On 2/12/25 07:23, Peter Eisentraut wrote:
I committed your patch on Sunday, and now it's about 72 hours later.
I've been observing this on cfbot for some time. Before the patch, you could go to cfbot at any
time and find between 5 and 10 test failures from this problem. And now there are none. So I'm
calling provisional success on this.
That's great! It's sort of unsatisfying though---and unnerving. I wish when the test failed we knew
what the oids were for the RESTRICT constraint and the just-dropped NO ACTION constraint. (There's
no way to get that after the fact, is there?) I probably won't keep putting time into this, but it
seems like there must still be a hard-to-hit bug in the code for caching query plans. Since the
behavior disappeared, it is more evidence that that's where the real problem lies.
Yours,
--
Paul ~{:-)
p...@illuminatedcomputing.com
