Michael Paquier <michael(at)paquier(dot)xyz> writes:
> I would rather keep the solution with client_min_messages, and the
> tests in vacuum.sql to keep those checks for the grammar parsing. So
> this basically brings us back to use the patch I proposed here:
> https://www.postgresql.org/message-id/20191107013942.GA1768@paquier.xyz
OK.
regards, tom lane