On 13.03.24 19:43, Tomas Vondra wrote:
> I looked at this patch today. I went through all the calls switched to
> equalRowTypes, and AFAIK all of them are correct - all the places
> switched to equalRowTypes() only need the weaker checks.
>
> There's only two places still calling equalTupleDescs() - relcache
> certainly needs that, and so does the assert in execReplication().
>
> As for attndims, I agree equalRowTypes() should not check that. We're
> not really checking that anywhere, it'd be quite weird to start with it
> here. Especially if the plan is to remove it entirely.
Thanks for checking this again. I have committed the patch as it was
presented then.