provide dependency marks

[thirtytwobits]

allow tests to declare ordering dependencies on other tests. For example, if test 'a' assumes you've modified the state of a fixture in 'b', require that test 'b' was run in the current session and was successful if a request to run test 'a' is found. This should be handled both by ensuring correct ordering in the pytest_collection_modifyitems hook and by ensuring that filters do not trim require prerequisites.

While one may argue that each test should work in isolation some hardware tests can take a very long time to run. Repeating such steps to ensure the attached hardware is in the correct state can become onerous.

Here's a hypothetical example: a linux computer requires that an OS image is freshly installed. Subsequent tests may pollute the installed environment so each test session starts with a fresh image. Reinstalling linux for every test would (in our hypothetical example) add hours to a test suite that would otherwise complete in 30 minutes. Instead we allow each test to assert something like @pytest.mark.after_success_of('test_install_linux')