Bug 13393: .focus() results in "Unspecified Error" in IE9

[dgalvez]

I couldn't find a better way to avoid IE9 crashing when evaluating document.activeElement. This patch sets documet.body as the activeElement in case elem.focus() is called before document.readyState !== "complete" and then redefines jQuery.event.special.focus.trigger to avoid re-running the extra check and body.focus() statements.

A test case has been added.

focus: {
            // Fire native event if possible so blur/focus sequence is correct
            trigger: function() {
                var focusTrigger = function() {
                    if ( this !== document.activeElement && this.focus ) {
                        this.focus();
                        return false;
                    }
                };

                if ( document.readyState !== "complete" ) {
                    document.body.focus();
                }

                jQuery.event.special.focus.trigger = focusTrigger;

                return focusTrigger.call( this );
            },
            delegateType: "focusin"
        }

The bug url: http://bugs.jquery.com/ticket/13393

[dgalvez]

I didn't add a check to run this for IE9 only or a check to make sure document.body.focus is a method. What do you think?

[dmethvin]

It's good to know that the problem only occurs before the document is ready. My Spidey senses tingle when it comes to actively forcing focus to anything though. If we're nested a couple of iframes deep for example, is this going to cause an issue if the outer iframe monkeys with the focus as well?

I wonder if we might be able to skip the document.activeElement check if the doc isn't ready and still get the focus to work.

I think the test should run on all browsers to make sure we didn't break them. On XML documents document.body.focus wouldn't be there but people should not be triggering events named "focus" on XML documents.

[dgalvez]

"is this going to cause an issue if the outer iframe monkeys with the focus as well?"
That's a good point Dave.

"I wonder if we might be able to skip the document.activeElement check if the doc isn't ready and still get the focus to work."
Good idea, let me try this and see if we get it working without breaking any existing test cases.

In case it's not possible to avoid the document.activeElement check I'll try to discard any outer iframe interaction with the current approach ( perhaps using this.ownerDocument.body.focus() instead and checking edge cases ).

[dmethvin]

Thanks for looking at this. The focus/blur stuff tends to be really fidgety in both IE and Firefox.

Another thought, maybe we could use the flag that @gibson042 is proposing in gh-1183 if focusing the body isn't reliable. We could avoid focusing the element if the document isn't ready but still run the handlers. Your solution is better but this could be a fallback.

[dmethvin]

@dgalvez, @timmywil said he'd take a look at this. It looks like we may be able to use document.documentElement.activeElement. @timmywil note that this only appears to happen before the doc is ready, in case that helps or is needed for the test case.

[timmywil]

check

[dgalvez]

Great! Sorry guys, busy week :(

[dmethvin]

ping @timmywil on this ... should I close the PR or do you plan to riff off this?

[dgalvez]

Yep, I'll poke @timmywil PR closed :)

[dgalvez]

@timmywil what about having document.documentElement.activeElement as a fallback for document.activeElement? Please take a look at my last commit. Perhaps we could redefine the focus method after the first call when doc is ready to avoid any overhead, although I don't think is needed.

Using document.documentElement.activeElement caused "focusin bubbles" tests to not pass and that's why I thought of the fallback approach.

@dmethvin reopening the PR was the way I found to ping @timmywil :)

[timmywil]

@dgalvez : Thanks! This is already in the works.

[dgalvez]

@timmywil cool!