``````
Encoding.default_external = Encoding.find(Encoding.locale_charmap)
Encoding.default_internal = __ENCODING__

[STDIN, STDOUT, STDERR].each do |io|
  io.set_encoding(Encoding.default_external, Encoding.default_internal)
end
``````

This solution works. The text is now correctly displayed in Windows console.

But my code crashes whenever Ruby faces a string that contains a
non-CP866 character (like em dash ""):

in `write': U+2014 from UTF-8 to IBM866
(Encoding::UndefinedConversionError)

How do i overcome this error? Not using UTF-8 is not an option.

Iconv has a nice solution: add "//IGNORE" to encoding name and error no
longer takes place. But IO does not recognize "//IGNORE", so this
solution i unacceptable.

I can't use Iconv instead of IO due to the following reason. I use
RubyMine. Unlike Windows console, RubyMine console works in pure UTF-8.
But whenever code run by RubyMine writes to disk, it seems to use
Windows encoding because text gets written in corrupt encoding. :( And i
do not know a way to distinguish RubyMine from Windows console so that
my program could perfrom only the conversions necessary for the current
environment.