fix missing mock for test

src/state/CallViewModel/localMember/Publisher.test.ts

@@ -14,13 +14,19 @@ import {
   type Mock,
   vi,
 } from "vitest";
-import { ConnectionState as LivekitConenctionState } from "livekit-client";
-import { type BehaviorSubject } from "rxjs";
+import {
+  ConnectionState as LivekitConenctionState,
+  LocalParticipant,
+  type LocalTrack,
+  type LocalTrackPublication,
+} from "livekit-client";
+import { BehaviorSubject } from "rxjs";
 import { logger } from "matrix-js-sdk/lib/logger";
 
 import { ObservableScope } from "../../ObservableScope";
 import { constant } from "../../Behavior";
 import {
+  flushPromises,
   mockLivekitRoom,
   mockLocalParticipant,
   mockMediaDevices,
@@ -33,8 +39,15 @@ import {
 import { type MuteStates } from "../../MuteStates";
 import { FailToStartLivekitConnection } from "../../../utils/errors";
 
+let scope: ObservableScope;
+
+beforeEach(() => {
+  scope = new ObservableScope();
+});
+
+afterEach(() => scope.end());
+
 describe("Publisher", () => {
-  let scope: ObservableScope;
   let connection: Connection;
   let muteStates: MuteStates;
   beforeEach(() => {
@@ -50,7 +63,6 @@ describe("Publisher", () => {
         setHandler: vi.fn(),
       },
     } as unknown as MuteStates;
-    scope = new ObservableScope();
     connection = {
       state$: constant({
         state: "ConnectedToLkRoom",
@@ -62,8 +74,6 @@ describe("Publisher", () => {
     } as unknown as Connection;
   });
 
-  afterEach(() => scope.end());
-
   it("throws if livekit room could not publish", async () => {
     const publisher = new Publisher(
       scope,
@@ -88,14 +98,19 @@ describe("Publisher", () => {
 
     (
       connection.livekitRoom.localParticipant.createTracks as Mock
-    ).mockResolvedValue([{}, {}]);
+    ).mockResolvedValue([
+      {
+        kind: "audio",
+        mute: vi.fn(),
+      },
+    ]);
 
     await expect(publisher.createAndSetupTracks()).resolves.not.toThrow();
     expect(
       connection.livekitRoom.localParticipant.createTracks,
     ).toHaveBeenCalledOnce();
 
-    // failiour due to localParticipant.publishTrack
+    // failure due to localParticipant.publishTrack
     (
       connection.livekitRoom.localParticipant.publishTrack as Mock
     ).mockRejectedValue(Error("testError"));
@@ -104,7 +119,7 @@ describe("Publisher", () => {
       new FailToStartLivekitConnection("testError"),
     );
 
-    // does not try other conenction after the first one failed
+    // does not try other connection after the first one failed
     expect(
       connection.livekitRoom.localParticipant.publishTrack,
     ).toHaveBeenCalledTimes(1);
@@ -135,6 +150,111 @@ describe("Publisher", () => {
 
     expect(
       connection.livekitRoom.localParticipant.publishTrack,
-    ).toHaveBeenCalledTimes(3);
+    ).toHaveBeenCalledTimes(2);
+  });
+});
+
+describe("Bug fix", () => {
+  // There is a race condition when creating and publishing tracks while the mute state changes.
+  // This race condition could cause tracks to be published even though they are muted at the
+  // beginning of a call coming from lobby.
+  // This is caused by our stack using manually the low level API to create and publish tracks,
+  // but also using the higher level setMicrophoneEnabled and setCameraEnabled functions that also create
+  // and publish tracks, and managing pending publications.
+  // Race is as follow, on creation of the Publisher we create the tracks then publish them.
+  // If in the middle of that process the mute state changes:
+  //  - the `setMicrophoneEnabled` will be no-op because it is not aware of our created track and can't see any pending publication
+  //  - If start publication is requested it will publish the track even though there was a mute request.
+  it("wrongly publish tracks while muted", async () => {
+    const audioEnabled$ = new BehaviorSubject(true);
+    const muteStates = {
+      audio: {
+        enabled$: audioEnabled$,
+        unsetHandler: vi.fn(),
+        setHandler: vi.fn(),
+      },
+      video: {
+        enabled$: constant(false),
+        unsetHandler: vi.fn(),
+        setHandler: vi.fn(),
+      },
+    } as unknown as MuteStates;
+
+    const mockSendDataPacket = vi.fn();
+    const mockEngine = {
+      client: {
+        sendUpdateLocalMetadata: vi.fn(),
+      },
+      on: vi.fn().mockReturnThis(),
+      sendDataPacket: mockSendDataPacket,
+    };
+
+    // cont mockRoomOptions = {} as InternalRoomOptions;
+
+    const localParticipant = new LocalParticipant(
+      "local-sid",
+      "local-identity",
+      // @ts-expect-error - for that test we want a real LocalParticipant to have the pending publications logic
+      mockEngine,
+      {},
+      new Map(),
+      {},
+    );
+
+    const connection = {
+      state$: constant({
+        state: "ConnectedToLkRoom",
+        livekitConnectionState$: constant(LivekitConenctionState.Connected),
+      }),
+      livekitRoom: mockLivekitRoom({
+        localParticipant,
+      }),
+    } as unknown as Connection;
+
+    const mediaDevices = mockMediaDevices({});
+
+    const mockTrack = vi.mocked<LocalTrack>({
+      kind: "audio",
+      mute: vi.fn(),
+    } as Partial<LocalTrack> as LocalTrack);
+    const createTrackLock = Promise.withResolvers<void>();
+    const createTrackSpy = vi.spyOn(localParticipant, "createTracks");
+    createTrackSpy.mockImplementation(async () => {
+      await createTrackLock.promise;
+      return [mockTrack];
+    });
+
+    const publishTrackSpy = vi.spyOn(localParticipant, "publishTrack");
+    publishTrackSpy.mockResolvedValue({} as unknown as LocalTrackPublication);
+
+    const publisher = new Publisher(
+      scope,
+      connection,
+      mediaDevices,
+      muteStates,
+      constant({ supported: false, processor: undefined }),
+      logger,
+    );
+
+    // Initially the audio is unmuted, so creating tracks should publish the audio track
+    const createTracks = publisher.createAndSetupTracks();
+    publisher.tracks$.subscribe(() => {
+      void publisher.startPublishing();
+    });
+    // now mute the audio before allowing track creation to complete
+    audioEnabled$.next(false);
+    // const publishing = publisher.startPublishing();
+    createTrackLock.resolve();
+    await createTracks;
+    // await publishing;
+
+    await flushPromises();
+
+    // It should not publish or instead call track.mute()
+    try {
+      expect(publishTrackSpy).not.toHaveBeenCalled();
+    } catch {
+      expect(mockTrack.mute).toHaveBeenCalled();
+    }
   });
 });

src/state/CallViewModel/localMember/Publisher.ts

@@ -144,6 +144,7 @@ export class Publisher {
           this.logger.error("Failed to create tracks", error);
         });
     }
+    // TODO why throw here? should we just do nothing?
     throw Error("audio and video is false");
   }
 
@@ -184,16 +185,36 @@ export class Publisher {
 
     for (const track of this.tracks$.value) {
       this.logger.info("publish ", this.tracks$.value.length, "tracks");
-      // TODO: handle errors? Needs the signaling connection to be up, but it has some retries internally
-      // with a timeout.
-      await lkRoom.localParticipant.publishTrack(track).catch((error) => {
-        this.logger.error("Failed to publish track", error);
-        throw new FailToStartLivekitConnection(
-          error instanceof Error ? error.message : error,
-        );
-      });
-      this.logger.info("published track ", track.kind, track.id);
 
+      // XXXX: Patch: Check if the track has been muted manually before publishing
+      // This is only a patch, the proper way would be to use livekit high-level enabled/disabled APIs
+      // or only use the low level create/publish APIs and have our own pending publication protection.
+      // Maybe we could change the livekit api to pre-load tracks without publishing them yet?
+      // Are we sure this is needed at all? What are the gains?
+      let isEnabled: boolean;
+      if (track.kind === Track.Kind.Audio) {
+        isEnabled = this.muteStates.audio.enabled$.value;
+      } else if (track.kind === Track.Kind.Video) {
+        isEnabled = this.muteStates.video.enabled$.value;
+      } else {
+        throw new Error("Unsupported track kind " + track.kind);
+      }
+
+      if (!isEnabled) {
+        // TODO should we also drop it?
+        // I believe the high-level LiveKit APIs will recreate a track?
+        await track.mute();
+      } else {
+        // TODO: handle errors? Needs the signaling connection to be up, but it has some retries internally
+        // with a timeout.
+        await lkRoom.localParticipant.publishTrack(track).catch((error) => {
+          this.logger.error("Failed to publish track", error);
+          throw new FailToStartLivekitConnection(
+            error instanceof Error ? error.message : error,
+          );
+        });
+        this.logger.info("published track ", track.kind, track.id);
+      }
       // TODO: check if the connection is still active? and break the loop if not?
     }
     this._publishing$.next(true);