Berend Tober <btober@seaworthysys.com> writes:
> Tom Lane wrote:
>> Hmm, it shouldn't do that ... and in a quick test here I couldn't
>> reproduce any such bug. What version of pg_dump are you using?
> Sorry I failed to specify. Production version is 7.3.1 (change is
> hard!), although I origianally worked out the implementation on version
> 8. I bet that is the problem.
The case I tested seems to work in 7.3 as well:
CREATE TABLE person (last_name varchar(24),
first_name varchar(24),
CONSTRAINT person_name_check CHECK (((last_name IS NOT NULL) OR
(first_name IS NOT NULL))));
CREATE TABLE person_change_history(
action VARCHAR(6),
update_date TIMESTAMP NOT NULL DEFAULT NOW(),
update_user NAME NOT NULL DEFAULT CURRENT_USER
) INHERITS (person);
pg_dump puts the CONSTRAINT only on person, as it should. I'm testing
7.3.10 but I don't see any changes in the 7.3 CVS log that look related.
Can you put together a reproducible test case?
regards, tom lane