Cleanup to keep panel from manipulating its panel strip assignment directly.

chrome/browser/ui/panels/panel_overflow_browsertest.cc

Index: chrome/browser/ui/panels/panel_overflow_browsertest.cc
diff --git a/chrome/browser/ui/panels/panel_overflow_browsertest.cc b/chrome/browser/ui/panels/panel_overflow_browsertest.cc
index 7b50a7e72ff2b4cd5bfabb04a05d62134f493491..6f4ea41c839a23507f9fec0488a8925ec5330118 100644
--- a/chrome/browser/ui/panels/panel_overflow_browsertest.cc
+++ b/chrome/browser/ui/panels/panel_overflow_browsertest.cc
@@ -254,9 +254,9 @@ IN_PROC_BROWSER_TEST_F(PanelOverflowBrowserTest, UpdateDraggableStatus) {
   PanelManager* panel_manager = PanelManager::GetInstance();
   Panel* panel = CreatePanel("panel");
   EXPECT_TRUE(panel->draggable());
-  panel->MoveToStrip(panel_manager->overflow_strip());
+  panel_manager->ChangePanelLayout(panel, PanelStrip::IN_OVERFLOW);
   EXPECT_FALSE(panel->draggable());
-  panel->MoveToStrip(panel_manager->docked_strip());
+  panel_manager->ChangePanelLayout(panel, PanelStrip::DOCKED);
   EXPECT_TRUE(panel->draggable());
   panel->Close();
 }
@@ -649,6 +649,7 @@ IN_PROC_BROWSER_TEST_F(PanelOverflowBrowserTest, CloseWithDelayedOverflow) {
   // CloseWindowAndWait() here because it will allow the delayed overflow
   // to complete.
   docked_strip->RemovePanel(panel1);
+  panel1->SetPanelStrip(NULL);

[Andrei, 2012/03/01 00:08:50]

That would nto be needed if RemovePanel() set it to NULL automatically.

[jennb, 2012/03/01 00:30:58]

Agree. This is the test that led me to try setting NULL in RemovePanel(). I
opted to add this one line rather than adding guards against NULL everywhere
that panel strip is accessed from the panel.

   EXPECT_EQ(1, docked_strip->num_panels());
   EXPECT_EQ(1, docked_strip->num_temporary_layout_panels());
   EXPECT_TRUE(overflow_panel->has_temporary_layout());