PlzNavigate: fix DevToolsProtocolTest.CrossSitePauseInBeforeUnload

content/browser/frame_host/navigation_request.cc

Index: content/browser/frame_host/navigation_request.cc
diff --git a/content/browser/frame_host/navigation_request.cc b/content/browser/frame_host/navigation_request.cc
index e45133bcfd436322384ff0de177721e129c16fab..be76acb9ccf913276245a64683240d3ccf66b1ae 100644
--- a/content/browser/frame_host/navigation_request.cc
+++ b/content/browser/frame_host/navigation_request.cc
@@ -6,6 +6,7 @@
 
 #include <utility>
 
+#include "content/browser/devtools/render_frame_devtools_agent_host.h"
 #include "content/browser/frame_host/frame_tree.h"
 #include "content/browser/frame_host/frame_tree_node.h"
 #include "content/browser/frame_host/navigation_controller_impl.h"
@@ -205,6 +206,7 @@ void NavigationRequest::BeginNavigation() {
   DCHECK(!loader_);
   DCHECK(state_ == NOT_STARTED || state_ == WAITING_FOR_RENDERER_RESPONSE);
   state_ = STARTED;
+  RenderFrameDevToolsAgentHost::OnBeforeNavigation(navigation_handle_.get());

[dgozman, 2016/02/25 18:11:24]

So, with PlzNavigate we do not ever send FrameMsg_Navigate to the frame (that's
where we got OnBeforeNavigation before)? Will we end up with two calls?
I'm asking because I don't really know how the new scheme works to verify
whether this is a right place to call OnBeforeNavigation.

[clamy, 2016/03/02 12:33:50]

Yes we never send FrameMsg_Navigate. The place here corresponds to the moment
where we received the BeforeUnload ACK and we are going to send the navigation
to the network stack. So OnBeforeNavigation will always be called just once.

 
   if (ShouldMakeNetworkRequestForURL(common_params_.url)) {
     // It's safe to use base::Unretained because this NavigationRequest owns