Re: "anyelement2" pseudotype

Tom Lane wrote:
> I realized that I can probably fix ATAddForeignKeyConstraint to do the
> right thing by having it pass the two actual column types to
> can_coerce_type, thus allowing check_generic_type_consistency to kick
> in and detect the problem.

Yeah, I came to the same conclusion. No amount of refactoring in 
parse_coerce.c is going to get the original concrete types back to 
compare. That should fix the problem with arrays, enums and any 
potential future generic types without mentioning them explicitly in 
there a la the hack there currently, thankfully.

Cheers

Tom