| From: | Greg Burd <greg(at)burd(dot)me> |
|---|---|
| To: | pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Cc: | Peter Geoghegan <pg(at)bowt(dot)ie>, Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com>, Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com>, Nathan Bossart <nathandbossart(at)gmail(dot)com> |
| Subject: | Re: Tepid: selective index updates for heap relations |
| Date: | 2026-11-20 18:26:13 |
| Message-ID: | awCQj064tyeIBiby@floki |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Here is v85, rebased on 0cb41ee8070. Nothing in it changes the
design; it follows two upstream changes and fixes a few small things
found in review of v84.
Rebase:
- 0006: 1e676702cac (MERGE transition tables for cross-partition
UPDATEs) restructured the two ExecARUpdateTriggers() calls in
ExecInsert() and ExecDeleteEpilogue() that 0006 had changed; they
keep upstream's shape and gain 0006's old-row argument.
- 0018: upstream added recovery/t/061_recovery_target_prefetch.pl, so
061_hot_indexed_vm.pl and 062_hot_indexed_recovery.pl are now 062
and 063.
Fixes:
- 0017: pg_relation_hot_indexed_stats() stopped walking a chain at
the first collapse-survivor stub, since a stub carries no
HEAP_HOT_UPDATED bit, so avg_chain_len and max_chain_len were too
small for any chain that VACUUM had collapsed. It now follows the
stub's forward link. 0018 adds a root, stub and live-member chain
to hot_indexed_updates, which reported length 1 before.
- 0018: amcheck reports a stub whose forward link points at a line
pointer that is not normal; it skipped those silently.
- 0015: static_assert() instead of StaticAssertDecl() in pg_upgrade.
- 0016: a prune comment still called stubs "PHOT-style key tuples".
The other patches differ from v84 only by the rebase.
-greg
| Attachment | Content-Type | Size |
|---|---|---|
| tepid-v85.tgz | application/gzip | 274.5 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Masahiko Sawada | 2026-11-20 18:30:32 | Re: Use an index scan in RemoveSubscriptionRel() |
| Previous Message | Melanie Plageman | 2026-11-20 18:08:24 | Re: WAL_LOG CREATE DATABASE strategy broken for non-standard page layouts |