Mule bugs: misidentification (Latin-1 vs. Chinese), revert issues

[stephen at xemacs.org]

Michael Sperber writes:

 > When I open a new file with an explicit coding system (say, UTF-8),
 > I sometimes get "UTF8" in the modeline, and sometimes "Auto", which
 > then seems to save in Latin-1.

Ouch!  Is there any chance of a reproducible recipe for the latter?
AFAIK you should never get "Auto" in the modeline; that should always
be replaced with a real coding system ("Auto" is short for
"autodetect").

Are you perchance on Windows?