Coding system in compilation buffer

[Enno Fennema]

I am using GNU Emacs 22.1.1 part of SuSE Linux 10.3 x86_64.
When doing a compile gcc writes its messages/warnings using utf-8 which 
does not help readability when the buffers assumes latin-1.

According to the documentation compilation-process-setup-function is 
called with compilation-buffer bound. I think that variable is actually 
called outbuf in compile.el That would be a documentation bug, however 
minor; the rest is a question.

I tried emacs initialization with:
>>
(defun compile-utf ()
  "Set compilation buffer for UTF-8"
  ( save-current-buffer
    (set-buffer outbuf)
    (set-buffer-process-coding-system "utf-8" "utf-8")))

(setq compilation-process-setup-function 'compile-utf)
<<

That doesn't work either complaining that there is no process yet.

I don't want to set a global utf-8 preference as normally I prefer
latin-1, which works fine for a West-European (for years I worked with 
ASCII only).

Does the above idea make sense, if so how do I make it work, or is there 
an alternative better approach?

Regards,
Enno Fennema