I wrote:
> Ouch. So this has been broken (by me, I think :-() since 8.0. Tells
> you something about how many people use hash indexes :-(
Actually it's not that bad --- this particular test case doesn't crash
in 8.0 or 8.1. I'm guessing we rearranged the order of transaction
abort processing in 8.2 in a way that exposed the problem. I patched
8.0 and 8.1 anyway, since it seems likely that some other more
complicated case might still show the problem there (eg something with
subtransactions).
regards, tom lane