Check that shared memory GpuMemoryBuffer opening succeeds

content/common/gpu/client/gpu_memory_buffer_impl.cc

Index: content/common/gpu/client/gpu_memory_buffer_impl.cc
diff --git a/content/common/gpu/client/gpu_memory_buffer_impl.cc b/content/common/gpu/client/gpu_memory_buffer_impl.cc
index e900829fda34756318a7a21ab8f7d4d55b8ffb51..072b9bc5be3861374d2c18fc6c6ffcb51402d6e7 100644
--- a/content/common/gpu/client/gpu_memory_buffer_impl.cc
+++ b/content/common/gpu/client/gpu_memory_buffer_impl.cc
@@ -44,9 +44,13 @@ scoped_ptr<GpuMemoryBufferImpl> GpuMemoryBufferImpl::CreateFromHandle(
     gfx::BufferUsage usage,
     const DestructionCallback& callback) {
   switch (handle.type) {
-    case gfx::SHARED_MEMORY_BUFFER:
-      return GpuMemoryBufferImplSharedMemory::CreateFromHandle(
-          handle, size, format, usage, callback);
+    case gfx::SHARED_MEMORY_BUFFER: {
+      scoped_ptr<GpuMemoryBufferImpl> buffer =
+          GpuMemoryBufferImplSharedMemory::CreateFromHandle(
+              handle, size, format, usage, callback);
+      CHECK(buffer);

[reveman, 2015/12/08 18:05:50]

How about changing the base::SharedMemory::IsHandleValid check in
GpuMemoryBufferImplSharedMemory::CreateFromHandle to a DCHECK and then we'll
fail in release builds properly with an OOM error when trying to map the buffer?
That would be consistent with how we handle this type of errors for shared
bitmaps and discardable memory.

+      return buffer.Pass();
+    }
 #if defined(OS_MACOSX)
     case gfx::IO_SURFACE_BUFFER:
       return GpuMemoryBufferImplIOSurface::CreateFromHandle(