Have the unload event execute in background on cross-site navigations

chrome/test/base/browser_with_test_window_test.cc

Index: chrome/test/base/browser_with_test_window_test.cc
diff --git a/chrome/test/base/browser_with_test_window_test.cc b/chrome/test/base/browser_with_test_window_test.cc
index 0ec7d96e8acaaeaf0e4096b97d2679d1c1e34d6d..88e4f086d465e4bab09a28064ada303ca4a33466 100644
--- a/chrome/test/base/browser_with_test_window_test.cc
+++ b/chrome/test/base/browser_with_test_window_test.cc
@@ -133,9 +133,6 @@ void BrowserWithTestWindowTest::CommitPendingLoad(
   RenderViewHost* test_rvh = pending_rvh ? pending_rvh : old_rvh;
   RenderViewHostTester* test_rvh_tester = RenderViewHostTester::For(test_rvh);
 
-  // Simulate a SwapOut_ACK before the navigation commits.
-  if (pending_rvh)
-    RenderViewHostTester::For(old_rvh)->SimulateSwapOutACK();
 
   // For new navigations, we need to send a larger page ID. For renavigations,
   // we need to send the preexisting page ID. We can tell these apart because
@@ -153,6 +150,9 @@ void BrowserWithTestWindowTest::CommitPendingLoad(
         controller->GetPendingEntry()->GetURL(),
         controller->GetPendingEntry()->GetTransitionType());
   }
+  // Simulate the SwapOut_ACK from the renderer.

[nasko, 2014/01/23 23:34:58]

Why did this move? 

[clamy, 2014/01/24 17:01:54]

It was causing a problem with some tests in unit_tests (the RVH would not end up
in the right state and would not be deleted). However, it must have come from
another underlying issue that was fixed, since removing this change does not
seem to affect tests at all. I undid the move.

+  if (pending_rvh)
+    RenderViewHostTester::For(old_rvh)->SimulateSwapOutACK();
 }
 
 void BrowserWithTestWindowTest::NavigateAndCommit(