Andreas Karlsson <andreas@proxel.se> writes:
> When working on a new feature for PostgreSQL I noticed that the
> pg_upgrade tests became flaky due to rules being dumped in a different
> order between the original cluster and the upgraded cluster. (For
> context: my regress scripts left a bunch of views with dependency
> circles which I had forgot to clean up.)
> So I propose that we sort rules on (schema, name, relation_name) instead
> of (schema, name, oid) like it currently does. The patch itself is small
> enough and makes it easier to test that we upgrade rules and views
> correctly with pg_upgrade.
Seems reasonable. Pushed with some trivial cosmetic adjustments
to make it look more like the identical adjacent cases for policies
and triggers.
regards, tom lane