Re: ALTER TABLE on system catalogs

On 28.06.18 10:14, Peter Eisentraut wrote:
> On 6/28/18 01:10, Michael Paquier wrote:
>> On Wed, Jun 27, 2018 at 01:37:33PM -0700, Andres Freund wrote:
>>> On 2018-06-27 22:31:30 +0200, Peter Eisentraut wrote:
>>>> I propose that we instead silently ignore attempts to add TOAST tables
>>>> to shared and system catalogs after bootstrapping.
>>>
>>> That seems like an extremely bad idea to me. I'd rather go ahead and
>>> just add the necessary toast tables than silently violate preconditions
>>> that code assumes are fulfilled.
>>
>> Agreed. Joe Conway was working on a patch to do exactly that. I was
>> personally looking for the possibility of having one with pg_authid in
>> v12 :)
>
> OK, that would change things a bit, in that the silent addition of a
> TOAST table would no longer be a problem, but it wouldn't fix the other
> scenarios that end up in an error. If such a patch is forthcoming, we
> can revisit this again afterwards.

After reviewing that thread, I think my patch would still be relevant.
Because the pending proposal is to not add TOAST tables to some catalogs
such as pg_attribute, so my original use case of allowing ALTER TABLE /
SET on system catalogs would still be broken for those tables.

--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services