Joe Conway <mail@joeconway.com> writes:
> OK, I'm on it now. Sorry I missed those issues. I guess my own testing
> was too myopic :(
Okay. I have patches to fix the domain-type issues, and will commit
as soon as I've finished testing 'em.
I would suggest that either gram.y or someplace early in the analyzer
should transpose the names from the coldeflist into the "user specified
alias" structure. That should fix the alias naming issues. The other
issues indicate that if a coldeflist is provided, you should check it
against the function return type in all cases not only RECORD. In the
non-RECORD cases it could be done in the parse analysis phase.
regards, tom lane