support generic tracks with multiple formats (https://github.com/aler9/rtsp-simple-server/issues/1103)

2025-10-05 23:26:54 +08:00 · 2022-10-28 19:10:54 +02:00
parent 394c2f0696
commit 6e6be34636
8 changed files with 243 additions and 157 deletions
--- a/client_read_test.go
+++ b/client_read_test.go
@@ -247,9 +247,13 @@ func TestClientRead(t *testing.T) {

 				track := &TrackGeneric{
 					Media: "application",
-					Formats: []string{"97"},
+					Payloads: []TrackGenericPayload{{
+						Type:   97,
 						RTPMap: "97 private/90000",
+					}},
 				}
+				err = track.Init()
+				require.NoError(t, err)

 				tracks := Tracks{track}
 				tracks.setControls()
--- a/track.go
+++ b/track.go
@@ -28,76 +28,110 @@ type Track interface {
 	url(*url.URL) (*url.URL, error)
 }

-func newTrackFromMediaDescription(md *psdp.MediaDescription) (Track, error) {
-	control := func() string {
-		for _, attr := range md.Attributes {
+func getControlAttribute(attributes []psdp.Attribute) string {
+	for _, attr := range attributes {
 		if attr.Key == "control" {
 			return attr.Value
 		}
 	}
 	return ""
-	}()
-
-	rtpmapPart1, payloadType := func() (string, uint8) {
-		rtpmap, ok := md.Attribute("rtpmap")
-		if !ok {
-			return "", 0
-		}
-		rtpmap = strings.TrimSpace(rtpmap)
-
-		rtpmapParts := strings.Split(rtpmap, " ")
-		if len(rtpmapParts) != 2 {
-			return "", 0
 }

-		tmp, err := strconv.ParseInt(rtpmapParts[0], 10, 64)
+func getRtpmapAttribute(attributes []psdp.Attribute, payloadType uint8) string {
+	for _, attr := range attributes {
+		if attr.Key == "rtpmap" {
+			v := strings.TrimSpace(attr.Value)
+			if parts := strings.SplitN(v, " ", 2); len(parts) == 2 {
+				if tmp, err := strconv.ParseInt(parts[0], 10, 8); err == nil && uint8(tmp) == payloadType {
+					return parts[1]
+				}
+			}
+		}
+	}
+	return ""
+}
+
+func getFmtpAttribute(attributes []psdp.Attribute, payloadType uint8) string {
+	for _, attr := range attributes {
+		if attr.Key == "fmtp" {
+			if parts := strings.SplitN(attr.Value, " ", 2); len(parts) == 2 {
+				if tmp, err := strconv.ParseInt(parts[0], 10, 8); err == nil && uint8(tmp) == payloadType {
+					return parts[1]
+				}
+			}
+		}
+	}
+	return ""
+}
+
+func getCodecAndClock(attributes []psdp.Attribute, payloadType uint8) (string, string) {
+	rtpmap := getRtpmapAttribute(attributes, payloadType)
+	if rtpmap == "" {
+		return "", ""
+	}
+
+	parts2 := strings.SplitN(rtpmap, "/", 2)
+	if len(parts2) != 2 {
+		return "", ""
+	}
+
+	return parts2[0], parts2[1]
+}
+
+func newTrackFromMediaDescription(md *psdp.MediaDescription) (Track, error) {
+	if len(md.MediaName.Formats) == 0 {
+		return nil, fmt.Errorf("no media formats found")
+	}
+
+	control := getControlAttribute(md.Attributes)
+
+	if len(md.MediaName.Formats) == 1 {
+		tmp, err := strconv.ParseInt(md.MediaName.Formats[0], 10, 8)
 		if err != nil {
-			return "", 0
+			return nil, err
 		}
 		payloadType := uint8(tmp)

-		return rtpmapParts[1], payloadType
-	}()
+		codec, clock := getCodecAndClock(md.Attributes, payloadType)

-	if len(md.MediaName.Formats) == 1 {
 		switch {
 		case md.MediaName.Media == "video":
 			switch {
-			case md.MediaName.Formats[0] == "26":
+			case payloadType == 26:
 				return newTrackJPEGFromMediaDescription(control)

-			case md.MediaName.Formats[0] == "32":
+			case payloadType == 32:
 				return newTrackMPEG2VideoFromMediaDescription(control)

-			case rtpmapPart1 == "H264/90000":
+			case codec == "H264" && clock == "90000":
 				return newTrackH264FromMediaDescription(control, payloadType, md)

-			case rtpmapPart1 == "H265/90000":
+			case codec == "H265" && clock == "90000":
 				return newTrackH265FromMediaDescription(control, payloadType, md)

-			case rtpmapPart1 == "VP8/90000":
+			case codec == "VP8" && clock == "90000":
 				return newTrackVP8FromMediaDescription(control, payloadType, md)

-			case rtpmapPart1 == "VP9/90000":
+			case codec == "VP9" && clock == "90000":
 				return newTrackVP9FromMediaDescription(control, payloadType, md)
 			}

 		case md.MediaName.Media == "audio":
 			switch {
-			case md.MediaName.Formats[0] == "0":
-				return newTrackPCMUFromMediaDescription(control, rtpmapPart1)
+			case payloadType == 0:
+				return newTrackPCMUFromMediaDescription(control, clock)

-			case md.MediaName.Formats[0] == "8":
-				return newTrackPCMAFromMediaDescription(control, rtpmapPart1)
+			case payloadType == 8:
+				return newTrackPCMAFromMediaDescription(control, clock)

-			case md.MediaName.Formats[0] == "14":
+			case payloadType == 14:
 				return newTrackMPEG2AudioFromMediaDescription(control)

-			case strings.HasPrefix(strings.ToLower(rtpmapPart1), "mpeg4-generic/"):
+			case strings.ToLower(codec) == "mpeg4-generic":
 				return newTrackMPEG4AudioFromMediaDescription(control, payloadType, md)

-			case strings.HasPrefix(rtpmapPart1, "opus/"):
-				return newTrackOpusFromMediaDescription(control, payloadType, rtpmapPart1, md)
+			case codec == "opus":
+				return newTrackOpusFromMediaDescription(control, payloadType, clock, md)
 			}
 		}
 	}
--- a/track_generic.go
+++ b/track_generic.go
@@ -8,152 +8,166 @@ import (
 	psdp "github.com/pion/sdp/v3"
 )

-func trackGenericGetClockRate(formats []string, rtpmap string) (int, error) {
-	if len(formats) < 1 {
-		return 0, fmt.Errorf("no formats provided")
-	}
+func findClockRate(track *TrackGeneric) (int, error) {
+	// RFC 4566
+	// When a list of
+	// payload type numbers is given, this implies that all of these
+	// payload formats MAY be used in the session, but the first of these
+	// formats SHOULD be used as the default format for the session
+	payload := track.Payloads[0]

 	// get clock rate from payload type
 	// https://en.wikipedia.org/wiki/RTP_payload_formats
-	switch formats[0] {
-	case "0", "1", "2", "3", "4", "5", "7", "8", "9", "12", "13", "15", "18":
+	switch payload.Type {
+	case 0, 1, 2, 3, 4, 5, 7, 8, 9, 12, 13, 15, 18:
 		return 8000, nil

-	case "6":
+	case 6:
 		return 16000, nil

-	case "10", "11":
+	case 10, 11:
 		return 44100, nil

-	case "14", "25", "26", "28", "31", "32", "33", "34":
+	case 14, 25, 26, 28, 31, 32, 33, 34:
 		return 90000, nil

-	case "16":
+	case 16:
 		return 11025, nil

-	case "17":
+	case 17:
 		return 22050, nil
 	}

 	// get clock rate from rtpmap
 	// https://tools.ietf.org/html/rfc4566
 	// a=rtpmap:<payload type> <encoding name>/<clock rate> [/<encoding parameters>]
-	if rtpmap == "" {
+	if payload.RTPMap == "" {
 		return 0, fmt.Errorf("attribute 'rtpmap' not found")
 	}

-	tmp := strings.Split(rtpmap, " ")
-	if len(tmp) < 2 {
-		return 0, fmt.Errorf("invalid rtpmap (%v)", rtpmap)
-	}
-
-	tmp = strings.Split(tmp[1], "/")
+	tmp := strings.Split(payload.RTPMap, "/")
 	if len(tmp) != 2 && len(tmp) != 3 {
-		return 0, fmt.Errorf("invalid rtpmap (%v)", rtpmap)
+		return 0, fmt.Errorf("invalid rtpmap (%v)", payload.RTPMap)
 	}

 	v, err := strconv.ParseInt(tmp[1], 10, 64)
 	if err != nil {
 		return 0, err
 	}
+
 	return int(v), nil
 }

+// TrackGenericPayload is a payload of a TrackGeneric.
+type TrackGenericPayload struct {
+	Type   uint8
+	RTPMap string
+	FMTP   string
+}
+
 // TrackGeneric is a generic track.
 type TrackGeneric struct {
 	Media    string
-	Formats []string
-	RTPMap  string
-	FMTP    string
+	Payloads []TrackGenericPayload

 	trackBase
+
+	clockRate int
 }

 func newTrackGenericFromMediaDescription(
 	control string,
 	md *psdp.MediaDescription,
 ) (*TrackGeneric, error) {
-	rtpmap := func() string {
-		for _, attr := range md.Attributes {
-			if attr.Key == "rtpmap" {
-				return attr.Value
-			}
-		}
-		return ""
-	}()
-
-	_, err := trackGenericGetClockRate(md.MediaName.Formats, rtpmap)
-	if err != nil {
-		return nil, fmt.Errorf("unable to get clock rate: %s", err)
-	}
-
-	fmtp := func() string {
-		for _, attr := range md.Attributes {
-			if attr.Key == "fmtp" {
-				return attr.Value
-			}
-		}
-		return ""
-	}()
-
-	return &TrackGeneric{
+	t := &TrackGeneric{
 		Media: md.MediaName.Media,
-		Formats: md.MediaName.Formats,
-		RTPMap:  rtpmap,
-		FMTP:    fmtp,
 		trackBase: trackBase{
 			control: control,
 		},
-	}, nil
+	}
+
+	for _, format := range md.MediaName.Formats {
+		tmp, err := strconv.ParseInt(format, 10, 8)
+		if err != nil {
+			return nil, err
+		}
+		payloadType := uint8(tmp)
+
+		t.Payloads = append(t.Payloads, TrackGenericPayload{
+			Type:   payloadType,
+			RTPMap: getRtpmapAttribute(md.Attributes, payloadType),
+			FMTP:   getFmtpAttribute(md.Attributes, payloadType),
+		})
+	}
+
+	err := t.Init()
+	if err != nil {
+		return nil, err
+	}
+
+	return t, nil
+}
+
+// Init initializes a TrackGeneric
+func (t *TrackGeneric) Init() error {
+	var err error
+	t.clockRate, err = findClockRate(t)
+	if err != nil {
+		return fmt.Errorf("unable to get clock rate: %s", err)
+	}
+
+	return nil
 }

 // ClockRate returns the track clock rate.
 func (t *TrackGeneric) ClockRate() int {
-	clockRate, _ := trackGenericGetClockRate(t.Formats, t.RTPMap)
-	return clockRate
+	return t.clockRate
 }

 func (t *TrackGeneric) clone() Track {
 	return &TrackGeneric{
 		Media:     t.Media,
-		Formats:   t.Formats,
-		RTPMap:    t.RTPMap,
-		FMTP:      t.FMTP,
+		Payloads:  append([]TrackGenericPayload(nil), t.Payloads...),
 		trackBase: t.trackBase,
+		clockRate: t.clockRate,
 	}
 }

 // MediaDescription returns the track media description in SDP format.
 func (t *TrackGeneric) MediaDescription() *psdp.MediaDescription {
-	return &psdp.MediaDescription{
-		MediaName: psdp.MediaName{
-			Media:   t.Media,
-			Protos:  []string{"RTP", "AVP"},
-			Formats: t.Formats,
-		},
-		Attributes: func() []psdp.Attribute {
-			var ret []psdp.Attribute
+	formats := make([]string, len(t.Payloads))
+	for i, pl := range t.Payloads {
+		formats[i] = strconv.FormatInt(int64(pl.Type), 10)
+	}

-			if t.RTPMap != "" {
-				ret = append(ret, psdp.Attribute{
+	var attributes []psdp.Attribute
+
+	for _, pl := range t.Payloads {
+		if pl.RTPMap != "" {
+			attributes = append(attributes, psdp.Attribute{
 				Key:   "rtpmap",
-					Value: t.RTPMap,
+				Value: strconv.FormatInt(int64(pl.Type), 10) + " " + pl.RTPMap,
 			})
 		}
-
-			if t.FMTP != "" {
-				ret = append(ret, psdp.Attribute{
+		if pl.FMTP != "" {
+			attributes = append(attributes, psdp.Attribute{
 				Key:   "fmtp",
-					Value: t.FMTP,
+				Value: strconv.FormatInt(int64(pl.Type), 10) + " " + pl.FMTP,
 			})
 		}
+	}

-			ret = append(ret, psdp.Attribute{
+	attributes = append(attributes, psdp.Attribute{
 		Key:   "control",
 		Value: t.control,
 	})

-			return ret
-		}(),
+	return &psdp.MediaDescription{
+		MediaName: psdp.MediaName{
+			Media:   t.Media,
+			Protos:  []string{"RTP", "AVP"},
+			Formats: formats,
+		},
+		Attributes: attributes,
 	}
 }
--- a/track_generic_test.go
+++ b/track_generic_test.go
@@ -10,11 +10,20 @@ import (
 func TestTrackGenericClone(t *testing.T) {
 	track := &TrackGeneric{
 		Media: "video",
-		Formats: []string{"100", "101"},
-		RTPMap:  "98 H265/90000",
-		FMTP: "98 profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
+		Payloads: []TrackGenericPayload{
+			{
+				Type:   98,
+				RTPMap: "H265/90000",
+				FMTP: "profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
 					"sprop-sps=QgEBAWAAAAMAAAMAAAMAAAMAlqADwIAQ5Za5JMmuWcBSSgAAB9AAAHUwgkA=; sprop-pps=RAHgdrAwxmQ=",
+			},
+			{
+				Type: 100,
+			},
+		},
 	}
+	err := track.Init()
+	require.NoError(t, err)

 	clone := track.clone()
 	require.NotSame(t, track, clone)
@@ -24,16 +33,26 @@ func TestTrackGenericClone(t *testing.T) {
 func TestTrackGenericMediaDescription(t *testing.T) {
 	track := &TrackGeneric{
 		Media: "video",
-		Formats: []string{"100", "101"},
-		RTPMap:  "98 H265/90000",
-		FMTP: "98 profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
+		Payloads: []TrackGenericPayload{
+			{
+				Type:   98,
+				RTPMap: "H265/90000",
+				FMTP: "profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
 					"sprop-sps=QgEBAWAAAAMAAAMAAAMAAAMAlqADwIAQ5Za5JMmuWcBSSgAAB9AAAHUwgkA=; sprop-pps=RAHgdrAwxmQ=",
+			},
+			{
+				Type: 100,
+			},
+		},
 	}
+	err := track.Init()
+	require.NoError(t, err)
+
 	require.Equal(t, &psdp.MediaDescription{
 		MediaName: psdp.MediaName{
 			Media:   "video",
 			Protos:  []string{"RTP", "AVP"},
-			Formats: []string{"100", "101"},
+			Formats: []string{"98", "100"},
 		},
 		Attributes: []psdp.Attribute{
 			{
--- a/track_opus.go
+++ b/track_opus.go
@@ -20,20 +20,20 @@ type TrackOpus struct {
 func newTrackOpusFromMediaDescription(
 	control string,
 	payloadType uint8,
-	rtpmapPart1 string,
+	clock string,
 	md *psdp.MediaDescription,
 ) (*TrackOpus, error) {
-	tmp := strings.SplitN(rtpmapPart1, "/", 3)
-	if len(tmp) != 3 {
-		return nil, fmt.Errorf("invalid rtpmap (%v)", rtpmapPart1)
+	tmp := strings.SplitN(clock, "/", 32)
+	if len(tmp) != 2 {
+		return nil, fmt.Errorf("invalid clock (%v)", clock)
 	}

-	sampleRate, err := strconv.ParseInt(tmp[1], 10, 64)
+	sampleRate, err := strconv.ParseInt(tmp[0], 10, 64)
 	if err != nil {
 		return nil, err
 	}

-	channelCount, err := strconv.ParseInt(tmp[2], 10, 64)
+	channelCount, err := strconv.ParseInt(tmp[1], 10, 64)
 	if err != nil {
 		return nil, err
 	}
--- a/track_pcma.go
+++ b/track_pcma.go
@@ -17,8 +17,8 @@ func newTrackPCMAFromMediaDescription(
 	rtpmapPart1 string) (*TrackPCMA, error,
 ) {
 	tmp := strings.Split(rtpmapPart1, "/")
-	if len(tmp) >= 3 && tmp[2] != "1" {
-		return nil, fmt.Errorf("PCMA tracks must have only one channel")
+	if len(tmp) == 2 && tmp[1] != "1" {
+		return nil, fmt.Errorf("PCMU tracks can have only one channel")
 	}

 	return &TrackPCMA{
--- a/track_pcmu.go
+++ b/track_pcmu.go
@@ -14,11 +14,11 @@ type TrackPCMU struct {

 func newTrackPCMUFromMediaDescription(
 	control string,
-	rtpmapPart1 string) (*TrackPCMU, error,
+	clock string) (*TrackPCMU, error,
 ) {
-	tmp := strings.Split(rtpmapPart1, "/")
-	if len(tmp) >= 3 && tmp[2] != "1" {
-		return nil, fmt.Errorf("PCMU tracks must have only one channel")
+	tmp := strings.SplitN(clock, "/", 2)
+	if len(tmp) == 2 && tmp[1] != "1" {
+		return nil, fmt.Errorf("PCMU tracks can have only one channel")
 	}

 	return &TrackPCMU{
--- a/track_test.go
+++ b/track_test.go
@@ -457,26 +457,41 @@ func TestTrackNewFromMediaDescription(t *testing.T) {
 				MediaName: psdp.MediaName{
 					Media:   "video",
 					Protos:  []string{"RTP", "AVP"},
-					Formats: []string{"98", "96"},
+					Formats: []string{"96", "98"},
 				},
 				Attributes: []psdp.Attribute{
 					{
 						Key:   "rtpmap",
-						Value: "98 H265/90000",
+						Value: "96 H264/90000",
+					},
+					{
+						Key:   "rtpmap",
+						Value: "98 MetaData",
+					},
+					{
+						Key: "rtcp-mux",
 					},
 					{
 						Key: "fmtp",
-						Value: "98 profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
-							"sprop-sps=QgEBAWAAAAMAAAMAAAMAAAMAlqADwIAQ5Za5JMmuWcBSSgAAB9AAAHUwgkA=; sprop-pps=RAHgdrAwxmQ=",
+						Value: "96 packetization-mode=1;profile-level-id=4d002a;" +
+							"sprop-parameter-sets=Z00AKp2oHgCJ+WbgICAgQA==,aO48gA==",
 					},
 				},
 			},
 			&TrackGeneric{
 				Media: "video",
-				Formats: []string{"98", "96"},
-				RTPMap:  "98 H265/90000",
-				FMTP: "98 profile-id=1; sprop-vps=QAEMAf//AWAAAAMAAAMAAAMAAAMAlqwJ; " +
-					"sprop-sps=QgEBAWAAAAMAAAMAAAMAAAMAlqADwIAQ5Za5JMmuWcBSSgAAB9AAAHUwgkA=; sprop-pps=RAHgdrAwxmQ=",
+				Payloads: []TrackGenericPayload{
+					{
+						Type:   96,
+						RTPMap: "H264/90000",
+						FMTP:   "packetization-mode=1;profile-level-id=4d002a;sprop-parameter-sets=Z00AKp2oHgCJ+WbgICAgQA==,aO48gA==",
+					},
+					{
+						Type:   98,
+						RTPMap: "MetaData",
+					},
+				},
+				clockRate: 90000,
 			},
 		},
 	} {
@@ -503,7 +518,7 @@ func TestTrackNewFromMediaDescriptionErrors(t *testing.T) {
 					Formats: []string{},
 				},
 			},
-			"unable to get clock rate: no formats provided",
+			"no media formats found",
 		},
 		{
 			"no rtpmap",
@@ -527,11 +542,11 @@ func TestTrackNewFromMediaDescriptionErrors(t *testing.T) {
 				Attributes: []psdp.Attribute{
 					{
 						Key:   "rtpmap",
-						Value: "96",
+						Value: "97 mpeg4-generic/48000/2",
 					},
 				},
 			},
-			"unable to get clock rate: invalid rtpmap (96)",
+			"unable to get clock rate: attribute 'rtpmap' not found",
 		},
 		{
 			"invalid clockrate 2",
@@ -548,7 +563,7 @@ func TestTrackNewFromMediaDescriptionErrors(t *testing.T) {
 					},
 				},
 			},
-			"unable to get clock rate: invalid rtpmap (96 mpeg4-generic)",
+			"unable to get clock rate: invalid rtpmap (mpeg4-generic)",
 		},
 		{
 			"invalid clockrate 3",
@@ -725,7 +740,7 @@ func TestTrackNewFromMediaDescriptionErrors(t *testing.T) {
 					},
 				},
 			},
-			"invalid rtpmap (opus/48000)",
+			"invalid clock (48000)",
 		},
 		{
 			"opus invalid 2",