Re: REVOKE FROM warning on grantor

On Thursday, March 14, 2024, Étienne BERSAC <etienne(dot)bersac(at)dalibo(dot)com>
wrote:

>
> However, I'd prefer if Postgres fails properly. Because the GRANT is
> actually not revoked. This prevent ldap2pg to report an issue in
> handling privileges on such roles.
>
> What do you think of make this warning an error ?
>
>
The choice of warning is made because after the command ends the grantmin
question does not exist. The revoke was a no-op and the final state is as
the user intended. Historically doing this didn’t give any message at all
which was confusing so we added a warning so the semantics of not failing
were preserved but there was some indication that something was amiss. I
don’t have a compelling argument to,change the long-standing behavior.
Client code can and probably should look for a show errors reported by the
backend. It is indeed possibly to treat this warning more serverly than
the server chooses to.

David J.