[Mac] Fix "Chrome windows use more 'hidden' space on Exposé/Mission Control".

chrome/browser/ui/cocoa/status_bubble_mac.mm

Index: chrome/browser/ui/cocoa/status_bubble_mac.mm
===================================================================
--- chrome/browser/ui/cocoa/status_bubble_mac.mm	(revision 133437)
+++ chrome/browser/ui/cocoa/status_bubble_mac.mm	(working copy)
@@ -118,7 +118,8 @@
   Hide();
 
   [[[window_ animationForKey:kFadeAnimationKey] delegate] invalidate];
-  Detach();
+  if (is_attached())

[Mark Mentovai, 2012/04/23 17:59:36]

Why not move this logic into Detach()?

Detach() is private and only called from within this class, and you’ve already
placed identical logic at each call site.

[asvitkine_google, 2012/04/23 18:06:24]

Done.

+    Detach();
   [window_ release];
   window_ = nil;
 }
@@ -447,8 +448,12 @@
   if (state == state_)
     return;
 
-  if (state == kBubbleHidden)
-    [window_ setFrame:NSMakeRect(0, 0, 1, 1) display:YES];
+  if (state == kBubbleHidden) {

[sail, 2012/04/23 18:08:53]

Hi Alexei. I think this change might reintroduce bug 54924.

In r77090 I changed the bubble so that it always remains attached. It seems like
attaching/reattaching causes a lot of things to mess up.

You should also check if you regress bugs 31821 and 61629.

[sail, 2012/04/23 18:10:22]

Also, you may decide that regressing the above bugs is an ok trade off in
exchange for fixing the expose bug.

I don't really have an opinion on that, it's your call.

+    if (is_attached())
+      Detach();
+  } else if (!is_attached()) {
+    Attach();
+  }
 
   if ([delegate_ respondsToSelector:@selector(statusBubbleWillEnterState:)])
     [delegate_ statusBubbleWillEnterState:state];
@@ -680,7 +685,8 @@
   DCHECK(parent);
   DCHECK(is_attached());
 
-  Detach();
+  if (is_attached())
+    Detach();
   parent_ = parent;
   Attach();
   UpdateSizeAndPosition();