Re: updates for handling optional argument in system functions

Fujii Masao <masao.fujii@gmail.com> writes:
> I have one question, though.

> Suppose a user executes GRANT or REVOKE on pg_get_ruledef(oid), then
> creates a dump with pg_dump. If that dump is restored into a newer
> server where these patches have been applied, the restore would fail
> because pg_get_ruledef(oid) no longer exists.

> Is that acceptable (this means that users need to handle the restore
> failure), or should pg_dump handle this case specially?

I wouldn't blink an eye at that.  There's no obvious use-case for
someone to mess with the permissions on that function.  Even
if there were, I don't think it'd be reasonable for pg_dump to
try to patch it up.  (I'm not sure that pg_dump would dump such
grants in the first place.)

			regards, tom lane