Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> writes:
> One user reported a issues with triggers related to dropped columns.
> When I checked his code, I found a more possible problems.
This is just the same kind of modified-rowtype problem we've discussed
before.
I continue to maintain that the only realistic fix is to stop using the
PLPGSQL_DTYPE_ROW infrastructure for variables of externally-defined
composite types. In the last go-round, this was rejected on the
(unproven) assumption that it would make things slower. But as a wise
man once said, you can make code arbitrarily fast if it doesn't have to
give the right answer.
regards, tom lane