command-lines tried so far

1. java -classpath jruby-complete.jar;. org.jruby.Main test.rb

    I would expect this to work, but JRuby for whatever reason converts Unicode strings
    to windows-1252 internally, so you get question marks. No surprises there, and I have
    already raised a separate ticket about that anyway.

    Output:
        encoding of the stream:
        encoding of the string: Windows-1252

2. java -Dfile.encoding=UTF-8 -classpath jruby-complete.jar;. org.jruby.Main test.rb

    This is supposed to be the workaround. Now you get non-question mark output, but not UTF-8 either.

    Output:
        encoding of the stream:
        encoding of the string: UTF-8

3. java -Dfile.encoding=UTF-8 -classpath jruby-complete.jar;. org.jruby.Main -Eutf-8 test.rb

    This is the suggested thing to try to see if it fixes it. Exactly the same behaviour as #2.

Test.java

public class Test
{
    public String getValue()
    {
        // forgive the Japanese South Park example...
        return "\u30D1\u30F3\u30C4\u3092\u96C6\u3081\u308B\u7B2C\u4E00\u6BB5\u968E\u30D1\u30F3\u30C4\u306E\u53CE\u96C6\u30FC\uFF01";
    }
}

test.rb

require 'java'

obj = Java::Test.new

File.open('out.txt', 'w') do |io|
  puts "encoding of the stream: #{io.external_encoding}"

  str = obj.value
  puts "encoding of the string: #{str.encoding}"

  io.puts(str)
end

Ok so something significant happened on master since JRuby 1.7.2 was released. I get this to work with:

jruby -J-Dfile.encoding=UTF-8 test.rb

So we fixed some internal encoding issue with Windows since 1.7.2 was released. The file.encoding is required because apparently windows uses Window-1252 by default. I suspect had you been on Windows-1252J this somehow would have been able to transcode these chars to Java Charset.

We (JRuby) really need to consider making JRuby run file.encoding=UTF-8 all the time on windows (at least as a default)....

tks, I spend several hours trying to find the bug, i was on 1.7.1, after 1.7.3 i can save the files with UTF-8 encoding.
tks