u"" is now used as defaut item id instead of node, as it needs to be serialised anyway. This is temporary solution, as empty string can be used for item id (not forbidden by XEP-0060), but for now this is not handled correctly in jp/SàT (see https://bugs.goffi.org/bugs/view/289). Correct solution will probably need to use a special key in extra to mark empty string in id.
Added/removed some required values for node or item when it made sense.
roster version is now stored after roster item has been added or removed. This way if something interrupting the process happens, former item update will be received and roster should be updated correctly.
Roster item removal was crashing because `load()` is not used (local PersistentDict cache is then not updated, and the key to remove is missing there). The exception is now catched and ignored to avoid loading the data only to remove an item.
rosterResync do a full resynchronisation of roster with server. This should not be needed, but may be useful if user suspect a sync issue for whatever reason.
A "clear" method has been added to delete all data for a specific namespace (and profile).
Internal cache (self._cache) is not initialised with an empty dict instead of None to allow using methods like __setitem__ or clear without having to do a potentially expansive `load()` before. The only counterpart is that we won't have an exception anymore if developer forget to load() data before using them, which can make debugging a bit more difficult.
use Android API to retrieve connectivity informations (connection availability + kind of connection).
Register to connectivity change broadcast, new client.networkEnabled/client.networkDisabled are called when suitable.
those methods can be called by platform specific plugins when network is known to be (un)available. This way, connection attempts can be cancelled when no network is available, saving resources (notably battery on mobile devices), or attempts can be restarted immediately when network is known to be available again.
factory's clientConnectionFailed and clientConnectionLost methods are monkey patched to allow client to tune termination:
- a warning when connection is lost in an unclean way
- connector is saved to allow to disable automatic reconnection and retry manually later
- new triggers connection_failed and connection_lost allow plugins to tune connection termination workflow
Client State Indication plugin is used to set state when pause/resume is called.
The state is changed on pause only after a delay (currently 30s), to avoid changing state if use is only checking something quickly on an other app.
req_timer may not have been called when session.reset() is called (e.g. on resuming failed after computer has been suspended), so reset() must cancel it and set its value to None.
encryptMessage is now called in a loop, because problems may happen by steps (e.g. trust issue after getting missing bundles).
Fixed a mix between "if" and "elif" in problems handling,
state handling is not using mmap anymore but an unix socket instead.
State change can now be managed.
Bridge signals are disabled when frontend is paused.
