Encoding issue when using JRuby

[samflores]

I'm facing an encoding compatibility error when trying to use faye-websocket with Goliath server in JRuby 1.7.3 (1.9 mode)

Encoding::CompatibilityError: incompatible character encodings: ASCII-8BIT and UTF-8
                    + at org/jruby/RubyString.java:1113
  handshake_signature at /home/samuel/.gem/jruby/1.9.3/gems/faye-websocket-0.4.7-java/lib/faye/websocket/draft76_parser.rb:35
                parse at /home/samuel/.gem/jruby/1.9.3/gems/faye-websocket-0.4.7-java/lib/faye/websocket/draft76_parser.rb:46
                parse at /home/samuel/.gem/jruby/1.9.3/gems/faye-websocket-0.4.7-java/lib/faye/websocket.rb:172
             __send__ at org/jruby/RubyBasicObject.java:1683
              receive at /home/samuel/.gem/jruby/1.9.3/gems/faye-websocket-0.4.7-java/lib/faye/websocket.rb:202
         receive_data at /home/samuel/.gem/jruby/1.9.3/gems/faye-websocket-0.4.7-java/lib/faye/adapters/goliath.rb:10
       event_callback at /home/samuel/.gem/jruby/1.9.3/gems/eventmachine-1.0.3-java/lib/eventmachine.rb:1484
        eventCallback at /home/samuel/.gem/jruby/1.9.3/gems/eventmachine-1.0.3-java/lib/jeventmachine.rb:92
          run_machine at /home/samuel/.gem/jruby/1.9.3/gems/eventmachine-1.0.3-java/lib/jeventmachine.rb:111
                  run at /home/samuel/.gem/jruby/1.9.3/gems/eventmachine-1.0.3-java/lib/eventmachine.rb:187
            synchrony at /home/samuel/.gem/jruby/1.9.3/gems/em-synchrony-1.0.3/lib/em-synchrony.rb:38
                start at /home/samuel/.gem/jruby/1.9.3/gems/goliath-1.0.1/lib/goliath/server.rb:73
           run_server at /home/samuel/.gem/jruby/1.9.3/gems/goliath-1.0.1/lib/goliath/runner.rb:296
                  run at /home/samuel/.gem/jruby/1.9.3/gems/goliath-1.0.1/lib/goliath/runner.rb:221
                 run! at /home/samuel/.gem/jruby/1.9.3/gems/goliath-1.0.1/lib/goliath/application.rb:111
              Goliath at /home/samuel/.gem/jruby/1.9.3/gems/goliath-1.0.1/lib/goliath/application.rb:129

[jcoglan]

This appears to be a bug in the http_parser gem, which is used by goliath. Take the following script: it contains the raw bytes of a hixie-76 handshake request, turns them into a string, and tells Http::Parser to parse them:

require 'http/parser'

bytes = [ 71, 69, 84, 32, 47, 32, 72, 84, 84, 80, 47, 49, 46, 49, 13, 10, 85,
          112, 103, 114, 97, 100, 101, 58, 32, 87, 101, 98, 83, 111, 99, 107,
          101, 116, 13, 10, 67, 111, 110, 110, 101, 99, 116, 105, 111, 110, 58,
          32, 85, 112, 103, 114, 97, 100, 101, 13, 10, 72, 111, 115, 116, 58,
          32, 108, 111, 99, 97, 108, 104, 111, 115, 116, 58, 55, 48, 48, 49,
          13, 10, 79, 114, 105, 103, 105, 110, 58, 32, 104, 116, 116, 112, 58,
          47, 47, 108, 111, 99, 97, 108, 104, 111, 115, 116, 58, 55, 48, 48,
          49, 13, 10, 83, 101, 99, 45, 87, 101, 98, 83, 111, 99, 107, 101, 116,
          45, 75, 101, 121, 49, 58, 32, 49, 57, 32, 111, 49, 55, 32, 54, 50,
          32, 52, 32, 53, 50, 13, 10, 83, 101, 99, 45, 87, 101, 98, 83, 111,
          99, 107, 101, 116, 45, 75, 101, 121, 50, 58, 32, 79, 52, 32, 48, 97,
          103, 39, 53, 121, 57, 32, 32, 72, 32, 54, 60, 32, 54, 101, 32, 57,
          49, 50, 13, 10, 13, 10,
          # these bytes are the upgrade data:
          241, 188, 139, 76, 196, 155, 103, 64 ]

string = bytes.pack('C*')
parser = Http::Parser.new

parser << string
p [parser.upgrade?, parser.upgrade_data.bytes.to_a]

This script prints the correct upgrade data on MRI, but not on JRuby:

$ rbenv shell 1.9.3-p392 
$ ruby hixie_76.rb 
[true, [241, 188, 139, 76, 196, 155, 103, 64]]

$ rbenv shell jruby-1.7.2 
$ ruby hixie_76.rb 
[true, [239, 191, 189, 76, 196, 155, 103, 64]]

[jcoglan]

I've filed a bug with Http::Parser: tmm1/http_parser.rb#23

[samflores]

Thanks a lot! ❤️

[jcoglan]

I'm closing this issue since the bug has been shown to originate in another library. Please let me know if this proves counterproductive and we need to work around the bug in this library.