Horton

Re: 'ERROR: attempted to update invisible tuple' from 'ALTER INDEX ... ATTACH PARTITION' on parent index

Robert Haas <robertmhaas@gmail.com>

View thread
From: Robert Haas <robertmhaas@gmail.com>
To: Shruthi Gowda <gowdashru@gmail.com>
Cc: Dilip Kumar <dilipbalaut@gmail.com>, Michael Paquier <michael@paquier.xyz>, PostgreSQL Development <pgsql-hackers@postgresql.org>
Date: 2023-07-12T20:02:23Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Add indisreplident to fields refreshed by RelationReloadIndexInfo()

  2. Fix updates of indisvalid for partitioned indexes

On Wed, Jul 12, 2023 at 12:28 PM Shruthi Gowda <gowdashru@gmail.com> wrote:
>   I reviewed the function RelationReloadIndexInfo() and observed that the 'indisreplident' field and the SelfItemPointer 't_self' are not refreshed to the pg_index tuple of the index.
>   Attached is the patch that fixes the above issue.

Oh, interesting. The fact that indisreplident isn't copied seems like
a pretty clear mistake, but I'm guessing that the fact that t_self
wasn't refreshed was deliberate and that the author of this code
didn't really intend for callers to look at the t_self value. We could
change our mind about whether that ought to be allowed, though. But,
like, none of the other tuple header fields are copied either... xmax,
xvac, infomask, etc.

-- 
Robert Haas
EDB: http://www.enterprisedb.com