[PATCH 0/1 Conductor] Decoupling specs from deltacloud

Jason Guiditta jguiditt at redhat.com
Fri Jun 29 19:53:41 UTC 2012


This patch is intended to fix only the case listed in $subject,
specifically in the non-bundler setup (though it should certainly
cause far fewer fails with no services running under bundler as well).
Note that I had to make somewhat of a temprary fix due to time
constraints, which is noted as such in the spec_helper.  If you run
specs from system gems before this patch, and turns off the deltacloud
service, you wil get in the vicinity of 180-some-odd failures.  After
the patch you should get 0.  I plan to keep poking at this next week,
but that will miss code freeze.  My plan is to try to sneak in fixes
for the 3 or so remaining fails that are caused when iwhd is not
running, and to see why controllers fail without the above-mentioned
temporary fix in spec_helper.

I noticed a couple major issues that need some though (and hopefully
will at least partly be covered by upcoming arch improvements, pending
list discussion):
* Our code is _really_ intertwined in a number of areas.  This makes
for a tough row to hoe with some of the tests.  When a developer has
to dig through 4-6 different models to figure out what needs to be
stubbed to get a single model's tests passing, that is a _bad
thing(tm)_.  We need better encapsulation of our code.
* A number of our controller tests are actually testing the model
behavior, which is flat out wrong.  Test models in model specs.
Controller specs should _only_ be testing against mock_models and
verifying things like 'method called', variable set, etc.  Otherwise
you are testing business logic in the wrong place.

-j



More information about the aeolus-devel mailing list