Martijn van Oosterhout <kleptog@svana.org> writes:
> The only thing I considered wrong with the previous patch was it's odd
> behaviour w.r.t. creation of types with the same name as one in
> pg_catalog. However, Tom said this is not a serious issue given the
> same problem exists in many other places in PostgreSQL.
> So, I've added some regression tests and here is the patch for
> inclusion.
Applied with corrections --- most notably, you weren't setting the
dependencies correctly for a shell type, and referencing numeric
function OIDs from C code is not done.
regards, tom lane