Fix fullscreen mode on aura

Fix fullscreen mode on aura
* Switching to fullscreen on aura is synchronous. Use the same path as windows.
* Use dispatcher to run all pending in in_process_browser_tests.
* Check if the window is already destroyed in NatveWidgetAura::Hide

BUG=103485
TEST=following tests should pass with this patch on aura: BrowserTest.TestFullscreenBubbleMouseLockState
BrowserTest.TestNewTabExitsFullscreen
BrowserTest.TestTabExitsItselfFromFullscreen


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=110337

[Ben Goodger (Google), 2011-11-16 15:48:24]

http://codereview.chromium.org/8571026/diff/2001/views/widget/native_widget_a...
File views/widget/native_widget_aura.cc (right):

http://codereview.chromium.org/8571026/diff/2001/views/widget/native_widget_a...
views/widget/native_widget_aura.cc:378: if (window_)
This seems kind of random. What other functions in NWA might be called after
window_ is randomly deleted?

[oshima, 2011-11-16 16:51:47]

http://codereview.chromium.org/8571026/diff/2001/views/widget/native_widget_a...
File views/widget/native_widget_aura.cc (right):

http://codereview.chromium.org/8571026/diff/2001/views/widget/native_widget_a...
views/widget/native_widget_aura.cc:378: if (window_)
On 2011/11/16 15:48:25, Ben Goodger (Google) wrote:
> This seems kind of random. What other functions in NWA might be called after
> window_ is randomly deleted?

On second thought, I think I should move this to Close, like
if (window_)
 Hide();

This happens when a window that has either child or transient child (similar to
tab contents detach fix i did a copule of weeks ago) is closing. Deleting parent
window deletes child & transient child before deleting child views, and deleting
view may try to close the NWA whose window has already been deleted.
 Since this should happen only when we're closing the window, Close is probably 
better place. This should also be possible when ownership is
WIDGET_OWNS_NATIVE_WIDGET, so I can add DCHECK. Please let me know what you
think.

BTW. my understanding of why this problem doesn't happen on gtk is because gtk
widget is refcounted. Not sure how this works on windows though.

[oshima, 2011-11-16 17:22:42]

uploaded new patch. PTAL

[Ben Goodger (Google), 2011-11-16 17:32:52]

lgtm

[commit-bot: I haz the power, 2011-11-16 18:20:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/oshima@chromium.org/8571026/6005

[commit-bot: I haz the power, 2011-11-16 19:50:04]

Change committed as 110337