Oilpan: fix build after r182644.

Source/modules/encryptedmedia/HTMLMediaElementEncryptedMedia.cpp

 // Copyright 2014 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "config.h"
 #include "modules/encryptedmedia/HTMLMediaElementEncryptedMedia.h"
 
 #include "bindings/core/v8/ExceptionState.h"
 #include "bindings/core/v8/ScriptPromise.h"
 #include "bindings/core/v8/ScriptPromiseResolver.h"
@@ skipping 45 matching lines @@
 private:
     SetMediaKeysHandler(ScriptState*, HTMLMediaElement&, MediaKeys*);
     void timerFired(Timer<SetMediaKeysHandler>*);
 
     void clearExistingMediaKeys();
     void setNewMediaKeys();
     void finish();
 
     void reportSetFailed(ExceptionCode, const String& errorMessage);
 
-    RawPtrWillBeMember<HTMLMediaElement> m_element;
-    Persistent<MediaKeys> m_newMediaKeys;
+    RawPtrWillBePersistent<HTMLMediaElement> m_element;

[haraken, 2014/09/25 03:35:27]

Are you sure that the persistent handle doesn't cause leaks?
RawPtrWillBePersistent is very likely to cause a memory leak.

In the beginning, I want to understand whether the m_element should be a RawPtr
or a RefPtr in non-oilpan builds. If it should be a RefPtr, using a
RefPtrWillBePersistent makes sense. If it should be a RawPtr (i.e., using a
RefPtr causes a leak), that implies that using a Persistent causes a leak as
well.

[DanielChow, 2014/09/25 03:48:32]

I am not sure that the persistent handle doesn't cause leaks. I am running test
to checkout it.

[sof, 2014/09/25 05:21:24]

Definitely worth checking, run them with --enable-leak-detection.

If there's a known upper bound of how long the promise will stay unfulfilled,
then a Persistent would be acceptable, I think. But clear the m_element
Persistent when done with it & at the earliest possible time.

So, is it known how long these Promises stay unfulfilled?

[haraken, 2014/09/25 05:45:02]

If that's the case, I think we should use a RefPtr in non-oilpan builds.

+    RawPtrWillBePersistent<MediaKeys> m_newMediaKeys;

[haraken, 2014/09/25 03:35:27]

MediaKeys are on-heap in both oilpan and non-oilpan, so it's wrong to use a
RawPtr in non-oilpan. I think Persistent<MediaKeys> is correct.

[DanielChow, 2014/09/25 03:48:32]

Done.

     Timer<SetMediaKeysHandler> m_timer;
 };
 
 typedef Function<void()> SuccessCallback;
 typedef Function<void(ExceptionCode, const String&)> FailureCallback;
 
 // Represents the result used when setContentDecryptionModule() is called.
 // Calls |success| if result is resolved, |failure| is result is rejected.
 class SetContentDecryptionModuleResult FINAL : public ContentDecryptionModuleResult {
 public:
@@ skipping 462 matching lines @@
     return thisElement.contentDecryptionModule();
 }
 
 void HTMLMediaElementEncryptedMedia::trace(Visitor* visitor)
 {
     visitor->trace(m_mediaKeys);
     WillBeHeapSupplement<HTMLMediaElement>::trace(visitor);
 }
 
 } // namespace blink