W3C home > Mailing lists > Public > public-webapi@w3.org > March 2006

Window 1.0 test suite

From: Maciej Stachowiak <mjs@apple.com>
Date: Mon, 20 Mar 2006 03:30:31 -0800
Message-Id: <57ECA798-2B7D-4127-868B-93D5431E9CFD@apple.com>
To: "Web APIs WG (public)" <public-webapi@w3.org>

Hi Everyone,

I started on a test suite for Window 1.0. Right now it only covers  
very basic things, but I tried to cover every conformance requirement  
in Section 2, Browsing Contexts, or at least put a comment about the  
need for coverage.

Here's the source:


I made a spiffy test generator that takes the .js source files and  
makes published versions of the test suite in various formats:


And here are generated versions of the tests in HTML, XHTML and SVG  


The HTML, XHTML and SVG versions mostly pass in Opera 9, Firefox 1.5,  
and current development builds of WebKit. Here are some issues I found:

- In Safari/WebKit, the tests relating to the Window being the  
defaultView for the document fail because we didn't make Window and  
AbstractView the same object. I acknowledge that this is a bug in  
WebKit and we plan to fix it.

- In Opera, "window" and "self" are both strict equal (===) to each  
other and equal (==) to "this" at global scope, but not strict equal  
to "this". I think this is a bug in Opera. But maybe the tests and  
spec should be written to say that to be the same object reference,  
it is sufficient to be merely equal and not strictly equal. Thoughts?

- IE fails all of the defaultView tests because it does not appear to  
support DOM 2 Views at all. It also fails a really strange subset of  
the strict equal tests. I am not sure if IE has some alternate way to  
get from a document to its corresponding window. It's pretty  
important to have this capability for CDR scenarios, so I would be  
against dropping it. I think from testing that IE might support this  
via the "parentWindow" property on document. Despite the name and  
what the docs say, it gives the actual window object of the document,  
not the window of the parent frame or anything like that. Should we  
specify parentWindow in addition, or instead of just relying on  
defaultView? Mozilla and Safari don't support it, but Opera does.


I encourage others to examine and contribute to the tests. It's  
probably a bad idea for one of the main authors of the spec to also  
write the test suite, especially if said author is also an  
implementor. It should be simple to add tests, or to add a new test  
template if anyone would like to test UAs for languages besides just  

Also, please test on other implementations. For instance, I'd love to  
hear the results on SVG-only implementations like ASV, or on browsers  
that use layout engines not covered here.

Received on Monday, 20 March 2006 11:30:44 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:16:20 UTC