Re: Order changes in PG16 since ICU introduction

On 6/8/23 17:15, Jeff Davis wrote:
> On Wed, 2023-06-07 at 20:52 -0400, Joe Conway wrote:
>> If the provider has no such thing, throw an error.
>
> Just to be clear, that implies that users (and buildfarm members) with
> LANG=C.UTF-8 in their environment would not be able to run a plain
> "initdb -D data"; they'd get an error. It's hard for me to estimate how
> many users might be inconvenienced by that, but it sounds like a risk.

Well, but only if their libc provider does not have C.UTF-8, correct?

I see
----------------
Linux Mint 21.1: /usr/lib/locale/C.utf8
RHEL 8: /usr/lib/locale/C.utf8
RHEL 9: /usr/lib/locale/C.utf8
AL2: /usr/lib/locale/C.utf8

However I do not see it on RHEL 7 :-(

> Perhaps for this specific case, and only in initdb, we change
> C.anything and POSIX.anything to the builtin provider?

Might be best, with some kind of warning perhaps?

> CREATE DATABASE and CREATE COLLATION could still reject such
> locales.

Seems to make sense.

--
Joe Conway
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com