[Buildbot-devel] buildbot tests fail on "master"

[Dustin J. Mitchell]

On Wed, Jan 7, 2009 at 12:06 PM, Jean-Paul Calderone <exarkun at divmod.com> wrote:
> How do you know you reordered them right if you don't have passing tests
> to tell you so?  Or do lots of tests which are largely unrelated to the
> changed code fail due to poor separation of concerns?  How do you do any
> further development with all those failing tests?  Why isn't keeping the
> test suite completely passing the top development process concern?
>
> Of course, it's not me doing any of this work, but it's rather discouraging
> to hear that this is how buildbot's future is going to be built.

As your last paragraph suggests you are aware, I resent the tone of your email.

If you look more deeply at the commits in question, you'd see that it
was the tests that were at fault for assuming a particular order.
Thus, there is no problem with the code itself, and none of the
patches I have merged since have affected those test cases.

Test suites can become an end in themselves, and not always with the
best outcomes for a project.  My top concern up until Monday was to
get some concerted hacking done on the one-oh branch, because I had
long stretches of uninterrupted time.  Now that my time is more
fragmented, I will soon find an appropriate length of time to fix
these tests.  We certainly won't see a release in which these tests
are failing.

Dustin

-- 
Storage Software Engineer
http://www.zmanda.com