We've got a similar problem, but the other way around. We pull proper UTF-8 data from a database, generate some XML out of it, and write it to standard out. The glassfish log (viewed from Netbeans) contains the ? placeholder in Windows, but the correct symbol (pi) when repeated in Mac OS/X.
I would love to pin down this character encoding to use UTF-8 on all systems. We develop on both Windows and OS/X, and deploy on Linux. It's a definite pain to have to deal with character set encoding differently in each environment.
Brian
[Message sent by forum member 'bleathem' ]
http://forums.java.net/jive/thread.jspa?messageID=373168