Support LoadSession() in MediaKeys and PPP_ContentDecryptor_Private interface.

media/base/media_keys.h

Index: media/base/media_keys.h
diff --git a/media/base/media_keys.h b/media/base/media_keys.h
index 0d86948564e32a350ebbed804292cd85f5787c36..de9ee5c52709af581984cde8d6e95db2ecd29bcc 100644
--- a/media/base/media_keys.h
+++ b/media/base/media_keys.h
@@ -42,16 +42,23 @@ class MEDIA_EXPORT MediaKeys {
   MediaKeys();
   virtual ~MediaKeys();
 
-  // Generates a key request with the |type| and |init_data| provided.
-  // Returns true if generating key request succeeded, false otherwise.
+  // Creates a session with the |content_type| and |init_data| provided.
+  // Returns true if a session is successfully created, false otherwise.
   // Note: UpdateSession() and ReleaseSession() should only be called after
-  // CreateSession() returns true.
+  // SessionCreatedCB is fired.
   // TODO(jrummell): Remove return value when prefixed API is removed.

[ddorwin, 2014/02/10 19:05:25]

Looking at the code, I'm not sure we actually need this return value.

PpapiDecryptor::CreateSession() is the only one that can return false, but this
is based on ContentDecryptorDelegate::CreateSession(), which only ever returns
true. In fact, I think we can remove the return value from all the EME functions
in ContentDecryptorDelegate.
PpapiDecryptor::CreateSession() also checks whether the ContentDecryptorDelegate
member is null. Is a keyerror sufficient in that case?

Android's ProxyMediaKeys::CreateSession() also returns false, but only in
response to the |type|, and it also reports a keyerror. I think we can remove
the return value there. We don't have such checks elsewhere.
Actually, where do we check those. In WD, these types of checks should be done
at a higher level before the task:
https://dvcs.w3.org/hg/html-media/raw-file/default/encrypted-media/encrypted-....
See below regarding piping things like this through.

Not for this CL, but something we should probably clean up.

[xhwang, 2014/02/10 22:30:55]

Created bug 342510 to track this.

   virtual bool CreateSession(uint32 session_id,
-                             const std::string& type,
+                             const std::string& content_type,
                              const uint8* init_data,
                              int init_data_length) = 0;
 
+  // Loads a session with the |web_session_id| provided.
+  // Returns true if a session is successfully loaded, false otherwise.
+  // Note: UpdateSession() and ReleaseSession() should only be called after
+  // SessionCreatedCB is fired.
+  virtual bool LoadSession(uint32 session_id,

[ddorwin, 2014/02/10 19:05:25]

Similar TODO about removing the return value.

Currently, this return value serves to say that CK does not support
loadSession(). However, this is not sufficient for the WD because this call is
in the task, but the algorithm specifies a synchronous exception in step 1:
https://dvcs.w3.org/hg/html-media/raw-file/default/encrypted-media/encrypted-...

We may need to pipe through some feature detection from Blink to
media::MediaKeys. Once that's done, we could remove this bool.

So maybe the TODO is different - and even a bug (or covered by the loadSession()
bug).

[xhwang, 2014/02/10 22:30:55]

Removed "bool" return value in this CL for LoadSession(), and fire
SessionError() if AesDecryptor::LoadSession() is called. Also add a TODO to
clean this later.

+                           const std::string& web_session_id) = 0;
+
   // Updates a session specified by |session_id| with |response|.
   virtual void UpdateSession(uint32 session_id,
                              const uint8* response,