Greg Stark <gsstark(at)mit(dot)edu> writes:
> Perhaps what should really be happening here is that there should be
> dependencies from the pg_inherit entry to the two tables rather than from one
> table to the other.
This seems unlikely to still have the correct semantics (DROP on child
is OK, DROP on parent is not unless CASCADE, in which case child is
dropped too, etc etc).
regards, tom lane