[Smart Lock, Settings Reconciliation] Histograms for tracking initial pref values.

chrome/browser/password_manager/password_manager_setting_migrator_service.h

Index: chrome/browser/password_manager/password_manager_setting_migrator_service.h
diff --git a/chrome/browser/password_manager/password_manager_setting_migrator_service.h b/chrome/browser/password_manager/password_manager_setting_migrator_service.h
index 3399b85035b04c692d5e120d8b29b9381b0311e7..e956f7ddc2b462a3ccc69ffa5efbff0002346ed5 100644
--- a/chrome/browser/password_manager/password_manager_setting_migrator_service.h
+++ b/chrome/browser/password_manager/password_manager_setting_migrator_service.h
@@ -21,6 +21,16 @@ namespace sync_driver {
 class SyncService;
 }
 
+// Enum used for histogram tracking of the initial values for the legacy and new
+// preferences.
+enum PasswordManagerPreferencesInitialValues {

[engedy, 2015/09/24 14:26:18]

What do you think about not defining this enum here at all, and instead simply
recording (N << 1 | L)? At the moment, I cannot really think of any benefits of
an explicit enum -- except for readability in the unit test, but I guess we
could just define the enum in the unittest.cc in an anonymous namespace.

[melandory, 2015/09/28 07:10:12]

Sure, I thought that enum is clearer, but it sounds good to have it just in unit
tests.

+  N0L0,
+  N0L1,
+  N1L0,
+  N1L1,
+  NUM_INITIAL_VALUES,
+};
+
 // Service that is responsible for reconciling the legacy "Offer to save your
 // web passwords" setting (henceforth denoted 'L', for legacy) with the new
 // "Enable Smart Lock for Passwords" setting (henceforth denoted 'N', for new).