On 2023-Apr-04, Tom Lane wrote:
> PG Bug reporting form <noreply@postgresql.org> writes:
> > result:
> > postgres=# alter table public.test DISABLE TRIGGER USER;
> > ERROR: 42704: trigger "trigger_test" for table "test_def" does not exist
> > LOCATION: EnableDisableTriggerNew, trigger.c:1658
>
> > Expected Result (Postgresql 15.2 and 14.4 - fine):
> > testdb=> alter table public.test DISABLE TRIGGER USER;
> > ALTER TABLE
>
> Commit ec0925c22 seems to have been quite snakebit.
No kidding :-( Clearly, commit 86f575948c77 ("Allow FOR EACH ROW
triggers on partitioned tables") should have included more tests.
> I already fixed
> a deficiency in it in v15/HEAD, but here we have a different symptom
> in the older branches. What's happening is that EnableDisableTrigger
> is ignoring the child trigger because it has tgisinternal set to true
> and the command passes skip_system = true.
Hmm, right.
> I'm inclined to think that in the older branches (pre f4566345c)
> we need to do
>
> - if (oldtrig->tgisinternal)
> + if (oldtrig->tgisinternal && !OidIsValid(oldtrig->tgparentid))
> {
> /* system trigger ... ok to process? */
>
> but I've not tested that. Alvaro, what do you think?
Oh, that's a nice and clean solution. I tested it (on 13 and 14) and it
does solve the problem, and no regression tests fail, but I didn't try
to break it further. 12 and back are unaffected, for lack of
86f575948c77.
--
Álvaro Herrera 48°01'N 7°57'E — https://www.EnterpriseDB.com/
"Nadie está tan esclavizado como el que se cree libre no siéndolo" (Goethe)