The scripts should be run with py.test with installed plugin pytest-regest (see https://pypi.python.org/pypi/pytest-regtest).

The plugin allows recording of approved output so that later test runs will check if the output is still the same. It is simple to use as you can see in test_via_regression.py.

In order to record output you have to use the regtest fixture like in the following example. This regtest behaves like a file handle, so you can write to it as usual:

    def test_0(regtest):
        print >> regtest, "this is the recorded output"

If you now create a new test function test_0 in a file test_xyz.py, first run

    $ py.test tests/test_xyz.py::test_0

which will show you the yet not approved output. You can approve this output using

    $ py.test --regtest-reset tests/test_xyz.py::test_0

Which will create a file in tests/_regtest_outputs/test_xyz.test_0.out which you should not forget to commit with git.

Later runs like

    $ py.test tests/test_xyz.py

will then check if the recorded output is still the same.