Correct bad WebKit merge....

webkit/glue/webframeloaderclient_impl.cc

Index: webkit/glue/webframeloaderclient_impl.cc
===================================================================
--- webkit/glue/webframeloaderclient_impl.cc	(revision 20246)
+++ webkit/glue/webframeloaderclient_impl.cc	(working copy)
@@ -241,6 +241,16 @@
     request.setTargetType(DetermineTargetTypeFromLoader(loader));
   }
 
+  // Inherit the first party for cookies from the request's frame.  However, if
+  // the request is for a main frame, the current document's
+  // firstPartyForCookies is the old document, so we leave firstPartyForCookies
+  // empty to indicate that the request is a first-party request.
+  if (request.targetType() != ResourceRequest::TargetIsMainFrame &&

[darin (slow to review), 2009/07/12 03:45:40]

Shouldn't this kind of logic live in WebCore?  Or, is this really Chromium
specific?

[abarth-chromium, 2009/07/12 08:10:51]

This patch is just added back some code I accidentally deleted in a previous
patch.  If this fixes the regression, I'd be inclined to land the patch as is
(to fix the Mstone-3 P1) and then file a followup bug to move to logic to
WebCore.

[darin (slow to review), 2009/07/12 13:05:04]

OK,LGTM

+      webframe_->frame()->document()) {
+    request.setFirstPartyForCookies(
+        webframe_->frame()->document()->firstPartyForCookies());
+  }
+
   // FrameLoader::loadEmptyDocumentSynchronously() creates an empty document
   // with no URL.  We don't like that, so we'll rename it to about:blank.
   if (request.url().isEmpty())