Fix screenshare failing after several attempts (#2771)

* Fix screenshare failing after several attempts

Re-use any existing transceivers when screen sharing. This prevents
transceivers accumulating and making the SDP too big: see linked bug.

This also switches from `addTrack()` to `addTransceiver ()` which is
not that large of a change, other than having to explicitly find the
transceivers after an offer has arrived rather than just adding tracks
and letting WebRTC take care of it.

Fixes https://github.com/vector-im/element-call/issues/625

* Fix tests

* Unused import

* Use a map instead of an array

* Add comment

* more comment

* Remove commented code

* Remove unintentional debugging

* Add test for screenshare transceiver re-use

* Type alias for transceiver map

spec/test-utils/webrtc.ts

@@ -104,12 +104,12 @@ export class MockRTCPeerConnection {
     private negotiationNeededListener: () => void;
     public iceCandidateListener?: (e: RTCPeerConnectionIceEvent) => void;
     public onTrackListener?: (e: RTCTrackEvent) => void;
-    private needsNegotiation = false;
+    public needsNegotiation = false;
     public readyToNegotiate: Promise<void>;
     private onReadyToNegotiate: () => void;
     localDescription: RTCSessionDescription;
     signalingState: RTCSignalingState = "stable";
-    public senders: MockRTCRtpSender[] = [];
+    public transceivers: MockRTCRtpTransceiver[] = [];
 
     public static triggerAllNegotiations(): void {
         for (const inst of this.instances) {
@@ -169,12 +169,23 @@ export class MockRTCPeerConnection {
     }
     close() { }
     getStats() { return []; }
-    addTrack(track: MockMediaStreamTrack): MockRTCRtpSender {
+    addTransceiver(track: MockMediaStreamTrack): MockRTCRtpTransceiver {
         this.needsNegotiation = true;
         this.onReadyToNegotiate();
+
         const newSender = new MockRTCRtpSender(track);
-        this.senders.push(newSender);
+        const newReceiver = new MockRTCRtpReceiver(track);
-        return newSender;
+
+        const newTransceiver = new MockRTCRtpTransceiver(this);
+        newTransceiver.sender = newSender as unknown as RTCRtpSender;
+        newTransceiver.receiver = newReceiver as unknown as RTCRtpReceiver;
+
+        this.transceivers.push(newTransceiver);
+
+        return newTransceiver;
+    }
+    addTrack(track: MockMediaStreamTrack): MockRTCRtpSender {
+        return this.addTransceiver(track).sender as unknown as MockRTCRtpSender;
     }
 
     removeTrack() {
@@ -182,9 +193,8 @@ export class MockRTCPeerConnection {
         this.onReadyToNegotiate();
     }
 
-    getSenders(): MockRTCRtpSender[] { return this.senders; }
+    getTransceivers(): MockRTCRtpTransceiver[] { return this.transceivers; }
-
+    getSenders(): MockRTCRtpSender[] { return this.transceivers.map(t => t.sender as unknown as MockRTCRtpSender); }
-    getTransceivers = jest.fn().mockReturnValue([]);
 
     doNegotiation() {
         if (this.needsNegotiation && this.negotiationNeededListener) {
@@ -198,7 +208,23 @@ export class MockRTCRtpSender {
     constructor(public track: MockMediaStreamTrack) { }
 
     replaceTrack(track: MockMediaStreamTrack) { this.track = track; }
-    setCodecPreferences(prefs: RTCRtpCodecCapability[]): void {}
+}
+
+export class MockRTCRtpReceiver {
+    constructor(public track: MockMediaStreamTrack) { }
+}
+
+export class MockRTCRtpTransceiver {
+    constructor(private peerConn: MockRTCPeerConnection) {}
+
+    public sender: RTCRtpSender;
+    public receiver: RTCRtpReceiver;
+
+    public set direction(_: string) {
+        this.peerConn.needsNegotiation = true;
+    }
+
+    setCodecPreferences = jest.fn<void, RTCRtpCodecCapability[]>();
 }
 
 export class MockMediaStreamTrack {

spec/unit/webrtc/call.spec.ts

@@ -41,7 +41,6 @@ import {
     installWebRTCMocks,
     MockRTCPeerConnection,
     SCREENSHARE_STREAM_ID,
-    MockRTCRtpSender,
 } from "../../test-utils/webrtc";
 import { CallFeed } from "../../../src/webrtc/callFeed";
 import { EventType, IContent, ISendEventResponse, MatrixEvent, Room } from "../../../src";
@@ -370,17 +369,15 @@ describe('Call', function() {
             ).typed(),
         );
 
-        const usermediaSenders: Array<RTCRtpSender> = (call as any).usermediaSenders;
+        // XXX: Lots of inspecting the prvate state of the call object here
+        const transceivers: Map<string, RTCRtpTransceiver> = (call as any).transceivers;
 
         expect(call.localUsermediaStream.id).toBe("stream");
         expect(call.localUsermediaStream.getAudioTracks()[0].id).toBe("new_audio_track");
         expect(call.localUsermediaStream.getVideoTracks()[0].id).toBe("video_track");
-        expect(usermediaSenders.find((sender) => {
+        // call has a function for generating these but we hardcode here to avoid exporting it
-            return sender?.track?.kind === "audio";
+        expect(transceivers.get("m.usermedia:audio").sender.track.id).toBe("new_audio_track");
-        }).track.id).toBe("new_audio_track");
+        expect(transceivers.get("m.usermedia:video").sender.track.id).toBe("video_track");
-        expect(usermediaSenders.find((sender) => {
-            return sender?.track?.kind === "video";
-        }).track.id).toBe("video_track");
     });
 
     it("should handle upgrade to video call", async () => {
@@ -400,16 +397,13 @@ describe('Call', function() {
         // setLocalVideoMuted probably?
         await (call as any).upgradeCall(false, true);
 
-        const usermediaSenders: Array<RTCRtpSender> = (call as any).usermediaSenders;
+        // XXX: More inspecting private state of the call object
+        const transceivers: Map<string, RTCRtpTransceiver> = (call as any).transceivers;
 
         expect(call.localUsermediaStream.getAudioTracks()[0].id).toBe("usermedia_audio_track");
         expect(call.localUsermediaStream.getVideoTracks()[0].id).toBe("usermedia_video_track");
-        expect(usermediaSenders.find((sender) => {
+        expect(transceivers.get("m.usermedia:audio").sender.track.id).toBe("usermedia_audio_track");
-            return sender?.track?.kind === "audio";
+        expect(transceivers.get("m.usermedia:video").sender.track.id).toBe("usermedia_video_track");
-        }).track.id).toBe("usermedia_audio_track");
-        expect(usermediaSenders.find((sender) => {
-            return sender?.track?.kind === "video";
-        }).track.id).toBe("usermedia_video_track");
     });
 
     it("should handle SDPStreamMetadata changes", async () => {
@@ -479,6 +473,23 @@ describe('Call', function() {
     });
 
     describe("should deduce the call type correctly", () => {
+        beforeEach(async () => {
+            // start an incoming  call, but add no feeds
+            await call.initWithInvite({
+                getContent: jest.fn().mockReturnValue({
+                    version: "1",
+                    call_id: "call_id",
+                    party_id: "remote_party_id",
+                    lifetime: CALL_LIFETIME,
+                    offer: {
+                        sdp: DUMMY_SDP,
+                    },
+                }),
+                getSender: () => "@test:foo",
+                getLocalAge: () => 1,
+            } as unknown as MatrixEvent);
+        });
+
         it("if no video", async () => {
             call.getOpponentMember = jest.fn().mockReturnValue({ userId: "@bob:bar.uk" });
 
@@ -1057,9 +1068,24 @@ describe('Call', function() {
     });
 
     describe("Screen sharing", () => {
+        const waitNegotiateFunc = resolve => {
+            mockSendEvent.mockImplementationOnce(() => {
+                // Note that the peer connection here is a dummy one and always returns
+                // dummy SDP, so there's not much point returning the content: the SDP will
+                // always be the same.
+                resolve();
+                return Promise.resolve({ event_id: "foo" });
+            });
+        };
+
         beforeEach(async () => {
             await startVoiceCall(client, call);
 
+            const sendNegotiatePromise = new Promise<void>(waitNegotiateFunc);
+
+            MockRTCPeerConnection.triggerAllNegotiations();
+            await sendNegotiatePromise;
+
             await call.onAnswerReceived(makeMockEvent("@test:foo", {
                 "version": 1,
                 "call_id": call.callId,
@@ -1090,12 +1116,7 @@ describe('Call', function() {
             ).toHaveLength(1);
 
             mockSendEvent.mockReset();
-            const sendNegotiatePromise = new Promise<void>(resolve => {
+            const sendNegotiatePromise = new Promise<void>(waitNegotiateFunc);
-                mockSendEvent.mockImplementationOnce(() => {
-                    resolve();
-                    return Promise.resolve({ event_id: "foo" });
-                });
-            });
 
             MockRTCPeerConnection.triggerAllNegotiations();
             await sendNegotiatePromise;
@@ -1130,29 +1151,52 @@ describe('Call', function() {
                 headerExtensions: [],
             });
 
-            const prom = new Promise<void>(resolve => {
+            mockSendEvent.mockReset();
-                const mockPeerConn = call.peerConn as unknown as MockRTCPeerConnection;
+            const sendNegotiatePromise = new Promise<void>(waitNegotiateFunc);
-                mockPeerConn.addTrack = jest.fn().mockImplementation((track: MockMediaStreamTrack) => {
-                    const mockSender = new MockRTCRtpSender(track);
-                    mockPeerConn.getTransceivers.mockReturnValue([{
-                        sender: mockSender,
-                        setCodecPreferences: (prefs: RTCRtpCodecCapability[]) => {
-                            expect(prefs).toEqual([
-                                expect.objectContaining({ mimeType: "video/somethingelse" }),
-                            ]);
-
-                            resolve();
-                        },
-                    }]);
-
-                    return mockSender;
-                });
-            });
 
             await call.setScreensharingEnabled(true);
             MockRTCPeerConnection.triggerAllNegotiations();
 
-            await prom;
+            await sendNegotiatePromise;
+
+            const mockPeerConn = call.peerConn as unknown as MockRTCPeerConnection;
+            expect(
+                mockPeerConn.transceivers[mockPeerConn.transceivers.length - 1].setCodecPreferences,
+            ).toHaveBeenCalledWith([expect.objectContaining({ mimeType: "video/somethingelse" })]);
+        });
+
+        it("re-uses transceiver when screen sharing is re-enabled", async () => {
+            const mockPeerConn = call.peerConn as unknown as MockRTCPeerConnection;
+
+            // sanity check: we should start with one transciever (user media audio)
+            expect(mockPeerConn.transceivers.length).toEqual(1);
+
+            const screenshareOnProm1 = new Promise<void>(waitNegotiateFunc);
+
+            await call.setScreensharingEnabled(true);
+            MockRTCPeerConnection.triggerAllNegotiations();
+
+            await screenshareOnProm1;
+
+            // we should now have another transciever for the screenshare
+            expect(mockPeerConn.transceivers.length).toEqual(2);
+
+            const screenshareOffProm = new Promise<void>(waitNegotiateFunc);
+            await call.setScreensharingEnabled(false);
+            MockRTCPeerConnection.triggerAllNegotiations();
+            await screenshareOffProm;
+
+            // both transceivers should still be there
+            expect(mockPeerConn.transceivers.length).toEqual(2);
+
+            const screenshareOnProm2 = new Promise<void>(waitNegotiateFunc);
+            await call.setScreensharingEnabled(true);
+            MockRTCPeerConnection.triggerAllNegotiations();
+            await screenshareOnProm2;
+
+            // should still be two, ie. another one should not have been created
+            // when re-enabling the screen share.
+            expect(mockPeerConn.transceivers.length).toEqual(2);
         });
     });
 

src/webrtc/call.ts

@@ -308,6 +308,16 @@ export type CallEventHandlerMap = {
     [CallEvent.SendVoipEvent]: (event: Record<string, any>) => void;
 };
 
+// The key of the transceiver map (purpose + media type, separated by ':')
+type TransceiverKey = string;
+
+// generates keys for the map of transceivers
+// kind is unfortunately a string rather than MediaType as this is the type of
+// track.kind
+function getTransceiverKey(purpose: SDPStreamMetadataPurpose, kind: TransceiverKey): string {
+    return purpose + ':' + kind;
+}
+
 /**
  * Construct a new Matrix Call.
  * @constructor
@@ -345,8 +355,10 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
     private candidateSendTries = 0;
     private candidatesEnded = false;
     private feeds: Array<CallFeed> = [];
-    private usermediaSenders: Array<RTCRtpSender> = [];
+
-    private screensharingSenders: Array<RTCRtpSender> = [];
+    // our transceivers for each purpose and type of media
+    private transceivers = new Map<TransceiverKey, RTCRtpTransceiver>();
+
     private inviteOrAnswerSent = false;
     private waitForLocalAVStream: boolean;
     private successor: MatrixCall;
@@ -634,6 +646,18 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
             audioMuted,
             videoMuted,
         }));
+
+        // gather transceivers from the new tracks so that we can use the same ones for tracks that
+        // we add later. We only do this for user media streams though: screenshare streams just always
+        // get their own unidirectional transceiver since a bidirectional screen share is pretty rare
+        // (we *could* re-use an existing recvonly transceiver for this, but it's simpler to just not).
+        if (purpose == SDPStreamMetadataPurpose.Usermedia) {
+            for (const track of stream.getTracks()) {
+                const transceiver = this.peerConn.getTransceivers().find(t => t.receiver.track == track);
+                this.transceivers.set(getTransceiverKey(purpose, track.kind), transceiver);
+            }
+        }
+
         this.emit(CallEvent.FeedsChanged, this.feeds);
 
         logger.info(
@@ -675,6 +699,12 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
             stream,
             purpose,
         }));
+
+        for (const track of stream.getTracks()) {
+            const transceiver = this.peerConn.getTransceivers().find(t => t.receiver.track == track);
+            this.transceivers.set(getTransceiverKey(purpose, track.kind), transceiver);
+        }
+
         this.emit(CallEvent.FeedsChanged, this.feeds);
 
         logger.info(`Call ${this.callId} pushed remote stream (id="${stream.id}", active="${stream.active}")`);
@@ -722,11 +752,6 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
         this.feeds.push(callFeed);
 
         if (addToPeerConnection) {
-            const senderArray = callFeed.purpose === SDPStreamMetadataPurpose.Usermedia ?
-                this.usermediaSenders : this.screensharingSenders;
-            // Empty the array
-            senderArray.splice(0, senderArray.length);
-
             for (const track of callFeed.stream.getTracks()) {
                 logger.info(
                     `Call ${this.callId} ` +
@@ -738,7 +763,27 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
                     `enabled=${track.enabled}` +
                     `) to peer connection`,
                 );
-                senderArray.push(this.peerConn.addTrack(track, callFeed.stream));
+
+                const tKey = getTransceiverKey(callFeed.purpose, track.kind);
+                if (this.transceivers.has(tKey)) {
+                    // we already have a sender, so we re-use it. We try to re-use transceivers as much
+                    // as possible because they can't be removed once added, so otherwise they just
+                    // accumulate which makes the SDP very large very quickly: in fact it only takes
+                    // about 6 video tracks to exceed the maximum size of an Olm-encrypted
+                    // Matrix event.
+                    const transceiver = this.transceivers.get(tKey);
+
+                    transceiver.sender.replaceTrack(track);
+                    // set the direction to indicate we're going to start sending again
+                    // (this will trigger the re-negotiation)
+                    transceiver.direction = transceiver.direction === "inactive" ? "sendonly" : "sendrecv";
+                } else {
+                    // create a new one: pass the track in and everything happens automatically
+                    this.transceivers.set(tKey, this.peerConn.addTransceiver(track, {
+                        streams: [callFeed.stream],
+                        direction: callFeed.purpose === SDPStreamMetadataPurpose.Usermedia ? "sendrecv" : "sendonly",
+                    }));
+                }
             }
         }
 
@@ -759,20 +804,23 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
      * @param callFeed to remove
      */
     public removeLocalFeed(callFeed: CallFeed): void {
-        const senderArray = callFeed.purpose === SDPStreamMetadataPurpose.Usermedia
+        const audioTransceiverKey = getTransceiverKey(callFeed.purpose, "audio");
-            ? this.usermediaSenders
+        const videoTransceiverKey = getTransceiverKey(callFeed.purpose, "video");
-            : this.screensharingSenders;
 
-        for (const sender of senderArray) {
+        for (const transceiverKey of [audioTransceiverKey, videoTransceiverKey]) {
-            this.peerConn.removeTrack(sender);
+            // this is slightly mixing the track and transceiver API but is basically just shorthand.
+            // There is no way to actually remove a transceiver, so this just sets it to inactive
+            // (or recvonly) and replaces the source with nothing.
+            if (this.transceivers.has(transceiverKey)) {
+                const transceiver = this.transceivers.get(transceiverKey);
+                if (transceiver.sender) this.peerConn.removeTrack(transceiver.sender);
+            }
         }
 
         if (callFeed.purpose === SDPStreamMetadataPurpose.Screenshare) {
             this.client.getMediaHandler().stopScreensharingStream(callFeed.stream);
         }
 
-        // Empty the array
-        senderArray.splice(0, senderArray.length);
         this.deleteFeed(callFeed);
     }
 
@@ -1139,9 +1187,19 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
                 return false;
             }
         } else {
-            for (const sender of this.screensharingSenders) {
+            const audioTransceiver = this.transceivers.get(getTransceiverKey(
-                this.peerConn.removeTrack(sender);
+                SDPStreamMetadataPurpose.Screenshare, "audio",
+            ));
+            const videoTransceiver = this.transceivers.get(getTransceiverKey(
+                SDPStreamMetadataPurpose.Screenshare, "video",
+            ));
+
+            for (const transceiver of [audioTransceiver, videoTransceiver]) {
+                // this is slightly mixing the track and transceiver API but is basically just shorthand
+                // for removing the sender.
+                if (transceiver && transceiver.sender) this.peerConn.removeTrack(transceiver.sender);
             }
+
             this.client.getMediaHandler().stopScreensharingStream(this.localScreensharingStream);
             this.deleteFeedByStream(this.localScreensharingStream);
             return false;
@@ -1167,9 +1225,11 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
                 const track = stream.getTracks().find((track) => {
                     return track.kind === "video";
                 });
-                const sender = this.usermediaSenders.find((sender) => {
+
-                    return sender.track?.kind === "video";
+                const sender = this.transceivers.get(getTransceiverKey(
-                });
+                    SDPStreamMetadataPurpose.Usermedia, "video",
+                )).sender;
+
                 sender.replaceTrack(track);
 
                 this.pushNewLocalFeed(stream, SDPStreamMetadataPurpose.Screenshare, false);
@@ -1183,9 +1243,9 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
             const track = this.localUsermediaStream.getTracks().find((track) => {
                 return track.kind === "video";
             });
-            const sender = this.usermediaSenders.find((sender) => {
+            const sender = this.transceivers.get(getTransceiverKey(
-                return sender.track?.kind === "video";
+                SDPStreamMetadataPurpose.Usermedia, "video",
-            });
+            )).sender;
             sender.replaceTrack(track);
 
             this.client.getMediaHandler().stopScreensharingStream(this.localScreensharingStream);
@@ -1219,15 +1279,12 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
             this.localUsermediaStream.addTrack(track);
         }
 
-        const newSenders = [];
-
         for (const track of stream.getTracks()) {
-            const oldSender = this.usermediaSenders.find((sender) => {
+            const tKey = getTransceiverKey(SDPStreamMetadataPurpose.Usermedia, track.kind);
-                return sender.track?.kind === track.kind;
-            });
-
-            let newSender: RTCRtpSender;
 
+            const oldSender = this.transceivers.get(tKey)?.sender;
+            let added = false;
+            if (oldSender) {
                 try {
                     logger.info(
                         `Call ${this.callId} `+
@@ -1239,8 +1296,13 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
                         `) to peer connection`,
                     );
                     await oldSender.replaceTrack(track);
-                newSender = oldSender;
+                    added = true;
                 } catch (error) {
+                    logger.warn(`replaceTrack failed: adding new transceiver instead`, error);
+                }
+            }
+
+            if (!added) {
                 logger.info(
                     `Call ${this.callId} `+
                     `Adding track (` +
@@ -1250,13 +1312,13 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
                     `streamPurpose="${callFeed.purpose}"` +
                     `) to peer connection`,
                 );
-                newSender = this.peerConn.addTrack(track, this.localUsermediaStream);
-            }
 
-            newSenders.push(newSender);
+                this.transceivers.set(tKey, this.peerConn.addTransceiver(track, {
+                    streams: [this.localUsermediaStream],
+                    direction: "sendrecv",
+                }));
+            }
         }
-
-        this.usermediaSenders = newSenders;
     }
 
     /**
@@ -2109,17 +2171,10 @@ export class MatrixCall extends TypedEventEmitter<CallEvent, CallEventHandlerMap
             }
         }
 
-        for (const trans of this.peerConn.getTransceivers()) {
+        const screenshareVideoTransceiver = this.transceivers.get(getTransceiverKey(
-            if (
+            SDPStreamMetadataPurpose.Screenshare, "video",
-                this.screensharingSenders.includes(trans.sender) &&
+        ));
-                    (
+        if (screenshareVideoTransceiver) screenshareVideoTransceiver.setCodecPreferences(codecs);
-                        trans.sender.track?.kind === "video" ||
-                        trans.receiver.track?.kind === "video"
-                    )
-            ) {
-                trans.setCodecPreferences(codecs);
-            }
-        }
     }
 
     private onNegotiationNeeded = async (): Promise<void> => {

src/webrtc/groupCall.ts

@@ -607,7 +607,9 @@ export class GroupCall extends TypedEventEmitter<
                 return false;
             }
         } else {
-            await Promise.all(this.calls.map(call => call.removeLocalFeed(call.localScreensharingFeed)));
+            await Promise.all(this.calls.map(call => {
+                if (call.localScreensharingFeed) call.removeLocalFeed(call.localScreensharingFeed);
+            }));
             this.client.getMediaHandler().stopScreensharingStream(this.localScreenshareFeed.stream);
             this.removeScreenshareFeed(this.localScreenshareFeed);
             this.localScreenshareFeed = undefined;