On Sat, Sep 17, 2011 at 05:49:54PM -0400, Ken Brown wrote:
> On 9/17/2011 4:40 PM, David Sastre wrote:
> >On Tue, Sep 13, 2011 at 09:45:37AM -0600, Eric Blake wrote:
> >>On 09/09/2011 08:59 AM, Corinna Vinschen wrote:
> >>>On Sep  9 13:33, Andy Koppe wrote:
> >>>>The 'C.UTF-8' default locale is not a bug, it was a deliberate design decision.
> >>>
> >>>Exactly.  And it has been discussed a lot on the cygwin-apps mailing
> >>>list.
> >>>
> >>>And above all, there *is* an official way for the user to align the
> >>>Cygwin locale with the Windows locale, see the -s and -u options
> >>>of the locale(1) command:
> >>>
> >>>   http://cygwin.com/cygwin-ug-net/using-utils.html#locale
> >>
> >>On 09/09/2011 09:09 AM, Corinna Vinschen wrote:
> >>>>OK, then the following four facilities are needed in Cygwin.
> >>>>
> >>>>1) We need the name of the locale which is in effect when the user has
> >>>>     not specified environment variables.
> >>>
> >>>In Fedora, for instance, the fallback is what is set as system default
> >>>in /etc/sysconfig/i18n.
> >>>
> >>>In Cygwin the fallback is the system default set in
> >>/etc/profile.d/lang.sh
> >>>or /etc/profile.d/lang.csh.
> >>>
> >>>Why should libintl use anything else on Cygwin, but not on Linux?
> >>>
> >>
> >>Given this, I think the bug is in cygwin for having base files
> >>/etc/profile.d/lang.{sh,csh} which hardcode LANG to C.UTF-8 instead
> >>of using locale -s -u to default LANG to the preferred Windows
> >>settings. Libintl should NOT be second-guessing an explicit setting
> >>of LANG, but cygwin should NOT be explicitly setting LANG to C.UTF-8
> >>in its default startup scripts without any regards to the Windows
> >>settings.  Whether setlocale(LC_ALL,"") returns C.UTF-8 or a
> >>Windows-appropriate string _when LANG is undefined_ is still worth
> >>solving, but right now, an out-of-the-box cygwin installation
> >>_always has LANG defined_ by the default startup scripts.  So our
> >>first focus should be to get that setting of LANG fixed to honor
> >>Windows, and to teach libintl that when LANG is set we really meant
> >>it.
> >
> >WRT the base-files package, would it be acceptable/does it make sense to set:
> >
> >test -z "${LC_ALL:-${LC_CTYPE:-$LANG}}"&&  export LANG=${locale -sU}
> >
> >in /etc/profile.d/lang.sh and
> >
> >if ( $?LC_ALL == 0&&  $?LC_CTYPE == 0&&  $?LANG == 0 ) setenv LANG = `locale -sU`
> >
> >in /etc/profile.d/lang.csh, both as proposed, _and_ a (possibly) commented-out
> >
> >test -z "${LC_ALL:-${LC_CTYPE:-$LANG}}"&&  export LANG=${locale -uU}
> >
> >in the skeletal .bash_profile and .profile (i.e. both system-wide and
> >user defined settings)?
> 
> If you want the user-defined setting to take effect, wouldn't you
> have to omit the `test -z ...'?  LANG will already be set when
> .bash_profile is processed.

Indeed, excuse the fast copy-paste. The proposal still stands.

-- 
Huella de clave primaria: AD8F BDC0 5A2C FD5F A179  60E7 F79B AB04 5299 EC56