On Thu Nov 11, 2021 at 3:36 PM EST, Tom Lane wrote:
> I thought about this some more and realized exactly why I wanted to
> restrict the change to cases where the other side is a plain foreign
> Var: that way, if anything surprising happens, we can blame it
> directly on the user having declared a local column with a different
> type from the remote column.
>
> That being the case, I took a closer look at deparseVar and realized
> that we can't simply check "IsA(node, Var)": some Vars in the
> expression can belong to local tables. We need to verify that the Var
> is one that will print as a remote column reference.
Eminently reasonable all around! `git apply` insisted that the v8 patch
didn't (apply, that is), but `patch -p1` liked it fine. I've put it
through a few paces and it seems good; what needs to happen next?