RE: Review of the test failures

Thanks for pulling this analysis together Robin!

I'll be going over them in the next week or so, and hopefully will have at least one implementer's opinion in the places indicated :)
 
-----Original Message-----
From: Robin Berjon [mailto:robin@w3.org] 
Sent: Tuesday, January 6, 2015 8:32 AM
To: www-dom@w3.org
Cc: HTML WG (public-html@w3.org)
Subject: Review of the test failures

Hi all, and happy new year!


As you know the Web Platform Tests people have been putting together 
tests and test results for a lot of the platform, including the DOM. 
Since we have pretty decent results for the DOM, I decided to take a 
look at them and see what problems they may be pointing to (notably 
trying to catch bugs in the TS). Here are the details.

In a number of cases it would not be entirely useless for implementers 
to weigh in.

http://w3c-test.org/dom/nodes/Node-properties.html

   • There is a known spec-bug on this: 
https://www.w3.org/Bugs/Public/show_bug.cgi?id=19431. It actually has 
ramifications in other places.
   • It would be useful for implementers to weigh in.

http://w3c-test.org/dom/nodes/Node-parentNode.html

   • Cross-origin access to 'data:' causes the problem, test was wrong 
because of that, has been merged.
   • https://github.com/w3c/web-platform-tests/pull/1477


http://w3c-test.org/dom/nodes/Node-isEqualNode.xhtml

   • Same cross-origin issue, still under discussion.
   • https://github.com/w3c/web-platform-tests/pull/1478


http://w3c-test.org/dom/nodes/Document-createElement-namespace.html

   • Browsers disagree on synthetic media types. This is the same old 
Document bag of knots. It is also the heart of the discussion from 
http://lists.w3.org/Archives/Public/public-html/2014Oct/0000.html. Some 
weighing in from implementers would be of use.
   • https://www.w3.org/Bugs/Public/show_bug.cgi?id=22960



http://w3c-test.org/dom/ranges/Range-insertNode.html

   • Painful to debug, I need to return to this one.
   • Document normalisation fixes lots of issues, but I don't think it 
should be called for.
   • Clearly not an area of great interoperability.

http://w3c-test.org/dom/ranges/Range-surroundContents.html

   • Same as previous

http://w3c-test.org/dom/nodes/attributes.html

   • The exodus away from Attr as a Node has not (yet?) happened. Still 
in flux.

http://w3c-test.org/dom/nodes/MutationObserver-childList.html

   • "This step does intentionally not pay attention to the suppress 
observers flag." in insert, which runs for replace before the insertion, 
is only respected in Gecko.

http://w3c-test.org/dom/lists/DOMTokenList-stringifier.html

   • No one serialises as per spec, but the serialisation makes sense.

http://w3c-test.org/dom/nodes/Element-children.html

   • Implementations differ on what needs to be enumerated on instances.

http://w3c-test.org/dom/nodes/Document-createElement.html

   • Some browsers don't restrict lowercasing (for localName) or 
uppercasing (for tagName) to the ASCII range

http://w3c-test.org/dom/nodes/Document-getElementsByTagName.html

   • similar issues to /dom/nodes/Element-children.html
   • similar issues to /dom/nodes/Document-createElement.html

http://w3c-test.org/dom/nodes/MutationObserver-attributes.html

   • Implementations don't properly default attribute: true.

http://w3c-test.org/dom/traversal/NodeIterator.html

   • Implementations don't seem to be following WebIDL in making use of 
a single operation callback interface for NodeFilter.

http://w3c-test.org/dom/events/Event-propagation.html
   • Implementations don't properly reset the stop propagation flag upon 
re-init.

http://w3c-test.org/dom/nodes/Document-createElementNS.html

   • Minor problem casting undefined to null.

http://w3c-test.org/dom/nodes/MutationObserver-characterData.html

   • Implementations don't properly default characterData: true.

http://w3c-test.org/dom/nodes/Document-constructor.html

   • Constructor not supported.

http://w3c-test.org/dom/nodes/Node-replaceChild.html

   • Wrong exception being thrown.
   • Some undetected hierarchy problems with doctypes.

http://w3c-test.org/dom/errors/exceptions.html

   • IDL stuff

http://w3c-test.org/interfaces.html

   • the usual WebIDL stuff

http://w3c-test.org/nodes/DOMImplementation-createDocument.html

   • Undefined -> null issues
   • Wrong exception

http://w3c-test.org/nodes/Element-classlist.html

   • Mostly white space issues. Since before your sun burned hot in 
space, and before your race was born, there have been white space issues.

http://w3c-test.org/interface-objects.html

   • IDL stuff

http://w3c-test.org/nodes/case.html

   • Implementation variations on corner cases.

http://w3c-test.org/nodes/Document-characterSet-normalization.html

   • Minor implementation alignment issues.

http://w3c-test.org/nodes/Document-createTreeWalker.html

   • Implementations haven't yet caught up to option arguments.

http://w3c-test.org/nodes/DOMImplementation-createHTMLDocument.html

   • Minor implementation difference.

http://w3c-test.org/nodes/MutationObserver-inner-outer.html

   • Actual implementation failure.

http://w3c-test.org/events/Event-constructors.html

   • Event constructor isn't implemented yet.

http://w3c-test.org/ranges/Range-intersectsNode.html

   • The spec seems right (a document intersects any range in it), but 
implementations disagree.

http://w3c-test.org/nodes/Node-insertBefore.html

   • Some issues with doctype insertion.

-- 
Robin Berjon - http://berjon.com/ - @robinberjon

Received on Thursday, 8 January 2015 19:51:15 UTC