There are a few issues that cropped up in the RBTools test suite,

especially when running on Python 3:

A bad merge made it so we were no longer importing parse_qsl and

  urlparse where they were used.
Some assertions were checking that some data were six.text_type when

  they ought to have been checking for the native string type.
A typo had ItemResourceTests.tearDown calling the superclass'

setUp method instead of tearDown.
hgsubversion tests haven't worked for a while (the test infrastructure

  overrides the .hgrc but the only way to install the hgsubversion

  extension is via .hgrc). Unfortunately some change made it so that

  after the setUpClass was called, if we then skipped the test, we'd end

  up with a bad cwd for the rest of the test suite, resulting in a bunch

  of broken tests. I haven't yet fixed this, but I have improved the way

  that tests are skipped when hgsubversion isn't available, resulting in

  both faster execution and avoiding the cwd problems.
In Python 3, re.split behaves somewhat differently when there's a

* character in the regex. This was resulting in SVN path splitting

  creating a list with every single character split instead of properly

  splitting only on slashes. /+ is a more correct regex for this

  anyway.

Ran unit tests on Python 2 and 3.