On Thu, Nov 24, 2022 at 12:31:09PM +0100, Daniel Gustafsson wrote:
> Looking at a recent pg_upgrade thread I happened to notice that the check for
> roles with a pg_ prefix only reports the error, not the roles it found. Other
> similar checks where the user is expected to alter the old cluster typically
> reports the found objects in a textfile. The attached adds reporting to make
> that class of checks consistent (the check for prepared transactions which also
> isn't reporting is different IMO as it doesn't expect ALTER commands).
>
> As this check is only executed against the old cluster the patch removes the
> check when printing the error.
+1. A backpatch would be nice, though not strictly mandatory as
that's not a bug fix.
+ ntups = PQntuples(res);
+ i_rolname = PQfnumber(res, "rolname");
Would it be worth adding the OID on top of the role name in the
generated report? That would be a free meal.
--
Michael
