BDF improvements
* Change the semantics of the BdfList/Dictionary methods with default values so they return the default value if the value is missing, or throw an exception if it the value's present but has the wrong type - this is more useful for validators, which have to deal with optional values without accepting values of the wrong type
* Reuse BdfReader/Writer for MetadataParser/Encoder so we don't have two implementations of BDF
See merge request !117
Helper class to reduce client boilerplate
* Renamed BdfReader methods for consistency with BdfList/BdfDictionary
* Added readList() and readDictionary() methods to BdfReader
* Added ClientHelper to reduce boilerplate when converting messages and metadata to and from BDF
* Moved PrivateGroupFactory to the same package as ClientHelper
See merge request !114
Unit tests for ValidationManagerImpl
Unit test for the validation manager. I also changed the way the validation manager loads unvalidated messages - instead of using a single DB task to load all unvalidated messages, it loads a list of message IDs and then loads each message in a separate task. This prevents the DatabaseExecutor from being blocked by a long-running task if there are lots of messages to validate.
See merge request !113
Add contact and transport keys in the same transaction
This avoids a potential problem where the app crashes after adding the contact but before adding the transport keys, leaving the contact unusable.
See merge request !112
New polling logic for Tor. #250
Poll for outgoing connections each time the network is enabled, then disable polling once our hidden service descriptor has been published. If our contacts do the same, any contacts that are offline when we poll will connect to us when they come online.
This should reduce bandwidth and battery usage. The idea comes from Ricochet:
https://github.com/ricochet-im/ricochet/issues/68Fixes#250.
See merge request !103
If two contacts publish their descriptors simultaneously, they may both stop polling without retrieving each other's descriptors. Continue polling for 2 intervals after publishing the descriptor.
New polling logic for LAN. #252
Same approach as !104. Modified the poller to cancel any scheduled poll when pollNow() is called and randomise the next polling interval so plugins that call pollNow() at the same time don't end up polling in sync.
Depends on !104. Fixes#252.
See merge request !105
New polling logic for Bluetooth. #251
Increase the polling interval exponentially each time polling is unsuccessful, up to a maximum of 60 minutes. The interval is reset to 2 minutes whenever a connection is made and whenever Bluetooth is re-enabled.
This should reduce battery usage when there are no contacts nearby.
Fixes#251.
See merge request !104
The polling interval increases exponentially each time polling is unsuccessful, up to a maximum of 60 minutes. The interval is reset to 2 minutes whenever a connection is made and whenever Bluetooth is re-enabled.
Don't communicate with inactive contacts
When we're introduced to a new contact, we need to store the transport keys and start key rotation before activating the contact. During this period we shouldn't use the transport keys for outgoing streams until either the introduction protocol is complete or we receive an incoming stream from the contact, indicating that they've also started key rotation.
See merge request !106
Database efficiency improvements
Some tweaks to the DB schema to improve efficiency:
* Only keep status rows for messages that are visible - this saves space and avoids the need to join the groupVisibilities table when selecting messages to offer or send
* Use adjacent columns for the composite primary key on the settings table
This MR depends on !101.
See merge request !102
Don't keep message status rows for groups that are invisible to the contact - this avoids the need to join the groupVisibilities table when selecting messages to offer or send. Add or remove status rows when group visibility changes.
