Hi Richard,
> As stated in [1], all paths arriving here are parameterized by top
> parents, so we should check against top_parent_relids if it exists in
> the two Asserts.
>
> Attached is a patch fixing that.
Probably it's just because of my limited experience with the optimizer
but I don't find the proposed change particularly straightforward. I
would suggest adding a comment before the Assert's and/or a detailed
commit message would be helpful.
Other than that I can confirm that both branches in the Assert's are
executed and the tests pass in different test environments.
--
Best regards,
Aleksander Alekseev