Transfer user agent override info between browser and renderer

content/browser/web_contents/navigation_controller_impl.cc

Index: content/browser/web_contents/navigation_controller_impl.cc
diff --git a/content/browser/web_contents/navigation_controller_impl.cc b/content/browser/web_contents/navigation_controller_impl.cc
index b24eb5b1390c3071230174dcd27948bd9c4f3d93..f475eaf94e2367fc33666f877a97b946d4034aba 100644
--- a/content/browser/web_contents/navigation_controller_impl.cc
+++ b/content/browser/web_contents/navigation_controller_impl.cc
@@ -574,6 +574,25 @@ void NavigationControllerImpl::LoadURL(
     const content::Referrer& referrer,
     content::PageTransition transition,
     const std::string& extra_headers) {
+  bool override = false;
+
+  // Carry over the user agent override if we're doing a regular navigation.
+  bool carry_over = transition == content::PAGE_TRANSITION_TYPED ||
+      transition == content::PAGE_TRANSITION_AUTO_BOOKMARK ||
+      transition == content::PAGE_TRANSITION_GENERATED;

[gone, 2012/05/24 18:26:21]

These conditions were the only ones we encountered on Android, save for the new
tab page.  This will likely have to be generalized, but I don't know the right
ones to use.

[jochen (gone - plz use gerrit), 2012/05/31 14:09:16]

What is the intention? When should the UA override be dropped?

[gone, 2012/05/31 21:43:25]

We intend to carry the override around until a call is made to explicitly turn
it off.  I'm not sure if this should include subframe navigations and the like,
though.

[Charlie Reis, 2012/06/01 00:36:08]

As noted before, I don't understand why it's per-RenderView.  What happens if
you open a link in a new tab?  What happens if you navigate cross-site to a
different process?

It feels odd to me to preserve the setting for a single tab, since we don't do
that for other similar settings like zoom level.

+  if (GetLastCommittedEntry() && carry_over)
+    override = GetLastCommittedEntry()->GetIsOverridingUserAgent();
+
+  LoadURLWithUserAgentOverride(url, referrer, transition, extra_headers,
+      override);
+}
+
+void NavigationControllerImpl::LoadURLWithUserAgentOverride(
+    const GURL& url,
+    const content::Referrer& referrer,
+    content::PageTransition transition,
+    const std::string& extra_headers,
+    bool is_overriding_user_agent) {
   if (content::HandleDebugURL(url, transition))
     return;
 
@@ -583,6 +602,7 @@ void NavigationControllerImpl::LoadURL(
   NavigationEntryImpl* entry = NavigationEntryImpl::FromNavigationEntry(
       CreateNavigationEntry(
           url, referrer, transition, false, extra_headers, browser_context_));
+  entry->SetIsOverridingUserAgent(is_overriding_user_agent);
 
   LoadEntry(entry);
 }
@@ -868,11 +888,8 @@ void NavigationControllerImpl::RendererDidNavigateToNewPage(
       static_cast<SiteInstanceImpl*>(web_contents_->GetSiteInstance()));
   new_entry->SetHasPostData(params.is_post);
   new_entry->SetPostID(params.post_id);
-
-  if (params.redirects.size() > 0)
-    new_entry->SetOriginalRequestURL(params.redirects[0]);
-  else
-    new_entry->SetOriginalRequestURL(params.url);
+  new_entry->SetOriginalRequestURL(params.original_request_url);

[tony, 2012/05/31 17:25:00]

How are original_request_url and url/redirects[0] different? Is this change
intentional?

[gone, 2012/05/31 21:43:25]

It's intentional; the redirect chain doesn't always match the original request
URL: http://b/issue?id=6213026&#ISSUE_HistoryHeader24

[tony, 2012/05/31 21:59:02]

I don't fully understand the issue, but is this bug specific to overriding user
agents? If not, it might be good to split this out into a separate change with a
test.

[gone, 2012/06/01 01:04:06]

Will do.

+  new_entry->SetIsOverridingUserAgent(params.is_overriding_user_agent);
 
   InsertOrReplaceEntry(new_entry, *did_replace_entry);
 }