Reset RC per-instance userdefaults if config database was not found. #4896
Conversation
|
LGTM on presubmit green |
| @@ -31,6 +31,7 @@ | |||
| #define RCNTableNameInternalMetadata "internal_metadata" | |||
| #define RCNTableNameExperiment "experiment" | |||
|
|
|||
| static BOOL gIsNewDatabase; | |||
There was a problem hiding this comment.
Does this need to be made thread-safe?
There was a problem hiding this comment.
it is. The variable is used on the database serial queue.
| @@ -102,6 +102,14 @@ - (instancetype)initWithDatabaseManager:(RCNConfigDBManager *)manager | |||
| _userDefaultsManager = [[RCNUserDefaultsManager alloc] initWithAppName:appName | |||
| bundleID:_bundleIdentifier | |||
| namespace:_FIRNamespace]; | |||
|
|
|||
| // Check if the config database is new. If so, clear the configs saved in userDefaults. | |||
| if ([_DBManager isNewDatabase]) { | |||
There was a problem hiding this comment.
How can this test ever fail? I don't see gIsNewDatabase ever set to NO.
There was a problem hiding this comment.
The test case setup method above creates a new test database. We just want to verify that this flag is set to true whenever an existing database is not found.
|
@ryanwilson nvm, I was adding you and wondering if NSUserDefault can opt out for backup, looks like it's not available. |
It happens on a restore of a previous backup of the device. The config database is gone, but the last eTag is still remembered (since it is in the user defaults). Hence, a new fetch call succeeds but only returns NO_CHANGE since the eTag was sent up. |
Addresses b/148386739, #4677 , #4734. Backgroubd: We explicitly disable backup of the sqlite database. Hence upon restore, we do not have any data. This causes a fetch to be made, but we send the last eTag (which is saved in userdefaults and cannot be opted out of backup) along with the fetch request which causes the RC backend to send a NO_CHANGE back. Fix is to identify if the database is gone, and if so, reset the userdefaults.