file name encodings

[Tom Tromey]

Right now libgcj assumes that strings used to interact with the system
are Latin-1 in some situations (e.g., command-line arguments), and
Unicode in others (e.g., file names are converted to utf-8).

Doesn't this seem wrong?  Instead it seems to me that we should assume
that interaction with the system (on Unix boxes; things differ
elsewhere) should be done using the charset of the current locale.

Any comments on this?  If this is the right thing to do I'll submit a
PR (I can't work on it in the near future).

I guess it would make to see what Classpath does, too.

Tom