Andres Freund <andres@anarazel.de> writes:
> I'm asking again, what exactly do we test by having these types in the
> dump? They're bog standard types, there's nothing new for pg_dump to
> test with them. No weird typmod rules, no weird parse-time type mapping,
> nothing?
That's a pretty fair point. The types' I/O functions will be exercised
well enough by the regression tests themselves, and it's hard to see what
more test coverage is gained by including them in pg_dump/pg_upgrade
testing. Maybe we should just drop those tables and be done with it.
regards, tom lane