Patch: gcj -vs- iconv

Zack Weinberg
Mon Mar 6 16:37:00 GMT 2000

On Mon, Mar 06, 2000 at 03:07:45PM -0800, Tom Tromey wrote:
> Zack> setlocale (LC_CTYPE, ""); 
> Zack> cset = nl_langinfo (CODESET);
> Zack> might give you what you want.  nl_langinfo is in <langinfo.h>.
> Do you mean _NL_CTYPE_CODESET_NAME?  That is in my langinfo.h, but it
> isn't documented in my glibc info manual.

My langinfo.h has this:

#define CODESET                 CODESET

I'm betting that CODESET is the standardized name, while
_NL_CTYPE_CODESET_NAME is the internal name.  I don't think you're
supposed to use the values that start with _NL_.  (Ulrich?)

Solaris langinfo.h has CODESET but not _NL_CTYPE_CODESET_NAME.

Also, from testing on Solaris, you need to deal with nl_langinfo
returning an empty string, or NULL.

> Anyway, thanks.  I'll update my patch to do this.


> Zack> I think there ought to be a way to specify the encoding of a
> Zack> file *inside* the file.  How does Mule do it?
> I don't know how Mule does it.  For Java this would have to be in a
> special comment or something like that.  Even then I don't see how it
> can work unless you pick a default encoding, like ASCII.

Mule extends the standard Emacs magic comment.

  /* -*- mode: java; coding: latin-1 -*- */

would force it to treat the file as ISO8859.1.  Of course, you need to
map those names to iconv names...

If it doesn't find such a comment, yes, it has a series of defaults to
fall back on.


More information about the Gcc-patches mailing list