On Sat, Mar 25, 2023 at 6:25 PM Peter Geoghegan <pg@bowt.ie> wrote:
> On Fri, Mar 24, 2023 at 8:13 AM Robert Haas <rhaas@postgresql.org> wrote:
> > If we're checking xmin and find that it is invalid (i.e. 0) just
> > report that as corruption, similar to what's already done in the
> > three cases that seem correct. If we're checking xmax and find
> > that's invalid, that's fine: it just means that the tuple hasn't
> > been updated or deleted.
>
> What about aborted speculative insertions? See
> heap_abort_speculative(), which directly sets the speculatively
> inserted heap tuple's xmin to InvalidTransactionId/zero.
Oh, dear. I didn't know about that case.
> It probably does make sense to keep something close to this check --
> it just needs to account for speculative insertions to avoid false
> positive reports of corruption. We could perform cross-checks against
> a tuple whose xmin is InvalidTransactionId/zero to verify that it
> really is from an aborted speculative insertion, to the extent that
> that's possible. For example, such a tuple can't be a heap-only tuple,
> and it can't have any xmax value other than InvalidTransactionId/zero.
Since this was back-patched, I think it's probably better to just
remove the error. We can introduce new validation if we want, but that
should probably be master-only.
--
Robert Haas
EDB: http://www.enterprisedb.com