[webkit-dev] Global constructors attributes

[Christophe Dumez - SISA]

Hi,

The code change is actually really small. This is basically, 2/3 lines in the bindings generator.

Alternatively, we could add [NotEnumerable, Deletable] before each *Constructor attribute in DOMWindow.idl but it seemed easier the patch the generator considering the number of such attributes in DOMWindow. Also, there is a risk of someone adding a new constructor attribute without the needed extended attributes.

Either way though, it seems easily reversible.

Kr,
Christophe DUMEZ.

________________________________
From: ryosuke.niwa at gmail.com [ryosuke.niwa at gmail.com] on behalf of Ryosuke Niwa [rniwa at webkit.org]
Sent: Wednesday, April 17, 2013 21:36
To: Christophe Dumez - SISA
Cc: webkit-dev at lists.webkit.org
Subject: Re: [webkit-dev] Global constructors attributes

This seems like a worthwhile change and I do support it but is it possible to make this fix behind a build flag or make it self-contained such that we can revert easily if needed?

While I do agree with you that the compatibility risk is probably small, I'd like to make sure we have an option of being able to revert the change or disabling the fix if we did find a compatibility issue later.

On Wed, Apr 17, 2013 at 11:32 AM, Christophe Dumez - SISA <ch.dumez at partner.samsung.com<mailto:ch.dumez at partner.samsung.com>> wrote:
Hi,

According to the Web IDL specification [1] (still valid in the latest Editor Draft [2]), global constructors should have the following attributes:
{ [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }

Quoting from the specification:
"""

For every interface<http://dev.w3.org/2006/webapi/WebIDL/#dfn-interface> that:

  *   is a callback interface<http://dev.w3.org/2006/webapi/WebIDL/#dfn-callback-interface> that has constants<http://dev.w3.org/2006/webapi/WebIDL/#dfn-constant> declared on it, or
  *   is a non-callback interface<http://dev.w3.org/2006/webapi/WebIDL/#dfn-interface> that is not declared with the [NoInterfaceObject]<http://dev.w3.org/2006/webapi/WebIDL/#NoInterfaceObject> extended attribute<http://dev.w3.org/2006/webapi/WebIDL/#dfn-extended-attribute>,

a corresponding property must exist on the ECMAScript global object. The name of the property is the identifier<http://dev.w3.org/2006/webapi/WebIDL/#dfn-identifier> of the interface, and its value is an object called theinterface object.

The property has the attributes { [[Writable]]: true, [[Enumerable]]: false, [[Configurable]]: true }. The characteristics of an interface object are described in section 4.4.1<http://dev.w3.org/2006/webapi/WebIDL/#interface-object>below.

"""

Currently, global constructors have the following attributes in WebKit, which is not according to spec:
{ [[Writable]]: true, [[Enumerable]]: true, [[Configurable]]: false }

This is causing some W3C compliance tests to fail, such as this one [3].

Behavior of other browsers:
- Firefox follows the specification
- Blink does not follow the specification
- IE9 follows partly the spec:
  * global constructors are not enumerable (according to spec)
  * global constructors are not deletable (not according to spec)

I don't think it is likely web sites rely on those global constructors being enumerable considering that they are not enumerable in both IE9 and Firefox. Making them deletable seems low risk to me as well.

I uploaded a patch proposal to bugzilla [4] to make WebKit follow this part of the spec. I would appreciate feedback from the rest of the community as this is a "web-exposed" behavior change.

Kr,
Christophe DUMEZ

[1] http://www.w3.org/TR/2012/CR-WebIDL-20120419/#es-interfaces
[2] http://dev.w3.org/2006/webapi/WebIDL/#es-interfaces
[3] http://w3c-test.org/webapps/ProgressEvents/tests/submissions/Ms2ger/interface.html (2 last tests)
[4] https://bugs.webkit.org/show_bug.cgi?id=110573


_______________________________________________
webkit-dev mailing list
webkit-dev at lists.webkit.org<mailto:webkit-dev at lists.webkit.org>
https://lists.webkit.org/mailman/listinfo/webkit-dev


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20130417/f3c00718/attachment.html>