Unified the use of BufferToTextureTargetMap for all the Processes

content/browser/gpu/compositor_util.h

 // Copyright 2013 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.
 
 #ifndef CONTENT_BROWSER_GPU_COMPOSITOR_UTIL_H_
 #define CONTENT_BROWSER_GPU_COMPOSITOR_UTIL_H_
 
 #include "base/values.h"
+#include "cc/output/buffer_to_texture_target_map.h"
 #include "content/common/content_export.h"
 
 namespace content {
 
 // Note: When adding a function here, please make sure the logic is not
 // duplicated in the renderer.
 
 // Returns true if zero-copy uploads is on (via flags, or platform default).
 // Only one of one-copy and zero-copy can be enabled at a time.
 CONTENT_EXPORT bool IsZeroCopyUploadEnabled();
@@ skipping 20 matching lines @@
 // Returns true if main thread can be pipelined with activation.
 CONTENT_EXPORT bool IsMainFrameBeforeActivationEnabled();
 
 // Returns true if images can be decode asynchronously from rasterization.
 CONTENT_EXPORT bool IsCheckerImagingEnabled();
 
 CONTENT_EXPORT base::DictionaryValue* GetFeatureStatus();
 CONTENT_EXPORT base::Value* GetProblems();
 CONTENT_EXPORT std::vector<std::string> GetDriverBugWorkarounds();
 
+// Populate buffer_to_texture_target_map for all buffer usage/formats.
+CONTENT_EXPORT cc::BufferToTextureTargetMap GetBufferToTextureTargetMap();

[reveman, 2017/06/09 14:28:46]

Can this be "void GetBufferToTextureTargetMap(cc::BufferToTextureTargetMap*
texture_targets)" so caller is not forced to make a copy of a map?

Also, would be nice to rename
settings.resource_settings.buffer_to_texture_target_map image_targets and change
the name of this a related functions to GetImageTargets. A better name for the
map type is BufferConfigurationTextureTargetMap imo. Fine to leave these last
two things to a follow up patch if preferred.

[sujith, 2017/06/14 10:30:18]

when i was trying the modifications for 
GetBufferToTextureTargetMap(cc::BufferToTextureTargetMap* texture_targets)
i think we need to create seperate variables each time to access it.
For example 

 GpuProcessTransportFactory::GpuProcessTransportFactory()
     : frame_sink_id_allocator_(kDefaultClientId),
       renderer_settings_(
          ui::CreateRendererSettings(GetBufferToTextureTargetMap())),

render_settings_ get initialized in constructor, so if we move it inside,
need to overload operator= 
so could you please suggest what approach should i follow.

Regarding the name change, as you suggest i will do a follow up patch.

[reveman, 2017/06/14 13:03:38]

You can remove it from the initialization list or just create a local wrapper
around GetBufferToTextureTargetMap for this specific use-case.

[danakj, 2017/06/29 16:06:54]

It's a move not a copy, and it would make use of RVO. Returning is the correct
way to do this in C++11.

+
 }  // namespace content
 
 #endif  // CONTENT_BROWSER_GPU_COMPOSITOR_UTIL_H_