Don't rely on device list being updated on invite/join for x-signing federation tests

[neilalexander]

The tests were using sync_until_user_in_device_list before, so even if a device list update is generated for the key/signature change, it will still fail if device list updates aren't generated for the invite/join, which seems like a very unfortunate side effect.

There are already separate tests for device list updates being generated on invites/joins, and that's not really what these specific tests should be trying to prove.

[richvdh]

For links: some of these lines were added in #749.

[richvdh]

The problem here is that it's presumably legitimate for a server to send the join event in one /sync and the device update in the next (or vice-versa)

If your only reason for wanting to change these things is "there are already separate tests for device list updates", I'd probably rather leave it as-is rather than risk making the tests flaky again.

[neilalexander]

Yeah — it was largely because I still wanted to test the x-signing behaviour but I don't think Dendrite always generates device change notifications when users are invited/join yet, but if it introduces a race then we may have to rethink.

[richvdh]

I'm not absolutely convinced that your change will introduce a race - note that #749 added a sync_until_user_in_device_list where there was nothing at all before - having a sync_until_join_event (isn't there a utility method we can use for that?) may well serve the purpose adequately, so if this is a blocker to you we can always try merging it and seeing what happens. But if it's just a matter of holding off until Dendrite implements a bit more stuff, I'd wait for that.