With the move to a central StateService we erroniously cut out search indexing from the process of setting decrypted ciphers to memory.
This commit calls the method responsible for setting decrypted ciphers and indexing when decrypting, instead of setting decrypted ciphers directly.
* Move Key Connector check to subclass
* Move authService.logout call to main program
* Move Key Connector migration check to unlock command
* Use get/setConvertAccountRequired flag
* Move Key Connector convert to own command, set usesKeyConnector after conversion
* Remove KC conversion check from syncCommand, fix callback
* Make class service private
* Fix naming convention
* Update jslib and deps
Some clients, like Directory Connector, use different key for their user identifier: entityId
We currently only check for userId in the migration service, but need to account for both.
A couple of helper methods were recently added to the StateMigrationService, but they were set to private and can't be used in children.
Some clients, like the Directory Connector, extend the StateMigrationService and need access to these methods.
The following data points are currently scoped to an account but are made global with this commit:
* Enable Menu Bar Icon
* Minimize To Menu Bar
* Close To Menu Bar
* Start To Menu Bar
Note: these are all electron specific fields
Currently the StateMigrationService depends on a userId key for running migrations, but if there is not an authenticated user saved to storage that userId is not present.
These changes allow for migrating state data even without an active user. For account specific settings like clearClipboard we now temporarily store those values together in disk state until an account is authed that they can be added to. Temp account state is then cleared.
Some notes:
* In order for this to work we need GlobalState.stateVersion to have a default value of StateVersion.One instead of StateVersion.Latest. Defaulting it to latest was causing migrations to not run on some clients (like desktop) that try to access storage before migrations have been run but save a version as if migrations did run.
* I also noticed we aren't clearing old state items from before migrating, and added a case for this to the migrator.
* I extracted a few bits of reused code into private methods in the stateMigration service. Things like get/set from storage, default options, etc.
With the account switching work a header was added to the desktop app that new acts as a home for the mac window controls.
Previously we needed a special home for these controls, but since moving them we are not just creating empty space.
Removing this class and the divs that use it corrects the behavior.
* Move CLI Key Connector check out of base class
* Add missing await
* Move safe operation out of try/catch block
* Move Key Connector migration check to unlock command
* Set convertAccountRequired flag in syncService
* Remove unneeded service
* [dep] Implement new StateService factory parameter from jslib
* [bug] Ensure setLastActive uses the correct userId
Sometimes, because of how often it fires, setLastActive can cause accounts to override each other. To make sure the correct userId is always used we now subscribe to activeUser in the appComponent and pass that value into any setLastActive calls.
* [bug] Show loader when logging out
When logging out of a large vault the application can appear to hang. This commit turns on the app component loader while logout is doing work.
* [bug] Stop tracking activity without an active user
* [style] Ran prettier
* [chore] Update jslib
* [bug] Improve state management performance
Large vaults see a clear degrade in performance using the state service, especially when multiple vaults are authed and unlocked at the same time.
Some changes made to address this:
1. Clearing in memory decrypted data for non active accounts. This really should have been something we were doing anyway, but letting go of that memory burden has a noticable performance boost.
2. Not loading a bunch of unecsassary data from disk accounts into memory on application startup. This was being done to initilize in memory accounts, but brought a lot of extra baggage with it like storing encrypted data in memory, even though it is never referenced that way.
3. Breaking the on disk state object up into seperate keys for accounts instead of storing everything together under a "state" key. This ensures there is less information fetched from disk each time we call for an account.
There were some restructuring changes needed to facilitate these items:
1. We need to be able to construct an account in the StateService, but typescript doesn't allow for new() constraints on generics so a factory needs to be created and passed into the StateService for this to work.
2. Since we can't reference an all-knowing "accounts" object for on disk state anymore we have to maintain a list of authenticated accounts, and this has been added.
3. The StateMigration service needed to be updated to break up the accounts object, so current dev and QA state will be broken and need to be reset.
Some other general refactorings that were helpful gettings this working:
1. Added a constant for keys to the StateService and StateMigrationService.
2. Bundling everything needed to deauthenticate a user into a dedicated method.
3. Bundling all the disk storage clear methods (that should be refactored later into client specific state services) into one helper method.
4. Bundling everything needed to dynamically select a new active user into a dedicated method.
* [bug] Set environmentUrls appropriatly on account add
* [bug] Stop tracking activity without an active user
* [bug] Remove lastActive from globalState and globalState migration
* [style] Ran prettier
* [Bug] VaultTimeout incorrectly defaults to "Never"
The default desktop vault timeout value is "On Restart", but there is no default set for this in the state service and account model.
This commit extends the StateService and Account model to consider the special vault timeout default requirements needed for desktop.
* [style] Lint fixes
* [chore] Update jslib
The client side storage restructuring work incorrectly checks if a vault has ever been unlocked to determine neverLock scenerios, but production does the opposite.
This creates an inability to never manually lock neverLock vaults.
This commit sets that condition back to the way it was.
