Re: UTF-8 and M-x shell?

[Kai Grossjohann <kai@emptydomain.de>]

Daniel Pfeiffer <occitan@esperanto.org> writes:

> I'm migrating my Gnu/Linux to LANG=de_DE.UTF-8 and finding that I need to
> recode my old Latin-9 filenames.
>
> But now Gnu Emacs (2-3 weeks old CVS) M-x shell gets problems.
> First off, the buffer still comes up in Latin-1.  That can
> apparently be fixed with (set-language-environment "utf-8"), but
> only interactively, not in .emacs.el!

Strange.  What happens if you put (set-language-environment "utf-8")
into ~/.emacs?  Is there an error message?

> But it doesn't help:
>
> - M-! umlaut-emitting-command displays old (1-byte) umlauts as octal
> and utf-8 correctly

That's how it is supposed to be.

> - but with the same commands in the shell buffer, with mode-line u:, it still
> displays everything as one-byte.

What does M-x describe-coding-system RET RET say when in the shell
buffer?  The codings for process i/o should be utf-8.  As a
workaround, you could try M-: (set-process-coding-system nil 'utf-8
'utf-8) RET from the *shell* buffer.

> What must I do for M-x shell to respect $LANG, like an xterm?
>
> (Btw. before I had LANG=de_DE@euro, which is Latin-9, but Emacs wrongly
> defaulted to Latin-1 -- which luckily is quite similar.)

That's also strange -- my Emacs defaulted to Latin-9 when I did this.
Maybe it was a font problem?  I had set "Emacs.font:
yadda-yadda-iso8859-15" originally, but had to revert to "Emacs.font:
yadda-yadda-iso8859-1" to get the right behavior.

Kai