PART4: Removed the owner cache special handling from UserCrosSettingsProvider.

Removed the owner cache special handling from UserCrosSettingsProvider.

Instead we simply never save this down to the policy blob.

BUG=chromium-os:14054
TEST=Existing tests should still pass.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=111157

[pastarmovj, 2011-10-06 14:22:21]

And the owner "sprecial" ops ;)

[Mattias Nissler (ping if slow), 2011-10-07 13:46:52]

LGTM if you can answer my question ;)

http://codereview.chromium.org/8168010/diff/1/chrome/browser/chromeos/login/s...
File chrome/browser/chromeos/login/signed_settings.cc (right):

http://codereview.chromium.org/8168010/diff/1/chrome/browser/chromeos/login/s...
chrome/browser/chromeos/login/signed_settings.cc:367: // We don't serialize this
one to the policy blob.
It's actually good to make it a NOTREACHED(), do you know why we had this branch
before?

[pastarmovj, 2011-10-13 12:00:41]

Commented on comment and rebased on fresh PART3.

http://codereview.chromium.org/8168010/diff/1/chrome/browser/chromeos/login/s...
File chrome/browser/chromeos/login/signed_settings.cc (right):

http://codereview.chromium.org/8168010/diff/1/chrome/browser/chromeos/login/s...
chrome/browser/chromeos/login/signed_settings.cc:367: // We don't serialize this
one to the policy blob.
On 2011/10/07 13:46:52, Mattias Nissler wrote:
> It's actually good to make it a NOTREACHED(), do you know why we had this
branch
> before?

Actually it is not NOTREACHED() it will be called whenever the user calls
CrosSettings.Set(kDeviceOwner, new_powner); which will put it in the local_state
cache it should however not be serialized down to the blob as this is not the
way to set the owner. 
The session_manager does that for us. Actually the cool thing now is that once
the session_manager writes its version of the owner we'll update the cache with
that and thus there is no way that a user can try to trick the browser into
thinking another user is the owner because this will be overwritten on policy
load. 
We don't trust this value for nothing but UI though so not really a security
issue.

[Chris Masone, 2011-10-13 23:55:15]

lgtm

[Denis Lagno, 2011-10-14 15:21:49]

lgtm