Fix crash on referencing deleted SignedInDevicesManager observer

Fix crash on referencing deleted SignedInDevicesManager observer

Each profile creates new instance of SignedInDevicesManager so, for example,
opening incognito window and then closing it, destroyes an instance of
SignedInDevicesManager created for this profile. The code has forgot to remove
itself from the list of observers in EventRouter.
That caused crash when that event was triggered later in a normal profile as
EventRouter would reference stale pointer.

In tests observer is not added and profile_ is null so null-checking it.

R=rdevlin.cronin@chromium.org
BUG=

Committed: https://crrev.com/d6dcf8725b262caae50f899ab7c47e8ff3980928
Cr-Commit-Position: refs/heads/master@{#319755}

[Rafał Chłodnicki, 2015-03-04 23:01:30]

This API is not used internally AFAIK (we use it in Opera) and I'm not sure of
any extension that uses it. If anyone is aware of one then it should be possible
to reproduce by opening incognito window, closing it and then trying to use
extension in a normal profile. I have not created a bug for this because I
thought it would be difficult to verify anyway (or maybe just because I'm
slightly lazy :/).

[Devlin, 2015-03-09 16:35:43]

Sorry for the delay.

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
File
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc
(right):

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc:119:
if (profile_) {
Will profile_ ever be null?  We don't seem to check it in the constructor...

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc:120:
extensions::EventRouter* router = extensions::EventRouter::Get(profile_);
nit: remove extensions:: prefix (you're in the extensions namespace here).

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc:122:
router->UnregisterObserver(this);
Shame we can't use a ScopedObserver for this... ah well.  (No action required.)

[Rafał Chłodnicki, 2015-03-09 20:46:43]

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
File
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc
(right):

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc:119:
if (profile_) {
On 2015/03/09 16:35:43, Devlin wrote:
> Will profile_ ever be null?  We don't seem to check it in the constructor...

But there are two constructors. One is for testing and has null profile.

https://codereview.chromium.org/980663002/diff/1/chrome/browser/extensions/ap...
chrome/browser/extensions/api/signed_in_devices/signed_in_devices_manager.cc:120:
extensions::EventRouter* router = extensions::EventRouter::Get(profile_);
On 2015/03/09 16:35:43, Devlin wrote:
> nit: remove extensions:: prefix (you're in the extensions namespace here).

Done.

[Devlin, 2015-03-09 20:49:47]

lgtm

[Rafał Chłodnicki, 2015-03-09 21:06:02]

The CQ bit was checked by rchlodnicki@opera.com

[commit-bot: I haz the power, 2015-03-09 21:06:45]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/980663002/20001

[commit-bot: I haz the power, 2015-03-09 23:10:00]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-03-09 23:10:56]

Patchset 2 (id:??) landed as
https://crrev.com/d6dcf8725b262caae50f899ab7c47e8ff3980928
Cr-Commit-Position: refs/heads/master@{#319755}