On 2019-12-04 15:14, Tom Lane wrote:
> Run the regression tests with "make installcheck", then:
>
> $ pg_dump -Fc regression >r.dump
> $ createdb r2
> $ pg_restore -d r2 r.dump
> pg_restore: while PROCESSING TOC:
> pg_restore: from TOC entry 6005; 2604 24821 DEFAULT gtest1_1 b postgres
> pg_restore: error: could not execute query: ERROR: column "b" of relation "gtest1_1" is a generated column
> Command was: ALTER TABLE ONLY public.gtest1_1 ALTER COLUMN b SET DEFAULT (a * 2);
>
>
> pg_restore: warning: errors ignored on restore: 1
> $
>
> It looks like gtest1_1 inherits column "b" from gtest1, so possibly
> pg_dump is just confused about the combination of inheritance and
> generated columns.
Yeah, there was some stuff about the "separate" dumping of defaults that
I apparently forgot to address. The attached patch fixes it. I'll see
about adding a test case for it, too.
> I see this in v12 as well as HEAD. One interesting question is how come
> the pg_upgrade test isn't failing --- maybe binary-upgrade mode handles
> this case differently?
Binary upgrade dumps out even inherited columns, so it won't run into
the "separate" case that's the issue here.
--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services