Encode frame number in pixel data for pose sync

third_party/WebKit/Source/modules/vr/VRPose.h

Index: third_party/WebKit/Source/modules/vr/VRPose.h
diff --git a/third_party/WebKit/Source/modules/vr/VRPose.h b/third_party/WebKit/Source/modules/vr/VRPose.h
index fc7d85db07707ff83d58afabf2c599057e7ce784..84e2319a59eb8d4e4a161c610124a39a12c29ae2 100644
--- a/third_party/WebKit/Source/modules/vr/VRPose.h
+++ b/third_party/WebKit/Source/modules/vr/VRPose.h
@@ -26,6 +26,10 @@ class VRPose final : public GarbageCollected<VRPose>, public ScriptWrappable {
   DOMFloat32Array* angularAcceleration() const { return m_angularAcceleration; }
   DOMFloat32Array* linearAcceleration() const { return m_linearAcceleration; }
 
+  // TODO(klausw): expose timestamp and poseNum to JS? We have those in the

[bajones, 2016/10/04 06:33:57]

timestamp is exposed in the VRFrameData object. I don't think poseNum is a good
thing to surface to JS, because it's use is really just for tracking render
state internally and we don't want developers relying on something that opaque.

[klausw, 2016/10/04 06:47:04]

OK, I'll remove the comment. When you mentioned Oculus being interested, I
wasn't sure if you meant via JS or just internally in the WebVR implementation
itself.

Side note, I guess it would be possible to use timestamps as keys for a pose map
to avoid the need for separate pose numbers, but using doubles as keys seems
weird even if it should technically work.

[klausw, 2016/10/04 17:52:47]

Removed the comment.

+  // underlying VRPosePtr, and they could be useful to scripts. If yes,
+  // they would need to be added to VRPose in the WebVR spec.
+
   void setPose(const device::blink::VRPosePtr&);
 
   DECLARE_VIRTUAL_TRACE();