Don't recreate feature policy on fragment navigation

content/browser/frame_host/navigator_impl.cc

Index: content/browser/frame_host/navigator_impl.cc
diff --git a/content/browser/frame_host/navigator_impl.cc b/content/browser/frame_host/navigator_impl.cc
index 07f3cc9d1eb6137155132c4f146032ea6b1efe7d..24dc9aa09df1400c29861f7eb7d3d4ec4e592fea 100644
--- a/content/browser/frame_host/navigator_impl.cc
+++ b/content/browser/frame_host/navigator_impl.cc
@@ -685,7 +685,8 @@ void NavigatorImpl::DidNavigate(
 
   // After setting the last committed origin, reset the feature policy in the
   // RenderFrameHost to a blank policy based on the parent frame.
-  render_frame_host->ResetFeaturePolicy();
+  if (!did_navigate)

[alexmos, 2017/02/24 21:21:12]

A couple of concerns with this.  First, I think you meant "if (did_navigate)",
as the negation will skip the ResetFeaturePolicy() on almost every navigation,
including normal cross-site navigations.

But also to skip fragments and other in-page navigations, "if
(!is_navigation_within_page)" seems to be the better thing to check. 
|did_navigate| returns false only for auto_subframe and ignored navigations, but
I think it's possible to navigate to a hash on the same page for other types of
navigations. (E.g., from comment on NAVIGATION_TYPE_NEW_PAGE: "cases like
navigations to a hash on the same page are NEW_PAGE").

(I also ran this by creis@ to double check, and he agreed.)

[iclelland, 2017/03/03 15:26:31]

Thanks for that -- I've updated the condition, to include both |did_navigate|
and !is_navigation_within_page, and added tests to ensure that the policy is not
reset on fragment navigation.

+    render_frame_host->ResetFeaturePolicy();
 
   // Send notification about committed provisional loads. This notification is
   // different from the NAV_ENTRY_COMMITTED notification which doesn't include