Pull up initFrameBuffer to ImageDecoder.

third_party/WebKit/Source/platform/image-decoders/ImageDecoder.h

 /*
  * Copyright (C) 2006 Apple Computer, Inc.  All rights reserved.
  * Copyright (C) Research In Motion Limited 2009-2010. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
@@ skipping 283 matching lines @@
 
   virtual void clearFrameBuffer(size_t frameIndex);
 
   // Decodes the image sufficiently to determine the image size.
   virtual void decodeSize() = 0;
 
   // Decodes the image sufficiently to determine the number of frames and
   // returns that number.
   virtual size_t decodeFrameCount() { return 1; }
 
+  // Called to initialize the frame buffer with the given index, based on the
+  // provided and previous frame's characteristics. Returns true on success. On
+  // failure, this will mark the image as failed. Before calling this method,
+  // the callee must verify that the frame exists.

[scroggo_chromium, 2016/11/14 13:51:45]

caller*

[joostouwerling, 2016/11/14 16:25:09]

Done.

+  virtual bool initFrameBuffer(size_t);
+
+  // Called by initFrameBuffer to determine if it can take the bitmap of the
+  // previous frame. This condition is different for GIF and WEBP.
+  virtual bool canReusePreviousFrameBuffer(size_t) const { return false; }

[scroggo_chromium, 2016/11/14 13:51:45]

This one doesn't actually need to be virtual. We never want to reuse for
DisposeOverwritePrevious or ImageFrame::BlendAtopPreviousFrame; it just so
happens that GIF doesn't use one, and WEBP doesn't use the other.

Once it's no longer virtual, it can be written in line or as a static method in
ImageDecoder.cpp

[joostouwerling, 2016/11/14 16:25:09]

For only GIF and WebP, this would be fine, but for APNG we can reuse
BlendAtopPreviousFrame. So with that it in mind, I think it is better to keep it
like this. 

I made it private though, since there is no need for subclasses to call this
method.

+
   // Performs any additional setup of the requested frame after it has been
   // initially created, e.g. setting a duration or disposal method.
   virtual void initializeNewFrame(size_t) {}
 
   // Decodes the requested frame.
   virtual void decode(size_t) = 0;
 
   RefPtr<SegmentReader> m_data;  // The encoded data.
   Vector<ImageFrame, 1> m_frameBufferCache;
   const bool m_premultiplyAlpha;
@@ skipping 32 matching lines @@
   bool m_isAllDataReceived = false;
   bool m_failed = false;
 
   sk_sp<SkColorSpace> m_embeddedColorSpace = nullptr;
   std::unique_ptr<SkColorSpaceXform> m_sourceToOutputDeviceColorTransform;
 };
 
 }  // namespace blink
 
 #endif