Re: ICU for global collation - Mailing list pgsql-hackers

From Andres Freund
Subject Re: ICU for global collation
Date
Msg-id 20220318000140.vzri3qw3p4aebn5p@alap3.anarazel.de
Whole thread Raw
In response to Re: ICU for global collation  (Peter Eisentraut <peter.eisentraut@enterprisedb.com>)
List pgsql-hackers
Hi,

On 2022-03-17 14:14:52 +0100, Peter Eisentraut wrote:
> On 17.03.22 13:01, Shinoda, Noriyoshi (PN Japan FSIP) wrote:
> > Thank you to all the developers.
> > I found that the description of the pg_database.daticulocale column was not written in the documentation.
> > The attached small patch adds a description of the daticulocale column to catalogs.sgml.
> 
> committed, thanks

Wee! That's a long time weakness addressed...


Just saw a weird failure after rebasing my meson branch ontop of this. Tests
passed on debian, suse, centos 8 stream, fedora rawhide (failed due to an
independent reason), but not on centos 7.


all runs: https://cirrus-ci.com/build/5190538184884224
centos 7: https://cirrus-ci.com/task/4786632883699712?logs=tests_world#L204
centos 7 failure:
https://api.cirrus-ci.com/v1/artifact/task/4786632883699712/log/build/testrun/icu/t/010_database/log/regress_log_010_database

not ok 1 - sort by database default locale
#   Failed test 'sort by database default locale'
#   at /tmp/cirrus-ci-build/src/test/icu/t/010_database.pl line 28.
#          got: 'a
# A
# b
# B'
#     expected: 'A
# a
# B
# b'
ok 2 - sort by explicit collation standard
not ok 3 - sort by explicit collation upper first
#   Failed test 'sort by explicit collation upper first'
#   at /tmp/cirrus-ci-build/src/test/icu/t/010_database.pl line 42.
#          got: 'a
# A
# b
# B'
#     expected: 'A
# a
# B
# b'
ok 4 - ICU locale must be specified for ICU provider: exit code not 0
ok 5 - ICU locale must be specified for ICU provider: error message
1..5

This is a run building with meson. But I've now triggered builds with autoconf
on centos 7 as well and that also failed. See
https://cirrus-ci.com/task/6194007767252992?logs=test_world#L378

So it looks like older ICU versions don't work?

Greetings,

Andres Freund

PS: I had not yet passed with_icu in the initdb tests for meson, that's why
there's two failures with autoconf but only one with meson.



pgsql-hackers by date:

Previous
From: Nathan Bossart
Date:
Subject: Re: XMAX_LOCK_ONLY and XMAX_COMMITTED (fk/multixact code)
Next
From: Masahiko Sawada
Date:
Subject: Re: Skipping logical replication transactions on subscriber side