On 12.01.23 18:40, Nathan Bossart wrote:
> On Thu, Jan 12, 2023 at 12:20:50PM -0500, Tom Lane wrote:
>> Peter Eisentraut <peter(dot)eisentraut(at)enterprisedb(dot)com> writes:
>>> On 12.01.23 01:04, Nathan Bossart wrote:
>>> - classoid == AggregateRelationId ||
>>>> I noticed that AggregateRelationId isn't listed in the ObjectProperty
>>>> array, so I think recordExtObjInitPriv() will begin erroring for that
>>>> classoid instead of ignoring it like we do today.
>>
>>> Hmm, we do have some extensions in contrib that add aggregates (citext,
>>> intagg). I suspect that the aggregate function is actually registered
>>> into the extension via its pg_proc entry, so this wouldn't actually
>>> matter. But maybe the commenting should be clearer?
>>
>> Yeah, I don't believe that AggregateRelationId is used in object
>> addresses; we just refer to pg_proc for any kind of function including
>> aggregates. Note that there is no "oid" column in pg_aggregate.
>
> Got it, thanks for clarifying.
I have updated the patch as you suggested and split out the aggregate
issue into a separate patch for clarity.