arc: Skip GMS Sign-In in case Android is already signed-in.

chrome/browser/chromeos/arc/arc_optin_uma.h

 // Copyright 2016 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef CHROME_BROWSER_CHROMEOS_ARC_ARC_OPTIN_UMA_H_
 #define CHROME_BROWSER_CHROMEOS_ARC_ARC_OPTIN_UMA_H_
 
 #include <ostream>
 
 namespace base {
@@ skipping 101 matching lines @@
   // Network connection is unavailable in ARC.
   NO_NETWORK_CONNECTION = 19,
 
   // ARC is not enabled.
   ARC_DISABLED = 20,
 
   // The size of this enum; keep last.
   SIZE,
 };
 
+enum class AccountCheckStatus : int {

[Ilya Sherman, 2017/03/24 23:22:08]

Please document that this enum is used to back an UMA histogram, and should
therefore be treated as append-only.

[khmel, 2017/03/24 23:51:53]

Done.

+  // Checked account exists and is up to date.
+  UP_TO_DATE = 0,
+
+  // Checked account is new in the system.
+  NEW = 1,
+
+  // Checked account exists but need to be reauthorized.
+  NEED_REAUTH = 2,
+
+  // Checked account exists but it is in unknown state.
+  UNKNOWN = 3,
+
+  // Account check has failed.
+  CHECK_FAILED = 4,
+
+  // The size of this enum; keep last.
+  SIZE,

[dcheng, 2017/03/25 00:55:04]

Another possibility is to use the mojo enum throughout, but I guess this makes
it hard. +yzshen, do you think it would make sense to make it easier to use
enums for UMA? The tricky part is that the UMA histogram macro expects size to
be strictly greater than the max possible value, rather than greater than or
equal to...

[khmel, 2017/03/25 01:24:14]

This good discussion. But we already use similar mapping technique. Would it
possible to keep it similar style so far?

[yzshen1, 2017/03/27 16:53:05]

(Sorry for late reply! I didn't notice this earlier. Please feel free to ping
next time if I didn't reply in 24 hours.)

It would be nice to have one less type. But I noticed the two relevant bug
reports (cannot use enum class with UMA macro; max needs to be larger). It
probably makes sense to have a separate type before those are resolved.
Otherwise, we would have to do static_casts, and also have a unnecessary value
in the mojom enum.

[Ilya Sherman, 2017/03/27 17:00:40]

FWIW, you don't have to add a new value to the mojom enum, you can just write
"max + 1" for the histogram macro invocation.  The static casts would indeed be
needed, though.

[yzshen1, 2017/03/27 17:07:46]

WRT the "max + 1" approach: if "max" is not named explicitly as "max", I imagine
it is possible that people update the mojom definition to append new enum
values, but forget to update the macro accordingly.

[khmel, 2017/03/27 17:17:38]

Thank you for you comments! What is about including SIZE into mojom? This way 
we can protect extending enum. WDYT?

+};
+
 void UpdateOptInActionUMA(OptInActionType type);
 void UpdateOptInCancelUMA(OptInCancelReason reason);
 void UpdateEnabledStateUMA(bool enabled);
 void UpdateProvisioningResultUMA(ProvisioningResult result, bool managed);
 void UpdateProvisioningTiming(const base::TimeDelta& elapsed_time,
                               bool success,
                               bool managed);
 void UpdateSilentAuthCodeUMA(OptInSilentAuthCode state);
 void UpdateAuthTiming(const char* histogram_name, base::TimeDelta elapsed_time);
 void UpdateAuthCheckinAttempts(int32_t num_attempts);
+void UpdateAuthAccountCheckStatus(AccountCheckStatus status);
 
 // Outputs the stringified |result| to |os|. This is only for logging purposes.
 std::ostream& operator<<(std::ostream& os, const ProvisioningResult& result);
 
 }  // namespace arc
 
 #endif  // CHROME_BROWSER_CHROMEOS_ARC_ARC_OPTIN_UMA_H_