Message-ID: <d4fcacbf4ccd829b5d138590bfe8f6d5e72cced8.camel@j-davis.com>
Subject: Re: Order changes in PG16 since ICU introduction
From: Jeff Davis <pgsql@j-davis.com>
To: Peter Eisentraut <peter@eisentraut.org>, Daniel Verite
	 <daniel@manitou-mail.org>
Cc: Andrew Gierth <andrew@tao11.riddles.org.uk>, Tom Lane
 <tgl@sss.pgh.pa.us>,  pgsql-hackers@lists.postgresql.org
Date: Wed, 07 Jun 2023 16:26:12 -0700
In-Reply-To: <0f03fa26-741c-b1ee-12ab-566116451ab1@eisentraut.org>
References: <543e4ce6-8205-4aea-b54b-12b9ef747bf6@manitou-mail.org>
	 <654a49f7ff7461bcf47be4181430678d45f93858.camel@j-davis.com>
	 <37520ec1ae9591f83132f82dbd625f3fc2d69c16.camel@j-davis.com>
	 <daa9f060aa2349ebc84444515efece49e7b32c5d.camel@j-davis.com>
	 <c840107b-4cb9-c8e9-abb7-1d8c5e0d51df@enterprisedb.com>
	 <a4388fa3acabf7794ac39fdb471ad97eebdfbe11.camel@j-davis.com>
	 <5111ffef640ec7ed5b75d9791dd18d821e96c31e.camel@j-davis.com>
	 <0f03fa26-741c-b1ee-12ab-566116451ab1@eisentraut.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.44.4-0ubuntu1 
MIME-Version: 1.0
Archived-At: 
 <https://www.postgresql.org/message-id/d4fcacbf4ccd829b5d138590bfe8f6d5e72cced8.camel%40j-davis.com>
Precedence: bulk

On Thu, 2023-06-08 at 00:11 +0200, Peter Eisentraut wrote:
> On 05.06.23 19:54, Jeff Davis wrote:
> > New patch series attached.
>=20
> Could you clarify what here is intended for 16 and what is for later?

I apologize about the patch churn here. I implemented several
approaches to see what feedback I get, and now it looks like we're
returning to a previous idea (the "builtin" provider).

In v16:

1. We need LOCALE to apply to all providers.

2. We need LOCALE=3DC to give the memcmp/pg_ascii behavior in all cases
(unless overridden by a separate LC_COLLATE or LC_CTYPE parameter).

Those are the biggest problems raised in this thread, and the patches
to accomplish those things are in scope for v16.

After we sort those out, there are two loose ends:

* What do we do in the case where the environment has LANG=3DC.UTF-8 (as
some buildfarm members do)? Is an error acceptable in that case?

* Do we move icu_validation_level back to ERROR?

Regards,
	Jeff Davis