Re: WebIDL test suite time

On Wednesday, April 18, 2012 at 7:18 PM, Boris Zbarsky wrote:

> On 4/18/12 2:12 PM, Marcos Caceres wrote:
> > I guess this really applies to Section 4 in the spec (ECMAScript binding). The rest can't be tested because the spec does not really define WebIDL Parsers as a conformance class (or it's targeted at spec Editors).
> 
> Though one could write an actual parser, throw various WebIDL at it, and 
> see if it's valid....
> 
> That will need to happen anyway, for the whole "two implementations" bit.

 

> > This could be made trivial is the Editor marked up the RFC2119 keywords. For example:
> > 
> > <p>The foo<em class="ct">MUST</em> do something.</p>
> 
> Not exactly. A single MUST can correspond to a long list of testable 
> assertions.

That's fine. It just gives us a hook to go from. Like if it says:

 <p>The foo<em class="ct">MUST</em> run the bar algorithm.</p> 

And the bar algorithm contains 50 steps. At least you have a good collection of conformance requirements and know how to hook them up to algorithms. Naturally, you then need to write the tests for each algorithm. If anything, it saves a bit of time.  

 


-- 
Marcos Caceres

Received on Wednesday, 18 April 2012 18:24:34 UTC