Re: Minutes of the framework discussion

Philippe Le Hegaret wrote:
> Framework license
> 
> Philippe: JSUnit is not under the W3C Software license. Got feedback
> that this may be a problem for compagnies in order for them to look at
> it and fix it.
> 
> Bob: originally, Edward was using GPL. He is now using LGPL/Mozilla
> Public license/GPL.  Edward is quite open on that.
> 
> Johnny: it may not satisfy those companies if it's under several licences.
> 
> Philippe: can talk to them again and see if they can be satisfied.
> 
> ACTION: Dimitris will follow up on the licence and if something can be
> done to satisfy those companies (but they may have conflicting issues).

It should be possible to run the ECMAScript tests on a different or 
custom test framework with an adaption layer and little no changes to 
the generated tests.  The Java tests initially supported multiple test 
frameworks (JUnit and Avalon), but since Avalon was abandoned by ASF 
there hasn't been any continuing effort to support multiple test frameworks.

I haven't been able to run JSUnit on Palm's Web Browser 2.0 emulator on 
Windows and was thinking that might require doing a stripped down test 
runner.  The Internet Explorer 5.5 for Windows CE looked like it might 
be able to run JSUnit as is if you increased the timeouts substantially.

The layout of JSUnit on Palm's Web Browser 2.0 is pretty bad.  See 
http://groups.yahoo.com/group/jsunit/message/157 for setup instructions.

> Framework
> 
> Philippe: Where are we? We have 2 versions: Bob and Edward
> versions. Heard concerns during the technical plenary regarding the
> JSUnit framework (browser sniffing, exceptions handling).
> 
> Bob: during last summer, modifications were made to JSUnit 1.3.2 in such
> way that the DOM Test Suite wasn't working anymore. Until recently, it
> wasn't fixed. The current CVS version from Edward has been fixed. My
> version does have changes and contains other proposed changes as well,
> including selection of content type and test implementation
> attributes. Except for a bug, we are waiting for feedback from Apple and
> Opera before releasing the next version of JSUnit. JSUnit 1.3.4 will be
> identical to my current version. The current bug only affects Mozilla on
> Apple and Opera as well. Currently, the DOM test suite does not work in
> Opera. trying to get their feedback.
> 
> Philippe: can help for Opera.
> 
> Dimitris: can help for Apple.
> 
> Bob: we modified the test suite to remove dependencies on HTML
> getElementById/innherHTML. Nothing browser specific. only relies on
> ECMAScript and some features of DOM Level 0.

I believe both JSUnit 1.3.3 and Bob's version have a annoying problem 
around line 293 of JsUnitCore.js that raised an exception when 
displaying the test files in the browser.  The following change will 
avoid the problem.

-if (top.xbDEBUG.on && top.testManager)
+if (top.xbDEBUG != null && top.xbDEBUG.on && top.testManager)

> Philippe: what about exceptions? IE does not support the ECMAScript
> exceptions.
> 
> Bob: IE does support exceptions but they return incorrect error codes.
> 
> Philippe: do we do browser sniffing to avoid those troubles in IE?
> 
> Bob: not sure... Don't remember if the exception mapping from last
> spring is still in the test suite.

Earlier adapters for MSXML did "adapt" for COM exceptions as did the 
original NIST JavaScript test suite.  Per discussions on the list from 
last year (or before), that adaptation was removed.

The only remaining "adaption" for MSXML is to remove the processing 
instruction used to represent the XML declaration.  Using PI's to 
represent the XML declaration was stated to be acceptible by the WG. 
However, since only MSXML displays the behavior to my knowledge and it 
would add a great deal of complexity to a lot of tests to adapt for this 
behavior, the simplest way to address the issue is just to have the 
loader remove it.

> Philippe: do we do browser sniffing for the framework?
> 
> Johnny: it may need to do so for the framework itself. there is nothing
> wring with that.
> 
> Bob: no user agent detection. only test for support of specific objects.
> 
> Philippe: as long as it doesn't change the result of the tests, it's
> fine.

I'll double check, but I don't believe there any remaining sniffing.

> Bob: following feedback from DOM TS and other browsers, we are very
> close to release a new version.  we need help from Opera and Apple. In
> contact with Hakon to fix the framework for Opera. Contacted David Hyatt
> from Apple as well. For Safari, they are moving rapidly.
> 
> ACTION: Philippe to contact Hakon regarding the test suite, cc Bob
> 

Would be great to have JSUnit and DOM TS support the new PDA browsers. 
I've done a little exploration, but it might require some vendor support 
too.

I know that the Konqueror team has been able to run the test suite, but 
I haven't been able to get Konqueror 3.1 on SuSE 8.0 to run it.  It 
would be great if they could provide pointers to getting things set up.

> ---
> 
> Philippe: We currently have a DOM Level 1 Core Test Suite, but DOM Level
> 1 doesn't exist a Core module. It only references Core interfaces. DOM
> Level 1 XML and DOM Level 1 HTML do use those Core interfaces though.
> 
> Johnny: seems a reorganization issue
> 
> Dimitris: should focus on DOM Level 2 instead.
> 
> -> Once the DOM Level 2 test suite is released, we may think about DOM
> Level 1.
> 
> ---
> 
> TF and Chair
> 
> Philippe: Following Dimitris's concerns, we discussed the idea of having
> a kind of Task Force for the DOM Test Suite. It will provide an
> organization of the work. However, neither Dimitris or myself do have
> the cycles for that unfortunately. We need someone who can keep track on
> what's going on, follow up on test submissions, provide a roadmap,
> organize teleconferences, ...
> 
> Bob will try to get an answer by the end of the week.
> 
> (Philippe will be on the road for a week starting end of this week)


I hope to have an iteration into the CVS in a few hours.

Received on Wednesday, 12 March 2003 22:33:06 UTC