Re: gettext binaries for Cygwin

[Brian Inglis]

On 2024-12-04 09:21, Bruno Haible wrote:
> Brian Inglis wrote:
> > (and the patch history there), all available in the build repo:
> >
> >          https://cygwin.com/git/?p=git/cygwin-packages/gettext.git
>
> The two patches
>    gettext-0.22.5-gettext-runtime-intl-gnulib-lib-localename-unsafe-c.patch
>    gettext-0.22.5-gettext-tools-gnulib-lib-localename-unsafe-c.patch
> in my understanding have the effect that when a user
>    - has set their language preference in the Windows Control Panel,
>    - has not any LC_* or LANG environment variable set,
>    - uses a Cygwin program that makes use of gettext
>      (with the usual setlocale(LC_ALL,"") in main()),
> no translations will be produced, thus ignoring the Windows Control Panel
> setting.
>
> Is my understanding correct?
>
> If so, why is this patch present? Is it considered to be a feature?

The three patches to {tools,runtime{,/intl}/gnulib/lib/localename_unsafe.c 
remove the condition on Cygwin as locale is always set properly (unless users 
work to do something bad!)

Cygwin Base package base-files installs usual skeleton and default profile 
scripts:

https://cygwin.com/git/?p=cygwin-apps/base-files.git;a=tree;f=etc/defaults/etc/profile.d;hb=HEAD

including:

$ tail /etc/profile.d/lang.sh
# The latest version as installed by the Cygwin Setup program can
# always be found at /etc/defaults/etc/profile.d/lang.sh

# Modifying /etc/profile.d/lang.sh directly will prevent
# setup from updating it.

# System-wide lang.sh file

# if no locale variable is set, indicate terminal charset via LANG
test -z "${_LC_ALL_SET_:-${LC_CTYPE:-$LANG}}" && export LANG=$(/usr/bin/locale 
-uU)

Also /etc/profile.d/cygwin.sh sets up PATH TMP TEMP PRINTER HOME, and some other 
skeletons also do similar setup:

$ grep LANG= /etc/skel/.*
/etc/skel/.mkshrc:#\\builtin export LANG=C LC_CTYPE=$p LC_MEASUREMENT=$p 
LC_MESSAGES=$p LC_PAPER=$p
/etc/skel/.profile:export LANG=$(locale -uU)

$ locale --help
...
Default locale information:

  -i, --input          Print current input locale
  -u, --user           Print locale of user's default UI language
  -s, --system         Print locale of system default UI language
  -f, --format         Print locale of user's regional format settings
                       (time, numeric & monetary)
  -n, --no-unicode     Print system default locale for non-Unicode programs
  -U, --utf            Attach ".UTF-8" to the result
...
$ for o in s u f i n; do export LANG=`locale -${o}U`; \
        echo "locale -$o: $LANG `date`"; done
locale -s: en_US.UTF-8 Wed, Dec  4, 2024 12:05:07 PM
locale -u: en_GB.UTF-8 04 Dec 2024 12:05:07
locale -f: en_CA.UTF-8 2024 Dec 04 Wed 12:05:07
locale -i: en_CA.UTF-8 2024 Dec 04 Wed 12:05:07
locale -n: en_CA.UTF-8 2024 Dec 04 Wed 12:05:07

Although there was no match in the default skeleton search above, I personally 
use `locale -fU` to set my *Regional* settings, and have argued for "-f" to be 
used instead of "-u":

$ grep -C1 LANG ~/.bash_profile
# language
export LANGUAGE=${LANGUAGE:-'en_CA:en_GB:en'}

--
/usr/bin/locale -fU > /dev/null 2>&1  \
        && LANG=$(/usr/bin/locale -fU)  \
        || LANG=$(/usr/bin/locale |     \
           /bin/sed -e '/^LANG=\|^LC_CTYPE=\|^LC_ALL=/{s///;h};$!d;x;s/"//g')
[ "${LANG:0:2}" = "en" ] || LANG=en_CA.UTF-8
export LANG

Also for time zones Cygwin tzset builds and uses a table based on upstream CLDR:

        /usr/share/unicode/cldr/common/supplemental/windowsZones.xml

to convert from MS Windows to IANA TZ DB time zones.

$ tail /etc/profile.d/tzset.sh
# Modifying /etc/profile.d/tzset.sh directly will prevent
# setup from updating it.

# System-wide tzset.sh file

#Uses the geographical location setting of the user to find the right
#mapping, rather than the locale setting.  Only on Windows 2000 which
#doesn't know about the user's geographical location, or if fetching
#the geographical location fails, it falls back to the user's locale.
test -z "$TZ" && export TZ=$(/usr/bin/tzset)

$ [ -n "$TZ" ] && zdump
zdump -Vc2024,2025 America/Edmonton
America/Edmonton  Sun Mar 10 08:59:59 2024 UT = Sun Mar 10 01:59:59 2024 MST 
isdst=0 gmtoff=-25200
America/Edmonton  Sun Mar 10 09:00:00 2024 UT = Sun Mar 10 03:00:00 2024 MDT 
isdst=1 gmtoff=-21600
America/Edmonton  Sun Nov  3 07:59:59 2024 UT = Sun Nov  3 01:59:59 2024 MDT 
isdst=1 gmtoff=-21600
America/Edmonton  Sun Nov  3 08:00:00 2024 UT = Sun Nov  3 01:00:00 2024 MST 
isdst=0 gmtoff=-25200

--
Take care. Thanks, Brian Inglis              Calgary, Alberta, Canada

La perfection est atteinte                   Perfection is achieved
non pas lorsqu'il n'y a plus rien à ajouter  not when there is no more to add
mais lorsqu'il n'y a plus rien à retirer     but when there is no more to cut
                                -- Antoine de Saint-Exupéry