Add tests for EDUs during fast joins#465
Conversation
Signed-off-by: Mathieu Velten <mathieuv@matrix.org>
MatMaul
changed the title
MatMaul
force-pushed
the
mv/fast-join-edus
branch
from
8eb79db to
7778cca
Compare
MatMaul
changed the title
DMRobertson
left a comment
DMRobertson
left a comment
There was a problem hiding this comment.
Seems reasonable. I've left a few thoughts but nothing that I think you necessarily need to action.
DMRobertson
force-pushed
the
mv/fast-join-edus
branch
from
bea88eb to
9bf9498
Compare
DMRobertson
force-pushed
the
mv/fast-join-edus
branch
from
9bf9498 to
40978ec
Compare
|
I think some of these failures are expected per Sean's comment here. For each failing test, I need to
|
|
In the light of Sean's comment #465 (comment) I'd like to write a test which checks you can still see incoming messages during resync after the server receives a presence EDU. But reviews welcome nonetheless. Edit: oh, that was bodged into an existing test in 7331ebc. Let's get this reviewed then. |
| }) | ||
|
|
||
| // we should be able to receive device list update EDU over federation during the resync | ||
| t.Run("CanReceiveDeviceListUpdateDuringPartialStateJoin", func(t *testing.T) { |
There was a problem hiding this comment.
I'm a bit dubious about this test TBH, I don't think it tests anything particularly useful as is and may just cause confusion.
There was a problem hiding this comment.
I don't fully follow this either. Shouldn't the device list update be send to Alice, even if she and Derek aren't in an encrypted room?
There was a problem hiding this comment.
As in: Alice should get told via /sync that Derek's device list has changed.
There was a problem hiding this comment.
In Synapse currently, yes, though its a bit of a grey area and we've certainly talked about only sending device list updates for E2EE enabled devices and/or users who share rooms.
OTOH, I think all the current integration tests assume this, so I don't see a problem assuming it here
4 participants
The typing test will fail until matrix-org/synapse#13830 is merged.
Presence test is not testing the sync because it seems to be lost sometimes, I haven't investigated yet.
At least it doesn't fail and block the transaction like with typing.