BrowserMainLoop owns FrameSinkManagerHost

content/browser/browser_main_loop.cc

Index: content/browser/browser_main_loop.cc
diff --git a/content/browser/browser_main_loop.cc b/content/browser/browser_main_loop.cc
index aa8fa9b89f30b8b5346cb10ad3302ffa60dbb203..8de1892d4ecc640a9a0b457202396950bfa96fdc 100644
--- a/content/browser/browser_main_loop.cc
+++ b/content/browser/browser_main_loop.cc
@@ -53,6 +53,7 @@
 #include "components/tracing/common/trace_to_console.h"
 #include "components/tracing/common/tracing_switches.h"
 #include "components/viz/display_compositor/host_shared_bitmap_manager.h"
+#include "components/viz/host/frame_sink_manager_host.h"
 #include "content/browser/browser_thread_impl.h"
 #include "content/browser/child_process_security_policy_impl.h"
 #include "content/browser/dom_storage/dom_storage_area.h"
@@ -1224,6 +1225,8 @@ void BrowserMainLoop::ShutdownThreadsAndCleanUp() {
   BrowserCompositorMac::DisableRecyclingForShutdown();
 #endif
 
+  frame_sink_manager_host_.reset();
+
 #if defined(USE_AURA) || defined(OS_MACOSX)
   {
     TRACE_EVENT0("shutdown",
@@ -1455,6 +1458,10 @@ int BrowserMainLoop::BrowserThreadsStarted() {
   }
 #endif  // defined(USE_AURA)
 #endif  // defined(OS_ANDROID)
+  if (!service_manager::ServiceManagerIsRemote()) {

[danakj, 2017/06/09 22:54:09]

Don't we need a FrameSinkManagerHost when the ServiceManagerIsRemote too?

[Fady Samuel, 2017/06/09 22:55:00]

No, the FrameSinkManagerHost is in the window server not in the browser process
when running in Mus+Ash.

[danakj, 2017/06/09 22:57:45]

Oh I see, this is different than FrameSinkManagerIsRemote :)

+    frame_sink_manager_host_ = base::MakeUnique<viz::FrameSinkManagerHost>();
+    frame_sink_manager_host_->ConnectToFrameSinkManager();
+  }
 
   // Enable the GpuMemoryBuffer dump provider with IO thread affinity. Note that
   // unregistration happens on the IO thread (See