In new-style hit-testing, there is always a possibility that the data provided to the HitTestQuery system is not sufficient to determine where the event should be routed to. In this case, the HitTestQuey component sends an inquiry to a WindowTreeClient (typically a renderer) for the FramdSinkId that should be the event recipient.

The WindowTreeClient provides this resolution via a mojo interface ( http://crbug.com/732405 )

But there needs to be a way to provide a mojo pipe corresponding to the implementation of this targeting-inquiry interface to the WindowTreeHost implementation. This bug is to implement that facility.