map.removeControl sometimes chokes in IE

[tschaub]

When a page unloads, map.removeControl is called for all controls. We check if control.outsideViewport before trying to remove control.div from map.viewportDiv. Though this seems like a sufficient check, the real check is to see if control.div.parentNode is map.viewportDiv before trying to remove child from parent.

[tschaub]

safer dom cleanup

[tschaub]

Aside from the failure of the OSM tests in IE, all tests pass in FF, IE6, and IE7.

Please review.

[crschmidt]

Yeah, this probably means that our displayOutsideViewport check is wrong in some cases, but whatev. This is the right check to make, rather than depending on a hueristic that people can break.

[tschaub]

Yeah, that was my initial concern, and I dug around a bit to see if we were messing up in an obvious way. The alternative explanation is that the div might already be removed from the viewport by other means (the page is closing, who knows how the browser deconstructs the dom).

[tschaub]

(In [5973]) Explicitly check that we know who the parent is before removing child from parent. r=crschmidt (closes #1316)