On Wed, Oct 28, 2026 at 15:23 UTC, Salma Elsayed <[email protected]> wrote: > This is v4 of the B-tree leaf page merge patch series (3 patches > attached), rebased on current master. There are a lot of changes since > v3 [1].
Thanks, Salma. My AI harness for PG hacking found a test issue in the posted v4: 007 passed 16 of 21 checks when TESTDATADIR contains a space; five checks get empty pg_waldump output. Could you check how waldump_range() invokes pg_waldump and handles failures? With a test-only correction, all 21 checks passed, without a server change. Targeted reruns for #103, #104 and #105, and adapted reproducers for #121 and #122, passed in the tested scenarios; 008 passed 703/703. The exact setup, results and limits are here: https://github.com/NikolayS/postgres/issues/101#issuecomment-6489250891 #110 and the reported backward parallel deadlock remain open. I have not reproduced the latter, nor reviewed these results by hand. Nik
