MediaStream API: Implement Navigator.getMediaDevices

MediaStream API: Implement Navigator.getMediaDevices
This is the new hotness; the old busted MediaStreamTrack.getSources()
will be removed later on.

BUG=338511

[Tommy Widenflycht, 2014-01-28 04:12:19]

Adam,
I know this is a big patch without tests but could you just have a quick look
and see if this is still the correct way of adding methods to the Navigator
object?

[abarth-chromium, 2014-01-28 06:49:48]

Seems generally fine.  I didn't review in super detail because we're missing
tests and such.

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDeviceInfo.idl (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDeviceInfo.idl:26: enum MediaDeviceKind {
"audioinput", "audiooutput", "videoinput" };
Maybe split these onto separate lines?

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDevicesRequest.cpp (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDevicesRequest.cpp:72:
m_callback->handleEvent(mediaDevices);
Generally its better to use ActiveDOMObject for these sorts of classes.  We
don't want to reveal to script when we've garbage collected the document. 
Instead, we want to stop delivering callbacks when the document has called
stop() on all the active DOM objects.

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDevicesRequest.h (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDevicesRequest.h:61: UserMediaController*
m_controller;
How do we know that this object outlives the controller?

[Tommy Widenflycht, 2014-01-28 22:35:02]

I attended your comments and added a simple test. However you might want to hold
off on further reviews since I have to split this into three patches
blink-chromium-blink for the test to work.

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDeviceInfo.idl (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDeviceInfo.idl:26: enum MediaDeviceKind {
"audioinput", "audiooutput", "videoinput" };
On 2014/01/28 06:49:49, abarth wrote:
> Maybe split these onto separate lines?

Done.

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDevicesRequest.cpp (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDevicesRequest.cpp:72:
m_callback->handleEvent(mediaDevices);
On 2014/01/28 06:49:49, abarth wrote:
> Generally its better to use ActiveDOMObject for these sorts of classes.  We
> don't want to reveal to script when we've garbage collected the document. 
> Instead, we want to stop delivering callbacks when the document has called
> stop() on all the active DOM objects.

Done.

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
File Source/modules/mediastream/MediaDevicesRequest.h (right):

https://codereview.chromium.org/145583015/diff/20001/Source/modules/mediastre...
Source/modules/mediastream/MediaDevicesRequest.h:61: UserMediaController*
m_controller;
On 2014/01/28 06:49:49, abarth wrote:
> How do we know that this object outlives the controller?

The controller is owned to by the page and I now clear this pointer when
ActiveDOMObject calls stop