Changes

* <s>Figure out why DOMMouseScroll events aren't being retargeted to the locked element</s> - humph, diogogmt

* <s>Look into roc's suggestion from bug 722449 comment 8, namely, using getBoundingClientRect vs. GetPrimaryFrame()</s> - diogogmt

* <s>Maybe use nsStubMutationObserver instead of nsIMutationObserver?</s> - humph

* <s>Extensions to the Document Interface. Add pointerLockElement attribute and exitPointerLock method </s> - diogogmt

* <s>Update Element interface to have requestPointerLock </s> - diogogmt

* <s>Update mochitests to use new pointerlock API</s> - diogogmt

* Update patch to trunk- diogogmt ==Mochitests for new PointerLock API== To have consistency between all the tests they were updated to the following:* Through out the test flags are set to represent success or failure* runTests function will run the assertions using the flags* All assertions are located in runTests()* body onload calls start()* start() waits for focus then start running the test* If test uses fullscreen API, then runTests() and SimpleTest.finish() are called when fullscreen mode is cancelled Structure of the tests(Top-Bottom):# Test description# SpecialPowers(if nedded)# SimpleTest.waitForExplicitFinish()# var declarations# function runTests()# Some code specific to the test# pointerlockchange event listener# pointerlockerror event listener# mozfullscreenchange event listener## runTests()## SimpleTest.finish()# function start()

| Checks if element is attached to the DOM Tree before locking before locking the pointer

| On the old spec the pointer object was attached to the navigator, so there was a possibility that an element belonging to a different DOM could request pointerlock. However, in the new API the element requests pointerlock on it self making this test redundant.

| Confirm that screenX/Y and clientX/Y are constant constants when the pointer is locked.