of doco.)

Otherwise it looks ok to me.
        regards, tom lane