[bind10-dev] release candidate progress

[JINMEI Tatuya / 神明達哉]

At Wed, 13 Feb 2013 21:50:01 +0530,
Mukund Sivaraman <muks at isc.org> wrote:

> > BTW, to be clear, it's not "additional tests": the new code had mostly
> > no tests, much less developed test-driven, so I simply asked for
> > "tests for the newly added code".
> 
> None of the code in cmdctl and bindctl (related to the login, and also
> the code that both calls out from bindctl and handles the API call in
> cmdctl) is tested.
> 
> One could throw something together as a test, but it is better if the HTTP
> API code is comprehensively tested. I proposed that this be implemented
> in a separate ticket (similar to the bug that you created and I fixed to
> add comprehensive tests for BoB -> #2353) for all of this API. It
> doesn't practically make a whole lot of difference if there was a test
> for the "/users-exist" API call alone, when nothing else is tested. For
> the POST calls, it would be better to write common test code for all the
> calls.
> 
> For the code introduced unrelated to the HTTP API, a test was added. So
> any new tests are additional tests. Does it have to be pointed out it
> was not developed test-driven when I have said as much on the ticket
> along with the reason for it?

In its current form of the branch, for example, it doesn't confirm if
the request is '/users-exist' then the existence of the users is
really checked.  And that should be testable even without handling
lower level APIs with minor refactoring.  I don't necessarily think an
explicit note whether it's test driven or not, but my point is that we
tend to easily conclude "we don't (can't) test this" but the bar
should be actually very high; humans are lazy, so we need some tighter
discipline to overcome the temptation.

---
JINMEI, Tatuya