cleanups not called when setUp fails with useFixture

[Cedric Brandily]

When we do:

  fixture1.useFixture(fixture2)

fixture2 cleanups are only called if fixture1.addCleanup(fixture2.cleanUp) is called ... but it's done in Fixture[1] only if fixture2.setUp() succeed.

[1] http://bazaar.launchpad.net/~testing-cabal/python-fixtures/master/view/head:/fixtures/fixture.py

[Cedric Brandily (cbrandily)]

https://code.launchpad.net/~cbrandily/python-fixtures/python-fixtures

[John Schwarz (jschwarz)]

I would like to comment that this sounds like a great idea. We've stumbled upon this issue here: [1], and specifically the discussion about this is here: [2].

I do think that this is quite an API change - developers might not except this change to happen (for example, a cleanUp code that was registered assumes that a setUp has been successful and won't check for it, causing the cleanUp to fail). I wonder if there's a non-invasive way to fix this gradually (maybe add a flag to useFixture that would, over time, become the default?)
This might be off base though, so I'd like to hear thoughts about this from others as well.

[1]: https://review.openstack.org/#/c/181991/
[2]: https://review.openstack.org/#/c/181991/1..4/neutron/tests/fullstack/fullstack_fixtures.py

[Robert Collins (lifeless)]

I wanted to capture the logic here for easy later review..., so I did a blog post. https://rbtcollins.wordpress.com/2015/06/22/revisiting-the-fixture-api-handling-leaky-resources/

This particular bug is invalid: useFixture isn't and shouldn't be responsible for figuring out the right way to deal with a failed fixture: since we need to solve the with: case, there's no need for the useFixture code to be any different to with:, other than that its spread out over two call sites.