Re: Tricky bugs in concurrent index build

Greg Stark <gsstark(at)mit(dot)edu> writes:
> What would happen if we just insert DELETE_IN_PROGRESS tuples normally? Would
> the only risk be that the index build would fail with a spurious unique
> constraint violation? I suppose it would be pretty common though given how
> updates work.

Yeah, that's the problem: if we can't support UPDATEs that don't change
the to-be-unique column, it ain't much of a feature.

> Incidentally does this point out a problem with the planner depending on
> unique constraints?

Good point. It doesn't depend on them yet, but we've been hoping to
make it do so once we have plan invalidation capability. We shall have
to think very carefully about timing semantics of all that.

regards, tom lane