To fix the cross-site post submission bug.

chrome/browser/ui/browser_navigator.h

Index: chrome/browser/ui/browser_navigator.h
diff --git a/chrome/browser/ui/browser_navigator.h b/chrome/browser/ui/browser_navigator.h
index 5df4136a0243f80a988448795cd27905cee9b319..dea6c2be0b5783c71ff9be630ab0a7cd14d2080c 100644
--- a/chrome/browser/ui/browser_navigator.h
+++ b/chrome/browser/ui/browser_navigator.h
@@ -7,6 +7,9 @@
 
 #include <string>
 
+#include "content/public/common/frame_navigate_params.h"
+
+#include "base/memory/ref_counted.h"
 #include "chrome/browser/ui/host_desktop.h"
 #include "content/public/browser/browser_context.h"
 #include "content/public/browser/global_request_id.h"
@@ -19,6 +22,7 @@
 class Browser;
 class Profile;
 class TabContents;
+using content::WebHTTPPOSTBodyParams;
 
 namespace chrome {
 
@@ -203,6 +207,9 @@ struct NavigateParams {
   // explicitly or inferred from an existing Browser instance.
   chrome::HostDesktopType host_desktop_type;
 
+  // post data
+  std::vector<content::WebHTTPPOSTBodyParams> post_data;

[michaeln, 2012/10/23 23:22:18]

How is this construct different than ResourceRequestBody::Element?

[irobert, 2012/10/24 00:51:17]

This is the exactly same as the ResourceRequestBody::Element. My purpose of
doing this is to pass all the elements in the post request to the new process
through IPC. Since you cannot pass an instance of a class through IPC (based on
my current understanding of the code.), I create this struct and convert each
element to this.

On 2012/10/23 23:22:18, michaeln wrote:

[michaeln, 2012/10/24 23:45:39]

Looks like ResourceRequestBody and it's Element data type do have ipc message
traits so they can be passed around in messages. If possible, it'd be better to
reuse that type instead of creating a new type here.

+
  private:
   NavigateParams();
 };