The current speech input plumbing in webkit and webcore does not work when there are more than 1 speech enabled input elements. The main issue is that webcore presents a single listener to webkit/embedder and this listener gets invoked without any knowledge of the target element for the events. This had forced the webcore speech code to allow only one speech recognition session at a time, which isn't useful because even though only one element can record audio at a time, multiple such recorded streams can be in recognition phase in parallel (since recognition takes time and may even involve a server roundtrip). This patch has the following changes to support the above (very useful) case: - WebCore::SpeechInputClient now requires a one time set for the WebCore::SpeechInputListener and takes in a 'requestId' for all calls. This requestId is returned back in all the listener callbacks for identifying which input element the event goes to. - WebCore::SpeechInput class generates the above mentioned request ids as necessary when each speech enabled input element gets created/destroyed and multiplexes the listener callbacks to the appropriate input element based on the request id. - All methods of WebKit::WebSpeechInputController (implemented by embedder) and WebKit::WebSpeechInputListener now require the requestId parameter as well. - The relevant mocks updated.