Replace RefCountImpl with rtc::RefCountedObject.

webrtc/test/vcm_capturer.cc

Index: webrtc/test/vcm_capturer.cc
diff --git a/webrtc/test/vcm_capturer.cc b/webrtc/test/vcm_capturer.cc
index 0a82236c9880487968c147938ecca7c8b83f26f9..792b97ff0883f7157a9f11f9cff622a9885b23b3 100644
--- a/webrtc/test/vcm_capturer.cc
+++ b/webrtc/test/vcm_capturer.cc
@@ -79,17 +79,13 @@ void VcmCapturer::Stop() {
 }
 
 void VcmCapturer::Destroy() {
-  if (vcm_ == NULL) {
+  if (!vcm_)
     return;
-  }
 
   vcm_->StopCapture();
   vcm_->DeRegisterCaptureDataCallback();
-  vcm_->Release();
-
-  // TODO(pbos): How do I destroy the VideoCaptureModule? This still leaves

[tommi, 2016/03/17 09:15:12]

is this still true? (if so, please leave the comment in)

[pbos-webrtc, 2016/03/17 15:04:00]

Not true anymore, since the creation does AddRef() now the assignment Release()s
properly.

[tommi, 2016/03/17 15:10:56]

I'm not following... the previous code already did vcm_->Release(), what's
different?

[pbos-webrtc, 2016/03/21 10:04:31]

That it did not do AddRef() which is now done implicitly through the
scoped_refptr.

[tommi, 2016/03/21 11:23:21]

Oh I see!  Yikes, that could also be an invite to a use-after-free issue (e.g.
if someone else AddRef+Releases the object). Do you know if there are more cases
of this (specifically in non-test code)?

-  //             non-freed memory.
-  vcm_ = NULL;
+  // Release reference to VCM.
+  vcm_ = nullptr;
 }
 
 VcmCapturer::~VcmCapturer() { Destroy(); }