Rosser Schwarz <rosser(dot)schwarz(at)gmail(dot)com> writes:
> We've opted to simply script issuing the statements serially, rather
> than as a function for this one. Your explanation makes sense, Tom,
> though I'm curious why it works in our long-running case (beyond the
> issue of fewer indices there).
Does the other case have any partial indexes? The particular error
was about an index predicate condition; it might be that you don't
happen to fall foul of the issue when dealing with simple CREATE
INDEX commands. I'd describe that as "miraculously managing not to
fail", though, not as a behavior I'd care to rely on.
regards, tom lane