Chrome OS multi-profiles backend and UI.

Chrome OS multi-profiles backend and UI.
* Tray - launch login UI for multi-profiles
* UserManager: GetLoggedInUsers(), SwitchActiveUser(),
  add GetActiveUser() which will later replace GetLoggedInUser()
* Login UI: support "sign in to add" mode

Notifications:
* Pass chromeos::User* in details for 
  - NOTIFICATION_LOGIN_USER_CHANGED
  - NOTIFICATION_ACTIVE_USER_CHANGED
  - NOTIFICATION_SESSION_STARTED
* Add NOTIFICATION_ACTIVE_USER_CHANGED (only when switching users for now)

Multi-profile hacks
* Initialize BrowserPolicyConnector only for primary user (http://crbug.com/230349)
* Redirect logging only once (http://crbug.com/230345)
* OAuth2LoginManager tracks only last logged in user (http://crbug.com/230342)

Depends on:
* Changes in ProfileManager https://codereview.chromium.org/14028010/
* Adding concept of "signin profile" https://codereview.chromium.org/13633003/

BUG=180903, 217016
TEST=ProfileManager tests, manual (with all CLs applied and --multi-profiles)

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194185

[Nikita (slow), 2013-04-11 12:34:30]

dpolukhin@ for */chromeos/login/*

sky@ for ash/*, chrome/browser/ui/ash/* and chrome/common/*

sadrul@ for ash/system/user/*, ash/system/tray/* and
chrome/browser/chromeos/system/ash_system_tray_delegate.cc

[Nikita (slow), 2013-04-11 12:36:27]

Mocks: http://goto.google.com/cros-multi-profiles-mocks
Design doc: http://goto.google.com/cros-multi-profiles-mocks

[Nikita (slow), 2013-04-11 13:06:39]

On 2013/04/11 12:36:27, Nikita Kostylev wrote:
> Design doc: http://goto.google.com/cros-multi-profiles-mocks

Correct link for design doc http://goto.google.com/cros-multi-profiles

[Nikita (slow), 2013-04-11 13:12:55]

>> Multi-profile hacks 

Filed issues for each of those.

[Dmitry Polukhin, 2013-04-11 13:34:14]

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/login_utils.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/login_utils.cc:330: bool is_primary_user =
!UserManager::Get()->IsUserLoggedIn();
Please move definition closer to use.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/mock_user_manager.h (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/mock_user_manager.h:92: // You can't mock these
function easily because nobody can create User objects
s/function/functions/

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/user_manager_impl.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:298:
logged_in_users_.insert(logged_in_users_.begin(), active_user_);
Does order of user matter?

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:307: User* user =
const_cast<User*>(FindUser(email));
I would add another function that returns non-const User* instead of adding
const_cast here. Also it will make clear that we would like to modify user.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:543: return;
Do you know how it will look from UI point of view?

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/user_manager_impl.h (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.h:237: User* active_user_;
FYI, perhaps we will need to switch to linked_ptr instead of plain pointers. Now
when we can add/removes active users we need to simplify memory management.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/syste...
File chrome/browser/chromeos/system/ash_system_tray_delegate.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/syste...
chrome/browser/chromeos/system/ash_system_tray_delegate.cc:559:
desktop_background_controller()->MoveDesktopToLockedContainer();
Will it handle multiple monitors? I.e. it looks like a lock screen so it should
be lock screen.

[sadrul, 2013-04-11 13:42:10]

ash/system/, ash_system_tray_delegate.cc LGTM

https://codereview.chromium.org/14139003/diff/6001/chrome/browser/chromeos/sy...
File chrome/browser/chromeos/system/ash_system_tray_delegate.cc (right):

https://codereview.chromium.org/14139003/diff/6001/chrome/browser/chromeos/sy...
chrome/browser/chromeos/system/ash_system_tray_delegate.cc:561:
ShowLoginWizard(std::string(), size);
You could just do: ShowLoginWizard(std::string(), gfx::Size());

[sky, 2013-04-11 16:44:25]

How come we use notifications for all of these instead of more specific
observers?

[Nikita (slow), 2013-04-11 19:59:24]

Do you mean those who care about these notifications should be added as
observer for UserManager instead?

I would prefer to change that separately then as these changes would be
purely mechanical and are out of scope for this CL.

[Dmitry Polukhin, 2013-04-12 08:04:15]

LGTM for chromeos parts

The code has too many TODOs but we need this code committed to parallel further
work.

[sky, 2013-04-12 15:02:14]

On Thu, Apr 11, 2013 at 12:59 PM, Nikita Kostylev
<nkostylev@chromium.org> wrote:
> Do you mean those who care about these notifications should be added as
> observer for UserManager instead?

Exactly.

> I would prefer to change that separately then as these changes would be
> purely mechanical and are out of scope for this CL.

Fair enough. Please file a bug so it isn't missed.

  -Scott

[sky, 2013-04-12 15:03:20]

LGTM

[Nikita (slow), 2013-04-12 15:41:18]

On 2013/04/12 15:02:14, sky wrote:
> On Thu, Apr 11, 2013 at 12:59 PM, Nikita Kostylev
> <mailto:nkostylev@chromium.org> wrote:
> > Do you mean those who care about these notifications should be added as
> > observer for UserManager instead?
> 
> Exactly.
> 
> > I would prefer to change that separately then as these changes would be
> > purely mechanical and are out of scope for this CL.
> 
> Fair enough. Please file a bug so it isn't missed.

Done, thanks!
Filed http://crbug.com/230852

[Nikita (slow), 2013-04-12 15:41:25]

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/login_utils.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/login_utils.cc:330: bool is_primary_user =
!UserManager::Get()->IsUserLoggedIn();
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> Please move definition closer to use.

Done.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/mock_user_manager.h (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/mock_user_manager.h:92: // You can't mock these
function easily because nobody can create User objects
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> s/function/functions/

Done.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/user_manager_impl.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:298:
logged_in_users_.insert(logged_in_users_.begin(), active_user_);
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> Does order of user matter?

Discussed, for now using same ordering logic as for logged in users. We'll
figure out this one later when switch UI will be implemented. Maybe it will make
sense to reorder list of logged in users on switch too.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:307: User* user =
const_cast<User*>(FindUser(email));
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> I would add another function that returns non-const User* instead of adding
> const_cast here. Also it will make clear that we would like to modify user.

Done.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.cc:543: return;
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> Do you know how it will look from UI point of view?

Yes, nothing will happen.
User will click on [Remove] item in menu and menu will just close.

But this should never happen in UI. [Remove] item is not shown on lock/add user
UI.

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
File chrome/browser/chromeos/login/user_manager_impl.h (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/login...
chrome/browser/chromeos/login/user_manager_impl.h:237: User* active_user_;
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> FYI, perhaps we will need to switch to linked_ptr instead of plain pointers.
Now
> when we can add/removes active users we need to simplify memory management.

I'll do that in a separate CL as I've looked over places where it has to be
changed and it will inflate scope of this CL.

Filed http://crbug.com/230852

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/syste...
File chrome/browser/chromeos/system/ash_system_tray_delegate.cc (right):

https://codereview.chromium.org/14139003/diff/1/chrome/browser/chromeos/syste...
chrome/browser/chromeos/system/ash_system_tray_delegate.cc:559:
desktop_background_controller()->MoveDesktopToLockedContainer();
On 2013/04/11 13:34:15, Dmitry Polukhin wrote:
> Will it handle multiple monitors? I.e. it looks like a lock screen so it
should
> be lock screen.

Yes, this method is multi-monitor aware and is already used on lock/unlock.

https://codereview.chromium.org/14139003/diff/6001/chrome/browser/chromeos/sy...
File chrome/browser/chromeos/system/ash_system_tray_delegate.cc (right):

https://codereview.chromium.org/14139003/diff/6001/chrome/browser/chromeos/sy...
chrome/browser/chromeos/system/ash_system_tray_delegate.cc:561:
ShowLoginWizard(std::string(), size);
On 2013/04/11 13:42:11, sadrul wrote:
> You could just do: ShowLoginWizard(std::string(), gfx::Size());

Done.

[Nikita (slow), 2013-04-12 16:21:29]

Filed bunch of new issues tracking new TODOs.

[Nikita (slow), 2013-04-15 16:02:08]

Committed patchset #8 manually as r194185 (presubmit successful).