Delete IsMutable and IsExclusive methods.

webrtc/common_video/include/i420_buffer_pool.h

 /*
  *  Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 12 matching lines @@
 // When the I420Buffer is destructed, the memory is returned to the pool for use
 // by subsequent calls to CreateBuffer. If the resolution passed to CreateBuffer
 // changes, old buffers will be purged from the pool.
 class I420BufferPool {
  public:
   I420BufferPool() : I420BufferPool(false) {}
   explicit I420BufferPool(bool zero_initialize);
 
   // Returns a buffer from the pool, or creates a new buffer if no suitable
   // buffer exists in the pool.
-  rtc::scoped_refptr<VideoFrameBuffer> CreateBuffer(int width, int height);
+  rtc::scoped_refptr<I420Buffer> CreateBuffer(int width, int height);
   // Clears buffers_ and detaches the thread checker so that it can be reused
   // later from another thread.
   void Release();
 
  private:
+  class PooledI420Buffer : public webrtc::I420Buffer {

[magjed_webrtc, 2016/05/25 13:17:01]

I don't see the need for this class, why not just use I420Buffer directly?

[nisse-webrtc, 2016/05/25 13:39:47]

To use I420Buffer directly, I have to include HasOneRef in it's public
interface. Ok?

[magjed_webrtc, 2016/05/25 14:56:07]

HasOneRef is already part of the VideoFrameBuffer interface so you don't have to
do anything.

[pbos-webrtc, 2016/05/25 15:26:07]

Can you use a rtc::RefCountedObject<I420Buffer> ?

[nisse-webrtc, 2016/05/26 07:44:42]

This cl deletes it from the interface. I don't think it belongs there (and pbos
doesn't like it either, unless he has changed his mind since yesterday). A
public IsMutable (with a default implementation returning false) is more
reasonable, but we're deleting that method precisely because it's too tricky to
implement correctly for the buffer pool. 

If it's possible to let the buffer pool use the type
rtc::RefCountedObject<I420Buffer> (after all, that's the type it uses when
constructing new buffers, so in principle it should be easy to keep that type
and delay upcast until buffers are returned). But I'm not sure how to type the
corresponding scoped_refptr:s, I'd need a little help with that.

Otherwise, I'd prefer to keep the intermediate PooledI420Buffer class, for the
sole purpose to make HasOneRef accessible.

+   public:
+    PooledI420Buffer(int width, int height) : I420Buffer(width, height) {}
+
+    // Underlying refcount access, used by the pool to check exclusive access.
+    virtual bool HasOneRef() const = 0;
+  };
+
   rtc::ThreadChecker thread_checker_;
-  std::list<rtc::scoped_refptr<I420Buffer>> buffers_;
+  std::list<rtc::scoped_refptr<PooledI420Buffer>> buffers_;
   // If true, newly allocated buffers are zero-initialized. Note that recycled
   // buffers are not zero'd before reuse. This is required of buffers used by
   // FFmpeg according to http://crbug.com/390941, which only requires it for the
   // initial allocation (as shown by FFmpeg's own buffer allocation code). It
   // has to do with "Use-of-uninitialized-value" on "Linux_msan_chrome".
   bool zero_initialize_;
 };
 
 }  // namespace webrtc
 
 #endif  // WEBRTC_COMMON_VIDEO_INCLUDE_I420_BUFFER_POOL_H_