Delete desktop/shell instance correctly when chrome is shutting down

Delete desktop/shell instance correctly upon shutdown.
* Moved parts->PostMainMessageLoopRun call from PostDestroyThreads to the right place (PostMainMessageLoopRun() method)
* Renamed Shell::DeleteInstanceForTesting to DeleteInstance as this is no longer for testing.

BUG=106070
TEST=browser_tests on aura will not crash as described in the bug

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=112836

[oshima, 2011-12-02 00:45:27]

stevenjb -> chrome/browser/
ben -> ui/

[stevenjb, 2011-12-02 01:30:54]

http://codereview.chromium.org/8774029/diff/1001/chrome/browser/chrome_browse...
File chrome/browser/chrome_browser_main_extra_parts_aura.cc (right):

http://codereview.chromium.org/8774029/diff/1001/chrome/browser/chrome_browse...
chrome/browser/chrome_browser_main_extra_parts_aura.cc:34:
aura::Desktop::DeleteInstance();
Since aura::Shell::Init() calls aura::Desktop::GetInstance(), shouldn't
aura::Desktop::GetInstance() call aura::Desktop::DeleteInstance()?

http://codereview.chromium.org/8774029/diff/1001/ui/aura/desktop_host_linux.cc
File ui/aura/desktop_host_linux.cc (right):

http://codereview.chromium.org/8774029/diff/1001/ui/aura/desktop_host_linux.c...
ui/aura/desktop_host_linux.cc:598: aura::Desktop::DeleteInstance();
Do we still need do this here?

[oshima, 2011-12-02 01:45:43]

http://codereview.chromium.org/8774029/diff/1001/chrome/browser/chrome_browse...
File chrome/browser/chrome_browser_main_extra_parts_aura.cc (right):

http://codereview.chromium.org/8774029/diff/1001/chrome/browser/chrome_browse...
chrome/browser/chrome_browser_main_extra_parts_aura.cc:34:
aura::Desktop::DeleteInstance();
On 2011/12/02 01:30:55, Steven Bennetts wrote:
> Since aura::Shell::Init() calls aura::Desktop::GetInstance(), shouldn't
> aura::Desktop::GetInstance() call aura::Desktop::DeleteInstance()?

I assume you mean aura_shell::Shell::DeleteInstance() calls
aura::Desktop::DeleteInstance().
Yes, I thought about it. Technically speaking, desktop instance can outlive
shell instance and
that's the reason why I kept them separated, but I can be convinced either way.
Ben, what do you think?

http://codereview.chromium.org/8774029/diff/1001/ui/aura/desktop_host_linux.cc
File ui/aura/desktop_host_linux.cc (right):

http://codereview.chromium.org/8774029/diff/1001/ui/aura/desktop_host_linux.c...
ui/aura/desktop_host_linux.cc:598: aura::Desktop::DeleteInstance();
On 2011/12/02 01:30:55, Steven Bennetts wrote:
> Do we still need do this here?

yes, for tests that does not depend chrome

[oshima, 2011-12-02 16:35:19]

Uploaded new patch, which closes windows before shell's window observers are
destroyed. It looks a bit messy, which I don't like, but this works. We need
consistent and better way to handle shell shutdown, but it'd require more
changes to various controllers, so I'd like to do it in separate CL.

[stevenjb, 2011-12-02 20:10:27]

The ChromeBrowserMainParts changes look fine by me, but I am not sure I entirely
follow the changes in shell.cc, I will leave it to ben to examine those.

[Ben Goodger (Google), 2011-12-02 20:25:35]

lgtm

[commit-bot: I haz the power, 2011-12-02 20:36:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/oshima@chromium.org/8774029/4007

[commit-bot: I haz the power, 2011-12-03 00:24:26]

The commit queue went berserk retrying too often for a
seemingly flaky test. Builder is win_rel, revision is 112804, job name
was 8774029-4007 (retry) (retry) (previous was lost) (previous was lost).