Unmanaged devices should forget device id/token

Unmanaged devices should forget device id/token

Device token/id should be erased when device management sever reports "not managed" from policy request.

BUG=64613
TEST=DeviceManagementPolicyProviderTest.DeviceNoLongerManaged

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=67600

[danno, 2010-11-29 08:24:26]

please review

[Jakob Kummerow, 2010-11-29 08:58:18]

As discussed offline, calling token_fetcher_->Restart() to forget token+id leads
to a small inefficiency, as the token fetcher will first ask the token server
for a new token and will only forget its data once that request fails with the
appropriate error message. We should fix this some time in the future. However,
users won't notice this, and the additional network traffic and server load is
minimal, therefore:

For the beta: LGTM.

[danno, 2010-11-29 08:59:55]

Filed bug 64614 to make sure that we don't forget about your feedback.

On 2010/11/29 08:58:18, jkummerow wrote:
> As discussed offline, calling token_fetcher_->Restart() to forget token+id
leads
> to a small inefficiency, as the token fetcher will first ask the token server
> for a new token and will only forget its data once that request fails with the
> appropriate error message. We should fix this some time in the future.
However,
> users won't notice this, and the additional network traffic and server load is
> minimal, therefore:
> 
> For the beta: LGTM.

[markusheintz_, 2010-11-29 09:01:42]

LGTM with nit.

http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
File chrome/browser/policy/device_management_policy_provider_unittest.cc
(right):

http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
chrome/browser/policy/device_management_policy_provider_unittest.cc:59: FilePath
GetTokenPath() const {
Why was this introduced? It looks like it is only used once in the test below.
Inline?

[danno-g, 2010-11-29 09:32:39]

Problem is that this method is private on the provider, and the test cases
don't have access. The test base class is already a friend, so this let's me
avoid test-specific friend declarations, which are a real pain to maintain.

On Mon, Nov 29, 2010 at 10:01 AM, <markusheintz@chromium.org> wrote:

> LGTM with nit.
>
>
>
>
http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
> File chrome/browser/policy/device_management_policy_provider_unittest.cc
> (right):
>
>
>
http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
> chrome/browser/policy/device_management_policy_provider_unittest.cc:59:
> FilePath GetTokenPath() const {
> Why was this introduced? It looks like it is only used once in the test
> below. Inline?
>
>
> http://codereview.chromium.org/5346005/
>

[markusheintz_, 2010-11-29 09:55:42]

I see. Thanks a lot for the explanation. 

On 2010/11/29 09:32:39, danno-g wrote:
> Problem is that this method is private on the provider, and the test cases
> don't have access. The test base class is already a friend, so this let's me
> avoid test-specific friend declarations, which are a real pain to maintain.
> 
> On Mon, Nov 29, 2010 at 10:01 AM, <mailto:markusheintz@chromium.org> wrote:
> 
> > LGTM with nit.
> >
> >
> >
> >
>
http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
> > File chrome/browser/policy/device_management_policy_provider_unittest.cc
> > (right):
> >
> >
> >
>
http://codereview.chromium.org/5346005/diff/2001/chrome/browser/policy/device...
> > chrome/browser/policy/device_management_policy_provider_unittest.cc:59:
> > FilePath GetTokenPath() const {
> > Why was this introduced? It looks like it is only used once in the test
> > below. Inline?
> >
> >
> > http://codereview.chromium.org/5346005/
> >