From: Tom Lane <tgl@sss.pgh.pa.us>
To: Aleksander Alekseev <aleksander@timescale.com>
cc: pgsql-committers@lists.postgresql.org,
        Alexander Korotkov <aekorotkov@gmail.com>,
        Andrew Dunstan <andrew@dunslane.net>
Subject: Re: pgsql: psql: add an optional execution-count limit to \watch.
In-reply-to: 
 <CAJ7c6TOFPoHVSxaQXfbm38_QN1v4qaugZW0LBTB-Gkvk2UwBQw@mail.gmail.com>
References: <E1pkTFl-001hzA-LR@gemulon.postgresql.org>
 <CAPpHfdv+10Uk6FWjsh3+ju7kHYr76LaRXbYayXmrM7FBU-=Hgg@mail.gmail.com>
 <2929174.1680876023@sss.pgh.pa.us>
 <CAPpHfdtgQVRZOwov-Ei4Rh=gQtFE8Hb9uKkO1d_XYmQ2XEXt0g@mail.gmail.com>
 <CAJ7c6TO9KpYYxoVVseWEQB5KtjWDkt8NfyAeKPcHoe2Jq+ykpw@mail.gmail.com>
 <4032363.1681134882@sss.pgh.pa.us>
 <CAJ7c6TOFPoHVSxaQXfbm38_QN1v4qaugZW0LBTB-Gkvk2UwBQw@mail.gmail.com>
Comments: In-reply-to Aleksander Alekseev <aleksander@timescale.com>
	message dated "Mon, 10 Apr 2023 17:30:00 +0300"
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="----- =_aaaaaaaaaa0"
Content-ID: <4076835.1681139300.0@sss.pgh.pa.us>
Date: Mon, 10 Apr 2023 11:09:41 -0400
Message-ID: <4076950.1681139381@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/4076950.1681139381%40sss.pgh.pa.us>
Precedence: bulk

------- =_aaaaaaaaaa0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <4076835.1681139300.1@sss.pgh.pa.us>

Aleksander Alekseev <aleksander@timescale.com> writes:
>> Hmm, so psql is honoring the LC_NUMERIC setting in that environment,
>> but perl isn't.  For me, it appears that adding 'use locale;' to
>> the test script will fix it ... can you confirm if it's OK for you?

> Right, src/bin/psql/t/001_basic.pl has "use locale;" since cd82e5c7
> and it fails nevertheless.
> ...
> So it looks like what happens is LC_ALL overwrites LC_NUMERIC for perl
> but not for psql.

Oh, right, there already is one :-(.  After some more research,
I believe I see the problem: Utils.pm does

BEGIN
{
	# Set to untranslated messages, to be able to compare program output
	# with expected strings.
	delete $ENV{LANGUAGE};
	delete $ENV{LC_ALL};
	$ENV{LC_MESSAGES} = 'C';

Normally, with your settings, LC_ALL=en_US.UTF-8 would dominate
everything.  After removing that from the environment, the child
psql process will honor LC_NUMERIC=ru_RU.UTF-8 and expect \watch's
argument to be "0,01".  However, I bet that perl has already made
its decisions about what its internal locale is, so it still thinks
it should print "0.01".

I am betting that we need to make Utils.pm do

	setlocale(LC_ALL, "");

after the above-quoted bit, else it isn't doing what it is supposed to
if the calling script has already done "use locale;", as indeed
psql/t/001_basic.pl (and a small number of other places) do.

The attached makes check-world pass for me under these conflicting
environment settings, but it's kind of a scary change.  Thoughts?

			regards, tom lane


------- =_aaaaaaaaaa0
Content-Type: text/x-diff; name="fix-TAP-test-locale-settings.patch";
	charset="us-ascii"
Content-ID: <4076835.1681139300.2@sss.pgh.pa.us>
Content-Description: fix-TAP-test-locale-settings.patch
Content-Transfer-Encoding: quoted-printable

diff --git a/src/test/perl/PostgreSQL/Test/Utils.pm b/src/test/perl/Postgr=
eSQL/Test/Utils.pm
index 878e12b15e..4c6cde8974 100644
--- a/src/test/perl/PostgreSQL/Test/Utils.pm
+++ b/src/test/perl/PostgreSQL/Test/Utils.pm
@@ -43,6 +43,7 @@ package PostgreSQL::Test::Utils;
 =

 use strict;
 use warnings;
+use locale;
 =

 use Carp;
 use Config;
@@ -55,6 +56,7 @@ use File::Spec;
 use File::stat qw(stat);
 use File::Temp ();
 use IPC::Run;
+use POSIX qw(locale_h);
 use PostgreSQL::Test::SimpleTee;
 =

 # We need a version of Test::More recent enough to support subtests
@@ -102,6 +104,7 @@ BEGIN
 	delete $ENV{LANGUAGE};
 	delete $ENV{LC_ALL};
 	$ENV{LC_MESSAGES} =3D 'C';
+	setlocale(LC_ALL, "");
 =

 	# This list should be kept in sync with pg_regress.c.
 	my @envkeys =3D qw (

------- =_aaaaaaaaaa0--