Add a Kill switch for WebVR cardboard support through finch.

chrome/browser/android/chrome_feature_list.cc

Index: chrome/browser/android/chrome_feature_list.cc
diff --git a/chrome/browser/android/chrome_feature_list.cc b/chrome/browser/android/chrome_feature_list.cc
index d237ad8568adf7c7ffb4a08821225a572ea4eb5b..25cc547727c561fee97044a5783b5008a6a28d0d 100644
--- a/chrome/browser/android/chrome_feature_list.cc
+++ b/chrome/browser/android/chrome_feature_list.cc
@@ -60,6 +60,7 @@ const base::Feature* kFeaturesExposedToJava[] = {
     &kTabReparenting,
     &kWebPaymentsModifiers,
     &kWebPaymentsSingleAppUiSkip,
+    &kWebVRCardboardSupport,
     &ntp_snippets::kIncreasedVisibility,
     &ntp_snippets::kForeignSessionsSuggestionsFeature,
     &ntp_snippets::kOfflineBadgeFeature,
@@ -144,6 +145,9 @@ const base::Feature kWebPaymentsModifiers{"WebPaymentsModifiers",
 const base::Feature kWebPaymentsSingleAppUiSkip{
     "WebPaymentsSingleAppUiSkip", base::FEATURE_ENABLED_BY_DEFAULT};
 
+const base::Feature kWebVRCardboardSupport{
+    "WebVRCardboardSupport", base::FEATURE_ENABLED_BY_DEFAULT};
+
 static jboolean IsEnabled(JNIEnv* env,
                           const JavaParamRef<jclass>& clazz,
                           const JavaParamRef<jstring>& jfeature_name) {
@@ -157,6 +161,19 @@ static jboolean IsEnabled(JNIEnv* env,
   return false;
 }
 
+static jlong GetFeature(JNIEnv* env,

[Ted C, 2017/01/31 18:24:30]

as a general practice, we really, really do not want to pass raw pointers back
to java (too easy to get wrong, incorrectly modified, etc.., even harder to
rationalize lifetime than raw native pointers IMO)

I think if we wanted to make this more java friendly, we'd create a Feature.java
class that wraps the raw pointer and maybe has the necessary apis to pull params
out of it.

But....that might be overkill for right now.  I'd suggest that instead of making
the changes to VariationsAssociatedData that you just add the method here that
calls GetFieldTrialParamByFeatureAsInt give a field trial name and param name.

I'm basically happy with any approach that doesn't pass a pointer.

[amp, 2017/01/31 18:41:43]

Duly noted.  I was trying to work out how to keep variation related code in the
/components/variations while still being able to get a reference to the Feature
and passing a pointer was all I could come up with.

I'm fine with adding the call directly here in chrome_feature_list.  It looks
like a lot of the methods in /components/variations files are being deprecated
in favor of /base/metrics anyway.

@rkaplow, any concerns with this path?

+                          const JavaParamRef<jclass>& clazz,
+                          const JavaParamRef<jstring>& jfeature_name) {
+  const std::string feature_name = ConvertJavaStringToUTF8(env, jfeature_name);
+  for (size_t i = 0; i < arraysize(kFeaturesExposedToJava); ++i) {
+    if (kFeaturesExposedToJava[i]->name == feature_name)
+      return reinterpret_cast<intptr_t>(kFeaturesExposedToJava[i]);
+  }
+  // Features queried via this API must be present in |kFeaturesExposedToJava|.
+  NOTREACHED();
+  return 0l;
+}
+
 bool RegisterChromeFeatureListJni(JNIEnv* env) {
   return RegisterNativesImpl(env);
 }