Re: Add --{no-,}bypassrls flags to createuser

On 2022-04-13 17:35, Kyotaro Horiguchi wrote:
> At Wed, 13 Apr 2022 16:10:01 +0900, Michael Paquier
> <michael(at)paquier(dot)xyz> wrote in
>> On Wed, Apr 13, 2022 at 03:46:25PM +0900, Kyotaro Horiguchi wrote:
>> > It is sensible to rig createuser command with full capability of
>> > CREATE ROLE is reasonable.
>> >
>> > Only --replication is added by commit 9b8aff8c19 (2010) since
>> > 8ae0d476a9 (2005). BYPASSRLS and NOBYPASSRLS were introduced by
>> > 491c029dbc (2014) but it seems to have forgotten to add the
>> > corresponding createuser options.
>> >
>> > By a quick search, found a few other CREATE ROLE optinos that are not
>> > supported by createuser.
>>
>> My question is: is BYPASSRLS common enough to justify having a switch
>> to createuser? As the development cycle of 15 has just finished and
>> that we are in feature freeze, you may want to hold on new patches for
>> a bit. The next commit fest is planned for July.
>
> I don't think there's a definitive criteria (other than feasibility)
> for whether each CREATE ROLE option should have the correspondent
> option in the createuser command. I don't see a clear reason why
> createuser command should not have the option.

Thank you for the review!
I created a new patch containing 'VALID UNTIL', 'ADMIN', and 'ROLE'.

To add the ROLE clause, the originally existing --role option
(corresponding to the IN ROLE clause) is changed to the --in-role
option. Would this not be good from a backward compatibility standpoint?

> As far as schedules are concerned, I don't think this has anything to
> do with 15.

I have registered this patch for the July commit fest.

--
Regards,

--
Shinya Kato
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION