Move multiple UI surfaces check to PageActionHandler.

chrome/common/extensions/api/extension_action/page_action_handler.cc

Index: chrome/common/extensions/api/extension_action/page_action_handler.cc
diff --git a/chrome/common/extensions/api/extension_action/page_action_handler.cc b/chrome/common/extensions/api/extension_action/page_action_handler.cc
index 136693b444f5f2527660dc8c879a6c950ba211a2..b39fe78d33bdb9fa89a3d0be99ea7d8929c9a30e 100644
--- a/chrome/common/extensions/api/extension_action/page_action_handler.cc
+++ b/chrome/common/extensions/api/extension_action/page_action_handler.cc
@@ -56,6 +56,12 @@ bool PageActionHandler::Parse(Extension* extension, string16* error) {
     }
   }
 
+  // An extension cannot have both browser and page actions.
+  if (extension->manifest()->HasKey(keys::kBrowserAction)) {

[Matt Perry, 2013/05/30 23:16:14]

The asymmetry of this bothers me (as in, why not put this in the
BrowserActionHandler? What would we do if we had a third UI surface?)... but I
don't have a better solution. 

[Yoyo Zhou, 2013/05/30 23:24:11]

It's kind of a hack. A better or worse hack would be to put the check in
BrowserActionHandler too.

Less of a hack would be a post-Parse step where we introduce registered
constraints checkers. But this check is simple enough that it doesn't warrant
that kind of complexity.

+    *error = ASCIIToUTF16(errors::kOneUISurfaceOnly);
+    return false;
+  }
+
   // If page_action_value is not NULL, then there was a valid page action.
   if (page_action_value) {
     page_action_info = ActionInfo::Load(extension, page_action_value, error);