Remove sizzle expando after clone action

[mgol]

Originally reported by Jogi: http://bugs.jquery.com/ticket/15104

Issue with space in selector with jquery 1.11.1 and Internet explorer.

exemple:

$(".testCls [name='blabla']")

cannot find anything when with older version, it work.

jsfiddle: http://jsfiddle.net/kwfHZ/1/

[mgol]

Jogi:

wrong jsfiddle http://jsfiddle.net/kwfHZ/3/

i'm trying to reproduce my case correctly ...

[mgol]

@gibson042:

Your examples seem to be working correctly. Can you describe what you're trying to do, what behavior you're expecting, and what behavior you're seeing?

[mgol]

@gibson042:

http://bugs.jquery.com/ticket/15103 is a duplicate of this ticket.

[mgol]

Jogi:

0k So there is my case:

http://jsfiddle.net/xorkk/JTtu3/

I don't understand why on IE8:

$("#tblMF .numasaCmp_1 [name='NUMASA_CMP']") => undefined

but $("#tblMF .numasaCmp_1").find("[name='NUMASA_CMP']") => works

On Firefox, both works.

[mgol]

@gibson042:

We'll need something more reduced than 1.4 KB of HTML plus a script unrelated to this issue.

[mgol]

Jogi:

Ok i've reduced the exemple:

http://jsfiddle.net/xorkk/JTtu3/

[mgol]

Jogi:

i See one more thing, if i remove the line with .prop("disabled", false) , my selector works correclty.

[mgol]

Jogi:

other observation:

$("#tblMF .numasaCmp_1 [name='NUMASA_CMP']") => undefined

but $("#tblAA .numasaCmp_1 [name='NUMASA_CMP']") => work

[mgol]

@gibson042:

I removed superfluous markup and script from your fiddle and changed .val() to .length for better insight: http://jsfiddle.net/JTtu3/57/

There is something to this that seems to be related to Sizzle combinator caching in oldIE.

[mgol]

Jogi:

Same problem with ":input" instead of [name...]
$(".nonStructAssetTemplate :input").length == 0
$(".nonStructAssetTemplate").find(":input").length != 0

[mgol]

@gibson042:

This is actually a problem with .clone: http://jsfiddle.net/JTtu3/59/

In IE8 (and other versions as well, I'm sure), the clone still has a reference to the Sizzle cache object of its referent, which wreaks all kinds of havoc with any operation using it. Maybe we can update fixCloneNodeIssues to look for and remove attributes associated with reference-equal clone/source properties.

[mgol]

@gibson042:

PR: #1652

[dmethvin]

Per the discussion on gh-1652 we decided there was no good solution but the least bad solution was that we'd clear the Sizzle expando. Is that correct?

[gibson042]

Correct.

[markelog]

@gibson042 please ping me up when you gonna update sizzle