Firefox case when screen and camera offer together

mediaengine.go

@@ -476,10 +476,10 @@ func (m *MediaEngine) updateFromRemoteDescription(desc sdp.SessionDescription) e
 	for _, media := range desc.MediaDescriptions {
 		var typ RTPCodecType
 		switch {
-		case !m.negotiatedAudio && strings.EqualFold(media.MediaName.Media, "audio"):
+		case strings.EqualFold(media.MediaName.Media, "audio"):
 			m.negotiatedAudio = true
 			typ = RTPCodecTypeAudio
-		case !m.negotiatedVideo && strings.EqualFold(media.MediaName.Media, "video"):
+		case strings.EqualFold(media.MediaName.Media, "video"):
 			m.negotiatedVideo = true
 			typ = RTPCodecTypeVideo
 		default:
@@ -561,18 +561,24 @@ func (m *MediaEngine) getRTPParametersByKind(typ RTPCodecType, directions []RTPT
 
 	m.mu.RLock()
 	defer m.mu.RUnlock()
-	if m.negotiatedVideo && typ == RTPCodecTypeVideo ||
-		m.negotiatedAudio && typ == RTPCodecTypeAudio {
-		for id, e := range m.negotiatedHeaderExtensions {
-			if haveRTPTransceiverDirectionIntersection(e.allowedDirections, directions) && (e.isAudio && typ == RTPCodecTypeAudio || e.isVideo && typ == RTPCodecTypeVideo) {
-				headerExtensions = append(headerExtensions, RTPHeaderExtensionParameter{ID: id, URI: e.uri})
-			}
-		}
+
+	var mediaHeaderExtensions map[int]mediaEngineHeaderExtension
+	if (m.negotiatedVideo && typ == RTPCodecTypeVideo) || (m.negotiatedAudio && typ == RTPCodecTypeAudio) {
+		mediaHeaderExtensions = m.negotiatedHeaderExtensions
 	} else {
+		mediaHeaderExtensions = make(map[int]mediaEngineHeaderExtension, len(m.headerExtensions))
 		for id, e := range m.headerExtensions {
-			if haveRTPTransceiverDirectionIntersection(e.allowedDirections, directions) && (e.isAudio && typ == RTPCodecTypeAudio || e.isVideo && typ == RTPCodecTypeVideo) {
-				headerExtensions = append(headerExtensions, RTPHeaderExtensionParameter{ID: id + 1, URI: e.uri})
-			}
+			mediaHeaderExtensions[id+1] = e
+		}
+	}
+
+	for id, e := range mediaHeaderExtensions {
+		if (typ == RTPCodecTypeVideo && !e.isVideo) || (typ == RTPCodecTypeAudio && !e.isAudio) {
+			continue
+		}
+
+		if haveRTPTransceiverDirectionIntersection(e.allowedDirections, directions) {
+			headerExtensions = append(headerExtensions, RTPHeaderExtensionParameter{ID: id, URI: e.uri})
 		}
 	}
 

peerconnection.go

@@ -1091,6 +1091,14 @@ func (pc *PeerConnection) SetRemoteDescription(desc SessionDescription) error {
 					_ = t.SetCodecPreferences(filteredCodecs)
 				}
 
+				if extensions, err := rtpExtensionsFromMediaDescription(media); err == nil {
+					headerExtensions := []RTPHeaderExtensionParameter{}
+					for uri, id := range extensions {
+						headerExtensions = append(headerExtensions, RTPHeaderExtensionParameter{URI: uri, ID: id})
+					}
+					_ = t.SetHeaderExtensions(headerExtensions)
+				}
+
 			case direction == RTPTransceiverDirectionRecvonly:
 				if t.Direction() == RTPTransceiverDirectionSendrecv {
 					t.setDirection(RTPTransceiverDirectionSendonly)
@@ -1203,7 +1211,7 @@ func (pc *PeerConnection) startReceiver(incoming trackDetails, receiver *RTPRece
 
 	for _, t := range receiver.Tracks() {
 		if t.SSRC() == 0 || t.RID() != "" {
-			return
+			continue
 		}
 
 		go func(track *TrackRemote) {

rtptransceiver.go

@@ -18,7 +18,8 @@ type RTPTransceiver struct {
 	receiver  atomic.Value // *RTPReceiver
 	direction atomic.Value // RTPTransceiverDirection
 
-	codecs []RTPCodecParameters // User provided codecs via SetCodecPreferences
+	codecs           []RTPCodecParameters // User provided codecs via SetCodecPreferences
+	headerExtensions []RTPHeaderExtensionParameter
 
 	stopped bool
 	kind    RTPCodecType
@@ -80,6 +81,33 @@ func (t *RTPTransceiver) getCodecs() []RTPCodecParameters {
 	return filteredCodecs
 }
 
+// SetHeaderExtensions stores header extension from SDP
+func (t *RTPTransceiver) SetHeaderExtensions(headerExtensions []RTPHeaderExtensionParameter) error {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	t.headerExtensions = headerExtensions
+	return nil
+}
+
+func (t *RTPTransceiver) hasHeaderExtension(uri string) bool {
+	t.mu.RLock()
+	defer t.mu.RUnlock()
+
+	// if nothing specific set, accept all
+	if len(t.headerExtensions) == 0 {
+		return true
+	}
+
+	for _, he := range t.headerExtensions {
+		if he.URI == uri {
+			return true
+		}
+	}
+
+	return false
+}
+
 // Sender returns the RTPTransceiver's RTPSender if it has one
 func (t *RTPTransceiver) Sender() *RTPSender {
 	if v := t.sender.Load(); v != nil {

sdp.go

@@ -340,7 +340,18 @@ func populateLocalCandidates(sessionDescription *SessionDescription, i *ICEGathe
 	}
 }
 
-func addTransceiverSDP(d *sdp.SessionDescription, isPlanB, shouldAddCandidates bool, dtlsFingerprints []DTLSFingerprint, mediaEngine *MediaEngine, midValue string, iceParams ICEParameters, candidates []ICECandidate, dtlsRole sdp.ConnectionRole, iceGatheringState ICEGatheringState, mediaSection mediaSection) (bool, error) {
+func addTransceiverSDP(
+	d *sdp.SessionDescription,
+	isPlanB, shouldAddCandidates bool,
+	dtlsFingerprints []DTLSFingerprint,
+	mediaEngine *MediaEngine,
+	midValue string,
+	iceParams ICEParameters,
+	candidates []ICECandidate,
+	dtlsRole sdp.ConnectionRole,
+	iceGatheringState ICEGatheringState,
+	mediaSection mediaSection,
+) (bool, error) {
 	transceivers := mediaSection.transceivers
 	if len(transceivers) < 1 {
 		return false, errSDPZeroTransceivers
@@ -392,6 +403,10 @@ func addTransceiverSDP(d *sdp.SessionDescription, isPlanB, shouldAddCandidates b
 
 	parameters := mediaEngine.getRTPParametersByKind(t.kind, directions)
 	for _, rtpExtension := range parameters.HeaderExtensions {
+		if !t.hasHeaderExtension(rtpExtension.URI) {
+			continue
+		}
+
 		extURL, err := url.Parse(rtpExtension.URI)
 		if err != nil {
 			return false, err