[Product-Developers] Functional testing with plone.app.testing and isolation

[Martin Aspeli]

On 14 July 2011 14:06, Rafael Oliveira <rafaelbco at gmail.com> wrote:
> Hi,
>
> I'm using plone.app.testing to do some functional testing using
> zope.testbrowser. According to the PyPI documentation of p.a.testing I
> have to call transaction.commit() whenever I add content to the Plone
> Site and want to inspect it in the testbrowser. However, when I do
> this it seems the changes I make to ZODB are not cleaned up when the
> test layer is thorn down.
>
> Example: I create a content item in a test and call
> transaction.commit(). In another test, of another package, belonging
> to a different layer, I can retrieve the same content item. All layers
> and tests are wrote in the standard way prescribed in p.a.testing
> documentation on PyPI.
>
> Is this the expected behavior or am I doing something wrong ? Do I
> have to manualy undo every change to ZODB if I use
> transaction.commit() ?

You shouldn't have to.

> Code: http://dev.plone.org/collective/browser/collective.cmfeditionsdexteritycompat/trunk/collective/cmfeditionsdexteritycompat

I wonder if this may be a sequencing issue: Instead of doing
transaction.commit() in setUp(), try to do it in the test_* method
before the call to browser.open().

It's also useful to look at the test runner output to see which order
it's setting up the layers in. The FunctionalTesting() layer basically
stacks a DemoStorage for each and every test (in its testSetUp()
method) and pops it in testTearDown(), which is how we get isolation.
If your own setUp() or any other transaction commit is running before
the stacking, you'll be committing to the underlying DemoStorage,
which would then be shared between tests.

Martin