Joe Conway <mail@joeconway.com> writes:
> I see what you did, and understand why, but it made me wonder if there
> is a more standard way of doing a test procedure for contrib. Should I
> be doing something differently?
Well, there is a standardized infrastructure for running ordinary
regression tests in contrib, but I dunno if it's of any use for dblink.
I suppose you could test in a single database by invoking dblink to
connect back to the same database you're in, but that'd seem like a
pretty cheesy test :-(.
I was planning to object to the "-U postgres" flags in your test script;
don't those just make it harder for people who aren't running the stuff
as user postgres? The 'pwd' invocations seem pointless as well.
regards, tom lane