Turns out that there is a valid usecase for retrieving event by id (notably
having received a push), but event ids should be scoped to room, so /event/{id}
is wrong.
The _get_joined_users_from_context cache stores a mapping from user_id
to avatar_url and display_name. Instead of storing those in a dict,
store them in a namedtuple as that uses much less memory.
We also try converting the string to ascii to further reduce the size.
We somehow specced APIs with reason strings, preserve the content
in the events and even have the clients display them, but failed
to actually pass the parameter through to the event content.
This returns the currently joined members in the room with their display
names and avatar urls. This is more efficient than /members for large
rooms where you don't need the full events.
This fixes a race whereby:
- User hits an endpoint.
- No cached transaction so executes main code.
- User hits same endpoint.
- No cache transaction so executes main code.
- Main code finishes executing and caches response and returns.
- Main code finishes executing and caches response and returns.
This race is common in the wild when Synapse is struggling under load.
This commit fixes the race by:
- User hits an endpoint.
- Caches the promise to execute the main code and executes main code.
- User hits same endpoint.
- Yields on the same promise as the first request.
- Main code finishes executing and returns, unblocking both requests.
It was always intended to allow custom keys on the join event, but this has
at some point been lost. Restore it.
If the user specifies keys like "avatar_url" then they will be clobbered.
hs.get_handlers() can not be invoked from split out processes. Moving
the invocations down a level means that we can slowly split out
individual servlets.
