unload from frame's window breaks in IE8

[brian.moschel@…]

If you have an iframe in the same domain and try to bind to unload:

$(someotherwindow).unload(fn)

This works in all browsers except IE8.

​http://jsfiddle.net/Y4Txu/1/

The problem is when you bind to unload, it uses the "one" method, so after unload, it unbinds itself and calls jquery.removeData, which has this:

// Browsers that fail expando deletion also refuse to delete expandos on
// the window, but it will allow it on all other JS objects; other browsers
// don't care
if ( jQuery.support.deleteExpando || cache != window ) {
  delete cache[ id ];
} else {
  cache[ id ] = null;
}

That first if is true for IE8 and below I believe. Cache isn't the current window in this case, its another frame's window object, so delete cache[id] throws an error, and the unload event handler isn't reached

The fix is to check if cache is any window, not just the current window

[anonymous]

I should say the error thrown is "Object doesn't support this action" and at least in 1.6.2 its on delete cache[id].

The error prevents the event handler from being called at all.

[Rick Waldron]

Duplicate of #8235.

[brian.moschel@…]

I don't think those 2 bugs are duplicates. In 8235, they said:

cache[ id ][ internalKey ] is checked without ensuring that cache[ id ] exists.

In the error I'm describing, the problem is

if ( jQuery.support.deleteExpando

cache != window )

is true if cache is another frame's window, but delete cache[id] breaks. That if statement should check if cache is ANY window, not just the current one. Those are not the same cases.

[brian.moschel@…]

That code sample should show:

if ( jQuery.support.deleteExpando || cache != window )

[Rick Waldron]

Or you could simply ensure that cache[ id ] actually exists...

Which is the same issue.

​https://github.com/jquery/jquery/pull/468

[Rick Waldron]

Aso, see my comment here: ​http://bugs.jquery.com/ticket/8235#comment:19

[dmethvin]

Is this a result of not having a test case for #8235? We fixed that problem without knowing the cause.

[brian.moschel@…]

That pull request doesn't actually address the issue I'm talking about. The core issue is that cache can be any window object.

I made a pull request with a test that currently breaks in IE8 and a fix.

​https://github.com/jquery/jquery/pull/470

[Rick Waldron]

@dmethvin, no they are actually unrelated.

[Rick Waldron]

Using jQuery.isWindow will be significantly slower and all calls to removeData will be affected, see: ​http://jsperf.com/iswindow-vs-inline-test

I've updated my branch to test for a window inference and have included your test with credit to you.

[dmethvin]

Yes .isWindow() is slower but look at the numbers. Even the "slow" path can run almost 2 *million* times a second. It's not likely the .removeData() path is critical enough that the difference matters.

[dmethvin]

Fixed by ​https://github.com/jquery/jquery/pull/468 .

[Dave Methvin]

Merge pull request #468 from rwldrn/10080

Fixes #10080. Test cache for window inference.

Changeset: f4811bfb04206b9d9c120001603eadb1b529c271

[dmethvin]

#8870 is a duplicate of this ticket.