Re: Getting vendors using W3C tests

On 05/31/2011 01:31 PM, L. David Baron wrote:

> For the second, what Mozilla has done is use a PAC (proxy
> autoconfig) file during many of our tests.  It maps hosts at the
> network layer all to one server.  The mistake we made there was not
> redirecting all other hosts to someplace that wouldn't work.  A
> similar solution seems like it could be useful here, though it's a
> bit more work to do it cross-browser.

That might work for us, although there might be some resistance to 
making all tests go through autoproxy since it isn't quite the standard 
codepath. I will see.

> For the third, there should be a review script that complains when
> tests don't include the necessary script, and it should be run
> regularly and problems reported someplace that's likely to lead to
> their being fixed.  (If we accumulate a collection of such scripts,
> it would be good if it were easy to run them before committing a
> test.)

Simon just suggested adding a simple dependency between testharness.js 
and testharnessreport.js so that testharness.js won't run any tests if 
testharnessreport.js hasn't loaded (e.g. by setting a global variable in 
testharnessreport.js and checking it is set in testharness.js). That 
seems like it should work going forward; the only problem is that a 
patch to simultaneously fix all currently-broken tests will be needed.

Received on Tuesday, 31 May 2011 11:49:37 UTC