ASCII-8bit encoding errors raises wrong class of exception

[ConradIrwin]

"\xc3".force_encoding('ascii-8bit').encode('utf-8') should raise:
Encoding::UndefinedConversionError: "\xC3" from ASCII-8BIT to UTF-8

However, in @tychobrailleur 's 0b8d74b it was changed to be Encoding::ConverterNotFoundError: code converter not found (ASCII-8BIT to UTF-8).

This broke the mail gem's handling of bad encodings on jruby-1.7.4 in ruby1.9 mode. Compare https://travis-ci.org/mikel/mail/jobs/8490907 and https://travis-ci.org/mikel/mail/jobs/7941638

[BanzaiMan]

Looks like this case was just pulled to RubySpec: rubyspec/rubyspec#146

[headius]

This has improved with the new CharsetTranscoder, but it still raises an incorrect error:

$ jruby -e '"\xc3".force_encoding("ascii-8bit").encode("utf-8")'
Encoding::InvalidByteSequenceError: ""\xC3"" on US-ASCII
  encode at org/jruby/RubyString.java:7619
  (root) at -e:1

This error is raised because the only way I could get Java transcoding logic to reject the high bytes of ASCII-8BIT was to require that they be in US-ASCII range...but that causes it to raise an "invalid" error rather than an "undefined" error.

[headius]

Fixed in 9k.

$ jruby -e '"\xc3".force_encoding("ascii-8bit").encode("utf-8")'
Encoding::UndefinedConversionError: "\xC3" from ASCII-8BIT to UTF-8
      encode at org/jruby/RubyString.java:6231
  __script__ at -e:1