On Mon, Apr 11, 2022 at 9:35 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > The other backend's page defragmentation step (from pruning)
> > would render our backend's HeapTuple pointer invalid. Presumably it
> > would just look like an invalid/non-matching xmin in our backend, at
> > the point of control flow that Valgrind complains about
> > (heapam_handler.c:509).
>
> Right, but there are other accesses below, and in any case match
> failure isn't necessarily the right thing.
That's what I meant -- it very likely would have been a match if the
same scenario played out, but without any concurrent pruning. With a
concurrent prune, xmin won't ever be a match (barring a
near-miraculous coincidence). That behavior is definitely wrong, but
also quite subtle (compared to what might happen if we got past the
xmin/xmax check). I think that that explains why it took this long to
notice the bug.
--
Peter Geoghegan