Use GpuPreferences to avoid directly accessing switches in gpu related code

content/gpu/gpu_child_thread.h

Index: content/gpu/gpu_child_thread.h
diff --git a/content/gpu/gpu_child_thread.h b/content/gpu/gpu_child_thread.h
index 7cbe0fa6f1c781b72c6607fea76cb8c73333dc24..8a9459bad9b175d1748930b41d6611edea53f63c 100644
--- a/content/gpu/gpu_child_thread.h
+++ b/content/gpu/gpu_child_thread.h
@@ -23,6 +23,7 @@
 #include "content/common/gpu/gpu_config.h"
 #include "content/common/gpu/x_util.h"
 #include "content/common/process_control.mojom.h"
+#include "gpu/command_buffer/service/gpu_preferences.h"
 #include "gpu/config/gpu_info.h"
 #include "mojo/public/cpp/bindings/binding_set.h"
 #include "mojo/public/cpp/bindings/interface_request.h"
@@ -50,14 +51,16 @@ class GpuChildThread : public ChildThreadImpl,
  public:
   typedef std::queue<IPC::Message*> DeferredMessages;
 
-  GpuChildThread(GpuWatchdogThread* gpu_watchdog_thread,
+  GpuChildThread(const gpu::GpuPreferences* gpu_preferences,

[boliu, 2016/02/26 22:17:32]

these should be const&

[Peng, 2016/02/29 16:09:22]

pointer is necessary, so the caller can use a nullptr and the GpuChildThread
will construct a GpuPreferences by itself.

[boliu, 2016/02/29 16:26:40]

Then maybe you want another constructor instead.

[Peng, 2016/02/29 19:29:50]

Do we really need another constructor with duplicate code in it, since const
pointer is allowed by our code style?

[boliu, 2016/02/29 19:48:22]

Just a general preference to not have parameters that can be missing if at all
possible. It's easy to have a mistake become a silent failure.

cs.chromium.org claims there is only a single callsite to each constructor, both
from content/gpu, so you don't have to duplicate more code by making
GpuPreference a required argument in these constructors.

[Peng, 2016/02/29 20:30:50]

I do agree it is better to avoid using const pointer in general. But in this
case, if use const reference,
1. we have to put GetGpuPreferences() in somewhere both gpu_main.cc and
in_process_gpu_thread.cc can access.
2. The InProcessGpuThread also need keep a duplicate GpuPreference instance.
3. With follow up CLs, the parameter gpu_preferences will be changed to const
reference.
So I think it is not necessary to do it in this short term CL.

[boliu, 2016/02/29 20:40:07]

content/gpu then? You can just expose that as a static method on GpuChildThread.
That's better than const* imo.

[Peng, 2016/02/29 21:18:29]

Done.

+                 GpuWatchdogThread* gpu_watchdog_thread,
                  bool dead_on_arrival,
                  const gpu::GPUInfo& gpu_info,
                  const DeferredMessages& deferred_messages,
                  GpuMemoryBufferFactory* gpu_memory_buffer_factory,
                  gpu::SyncPointManager* sync_point_manager);
 
-  GpuChildThread(const InProcessChildThreadParams& params,
+  GpuChildThread(const gpu::GpuPreferences* gpu_preferences,
+                 const InProcessChildThreadParams& params,
                  GpuMemoryBufferFactory* gpu_memory_buffer_factory,
                  gpu::SyncPointManager* sync_point_manager);
 
@@ -68,12 +71,12 @@ class GpuChildThread : public ChildThreadImpl,
   void Init(const base::Time& process_start_time);
   void StopWatchdog();
 
+ private:
   // ChildThread overrides.
   bool Send(IPC::Message* msg) override;
   bool OnControlMessageReceived(const IPC::Message& msg) override;
   bool OnMessageReceived(const IPC::Message& msg) override;
 
- private:
   // GpuChannelManagerDelegate implementation.
   void AddSubscription(int32_t client_id, unsigned int target) override;
   void ChannelEstablished(const IPC::ChannelHandle& channel_handle) override;
@@ -131,6 +134,8 @@ class GpuChildThread : public ChildThreadImpl,
   void BindProcessControlRequest(
       mojo::InterfaceRequest<ProcessControl> request);
 
+  gpu::GpuPreferences gpu_preferences_;
+
   // Set this flag to true if a fatal error occurred before we receive the
   // OnInitialize message, in which case we just declare ourselves DOA.
   bool dead_on_arrival_;