On Thu, 7 Mar 2024 at 22:50, David Rowley <dgrowleyml@gmail.com> wrote:
>
> On Thu, 7 Mar 2024 at 15:24, Tender Wang <tndrwang@gmail.com> wrote:
> >
> > Andrei Lepikhov <a.lepikhov@postgrespro.ru> 于2024年3月6日周三 11:37写道：
> >> I think, it is a bug. Should it be fixed (and back-patched) earlier?
> >
> > Agreed. Need David to review it as he knows this area best.
>
> This is on my list of things to do. Just not at the top yet.
I've gone over this patch and I'm happy with the changes to
nodeMemoize.c. The thing I did change was the newly added test. The
problem there was the test was passing for me with and without the
code fix. I ended up changing the test so the cache hits and misses
are reported. That required moving the test to above where the
work_mem is set to 64KB so we can be certain the values will all be
cached and the cache hits are predictable.
My other changes were just cosmetic.
Thanks for working on this fix. I've pushed the patch.
David