Re: [PATCH] Fix alter subscription concurrency errors

> "Gregory Stark (as CFM)" <stark(dot)cfm(at)gmail(dot)com> writes:
> > Hm. This patch is still waiting on updates. But it's a bug fix and it
> > would be good to get this in. Is someone else interested in finishing
> > this if Jeite isn't available?
>
> I think the patch as-submitted is pretty uninteresting, mainly because the
> design of adding bespoke lock code for subscription objects doesn't scale.
> I'm not excited about doing this just for subscriptions without at least a
> clear plan for working on other object types.
>
> Alvaro suggested switching it to use get_object_address() instead, which'd
> be better; but before going in that direction we might have to do more
> work on get_object_address's error reporting (note the para in its header
> comments saying it's pretty weak on that).

Sorry for not responding earlier in this thread. I'll be honest in
saying this was a small annoyance to me, so I ignored theresonses more
than I should have. It caused some test flakiness in the Citus test
suite, and it seemed that fixing the underlying issue in Postgres was
most appropriate. I addressed this in Citus its test suite by
disabling the relevant test (which was an edge case test anyway). So
my immidiate problem was fixed, and I stopped caring about this patch
very much. Definitely not enough to address this for all other DDLs
with the same issue.

All in all I'm having a hard time feeling motivated to work on a patch
that I don't care much about. Especially since I have two other
patches open for a few commit fests that I actually care about, but
those patches have received (imho) very little input. Which makes it
hard to justify to myself to spend time on this patch, given the
knowledge that if I would spend time on it, it might take away the
precious reviewer time from the patches I do care about.