Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
From: Andrew Dunstan <andrew@dunslane.net>
Mime-Version: 1.0 (1.0)
Subject: Re: pgsql: Add standard collation UNICODE
Date: Sat, 11 Mar 2023 17:42:59 -0500
Message-Id: <8CA29D75-B231-4CE2-AB44-B8673DD7E481@dunslane.net>
References: <20230311220515.x45vln6ogzyd57p7@awork3.anarazel.de>
Cc: Jeff Davis <pgsql@j-davis.com>,
 Peter Eisentraut <peter@eisentraut.org>,
 pgsql-committers@lists.postgresql.org
In-Reply-To: <20230311220515.x45vln6ogzyd57p7@awork3.anarazel.de>
To: Andres Freund <andres@anarazel.de>
Archived-At: 
 <https://www.postgresql.org/message-id/8CA29D75-B231-4CE2-AB44-B8673DD7E481%40dunslane.net>
Precedence: bulk


> On Mar 11, 2023, at 5:05 PM, Andres Freund <andres@anarazel.de> wrote:
>=20
> =EF=BB=BFHi,
>=20
>> On 2023-03-11 10:20:33 -0800, Jeff Davis wrote:
>> On Sat, 2023-03-11 at 08:31 -0500, Andrew Dunstan wrote:
>>>>=20
>>> It's fairly old:
>>>=20
>>> $ rpm -q -a | grep icu
>>>  libicu-50.2-4.0.amzn1.x86_64
>>>  libicu-devel-50.2-4.0.amzn1.x86_64
>>=20
>> FWIW I tried ICU 50.2 (built from sources) and the root locale still
>> sorts lowercase first.
>=20
> On one my hacking-on-meson-build trees ([1]) I saw the problem in CI, on
> centos-7. In case you want a way to reproduce the issue. You can merge tha=
t
> tree into a throway tree, none of the changes should affect your work. You=
 can
> even log into the machine if it's on your repo ("re-run with terminal
> access"), if you want to figure out the details.
>=20
> https://cirrus-ci.com/task/6015098225950720
> https://api.cirrus-ci.com/v1/artifact/task/6015098225950720/testrun/build/=
testrun/regress/regress/regression.diffs
>=20
> Centos 8, fedora rawhide, suse tumbleweed are all OK though.


I have a Centos 7 setup I can play with tomorrow.

Cheers=20

Andrew