I have a weird (at least for me) problem.
$rake test
Units: 50 tests, 155 assertions, 0 failures, 0 errors Functionals: 51 tests, 122 assertions, 0 failures, 0 errors Integration: 1 tests, 19 assertions, 0 failures, 0 errors
$rake test:functionals
51 tests, 109 assertions, 3 failures, 0 errors
$ruby test/functional/registrations_controller_test.rb
20 tests, 59 assertions, 3 failures, 0 errors
Why rake test passes everything ok and the other fails?
I imagine this could have something to do with test data. Check that all the require fixtures are included in your controller test. The data may be being loaded in other tests - hence the passes when you run the whole suite.
Steve
Now something weirder happens:
$rake test
# everything OK
$ruby test/functional/registrations_controller_test.rb
# everything OK
$rake test:functionals
# 3 failures
$ruby test/functional/registrations_controller_test.rb
# 3 failures LOL
It's a cache issue or what is happening here? i think its more like a bug...
I solved the problem adding missing fixtures as Steve told me, but hey! wait a second!, this means that i always have to test my functional tests after doing a ""complete"" test?
My conclusion is that some fixtures gets cached somewhere after running my unit tests, that shouldn't happen.
Is this meant to be this way or it's a bug?
You are likely using fixtures, right? If you look at the fixtures loaded for your RegistrationsControllerTest, be sure that all the fixture expectations are listed in that file. I've run into this very problem before. There are some who will recommend all fixtures be listed in the test_helper.rb, but I think that's just an excuse not to stop and think about which fixtures you really *need* for a test.
-Rob
Rob Biedenharn http://agileconsultingllc.com Rob@AgileConsultingLLC.com
Emilio Tagua wrote:
> $ruby test/functional/registrations_controller_test.rb
next, try:
$ruby test/functional/registrations_controller_test.rb -n test_one_specific_test
I solved the problem adding missing fixtures as Steve told me, but hey! wait a second!, this means that i always have to test my functional tests after doing a ""complete"" test?
I interpret this to mean "each time I change any test, must I run all the tests independently to verify they isolate correctly?"
The general answer is no - testing the tests is the road to madness. Your production code never had a bug here, and tests should fail _more_easily_ than production code gets bugs.
However, yes, each time you add a fixture to a test case you risk breaking test isolation with other test cases. A good way to catch this is test with rake test:recent or rake test:svn_modified, each time you change the code. That way, as you change different combinations of test cases, you get to see if they (apparently) isolate from each other, independent of the group. Many a Rails application can no longer run a partial test because only the total test worked when they integrated.
My conclusion is that some fixtures gets cached somewhere after running my unit tests, that shouldn't happen.
Yes, in the database. (-;
Is this meant to be this way or it's a bug?
You are lucky to have test fixtures. Some systems make them so slow, their programmers must avoid database calls in most tests! So any residual optimizations here are generally to your benefit. Perfect test isolation would restart and rebuild the database between each test run. We rely on less than perfect isolation in exchange for speed.