On Sat, 2009-04-04 at 22:09 +0000, Antoine Pitrou wrote:
> Antoine Pitrou <pitrou@free.fr> added the comment:
> 
> teardown
> 
> Why should they? It's only an implementation choice, and not a wise one
> I would say (precisely because people are used to the fact that the
> standard tearDown() method does nothing, and doesn't need to be called).
> 
> I explained my proposal in terms of actual use cases, but I don't see
> any actual use case of addCleanup() in your argument.

I was arguing by analogy: if we were to implement addCleanup as
something called at the end of the base class tearDown, then it would
clearly support LIFO tearing down of anything people have done [except
that we can't rely on them having called the base tearDown]. The next
best thing then is to call it from run() after calling tearDown.

If you want a worked example:
---
class TestSample(TestCase):

    def setUp(self):
        dirname = mkdtemp()
        self.addCleanup(shutils.rmtree, dirname, ignore_errors=True)
        db = make_db(dirname)
        self.addCleanup(db.tearDown)

    ....
---

This depends on db being torn down before the rmtree, or else the db
teardown will blow up (and it must be torn down to release locks
correctly on windows).

-Rob