Re: allow granting CLUSTER, REFRESH MATERIALIZED VIEW, and REINDEX

On Wed, Jun 14, 2023 at 11:17:11AM -0700, Nathan Bossart wrote:
> On Tue, Jun 13, 2023 at 04:54:42PM -0700, Nathan Bossart wrote:
> > On Wed, Jun 14, 2023 at 08:16:15AM +0900, Michael Paquier wrote:
> >> So, yes, agreed about the removal of has_partition_ancestor_privs().
> >> I am adding an open item assigned to you and Jeff.
> >
> > Thanks. I suspect there's more discussion incoming, but I'm hoping to
> > close this item one way or another by 16beta2.
>
> Concretely, I am proposing something like the attached patches.

The result after 0001 is applied is that a couple of
object_ownercheck() calls that existed before ff9618e are removed from
some ACL checks in the REINDEX, CLUSTER and VACUUM paths. Is that OK
for shared relations and shouldn't cluster_is_permitted_for_relation()
include that? vacuum_is_permitted_for_relation() is consistent on
this side.

Here are the paths that now differ:
cluster_rel
get_tables_to_cluster
get_tables_to_cluster_partitioned
RangeVarCallbackForReindexIndex
ReindexMultipleTables

0002 looks OK to retain the skip check for toast relations in the
VACUUM case.
--
Michael