Le lun. 29 nov. 2021 à 22:27, Tom Lane <tgl@sss.pgh.pa.us> a écrit :
Guillaume Lelarge <guillaume@lelarge.info> writes: > Le lun. 29 nov. 2021 à 20:39, Tom Lane <tgl@sss.pgh.pa.us> a écrit : >> [ looks closer ... ] Ah, that patch is a bit buggy: it fails to do the >> right thing in the cases where the loop does a "continue". The attached >> revision seems to behave properly.
> I've tried your patch with my test case. It still uses a lot of memory. > Actually even more.
Hmm ... I tried it with your test case, and I see the backend completing the query without going beyond 190MB used (which is mostly shared memory). Without the patch it blows past that point very quickly indeed.
I'm checking it in HEAD though; perhaps there's something else wrong in the back branches?
That's also what I was thinking. I was only trying with v14. I just checked with v15devel, and your patch works alright. So there must be something else with back branches.