Michael Paquier <michael@paquier.xyz> writes:
> On Wed, Mar 21, 2018 at 01:40:23AM -0400, Tom Lane wrote:
>> I don't mind making it an ereport, but I think it needs to be FATAL
>> for the reason stated in the comment.
> Okay for the FATAL. I can see that at this time of the day your patch
> 0002 has already been pushed as 846b5a5 with an elog(). Still, it seems
> to me that this is not an internal error but an error caused by an
> external cause which can be user-visible, so I would push forward with
> switching it to an ereport().
Well, after some consideration I pushed it like that because it's not
really a user-facing error but a developer-facing error. Should we ask
translators to spend time on that message? I doubt it. Also, the
adjacent test to refuse PGC_POSTMASTER variables is just an elog;
that seems like pretty much the same sort of situation, and nobody's
complained about it.
If we do something here we should change both of those calls, but
I really doubt it's worth the effort.
regards, tom lane