provide native branch for dojo.query() for latest WebKit and/or Safari 3.1

[alex]

as we've always planned for, browsers are starting to implement the native CSS selector APIs. Dojo 1.1's dojo.query() should be updated to deferr to these APIs when available.

[alex]

see ​http://webkit.org/blog/156/queryselector-and-queryselectorall/

[alex]

(In [12371]) update dojo.query to defer to [node].querySelectorAll() if it's available. Fixes #5832. !strict

[bill]

Looks like this breaks webkit, see ​http://www.dojotoolkit.org/forum/dijit-dijit-0-9/dijit-development-discussion/bordercontainer-webkit-webkit-dojo-nightly#comment-12190. Please fix or rollback.

[alex]

Apple's on it:

​http://bugs.webkit.org/show_bug.cgi?id=17313

I'm not keen to roll it back...it's their bug to fix.

[alex]

(In [13066]) update query() to move combinator-rooted queries off to a non-querySelectorAll branch until the webapis WG decides that they're the right thing to do (which they clearly are). Refs #5832. !strict

[nathan]

I'm think that this is related to bug #6252...maybe their fix didn't get in before safari 3.1 was released? I have verified this by rolling back these two changes and it fixes the crash in Safari 3.1

Would you potentially be a bit more keen on rolling it back, since it's now affecting the released version of Safari?

[alex]

lets not roll back. Other browsers are sporting QSA, and the right thing to do is to work around Webkit. A check like:

if((dojo.isSafari && dojo.isSafari > 3.1)&&(...)){
}

[nathan]

I agree. I think that would be an appropriate solution.

[alex]

verified w/ the webkit team that this bug:

​http://bugs.webkit.org/show_bug.cgi?id=17313

didn't land in time for Safari 3.1. WIll apply a patch to our side.

[alex]

(In [13142]) ensure that we don't blow up on Safari 3.1. Fixes #5832. !strict

[bill]

In [27941]:

Remove codepath for desupported safari 3.1. Refs #5832, #9045, #11844 !strict. Presumably doesn't fix #9045, which is awaiting a test case.