On 2019-Apr-21, Tom Lane wrote:
> ISTM that this is a bug, not a feature: if there's any point at
> all to saying ONLY in this context, it's that we're not supposed
> to be doing anything as expensive as adding a new constraint to
> a child partition. No? So I think that this should have failed.
Hmm, yeah, this is not intentional and I agree that it shouldn't be
doing this.
> We need to require the partition(s) to already have attnotnull set.
Sounds good to me, yes.
Do you want me to see about this?
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services