Bugfixes for 'If remote user leaves room we no longer receive device updates' (#1262)

* Bugfixes for 'If remote user leaves room we no longer receive device updates'

* Update whitelist and README

5 files changed, 34 insertions, 15 deletions

diff --git a/README.md b/README.md
index 168a864a..6c668a22 100644
--- a/README.md
+++ b/README.md
@@ -33,7 +33,7 @@ Then point your favourite Matrix client at `http://localhost:8008`. For full ins
 
 We use a script called Are We Synapse Yet which checks Sytest compliance rates. Sytest is a black-box homeserver
 test rig with around 900 tests. The script works out how many of these tests are passing on Dendrite and it
-updates with CI. As of July 2020 we're at around 48% CS API coverage and 50% Federation coverage, though check
+updates with CI. As of August 2020 we're at around 52% CS API coverage and 65% Federation coverage, though check
 CI for the latest numbers. In practice, this means you can communicate locally and via federation with Synapse
 servers such as matrix.org reasonably well. There's a long list of features that are not implemented, notably:
  - Receipts
@@ -42,7 +42,6 @@ servers such as matrix.org reasonably well. There's a long list of features that
  - User Directory
  - Presence
  - Guests
- - E2E keys and device lists
 
 We are prioritising features that will benefit single-user homeservers first (e.g Receipts, E2E) rather
 than features that massive deployments may be interested in (User Directory, OpenID, Guests, Admin APIs, AS API).
@@ -56,6 +55,7 @@ This means Dendrite supports amongst others:
  - Media APIs
  - Redaction
  - Tagging
+ - E2E keys and device lists
 
 
 # Contributing
diff --git a/are-we-synapse-yet.list b/are-we-synapse-yet.list
index 7655a3b8..cd2a5e74 100644
--- a/are-we-synapse-yet.list
+++ b/are-we-synapse-yet.list
@@ -831,6 +831,7 @@ psh Trying to get push rules with unknown scope fails with 400
 psh Trying to get push rules with unknown template fails with 400 
 psh Trying to get push rules with unknown attribute fails with 400 
 psh Trying to get push rules with unknown rule_id fails with 404
+psh Rooms with names are correctly named in pushes
 v1s GET /initialSync with non-numeric 'limit' 
 v1s GET /events with non-numeric 'limit' 
 v1s GET /events with negative 'limit' 
@@ -839,7 +840,7 @@ ath Event size limits
 syn Check creating invalid filters returns 4xx 
 f,pre New federated private chats get full presence information (SYN-115) 
 pre Left room members do not cause problems for presence 
-crm Rooms can be created with an initial invite list (SYN-205) 
+crm Rooms can be created with an initial invite list (SYN-205) (1 subtests)
 typ Typing notifications don't leak 
 ban Non-present room members cannot ban others 
 psh Getting push rules doesn't corrupt the cache SYN-390 
diff --git a/keyserver/internal/device_list_update.go b/keyserver/internal/device_list_update.go
index c4b098a4..85785b07 100644
--- a/keyserver/internal/device_list_update.go
+++ b/keyserver/internal/device_list_update.go
@@ -230,6 +230,7 @@ func (u *DeviceListUpdater) worker(ch chan gomatrixserverlib.ServerName) {
 	}
 
 	// on failure, spin up a short-lived goroutine to inject the server name again.
+	scheduledRetries := make(map[gomatrixserverlib.ServerName]time.Time)
 	inject := func(srv gomatrixserverlib.ServerName, duration time.Duration) {
 		time.Sleep(duration)
 		ch <- srv
@@ -237,13 +238,20 @@ func (u *DeviceListUpdater) worker(ch chan gomatrixserverlib.ServerName) {
 
 	for serverName := range ch {
 		if !shouldProcess(serverName) {
-			// do not inject into the channel as we know there will be a sleeping goroutine
-			// which will do it after the cooloff period expires
-			continue
+			if time.Now().Before(scheduledRetries[serverName]) {
+				// do not inject into the channel as we know there will be a sleeping goroutine
+				// which will do it after the cooloff period expires
+				continue
+			} else {
+				scheduledRetries[serverName] = time.Now().Add(cooloffPeriod)
+				go inject(serverName, cooloffPeriod) // TODO: Backoff?
+				continue
+			}
 		}
 		lastProcessed[serverName] = time.Now()
 		shouldRetry := u.processServer(serverName)
 		if shouldRetry {
+			scheduledRetries[serverName] = time.Now().Add(cooloffPeriod)
 			go inject(serverName, cooloffPeriod) // TODO: Backoff?
 		}
 	}
diff --git a/syncapi/internal/keychange.go b/syncapi/internal/keychange.go
index 7623fd9d..7d127aa8 100644
--- a/syncapi/internal/keychange.go
+++ b/syncapi/internal/keychange.go
@@ -96,7 +96,8 @@ func DeviceListCatchup(
 		return hasNew, nil
 	}
 	// QueryKeyChanges gets ALL users who have changed keys, we want the ones who share rooms with the user.
-	queryRes.UserIDs = filterSharedUsers(ctx, stateAPI, userID, queryRes.UserIDs)
+	var sharedUsersMap map[string]int
+	sharedUsersMap, queryRes.UserIDs = filterSharedUsers(ctx, stateAPI, userID, queryRes.UserIDs)
 	util.GetLogger(ctx).Debugf(
 		"QueryKeyChanges request p=%d,off=%d,to=%d response p=%d off=%d uids=%v",
 		partition, offset, toOffset, queryRes.Partition, queryRes.Offset, queryRes.UserIDs,
@@ -114,13 +115,20 @@ func DeviceListCatchup(
 	}
 	// if the response has any join/leave events, add them now.
 	// TODO: This is sub-optimal because we will add users to `changed` even if we already shared a room with them.
-	for _, userID := range membershipEvents(res) {
+	joinUserIDs, leaveUserIDs := membershipEvents(res)
+	for _, userID := range joinUserIDs {
 		if !userSet[userID] {
 			res.DeviceLists.Changed = append(res.DeviceLists.Changed, userID)
 			hasNew = true
 			userSet[userID] = true
 		}
 	}
+	for _, userID := range leaveUserIDs {
+		if sharedUsersMap[userID] == 0 {
+			// we no longer share a room with this user when they left, so add to left list.
+			res.DeviceLists.Left = append(res.DeviceLists.Left, userID)
+		}
+	}
 	// set the new token
 	to.SetLog(DeviceListLogName, &types.LogPosition{
 		Partition: queryRes.Partition,
@@ -221,7 +229,7 @@ func TrackChangedUsers(
 
 func filterSharedUsers(
 	ctx context.Context, stateAPI currentstateAPI.CurrentStateInternalAPI, userID string, usersWithChangedKeys []string,
-) []string {
+) (map[string]int, []string) {
 	var result []string
 	var sharedUsersRes currentstateAPI.QuerySharedUsersResponse
 	err := stateAPI.QuerySharedUsers(ctx, &currentstateAPI.QuerySharedUsersRequest{
@@ -229,7 +237,7 @@ func filterSharedUsers(
 	}, &sharedUsersRes)
 	if err != nil {
 		// default to all users so we do needless queries rather than miss some important device update
-		return usersWithChangedKeys
+		return nil, usersWithChangedKeys
 	}
 	// We forcibly put ourselves in this list because we should be notified about our own device updates
 	// and if we are in 0 rooms then we don't technically share any room with ourselves so we wouldn't
@@ -241,7 +249,7 @@ func filterSharedUsers(
 			result = append(result, uid)
 		}
 	}
-	return result
+	return sharedUsersRes.UserIDsToCount, result
 }
 
 func joinedRooms(res *types.Response, userID string) []string {
@@ -288,16 +296,16 @@ func membershipEventPresent(events []gomatrixserverlib.ClientEvent, userID strin
 // "For optimal performance, Alice should be added to changed in Bob's sync only when she adds a new device,
 // or when Alice and Bob now share a room but didn't share any room previously. However, for the sake of simpler
 // logic, a server may add Alice to changed when Alice and Bob share a new room, even if they previously already shared a room."
-func membershipEvents(res *types.Response) (userIDs []string) {
+func membershipEvents(res *types.Response) (joinUserIDs, leaveUserIDs []string) {
 	for _, room := range res.Rooms.Join {
 		for _, ev := range room.Timeline.Events {
 			if ev.Type == gomatrixserverlib.MRoomMember && ev.StateKey != nil {
 				if strings.Contains(string(ev.Content), `"join"`) {
-					userIDs = append(userIDs, *ev.StateKey)
+					joinUserIDs = append(joinUserIDs, *ev.StateKey)
 				} else if strings.Contains(string(ev.Content), `"leave"`) {
-					userIDs = append(userIDs, *ev.StateKey)
+					leaveUserIDs = append(leaveUserIDs, *ev.StateKey)
 				} else if strings.Contains(string(ev.Content), `"ban"`) {
-					userIDs = append(userIDs, *ev.StateKey)
+					leaveUserIDs = append(leaveUserIDs, *ev.StateKey)
 				}
 			}
 		}
diff --git a/sytest-whitelist b/sytest-whitelist
index e92e883c..4d37b3ee 100644
--- a/sytest-whitelist
+++ b/sytest-whitelist
@@ -142,6 +142,8 @@ Server correctly handles incoming m.device_list_update
 Device deletion propagates over federation
 If remote user leaves room, changes device and rejoins we see update in sync
 If remote user leaves room, changes device and rejoins we see update in /keys/changes
+If remote user leaves room we no longer receive device updates
+Get left notifs in sync and /keys/changes when other user leaves
 Can query remote device keys using POST after notification
 Can add account data
 Can add account data to room