Wrap MediaMetadata in base::Optional in content and chrome

content/browser/android/web_contents_observer_proxy.h

Index: content/browser/android/web_contents_observer_proxy.h
diff --git a/content/browser/android/web_contents_observer_proxy.h b/content/browser/android/web_contents_observer_proxy.h
index af0895dc6693f3ea23fdda5d44268333e38b0639..8888f538dbcfd5c143395318e9b7ec39d1fec316 100644
--- a/content/browser/android/web_contents_observer_proxy.h
+++ b/content/browser/android/web_contents_observer_proxy.h
@@ -72,9 +72,10 @@ class WebContentsObserverProxy : public WebContentsObserver {
   void DidStartNavigationToPendingEntry(
       const GURL& url,
       NavigationController::ReloadType reload_type) override;
-  void MediaSessionStateChanged(bool is_controllable,
-                                bool is_suspended,
-                                const MediaMetadata& metadata) override;
+  void MediaSessionStateChanged(
+      bool is_controllable,
+      bool is_suspended,
+      const base::Optional<MediaMetadata>& metadata) override;

[dcheng, 2016/09/06 21:11:16]

The documentation for base::Optional specifically recommends passing parameters
as T* rather than const base::Optional<T>&. Let's do that here and elsewhere,
except where we can't (e.g. the IPC message handler)

[mlamouri (slow - plz ping), 2016/09/07 08:55:53]

FWIW, the gist of this recommendation is to avoid methods taking an Optional<T>
instead of T* but I think it's fine to take an Optional<T> if it's expected that
the method is already  handling an Optional<T>. In other words, we should avoid
unwrapping the Optional<T> to pass it to a method taking T* then wrap T* back in
an Optional<T>.

[Zhiqiang Zhang (Slow), 2016/09/07 10:45:06]

I think passing `const base::Optional<T>&` is OK. We are not constructing new
Optionals, so should not have the problems described in the documentation.

   void SetToBaseURLForDataURLIfNeeded(std::string* url);
 
   void DidFailLoadInternal(bool is_provisional_load,