Re: Re: PHPT --SKIPIF--

Joe Stump wrote: >>> Isn't the point of testing to keep each part as a self contained beast ? >>> Anyway I've always believed that tests should be able to run by them self >>> and thus not having the exact problem you are describing ;-) > > Okay, I'll put it another way. Let's say 001-test-a.phpt tests class > Foo. It runs through and finds that Foo isn't behaving as expected. Now, > 002-test-b.phpt tests class Bar, which extends from Foo. Can you tell me > how to make a test that is "a self contained beast" on that use case > (I'd assume using classes, etc. is a somewhat common use case these days). > > My argument is that 002-test-b.phpt isn't in a failed state because it's > not in a position to be tested in the first place. Just let both tests fail :) Every time I change one small thing in PEAR, it breaks 200 tests, I either update all the tests, or change the thing that broke the test (usually change the thing that broke the test). It's a powerful incentive to maintain BC Greg