This introduces match_html, which converts actual and expected HTML strings into "canonical" HTML using Nokogiri with NOBLANKS and COMPACT, and then does a simple equality comparison.
This eliminates whitespace differences introduced by library changes (e.g. the libxml2 2.9.0 change).
pretty_text_spec.rb has been changed to use match_html where appropriate. and all tests pass under libxml2 2.7.6, 2.8.0 or 2.9.0