Re: DOM Testing framework?

On Thu, 16 Jul 2009 18:20:13 +0100, Stewart Brodie <stewart.brodie@antplc.com> wrote:
> Kartikaya Gupta <lists.w3@stakface.com> wrote:
> 
> > I've been running the Java versions of the L1 and L2 tests against our
> > implementation as part of our test suite. For the most part they're
> > decent. Some of the tests are incorrect, and they (obviously) don't test
> > absolutely everything, but they make a decent starting point. I can
> > compile a list of the tests that I've discovered to be wrong if that would
> > help.
> 
> I would be interested to see your list.
> 

In addition to the (presumably Java-only) problems I noted on the domts mailing list [1], here are the tests that we have currently commented out for various reasons:

In the DOM L2 Core:
createDocument08 - according to web DOM core passing an empty string for qualifiedName in createDocument() is allowed. see mozilla bug 369092 also.
localName02 - this expects a null return value when it should be non-null

In the DOM L2 HTML:
HTMLAppletElement05 - expects .codeBase to be relative 
HTMLBodyElement02 - expects .background to be relative 
HTMLDocument03 - expects "" instead of null 
HTMLDocument08 - counts objects as applets 
object06 - expects .data to be relative

> We use the JavaScript versions of them and have also found problems with
> some of the tests.  I imported the tests into our source repository and then
> have local changes to fix the problems (so that we pass them all :-)  For
> example, in DOM Level 2 HTML, HTMLDocument03 (document.domain is correctly
> non-empty when you run the tests over HTTP) and HTMLFrameElement09 (the
> expected data is the title of the wrong frame) are both wrong.
> 

What's wrong with HTMLFrameElement09? It references the frame.html file which has a title of "NIST DOM HTML Test - FRAME", which is correct (at least in the Java version).

> I also had to change a couple of the tests due to retrospective changes to
> the spec after the test suite was written.  The main two I can recall right
> now are the test that expects getAttribute() to return an empty string when
> the attribute doesn't exist (the spec was changed to say the return is null
> in this case),

Was it? I thought this was going into web DOM core. It's on the issues list at http://simon.html5.org/specs/web-dom-core

> and the Level 2 Events test that verifies that capturing
> listeners are not triggered during the at-target phase (which is correct,
> but the spec is being changed to match the buggy Firefox behaviour that has
> been replicated into Opera & WebKit (iirc) now, so we've had to follow suit
> as web content is starting to rely on it - this was discussed a few weeks
> ago, either here on on the whatwg mailing list, see archives for details)
> 

We haven't made this change yet and likely won't unless we get complaints about our current behavior.

Cheers,
kats

[1] http://lists.w3.org/Archives/Public/www-dom-ts/2007Nov/0000.html

Received on Thursday, 16 July 2009 20:22:30 UTC