This ensures that its resource usage metrics get recorded somewhere rather than
getting lost.
(It also fixes an error when called from a nested logging context which
completes before the bg process)
The existing deferred timeout helper function (and the one into twisted)
suffer from a bug when a deferred's canceller throws an exception, #3842.
The new helper function doesn't suffer from this problem.
Previously we queued up the poke correctly when the device was deleted,
but then the actual EDU wouldn't get sent, as the device was no longer known.
Instead, we now send EDUs for deleted devices too if there's a poke for them.
There were a bunch of places where we fire off a process to happen in the
background, but don't have any exception handling on it - instead relying on
the unhandled error being logged when the relevent deferred gets
garbage-collected.
This is unsatisfactory for a number of reasons:
- logging on garbage collection is best-effort and may happen some time after
the error, if at all
- it can be hard to figure out where the error actually happened.
- it is logged as a scary CRITICAL error which (a) I always forget to grep for
and (b) it's not really CRITICAL if a background process we don't care about
fails.
So this is an attempt to add exception handling to everything we fire off into
the background.
These processes take a long time compared to the request, so there is lots of
"Entering|Restoring dead context" in the logs. Let's try to shut it up a bit.
In `Notifier._on_new_room_event`, `preserve_fn` around its subroutines which
return deferreds, so that it is safe to call it with an active logcontext.
Remove `@preserve_fn` decorators on `on_new_room_event`,
`_notify_pending_new_room_events`, `_on_new_room_event`, `on_new_event`, and
`on_new_replication_data` - none of these functions return a deferred, and the
decorator does nothing unless the wrapped function returns a deferred, so the
decorator was a no-op.
There was a race condition that caused the notifier to 'miss' the
timeout notification, since there were no other checks for the timeout
this caused listeners to get stuck in a loop until something happened.
This is for two reasons:
1. Suppresses duplicates correctly, as the notifier doesn't do any
duplicate suppression.
2. Makes it easier to connect the AppserviceHandler to the replication
stream.
Change how the notifier updates the map from room_id to user streams on
receiving a join event. Make it update the map when it notifies for the
join event, rather than using the "user_joined_room" distributor signal
Access it directly from the homeserver itself. It already wasn't
inheriting from BaseHandler storing it on the Handlers object was
already somewhat dubious.
