Update the Chromium version of webrtc::VideoCaptureModule to be reference counted.

Update the Chromium version of webrtc::VideoCaptureModule to be reference counted.
The reason is to be able to share ownership between PeerConnection and Chromium.

TEST=
BUG=
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=102286

[perkj_chrome, 2011-09-13 07:55:34]

Can you please review this?

[scherkus (not reviewing), 2011-09-13 15:22:58]

http://codereview.chromium.org/7887002/diff/1/content/renderer/media/video_ca...
File content/renderer/media/video_capture_module_impl.h (right):

http://codereview.chromium.org/7887002/diff/1/content/renderer/media/video_ca...
content/renderer/media/video_capture_module_impl.h:27: virtual int32_t AddRef();
is this actually an interface that we're implementing? why does it need to be
virtual?

traditionally (since implementing refcounting is tricky), you extend
base::RefCountedThreadSafe<MyClassName>

[wjia(left Chromium), 2011-09-13 18:13:43]

webrtc::videocapturemodule::VideoCaptureImpl will have AddRef and ReleaseRef in
webrtc r580. 

lgtm, after scherkus's comments are addressed.

[wjia(left Chromium), 2011-09-13 18:18:35]

by the way, since you need add "OVERRIDE" to AddRef and ReleaseRef after
chromium has webrtc r580, you could hold the commit till webrtc roll in
chromium (hopefully on Friday) to avoid another trivial patch.

On Tue, Sep 13, 2011 at 11:13 AM, <wjia@chromium.org> wrote:

> webrtc::videocapturemodule::**VideoCaptureImpl will have AddRef and
> ReleaseRef in
> webrtc r580.
>
> lgtm, after scherkus's comments are addressed.
>
>
>
http://codereview.chromium.**org/7887002/<http://codereview.chromium.org/7887...
>

[perkj_chrome, 2011-09-19 11:14:21]

http://codereview.chromium.org/7887002/diff/1/content/renderer/media/video_ca...
File content/renderer/media/video_capture_module_impl.h (right):

http://codereview.chromium.org/7887002/diff/1/content/renderer/media/video_ca...
content/renderer/media/video_capture_module_impl.h:27: virtual int32_t AddRef();
Yes- it is part of the interface we are implementing. 
Same as webrtc_audio_device_impl.h.
We have not used base::RefCountedThreadSafe since the return type is not the
same.

On 2011/09/13 15:22:58, scherkus wrote:
> is this actually an interface that we're implementing? why does it need to be
> virtual?
> 
> traditionally (since implementing refcounting is tricky), you extend
> base::RefCountedThreadSafe<MyClassName>

[scherkus (not reviewing), 2011-09-19 17:42:54]

LGTM -- thanks for the explanation

I'm a little surprised that WebRTC doesn't provide a default implementation and
that we have to duplicate refcounting code

http://codereview.chromium.org/7887002/diff/4001/content/renderer/media/video...
File content/renderer/media/video_capture_module_impl.cc (right):

http://codereview.chromium.org/7887002/diff/4001/content/renderer/media/video...
content/renderer/media/video_capture_module_impl.cc:49: VLOG(1) <<
"VideoCaptureModuleImpl::AddRef()";
I'd get rid of these VLOGs -- they don't add much value and
WebRtcAudioDeviceImpl doesn't have any either

http://codereview.chromium.org/7887002/diff/4001/content/renderer/media/video...
content/renderer/media/video_capture_module_impl.cc:54: VLOG(1) <<
"VideoCaptureModuleImpl::Release()";
ditto