fail. The change had to do with the way setuptools was installed.
This seems to explain the changed test output, which is incidental to
what's tested. This test *was* only run for Python 2.5.

I've changed the test to run whenever Python 2.4 isn't used and
updated the test to reflect the new output.