Re: Patch for migration of the pg_commit_ts directory

On Thu, Oct 15, 2026 at 2:55 PM Hayato Kuroda (Fujitsu)
<kuroda(dot)hayato(at)fujitsu(dot)com> wrote:
>
> Hi hackers,
>
> I did a self reviewing for 0001 and 0002, and updated the doc a bit.
> PSA new patch set.
>

Hi Kuroda-san,

Thanks for splitting the patch set; it makes this much easier to
review. A few comments on v16.

1. The patch set no longer applies cleanly on HEAD. t/009 is now taken
by 009_composite_types.pl, and there are conflicts in check.c and
pg_upgrade.c.
2. 0003: There is already a separate patch that preserves replication
origins during upgrade [1], so I suggest removing 0003 from this patch
set.
3. 0001/0002: I agree with Maxim that 0001 shouldn't be committed
without 0002 (or without origins being preserved). Otherwise the wider
commit_ts range turns stale origin IDs into false
update/delete_origin_differs conflicts, as Sawada-san showed.
4. 0002: The option name --pg-commit-ts and the help text "copy
pg_commit_ts directory" describe the implementation rather than the
behaviour. How about something like --preserve-commit-timestamps?
Also, the help line is placed between -V and --clone, which breaks the
alphabetical order of the long options.

[1] https://www.postgresql.org/message-id/CAFPTHDbPXU7cs189J=BrzSa20EJzajOU25N7G7nT6UZqMkRnXg@mail.gmail.com

Regards,
Ajin Cherian
Fujitsu Australia