Regression with Drag&Drop

[derrell]

Describe the bug
When the drag target passes over a possible drop target enroute to a different drop target, and is dropped on the latter, the drop event is fired on the former instead of the latter as expected.

This regression was introduced with commit 4160586
@johnspackman This was your commit.

To Reproduce
In the following playground sample, drag the "Drag Me" label over "First Window" then continue on to "Second Window". When you release the mouse while over "Second Window", the alert will tell you that the drop event was actually wrongly fired on "First Window". If, on the other hand, you drag the round-about way, around the windows so that the drag never passes over "First Window" and goes immediately to "Second Window" where it is dropped, the correct object receives the drop event.

http://tinyurl.com/y7k8a2zq

Expected behavior
A drop on "Second Window" would be expected to fire a drop event on 'Second Window", not on "First Window" as it is doing with the regression commit through master.

[reedspool]

Found the cause:

qooxdoo/framework/source/class/qx/event/handler/DragDrop.js

Line 749 in 4160586

return self.__fireEvent("dragleave", self.__dropTarget, self.__dragTarget, false, e);

This returns after emitting an event. In the previous version, there was no return, and this.__dropTarget is subsequently updated. I think the solution is to just remove return.

[johnspackman]

I think this PR will fix your issue; thank you @reedspool, that bit of diagnostics work probably saved me a whole heap of work.

You were right in that the return was wrong because it meant the code was skipping essential stages, the final fix is slightly different in that it's necessary to chain the results together (the events system operates a kind of ultra lightweight promise system)