PlzNavigate: have renderer-initiated navigations be same-process

content/browser/frame_host/render_frame_host_manager.cc

Index: content/browser/frame_host/render_frame_host_manager.cc
diff --git a/content/browser/frame_host/render_frame_host_manager.cc b/content/browser/frame_host/render_frame_host_manager.cc
index b9328a40b2539c35bda23a47ff0299f2e67c4949..7fab96306c439b9d5dca4010475851f10fcec93d 100644
--- a/content/browser/frame_host/render_frame_host_manager.cc
+++ b/content/browser/frame_host/render_frame_host_manager.cc
@@ -720,9 +720,10 @@ RenderFrameHostImpl* RenderFrameHostManager::GetFrameHostForNavigation(
   // The appropriate RenderFrameHost to commit the navigation.
   RenderFrameHostImpl* navigation_rfh = nullptr;
 
-  // TODO(carlosk): do not swap processes for renderer initiated navigations
-  // (see crbug.com/440266).
+  // TODO(carlosk): Once there is support for cross-process scripting remove the

[clamy, 2015/03/06 13:43:09]

I think you should add a little more explanation on why this is safe to do.
Something like: "Renderer-initiated navigations that may require a SiteInstance
swap are sent to the browser via the OpenURL IPC, and are afterwards treated as
browser-initiated navigations. The only NavigationRequests marked as
renderer-initiated were created after receiving a BeginNavigation IPC, and
should proceed in the same renderer that sent the IPC."

[carlosk, 2015/03/06 15:44:15]

Done. I didn't realized/remembered that situation but indeed an "open in new
tab" renderer initiated navigation will follow that other path.

+  // check for renderer-initiated navigations. See crbug.com/440266.
   if (current_site_instance == dest_site_instance.get() ||
+      !request.browser_initiated() ||
       (!frame_tree_node_->IsMainFrame() &&
        !base::CommandLine::ForCurrentProcess()->HasSwitch(
            switches::kSitePerProcess))) {