Pop-up chrome....

chrome/browser/cocoa/browser_window_controller_unittest.mm

 // Copyright (c) 2009 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/scoped_nsobject.h"
 #include "base/scoped_nsautorelease_pool.h"
 #include "base/scoped_ptr.h"
 #include "chrome/browser/browser_process.h"
 #include "chrome/browser/cocoa/browser_test_helper.h"
 #include "chrome/browser/cocoa/browser_window_controller.h"
@@ skipping 59 matching lines @@
   EXPECT_FALSE([controller isFullscreen]);
   [controller setFullscreen:YES];
   EXPECT_TRUE([controller isFullscreen]);
   [controller setFullscreen:NO];
   EXPECT_FALSE([controller isFullscreen]);
 
   // Confirm the real fullscreen command doesn't return nil
   EXPECT_TRUE([controller_ fullscreenWindow]);
 }
 
+TEST_F(BrowserWindowControllerTest, TestNormal) {
+  // Make sure a normal BrowserWindowController is, uh, normal.
+  EXPECT_TRUE([controller_ isNormalWindow]);
+
+  // And make sure a controller for a pop-up window is not normal.
+  scoped_ptr<Browser> popup_browser(Browser::CreateForPopup(
+                                      browser_helper_.profile()));
+  controller_.reset([[BrowserWindowController alloc]
+                              initWithBrowser:popup_browser.get()
+                                takeOwnership:NO]);
+  EXPECT_FALSE([controller_ isNormalWindow]);
+
+  // The created BrowserWindowController gets autoreleased, so make
+  // sure we don't also release it.
+  // (Confirmed with valgrind).
+  controller_.release();

[pink (ping after 24hrs), 2009/07/01 12:53:00]

why is this necessary? won't it do that when the test finishes?

[John Grabowski, 2009/07/02 20:28:25]

Somehow the controller gets autoreleased.  I can confirm this by seeing it's
dealloc called when an autorelease pool gets drained.  I could not, however,
figure out when it gets autoreleased.  My best guess is [BrowserWindowController
destroyBrowser], but a breakpoint there wasn't hit.  Maybe gdb doesn't like me
today.

I can also confirm that valgrind is happy with this change, and that an extra
release of this object blows up (dbl-free).  

I wish I had a better answer.

+}
+
 /* TODO(???): test other methods of BrowserWindowController */