Manfred Koizar <mkoi-pg(at)aon(dot)at> writes:
> If the sequence is "first update xmin, then set the commit bit", we
> never have an inconsistent state. And if the change is lost, it can
> be redone by the next backend visiting the tuple.
Not if the subtransaction log state has been removed as no longer
needed. I think a WAL entry will be essential. (An alternative
might be to keep subtransaction state as long as we keep pg_clog
state, but that's pretty unpleasant too.)
I think we'd be a lot better off to design this so that we don't need to
alter heap tuple xmin values...
regards, tom lane