"Brian Hirt" <bhirt@mobygames.com> writes:
> Great, I'm also trying to create a reproducable test case for the original
> problem i reported with duplicate rows/oids/pkeys. Maybe both problems are
> a result of the same bug; i don't know.
Were the duplicate rows all in tables that had functional indexes based
on functions similar to developer_aka_search_name? The problem we're
seeing here seems to be due to VACUUM not being able to cope with the
side effects of the SELECT inside the index function.
regards, tom lane