On 21/01/2026 1:07 AM, Tom Lane wrote:
Nathan Bossart <[email protected]> writes:
On Tue, Jan 20, 2026 at 04:32:31PM -0500, Tom Lane wrote:
I don't think "let's make select_best_grantor even more magic"
is the right approach. IMO, if there is a GRANTED BY clause,
we should use exactly that grantor and not apply select_best_grantor
at all. This is, for example, certainly the behavior needed for
pg_dump.
I started on something like that here:
https://postgr.es/m/aRYLkTpazxKhnS_w%40nathan
Ah, I wonder if that discussion was lurking in my hindbrain.
I just posted a different take on how to do it in that thread,
but the behavioral change should be the same.
regards, tom lane
Thank you.
Fixing explicit grantor case is definitely the most critical thing.
And I completely agree with your patch.
But I wonder if we do refactoring of this revoke privileges stuff,
should we also provide correct (expected) behaviour in case of missing
grantor specification. i.e.
revoke all privileges on table <T> from <role>;
If privileges to access this table were granted to this role by multiple
grantors, then it is natural to expect that the statement above will
remove all such grants and so as a result <role> can not access this
table any more, rather than try to find best grantor and finally still
leave privileges for this role, isn't it?
