Re: Patch: dumping tables data in multiple chunks in pg_dump - Mailing list pgsql-hackers

From Hannu Krosing
Subject Re: Patch: dumping tables data in multiple chunks in pg_dump
Date
Msg-id CAMT0RQSudopnkyrk+k=5naHo2GiOcJsSGQBOfF1ovsbTs7N6Ew@mail.gmail.com
Whole thread Raw
In response to Re: Patch: dumping tables data in multiple chunks in pg_dump  (Andres Freund <andres@anarazel.de>)
List pgsql-hackers
On Wed, Jan 14, 2026 at 11:17 PM Andres Freund <andres@anarazel.de> wrote:
>
> Hi,
>
> On 2026-01-14 23:10:38 +0100, Hannu Krosing wrote:
> > Is this the whole guidelines we have for this ?
>
> I don't think we have more written down in a central place.
>
>
> > Did not recognise it at once, but I assume that this implies also
> > things about backporting, as only way to backport something is to put
> > it in a minor release.
>
> Correct.

Thanks for confirming!

> > I was hoping that there were some exceptions possible for things
> > affecting interactions between different versions that need also
> > support from older versions, but if this page is all we have then
> > likely not.
>
> I think we have made maybe a handful of exceptions over the years, not more.
>
> > Do we have the same strict no improvements in minor versions policy
> > for contrib/ extensions and tools?
>
> Yes.

I have confirmed that one can not load a -Fd dump from newer version
of pg_dump using an older version of pg_restore.
It plainly refuses without even trying.

~/tmp$ /usr/lib/postgresql/17/bin/pg_dump -p 5433 -Fd testload -f dump17
~/tmp$ /usr/lib/postgresql/16/bin/pg_restore -p 5432 -C -d postgres dump17
pg_restore: error: unsupported version (1.16) in file header

Have we thought of other ways than backporting for supporting clients
who may improvements for older versions ?
I know that 2ndQuardrant maintained its own distribution for a while
for enterprise customers who absolutely needed some improvements.
Could PGDG repos, for example, have a separate "unofficial" backports
section for this?

----
Hannu



pgsql-hackers by date:

Previous
From: Heikki Linnakangas
Date:
Subject: Re: refactor architecture-specific popcount code
Next
From: "Hayato Kuroda (Fujitsu)"
Date:
Subject: RE: Exit walsender before confirming remote flush in logical replication