>
> Yeah, verify_common.c does require the header, but what I meant was that
> verify_nbtree.c no longer needs it.
Oh, thanks, now I see..
Updated the patches.
I additionally tested the latest version (v2026-01-12-v2) -- and all looked good to me.
testing performed:
1. Applied cleanly; built with --enable-cassert, make -C contrib/amcheck check — all tests pass
2. Tested for 9 core GiST opclasses, on larger indexes (10M records), with both heapallindexed=false and heapallindexed=true
(10M records, old macbook m1):
Opclass heapallindexed=false heapallindexed=true
-------------- -------------------- -------------------
point_ops 3.9s 5.9s
box_ops 6.3s 9.3s
circle_ops 5.5s 10.6s
poly_ops 5.8s 13.7s
inet_ops 0.6s 5.6s
range_ops 3.2s 6.9s
multirange_ops 2.4s 10.8s
tsvector_ops 3.8s 13.7s
tsquery_ops 0.3s 4.9s
3. Similarly, just to double-check, for extensions that use GiST:
- contrib modules tested: btree_gist, cube, hstore, ltree, pg_trgm, seg (tested a single opclass for each; so 8 out of 30+ opclasses)
- skipped: intarray because it was annoyingly slow to build (looks like O(n²)?)
