Starting with 2.14.0, the tokenizer in libxml2's HTML parser conforms to
HTML5 which means that some documents will be parsed differently. The
following changes affect the lxml test suite:
- <title> is an RCDATA element, meaning that it can only be closed
  with a "</title>" string. This is easy to fix by closing the element
  correctly. Unfortunately, libxml2's old behavior is used to document
  and test how the parser previously recovered from malformed HTML.
- Newlines are normalized now and U+000D CARRIAGE RETURN (CR) is
  stripped from the input.
- ASCII control chars are now allowed.
- Processing instructions are parsed as bogus comments.
Instead of trying to check for different libxml2 versions, I simply
deleted parts of some tests.

scoder pushed a commit that referenced this pull request

Starting with 2.14.0, the tokenizer in libxml2's HTML parser conforms to
HTML5 which means that some documents will be parsed differently. The
following changes affect the lxml test suite:
- <title> is an RCDATA element, meaning that it can only be closed
  with a "</title>" string. This is easy to fix by closing the element
  correctly. Unfortunately, libxml2's old behavior is used to document
  and test how the parser previously recovered from malformed HTML.
- Newlines are normalized now and U+000D CARRIAGE RETURN (CR) is
  stripped from the input.
- ASCII control chars are now allowed.
- Processing instructions are parsed as bogus comments.
Instead of trying to check for different libxml2 versions, I simply
deleted parts of some tests.

Merged