Re: pg_createsubscriber does not check output_plugin_libraries

On Fri, Sep 11, 2026 at 11:51 AM Hayato Kuroda (Fujitsu)
<kuroda(dot)hayato(at)fujitsu(dot)com> wrote:
>
> > As long as everyone is in agreement that --dry-run *should* be
> > performing those checks in the first place, I'm happy.
>
> Yes, it's my intention. Fujii-san, how do you feel? If you're OK, we can proceed.

I'm not sure if we have a clear policy on what --dry-run should check.
Personally, I'd like it to check as many prerequisites as practical.
Ideally, I just wish that a successful dry run would mean that the actual
run will succeed, but that's probably not always possible.

Some checks may be difficult to perform or have undesirable side effects,
especially if they require creating, dropping, or modifying objects. They
may also require complicated code or introduce performance overhead. So I
think it's reasonable to leave such checks out.

Therefore, I'd just favor adding checks that are straightforward and
practical to perform.

TBH, the code for checking output_plugin_libraries seems to be getting
more complicated than I initially expected. But, from my view, it hasn't
become complicated and difficult enough to justify leaving the check out,
at least *for now*.

> > Even for PG17, specifying a non-superuser in the connection string
> > leads to a successful --dry-run, then fails during normal operation.
>
> To confirm, is it allowed to add such a verification in back branches?
> I imagined running something like below after connecting to the publisher.

Adding such verification would be useful, but it looks like an improvement
rather than a bug fix to me. So, I don't think we can add it to the stable
branches. Thoughts?

Regards,

--
Fujii Masao