Add UMA metrics to track user action sequence in VR

third_party/WebKit/Source/modules/vr/VRDisplay.cpp

Index: third_party/WebKit/Source/modules/vr/VRDisplay.cpp
diff --git a/third_party/WebKit/Source/modules/vr/VRDisplay.cpp b/third_party/WebKit/Source/modules/vr/VRDisplay.cpp
index 31fbc5f152979223fcafbeaa97bdcbe2648c76b2..5122c99d42d313daa98f7a7ad95c259908966e61 100644
--- a/third_party/WebKit/Source/modules/vr/VRDisplay.cpp
+++ b/third_party/WebKit/Source/modules/vr/VRDisplay.cpp
@@ -220,6 +220,14 @@ ScriptPromise VRDisplay::requestPresent(ScriptState* script_state,
     return promise;
   }
 
+  // When we are requesting to start presentation with a user action or the
+  // display has activated, record the user action.
+  if (first_present &&

[mthiesse, 2017/05/10 14:58:13]

Why duplicate the logic for when we should (usually) successfully present?

Why not record the action when we call display_->RequestPresent to know that the
page successfully requested presentation.

If you want to also capture failures, why not record an action unconditionally
when the page calls RequestPresent, and another action when successful?

[billorr, 2017/05/10 18:47:13]

I'm trying to filter calls of "requestPresent" to only user actions - which
include putting the device in a headset and clicking to initiate presentation. 
I also filter to the first present because otherwise we are already presenting.

This is almost identitical to the logic above (could be reduced to just
"if(first_present)"), but this appears more clear and future proof if the logic
changes in this function.

I put it here because it is after first_present is defined.  This could be moved
up to the top of the function to get all calls, including those that will fail,
or moved further down to be more likely to indicate success at presenting.  I
don't have a strong preference for either of those, and I think realistically
they should give very similar results.

https://chromium.googlesource.com/chromium/src/+/72a5c913974f7c92e7dc393bf870...

Guidance is to have one event triggered per user action.  This is following that
if we consider clicking "enter vr" and putting the phone in the headset as two
separate actions.

+      (UserGestureIndicator::ProcessingUserGesture() || in_display_activate_)) {
+    Platform::Current()->RecordAction(
+        UserMetricsAction("VR.WebVR.requestPresent"));
+  }
+
   // A valid number of layers must be provided in order to present.
   if (layers.size() == 0 || layers.size() > capabilities_->maxLayers()) {
     ForceExitPresent();
@@ -671,7 +679,13 @@ void VRDisplay::StopPresenting() {
       // Can't get into this presentation mode, so nothing to do here.
     }
     is_presenting_ = false;
+
     OnPresentChange();
+
+    // Record user action for stop presenting.  Note that this could be
+    // user-triggered or not.
+    Platform::Current()->RecordAction(
+        UserMetricsAction("VR.WebVR.StopPresenting"));
   }
 
   rendering_context_ = nullptr;