On Sun, Jun 1, 2025 at 12:07 PM Peter Geoghegan <pg@bowt.ie> wrote:
>
> Commit 052026c9b9 made heap_vacuum_rel call RelationGetNumberOfBlocks
> before it calls vacuum_get_cutoffs -- it swapped the order. This is
> wrong, as explained by an intact comment above the call to
> vacuum_get_cutoffs.
>
> In short, there is now a brief window during which the relation can be
> extended that'll allow heap pages with tuple headers < VACUUM's
> OldestXmin to be created, which are overlooked by that same VACUUM
> (they're beyond the same VACUUM's rel_pages). As a result of all this,
> VACUUM might advance pg_class.relfrozenxid to a later/younger value
> than a remaining/unfrozen XID value from one of these unscanned heap
> pages.
Thanks for the report. That was a dumb mistake. There was no reason
for me to move the line up as you can see in the diff -- it must have
been unintentional. I'll push the fix tomorrow.
I started to feel like I ought to write a TAP test, but I'm hesitant
to add a whole new TAP test for a case when a comment should have been
sufficient deterrent.
- Melanie
