Persist prompt/block download limiter state on renderer-initiated loads.

content/browser/frame_host/navigation_handle_impl.h

Index: content/browser/frame_host/navigation_handle_impl.h
diff --git a/content/browser/frame_host/navigation_handle_impl.h b/content/browser/frame_host/navigation_handle_impl.h
index 07501267d8287e0f2dfbc7d0ff520239ba6792e0..66f3774d04f078c786fe2d338f8baa4f2bffd31c 100644
--- a/content/browser/frame_host/navigation_handle_impl.h
+++ b/content/browser/frame_host/navigation_handle_impl.h
@@ -69,6 +69,7 @@ class CONTENT_EXPORT NavigationHandleImpl : public NavigationHandle {
   static std::unique_ptr<NavigationHandleImpl> Create(
       const GURL& url,
       FrameTreeNode* frame_tree_node,
+      bool is_renderer_initiated,
       bool is_synchronous,
       bool is_srcdoc,
       const base::TimeTicks& navigation_start,
@@ -79,6 +80,7 @@ class CONTENT_EXPORT NavigationHandleImpl : public NavigationHandle {
   const GURL& GetURL() override;
   bool IsInMainFrame() override;
   bool IsParentMainFrame() override;
+  bool IsRendererInitiated() override;
   bool IsSynchronousNavigation() override;
   bool IsSrcdoc() override;
   bool WasServerRedirect() override;
@@ -120,6 +122,14 @@ class CONTENT_EXPORT NavigationHandleImpl : public NavigationHandle {
   // as modifications will not be reflected in the network stack.
   const net::HttpResponseHeaders* GetResponseHeaders();
 
+  // Update whether the current navigation is renderer-initiated. Currently
+  // required during transfer navigations.
+  // TODO(creis): Remove this when transfer navigations do not require pending
+  // entries. See https://crbug.com/495161.

[Charlie Reis, 2016/05/11 21:50:03]

Yikes, looks like most of this was copied from update_entry_id_for_transfer,
which has a stale comment.  I've recently fixed https://crbug.com/495161 and we
don't use a pending entry during transfer.  (I'll have to see if I can remove
update_entry_id_for_transfer now.)

I'm wondering if this is dead code as well.  I'll try to find time to
investigate if I can, but let me know if you're aware of a case this is needed
for.

[clamy, 2016/05/12 04:04:44]

Considering how the NavigationHandle is used (ie from start to stop), there
should be no reason to change is_renderer_initiated ever over the course of the
navigation. In transfer navigations we transfer the actual NavigationHandle, so
there's no need to update is_renderer_initiated.

[dominickn, 2016/05/12 04:22:58]

I did just copy this from the update_entry_id_for_transfer; I don't really know
enough about the navigation process to say whether or not this case is needed.
Based on these comments I've removed the method and its caller.

+  void set_is_renderer_initiated(bool is_renderer_initiated) {
+    is_renderer_initiated_ = is_renderer_initiated;
+  }
+
   // Get the unique id from the NavigationEntry associated with this
   // NavigationHandle. Note that a synchronous, renderer-initiated navigation
   // will not have a NavigationEntry associated with it, and this will return 0.
@@ -231,6 +241,7 @@ class CONTENT_EXPORT NavigationHandleImpl : public NavigationHandle {
 
   NavigationHandleImpl(const GURL& url,
                        FrameTreeNode* frame_tree_node,
+                       bool is_renderer_initiated,
                        bool is_synchronous,
                        bool is_srcdoc,
                        const base::TimeTicks& navigation_start,
@@ -255,6 +266,7 @@ class CONTENT_EXPORT NavigationHandleImpl : public NavigationHandle {
   bool is_external_protocol_;
   net::Error net_error_code_;
   RenderFrameHostImpl* render_frame_host_;
+  bool is_renderer_initiated_;

[clamy, 2016/05/12 04:04:44]

I don't see a good reason why this should change over the course of the
navigation. This should be const.

[dominickn, 2016/05/12 04:22:58]

Done.

   bool is_same_page_;
   const bool is_synchronous_;
   const bool is_srcdoc_;