From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Geoghegan <pg@bowt.ie>
cc: Hans Buschmann <buschmann@nidsa.net>,
        Amit Langote <amitlangote09@gmail.com>,
        "David G. Johnston" <david.g.johnston@gmail.com>,
        "pgsql-bugs@lists.postgresql.org" <pgsql-bugs@lists.postgresql.org>
Subject: Re: AW: AW: BUG #18147: ERROR: invalid perminfoindex 0 in RTE with
 relid xxxxx
In-reply-to: 
 <CAH2-WzkYBZ8qCsMs8i3BuKFLXO2UGKLiB+hhQZDC7=3Y+yf1pA@mail.gmail.com>
References: <18147-6fc796538913ee88@postgresql.org>
 <CAKFQuwYFq-Q2rxbXXBmdHCAEFm-Pu_qhPAEi-PO+LmjtYN2hhg@mail.gmail.com>
 <2796464.1696525626@sss.pgh.pa.us>
 <9fdfed847859435e8641e613198be9e3@nidsa.net>
 <3319880.1696778678@sss.pgh.pa.us>
 <3c9fe737f6aa4ae9b0dc19bffd97e106@nidsa.net>
 <408285.1698085850@sss.pgh.pa.us>
 <CAH2-WzmY-1GPgLV=3HXPMf0DPwZCjeVVceMSViaoMj=unL1Qtw@mail.gmail.com>
 <411084.1698087308@sss.pgh.pa.us>
 <CAH2-Wzk=D3X3fKPeg-u1VirLOKU64xcjSnqyuStiFw2kYJuuGw@mail.gmail.com>
 <CAH2-WzmU_BQ=-H9L+bxTSMQBqHMjp1DSwGypvL0gKs+dTOfkKg@mail.gmail.com>
 <458056.1698107313@sss.pgh.pa.us>
 <CAH2-WzkYBZ8qCsMs8i3BuKFLXO2UGKLiB+hhQZDC7=3Y+yf1pA@mail.gmail.com>
Comments: In-reply-to Peter Geoghegan <pg@bowt.ie>
	message dated "Mon, 23 Oct 2023 17:45:39 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <462248.1698109561.1@sss.pgh.pa.us>
Content-Transfer-Encoding: 8bit
Date: Mon, 23 Oct 2023 21:06:01 -0400
Message-ID: <462249.1698109561@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/462249.1698109561%40sss.pgh.pa.us>
Precedence: bulk

Peter Geoghegan <pg@bowt.ie> writes:
> On Mon, Oct 23, 2023 at 5:28 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Separately, I wonder if index_unchanged_by_update should actually just
>>> always give the hint with a non-HOT update, regardless of the
>>> specifics for each index/its columns -- just like on the v14 branch.

>> I'm confused.  Wouldn't that be the exact opposite of "unchanged"?

> Well, in practice "indexUnchanged = true" means "do bottom-up deletion
> if it's the only way to avoid a page split". The justification is that
> the incoming tuple is "logically unchanged" (actually it's more
> complicated than that, but that's our starting point).

But doesn't the need for a non-HOT update show that the tuple *was*
changed --- in index-relevant columns, even?  Maybe I'm still not
understanding exactly what condition we're detecting.

			regards, tom lane