On 06/24/2013 03:39 PM, Tom Lane wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> We probably do need to preserve attribute numbers across an upgrade,
>> even for foreign tables. I think those could make their way into
>> other places.
> Hm ... this argument would also apply to composite types; do we get it
> right for those?
>
>
Yes we do. It's handled at pg_dump.c starting about line 8936.
It looks like we need to add cases of foreign tables to the block that
starts around line 13117.
We should also have a test of this in the core regression tests so that
doing the wrong thing might be caught by regular upgrade testing.
cheers
andrew