Re: pgsql: Teach DSM registry to ERROR if attaching to an uninitialized ent - Mailing list pgsql-hackers

From Robert Haas
Subject Re: pgsql: Teach DSM registry to ERROR if attaching to an uninitialized ent
Date
Msg-id CA+TgmoaSpDa8dqWBwjRAittp+XTKN698Tpg3CEMpi=08sKc4Sw@mail.gmail.com
Whole thread Raw
In response to Re: pgsql: Teach DSM registry to ERROR if attaching to an uninitialized ent  (Nathan Bossart <nathandbossart@gmail.com>)
List pgsql-hackers
On Tue, Nov 25, 2025 at 4:16 PM Nathan Bossart <nathandbossart@gmail.com> wrote:
> My tests seem happy, so I will plan on committing these patches tomorrow.
> The only difference in v4 is that pg_get_dsm_registry_allocations() will no
> longer show partially-initialized entries.  I thought that was the best
> option in this case because such entries shouldn't have any allocated
> memory, and retrying GetNamed*() would set *found to false.

Without actually opening the new patch files, that seems OK to me. I
think the important thing about such a view is that it shouldn't make
any state changes; it should just show how things are. In my ideal
world, it would probably show partially-initialized entires in some
distinguishable way, like with a null size. But leaving them out is
also defensible. The thing that I think really matters for failure
cases is the ability to retry whatever's gone wrong, either to get it
fixed or to see the error, and the other changes already accomplish
that goal, so I'm happy.

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



pgsql-hackers by date:

Previous
From: Manni Wood
Date:
Subject: Re: Speed up COPY FROM text/CSV parsing using SIMD
Next
From: Masahiko Sawada
Date:
Subject: Re: POC: enable logical decoding when wal_level = 'replica' without a server restart