Order of selector expression returns different results for .find(selector) in Opera 10.63

[mechismo]

Working test case: ​http://jsfiddle.net/hd4NK/

In Opera 10.63 the following return different length collections:

    $("#con").find(".required, #bar, #btm :input");
    $("#con").find("#btm :input, .required, #bar");

when querying the following mark-up

<fieldset id="cont">
	<div id="top">
		<input type="text" class="required"  />
		<input type="text" id="bar"  />
	</div>
	<div id="btm">
		<input type="text" class="required2"  />
		<input type="text" id="foo"  />
	</div>
</fieldset>

All other browsers tested (FF 3.6, IE8, Safari5, Chrome8) return expected results. Same issue is in 1.4.4 too.

In 1.4.2 all browsers return the same results.

[jitter]

Thanks for taking the time to contribute to the jQuery project by writing a bug report and providing a testcase!

I checked in on your test case and yes Opera does indeed return a different result from all other browsers. It is a a bug in Opera's querySelectorAll implementation.

CC-ing miketaylr on this one. <-- ​test case for you

[miketaylr]

Yeah, the bug appears to be that qSA isn't throwing for an unsupported pseudo-class if there is a valid part first. So document.querySelectorAll(':foo, input') will throw, but not document.querySelectorAll('input, :foo').

[mechismo]

Fair enough. Interesting that is worked OK in 1.4.2 which indicates to me that it's maybe not a bug in Opera but I haven't dug deeply into this. Maybe jQuery wasn't using native qSA in 1.4.2 for Opera.

[jitter]

Replying to mechismo:

Fair enough. Interesting that is worked OK in 1.4.2 which indicates to me that it's maybe not a bug in Opera but I haven't dug deeply into this. Maybe jQuery wasn't using native qSA in 1.4.2 for Opera.

No it is a bug in Opera. There were some internal changes from jQuery 1.4.2 to 1.4.4 which caused this to become apparent.

[danheberden]

Still present in opera11 :/

[dmethvin]

I think we clearly established that this was an Opera bug we can't work around, and miketaylr's buddies are hot on the case.

[sindresorhus]

#10921 is a duplicate of this ticket.