Re: [Fwd: Re: Re: Request for review of XHR test Pull Request #91 [Was: Re: Intel submission: XHR timeout tests]]

On 06/04/2013 11:43 AM, Hallvord Reiar M. Steen wrote:
> On Fri, 17 May 2013 15:28:15 +0200, Dominik Röttsches
> <dominik.rottsches@intel.com> wrote:
>
>     as you might have seen, I updated the pull request with your
>     comments hopefully satisfactorily addressed.
>     https://github.com/w3c/web-platform-tests/pull/91
>
>
> Hi Dominik,
> I've merged the tests in but - just FYI - I plan to make some changes to
> them.
>
> One issue I'm wondering about is why
> http://w3c-test.org/web-platform-tests/submissions/91/XMLHttpRequest/xmlhttprequest-timeout-runner.js
>
> contains this code:
>
> // Abort test execution if an individual test case fails.
> add_result_callback(function (t) {
> if (t.status == t.FAIL)
> done();
> });

FWIW I think this pattern is very harmful and should be on a list of 
things that reviewers are expected to raise issues on. If you really 
have some state that means that running test B when test A failed is 
bad, just have a global flag that gets set when a test fails and make 
the first assertion in each test assert_true(prerequisites_passed). Not 
easy, but having shared state between tests isn't something that we want 
to encourage.

Received on Tuesday, 4 June 2013 09:59:01 UTC