Re: Emacs fails to use default locale

[Ken Brown]

On 11/27/2009 10:49 AM, Jan Djärv wrote:
> Ken Brown skrev:
> > According to
> >
> >   http://www.opengroup.org/onlinepubs/007908799/xbd/envvar.html
> >
> > if LC_ALL, LC_CTYPE, and LANG are all either unset or empty, then the 
> > implementation-dependent default locale is supposed to be used.  My 
> > understanding is that this should be done by using setlocale(LC_CTYPE, 
> > "").  But emacs's locale initialization seems to just use the 
> > environment variables, so it doesn't pick up the default locale if the 
> > variables are unset.
>
> setlocale(LC_ALL, "") is the way.  Emacs does that if LC_ALL is unset or 
> not "C".
>
> However, because of internal reasons, Emacs also does:
>   /* The Emacs Lisp reader needs LC_NUMERIC to be "C",
>      so that numbers are read and printed properly for Emacs Lisp.  */
>   setlocale (LC_NUMERIC, "C");
>
> Can you describe in more detail the problem you are seeing?  I.e. what 
> output is wrong?

I'm working in Cygwin 1.7, which has recently adopted a default locale 
that uses UTF-8 encoding.  If I start emacs with none of the locale 
variables set, emacs appears to assume an ASCII character set.  (I don't 
see "U" in the  mode line, for instance.)  Moreover, C-h C returns the 
error message "Unknown button type `help-xref'".

When I referred to emacs's locale initialization in my original post, I 
was referring to the code in lisp/international/mule-cmds.el.  I don't 
really understand how this interacts with the setlocale calls.

Here's one further bit of information:  If I evaluate (getenv "LANG") in 
 the scratch buffer, it returns nil (as expected), but then C-h C 
suddenly works.  And it tells me that the default coding system is nil, 
so it clearly hasn't picked up the system default.

Ken