Re: pgsql: Refactor ChangeVarNodesExtended() using the custom callback

On Wed, May 7, 2025 at 12:47 PM Aleksander Alekseev
<aleksander(at)timescale(dot)com> wrote:
> > How can this or other commits on master branch prevent creating new
> > tag REL_18_BETA2 or forking REL_18_STABLE from master? Perhaps, this
> > commit is not the only one on master branch after tag REL_18_BETA1.
>
> Right, but I see people fixing whitespaces and documentation for PG18
> in the `master` branch right now. My humble understanding is that
> `master` is basically REL_18_STABLE at the moment.
>
> If one wants to fork REL_18_STABLE for instance, either the branch
> will include your change (which we probably don't want), or one will
> have to manually revert it, or create a fork from BETA1 and
> cherry-pick all the necessary changes from `master`.
>
> This is the reason why I suspected this was committed by mistake.

OK, now I get your point: you think this should go to the next major
release. But no, it's not. It's intended to eventually land into
REL_18_STABLE, REL_18_BETA2 etc. It addresses one of PG18 Open Items:
Richard Guo found the way SJE changes ChangeVarNodes() unsatisfiable,
and this commit is intended to fix that. Also, note that it is not
the only PG18 Open Item left after beta1, and fixing Open Items
usually requires changes besides docs and whitespaces.

This commit changes ABI, but this should be OK. We do preserve ABI
only for stable releases.

------
Regards,
Alexander Korotkov
Supabase