Encoding::UndefinedConversionError: UTF-8 to US-ASCII regression

[mkristian]

the following script t.rb:

# encoding: utf-8
require 'stringio'
s = StringIO.new
s << 'ā'
puts s.string

gives an Encoding::UndefinedConversionError: ""\xC4\x81"" from UTF-8 to US-ASCII error when executed likes this

$ LANG= jruby t.rb
Encoding::UndefinedConversionError: ""\xC4\x81"" from UTF-8 to US-ASCII
   write at org/jruby/ext/stringio/StringIO.java:1140
      << at org/jruby/ext/stringio/StringIO.java:241
  (root) at t.rb:4

version jruby-1.7.8 and before as well as MRI works like this:

$ LANG= ruby t.rb
ā

the same way as

$ LANG=en_US.UTF-8 jruby t.rb
ā

the problem causes problems with jruby/rubygems-servlets#5 since the behaviour of StringIO#<< seems to be platform dependent in the sense it depends on platform encoding. maybe this platform dependency is a feature - have not looked it up.

[rtyler]

Does the work-around in maven-tools really mean that this issue is fixed for all JRuby users?

[mkristian]

no, only for maven-tools. thanx

[mkristian]

already fixed with jruby-1.7.19 :)