Umm.. I have something to add to this..

On 07/11/06, Phlip < [EMAIL PROTECTED]> wrote:

You shouldn't need to run them in any order. Each test case should set
up data, invoke a web browser, browse your site, assert, and shut down
your browser.

Google "Test Isolation" for lectures on why test cases should run in
any order. Your cases should not, for example, drive your website into
a given state, and then rely on that state in the next case. Doing
this is a sign you have not characterized that state.


Bah!  Lectures!  You want to lecture me on how to test!?

I see this style of testing (that you describe) as almost a complete waste of time (for me).  I think that fundamentally we are hitting on a difference in approach between System Testing and Unit/Integration Testing. 

If my focus were on Unit/Integration Testing, I would likely agree with you.  That is, you set up the tests, perform the required steps and checks and then tear them down to a known, good state.  If you do that, why should you care what order you run them in?  However, I am a System/Black-Box tester and during a particular test run, the system gets into fairly complex setups that would be rather stupid to close the web browser between each test.  There is no *reset* between tests.

Just as everyone's definition of what a "test" might be, so we should respect that the order in which the tests need to be executed may be different.  The addition of the feature to specify an order to the test cases is a valuable one that acknowledges that Watir is not only a tool for unit/integration/developer testing.

Paul C.

_______________________________________________
Wtr-general mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/wtr-general

Reply via email to