On Tue, Mar 17, 2020 at 10:42 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Argh, -ENOCAFFEINE. You're right, AttrNumber is not the relevant
> type here. There's still an issue of whether anyplace has tried
> to store a varno in less-than-int, but it seems less likely than
> I was thinking earlier.
That seems pretty unlikely to me, but even if it does happen, it's
just a bug which we could simply fix when we find it. I mean, I don't
want to have bugs, but there's no patch that is free of all risk. If
we go through and look around at how this is handled in various
places, and check as carefully as we can for possible problems, I
don't think it would be a particularly high-risk patch.
> > (It doesn't seem like we're all that careful about whether we're using
> > int, which is signed, or Index, which is unsigned, and we might want
> > to go around and tighten that up.
>
> Yeah, I'd be leery of trying to use that last bit. The cost-benefit
> ratio on that is *definitely* not attractive.
That seems clear, but trying to be consistent about the types we use
for things is a good idea on general principle, even if we have no
intention of doing anything that really relies on it any time in the
forseeable future. If nothing else, it might avoid some compiler
warnings.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company