:focusable and :tabbable are broken with jQuery 1.3.2

I've looked at the code. The interesting part seems to be this check:

// the element and all of its ancestors must be visible
// the browser may report that the area is hidden
&& !$(element)['area' == nodeName ? 'parents' : 'closest'](':hidden').length;

Seems like we could just check the visibility of the element itself instead of checking ancestors. So this should do the trick:

&& $(element).is(":visible");

Fixes 7 of 17 failing tests, thats something...

+1 to Joern's suggested patch - I definitely agree that checking if the element is visible makes more sense.

Concerning the 'visible, enabled elements' tests failing for the 'area with href' case, I've noticed that these <area> elements consistently have an offsetWidth and offsetHeight of 0. I believe this is because the test map doesn't have any <img usemap=""/> relationships attached to it, so all of its internal <area>s are essentially hidden. jQuery 1.3.2 checks for a positive offsetWidth or offsetHeight in its implementation of the ':visible' selector and so this is what causes that subtest to fail.

However, the <map> parent should still have a valid width/height, and the following tweak to Joern's suggest patch causes all of the 'visible, enabled elements' tests to pass 100% in Firefox 2/3:

&& ('area' == nodeName
    ? $(element).closest('map').is(':visible')
    : $(element).is(':visible'));

The problem is that the same is not true for either Chrome or IE 7. They both consistently report that the <map> and all nested <area>s have an offsetWidth/offsetHeight of 0, causing the is(':visible') check to report false every time.

So :focusable is actually used only by :tabbable, which in turned is used only by Dialog. Seems like we imagined more of a usecase where actually only one exist: focus the first element in a dialog.

Which reminds me of the code in my validation plugin, which basically just gets the first invalid input (not interesting here), then does .filter(":visible").focus();

The call is wrapped in a try-catch-swallow block for some IE-edgecases.

can someone please explain this bug in more detail, and where this applies? (i.e. what change?)

Mostly fixed in 3f070bd.

Area elements are fixed in fe13fba.

What's left?

The object tests still fail in IE. I cant seem to come up with a test that passes in all browsers at the same time.

Objects fixed in 0ab54d8.

Core tests: Fixed object element for focusable/tabbable tests. Fixes #4488 - :focusable and :tabbable are broken with jQuery 1.3.2.

Changeset: 0ab54d87dc521b73819a3ac480f7aa94365e52e2