Compilation problem with the unicode on win32

A problem occurs when compiling the file has a UTF-8 encoded string on win32.
The string name is used as a class name, and it's contained the encoded string
except a standard 7-bit ASCII set and Japanese character set in Japanese environment.

The problem can be reproduced by the following test case.
---------------------------------------------
class Test {
    public static void main(String args[]) {
    }
}

class Test_\u00e9 {
}

class Test_\u3042 {
}
---------------------------------------------

The class "Test_\u3042" is OK, but "Test_\u00e9" is NG.
A javac recognizes "Test_\u00e9" as "Test_?", so the compile error occurs.

kenichi.kurosaki@Japan 1999-10-01