I wrote:
> Richard Guo <guofenglinux@gmail.com> writes:
>> It seems to me there is oversight here. Actually in next level up this
>> othersj would null all the relids in its syn_righthand, not only the
>> relids in its min_righthand.
> Good point. I think this code originated before it was clear to me
> that nullingrels would need to follow the syntactic structure.
Although ... the entire point here is that we're trying to build quals
that don't match the original syntactic structure. I'm worried that
distribute_qual_to_rels will do the wrong thing (put the qual at the
wrong level) if we add more nullingrel bits than we meant to. This
might be less trivial than it appears.
regards, tom lane