Say guys, could you look into why btree_gist is failing its timetz
regression test in the 8.0 branch, if built with
--enable-integer-datetimes? Several of the buildfarm machines
are showing this, and I've reproduced it locally.
When I saw it I said to myself, "Oh, I forgot to back-patch the code
to make the timetz time + zone calculation do the Right Thing".
But it still fails after back-patching. The same regression test
passes in CVS HEAD, so I'm thinking there must be some other bug fix
that needs to get propagated back to 8.0. I don't have enough knowledge
of the GIST code to figure it out quickly, though.
regards, tom lane