On Mon, Dec 2, 2024 at 8:18 PM Peter Geoghegan <p...@bowt.ie> wrote: > Attached is a refined version of a test case I posted earlier on [2], > decisively proving that GiST index-only scans are in fact subtly > broken. Right now it fails, showing a wrong answer to a query. The > patch adds an isolationtest test case to btree_gist, based on a test > case of Andres'.
I can confirm that the same bug affects SP-GiST. I modified the original failing GiST isolation test to make it use SP-GiST instead, proving what I already strongly suspected. I have no reason to believe that there are any similar problems in core index AMs other than GiST and SP-GiST, though. Let's go through them all now: nbtree already does everything correctly, and all remaining core index AMs don't support index-only scans *and* don't support scans that don't just use an MVCC snapshot. -- Peter Geoghegan
