The OutputStream implementation returned by RubyIO.getOutputStream always treats incoming bytes as being ASCII-8BIT encoding. This breaks writes when the underlying RubyIO has been initialized with an incompatible external encoding, like UTF-8.

This led to one failure in the Psych tests as noted in ruby/psych#481 and fixed by #6587.

I believe this stream should handle the encoding better using one of these options (perhaps?):

• Assume the bytes are encoded like the IO's internal encoding, which may be the default UTF-8 of Ruby. The bytes will be transcoded to whatever the external encoding is supposed to be.
• Assume the bytes are encoded like the IO's external encoding. They will be written directly without any transcoding (or any verification that they are in that encoding).