During a rejection of an invite received over federation, we ask a remote
server to make us a `leave` event, then sign it, then send that with
`send_leave`.
We were saving the *unsigned* version of the event (which has a different event
id to the signed version) to our db (and sending it to the clients), whereas
other servers in the room will have seen the *signed* version. We're not aware
of any actual problems that caused, except that it makes the database confusing
to look at and generally leaves the room in a weird state.
Make sure that we accept join events from any server, rather than just the
origin server, to make the federation join dance work correctly.
(Fixes#1893).
Fix a bug in ``logcontext.preserve_fn`` which made it leak context into the
reactor, and add a test for it.
Also, get rid of ``logcontext.reset_context_after_deferred``, which tried to do
the same thing but had its own, different, set of bugs.
A few non-functional changes:
* A bunch of docstrings to document types
* Split `EventsStore._persist_events_txn` up a bit. Hopefully it's a bit more
readable.
* Rephrase `EventFederationStore._update_min_depth_for_room_txn` to avoid
mind-bending conditional.
* Rephrase rejected/outlier conditional in `_update_outliers_txn` to avoid
mind-bending conditional.
This just takes the existing `room_queues` logic and moves it out to
`on_receive_pdu` instead of `_process_received_pdu`, which ensures that we
don't start trying to fetch prev_events and whathaveyou until the join has
completed.
Unfortunately this significantly increases the size of the already-rather-big
FederationHandler, but the code fits more naturally here, and it paves the way
for the tighter integration that I need between handling incoming PDUs and
doing the join dance.
Other than renaming the existing `FederationHandler.on_receive_pdu` to
`_process_received_pdu` to make way for it, this just consists of the move, and
replacing `self.handler` with `self` and `self` with `self.replication_layer`.
When a server sends a third party invite another server may be the one
that the inviting user registers with. In this case it is that remote
server that will issue an actual invitation, and wants to do it "in the
name of" the original invitee. However, the new proper invite will not
be signed by the original server, and thus other servers would reject
the invite if it was seen as coming from the original user.
To fix this, a special case has been added to the auth rules whereby
another server can send an invite "in the name of" another server's
user, so long as that user had previously issued a third party invite
that is now being accepted.
Loading push rules now happens in the datastore, so we can remove
the methods that loaded them outside the datastore.
The ``waiting_for_join_list`` in federation handler is populated by
anything, so can be removed.
The ``_get_members_events_txn`` method isn't called from anywhere
so can be removed.
Move the functions inside the distributor and import them
where needed. This reduces duplication and makes it possible
for flake8 to detect when the functions aren't used in a
given file.
If rejecting a remote invite fails with an error response don't fail
the entire request; instead mark the invite as locally rejected.
This fixes the bug where users can get stuck invites which they can
neither accept nor reject.
pycharm supports them so there is no need to use the other format.
Might as well convert the existing strings to reduce the risk of
people accidentally cargo culting the wrong doc string format.
Move the checks for whether an event is new state inside persist
event itself.
This was harder than expected because there wasn't enough information
passed to persist event to correctly handle invites from remote servers
for new rooms.
