"Nick Fankhauser" <nickf@ontko.com> writes:
> I agree that this still may be a bug.
It definitely is a bug --- we fixed a similar problem around 7.1.1 or
so, but this test case appears to expose a different variant of the
mistake. (The planner is generating a plan that the executor can't
handle; it's supposed to know not to do that.)
I think I know where to fix it, but am not confident enough in
my powers of analysis today to want to actually commit anything.
(I've had a bad head-cold all week and am still unable to do anything
that requires more than a few minutes of sustained thought :-()
Will get back on it as soon as I feel better...
regards, tom lane