swingi(at)gmail(dot)com writes:
> I just noticed that when you add a not valid constraint to a table that is
> inherited, this will result in pg_dump emitting an add constraint operation
> for every table in the inheritance hierarchy. Therefore restoring a dump
> will cause errors. Below is an example, first with a valid constraint (goes
> well), and then the not valid constraint.
Seems to be a simple oversight in the code for handling not-valid
constraints. Thanks for the report!
regards, tom lane