Dean Rasheed <dean.a.rasheed@gmail.com> writes:
> I can't think of any practical uses for this kind of query, so I don't
> think it's worth worrying too much about its results until/unless
> someone comes up with a real use-case.
> However, given that we currently support queries like "select distinct
> * from nocols" (albeit with rather odd results), I don't think we
> should start throwing new errors for them. Perhaps the actual risk of
> a backwards-compatibility break is small, but so too is any benefit
> from adding such new errors.
> So +1 for the patch as-is, with no new errors.
How about as-is in the back branches, and throw the new errors only
in HEAD?
regards, tom lane