we tried ``ghost.py`` for a longer time since it integrates very well in integration tests and it seemed to be slim. unfortunately it was really flaky and we needed to spend *a lot* in fixing time (tests where running on some machines, on some not). i tried nodeunit for two hours, seems nice, but not a browser tool (actually this was the reason i tried it to be fast). but its not in your stack and somehow it was a bit difficult to setup everything dom-interactive.
in the end we use selenium and we are happy with that. it seems like a heavy solution, but its not compared to the effort for other tools. additionally these kind of tests should be used rarely anyway.
what i like about selenium:
- you can develop and debug tests in your favourite browser (setting a breakpoint will not cause any timeouts)
- you'll get it to run everywhere, also on your headless ci-server
- it is predictable and stable
- it integrates also very good into python/pyramid stack
- there are many resources around, even the java explanations are helpful since the api is similar
what i don't like about selenium:
- we still have some situations where the ci-server fails even though the local tests run. after some time you know where to wait actively; but in the end its somehow logic and predictable.
i cannot tell you about ``py.test``, but starting a selenium-test should be possible from anywhere.
hope this helps,
On 26.01.2013, at 17:06, tonthon wrote: