When upgrading to Ruby 2.0, a test case started to fail:
expected = "\xD1\x9B\x86"
assert_equal expected, actual
with the following message:
<"ћ\x86"> expected but was
<"\xD1\x9B\x86">.
The actual
variable contains a binary string obtained from an external library call.
The problem is that the default encoding of source files (and therefore string literals) changed in Ruby 2.0 from US-ASCII to UTF-8.