Re: BUG #17485: Records missing from Primary Key index when doing REINDEX INDEX CONCURRENTLY

On Tue, May 31, 2022 at 12:19 PM Andres Freund <andres(at)anarazel(dot)de> wrote:
> I don't think that assertion is correct.
>
> Consider transactions aborting concurrently with heap pruning. You could have
> done a HTSV for one chain element, a concurrent abort happened, then you did
> the HTSV for another chain element. If the HTSVs were not in the order of the
> HOT chain you could see HEAPTUPLE_DEAD for an earlier chain element, while
> seeing HEAPTUPLE_INSERT_IN_PROGRESS in a later one. There's several other
> scenarios with subtransaction aborts as well, I think.

I think that it probably was correct before I rebased the patch on top
of your bugfix commit 18b87b201f. The original version would have
actually called HTSV directly, at the point that it accessed each
tuple from a HOT chain. If nothing else this suggests that the patch
should be clear on this point about not calling HTSV in HOT chain
order.

Offhand I think that it probably would still work if it was limited to
HEAPTUPLE_LIVE (no more asserting in the HEAPTUPLE_INSERT_IN_PROGRESS
case). Not sure if that's worth it. A topic for another time.

--
Peter Geoghegan