Deprecate passing block to mock object constructor

[floehopper]

This mechanism is confusing, because the block is evaluated in the
context of the newly instantiated block and not in the context of the
current test.

The mechanism is also of dubious value, because it's possible to use
Object#tap or define stubs/expectations with the mock object as the
explicit receiver without the code becoming much more verbose.

It's unfortunate that the deprecations make the tests more complicated,
but we will be able to simplify them again when this functionality is
removed in a later version.

See #286 for further details.

[floehopper]

@chrisroos: I'd be grateful if you could review this at some point.

[floehopper]

I've rebased against master and force-pushed in the hope of fixing the Travis CI build for this branch.

[floehopper]

Build is now fixed.

[floehopper]

I've now introduced an earlier commit which improves the documentation of the existing behaviour, before that behaviour is then deprecated in the subsequent commit.

@chrisroos: It'd be great if you could have a look at this when you have a chance. I don't think it should take you long. Let me know if you want help on how to generate the documentation...

[chrisroos]

Out of interest, why have you used DeprecationDisabler.disable_deprecations (i.e. the module method) in mock_with_initializer_block_test.rb and disable_deprecations (i.e. the included instance method) in mockery_test.rb?

As an observation, I presume there aren't any tests for this now-deprecated behaviour for stub and stub_everything.

Aside from my question and observation this all looks good to me, @floehopper.

[floehopper]

@chrisroos:

Thanks for reviewing.

Out of interest, why have you used DeprecationDisabler.disable_deprecations (i.e. the module method) in mock_with_initializer_block_test.rb and disable_deprecations (i.e. the included instance method) in mockery_test.rb?

In the former, each test is instantiating a new instance of a test case (via TestRunner#run_as_test) and the DeprecationDisabler` module would need to be included into each instance. So I decided it was simpler to just use the module method in this case. Does that make sense?

As an observation, I presume there aren't any tests for this now-deprecated behaviour for stub and stub_everything.

You are correct. I think I'm OK with this on the basis I know that the behaviour is implemented within Mock#initialize, i.e. I'm using a pragmatic white-box testing approach.

I'm planning to get this merged, so do shout if you have any more comments or questions.