update gortsplib (#2390)

go.mod

@@ -8,7 +8,7 @@ require (
 	github.com/alecthomas/kong v0.8.0
 	github.com/aler9/writerseeker v1.1.0
 	github.com/bluenviron/gohlslib v1.0.3
-	github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230919201539-fc2a10a4999a
+	github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230921145131-44da79f72d5e
 	github.com/bluenviron/mediacommon v1.3.1-0.20230919191723-607668055ebe
 	github.com/datarhei/gosrt v0.5.4
 	github.com/fsnotify/fsnotify v1.6.0

go.sum

@@ -14,8 +14,8 @@ github.com/benburkert/openpgp v0.0.0-20160410205803-c2471f86866c h1:8XZeJrs4+ZYh
 github.com/benburkert/openpgp v0.0.0-20160410205803-c2471f86866c/go.mod h1:x1vxHcL/9AVzuk5HOloOEPrtJY0MaalYr78afXZ+pWI=
 github.com/bluenviron/gohlslib v1.0.3 h1:FMHevlIrrZ67uzCXmlTSGflsfYREEtHb8L9BDyf7lJc=
 github.com/bluenviron/gohlslib v1.0.3/go.mod h1:R/aIsSxLI61N0CVMjtcHqJouK6+Ddd5YIihcCr7IFIw=
-github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230919201539-fc2a10a4999a h1:2zaPnmdTRJyfySBbRF8mCQDfOFOL2qQHfQFrLpKeyS0=
+github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230921145131-44da79f72d5e h1:Y8b0vKPQLerALedmNNBmxrJR6sBcnge+fQeCH+Kfh3A=
-github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230919201539-fc2a10a4999a/go.mod h1:0rVtKDafUA14isZuaBTm5+X9NPqLYs/lY8JIww6+doM=
+github.com/bluenviron/gortsplib/v4 v4.1.1-0.20230921145131-44da79f72d5e/go.mod h1:0rVtKDafUA14isZuaBTm5+X9NPqLYs/lY8JIww6+doM=
 github.com/bluenviron/mediacommon v1.3.1-0.20230919191723-607668055ebe h1:8kvIJfRXvv1Za1hdArKjvd/l8WCHJF+d+oLtANdFbr8=
 github.com/bluenviron/mediacommon v1.3.1-0.20230919191723-607668055ebe/go.mod h1:/vlOVSebDwzdRtQONOKLua0fOSJg1tUDHpP+h9a0uqM=
 github.com/bytedance/sonic v1.5.0/go.mod h1:ED5hyg4y6t3/9Ku1R6dU/4KyJ48DZ4jPhfY1O2AihPM=

internal/core/hls_muxer.go

@@ -472,12 +472,12 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*description.Media,
 		}
 	}
 
-	var audioFormatMPEG4AudioGeneric *format.MPEG4AudioGeneric
+	var audioFormatMPEG4Audio *format.MPEG4Audio
-	audioMedia = stream.Desc().FindFormat(&audioFormatMPEG4AudioGeneric)
+	audioMedia = stream.Desc().FindFormat(&audioFormatMPEG4Audio)
 
 	if audioMedia != nil {
-		stream.AddReader(m.writer, audioMedia, audioFormatMPEG4AudioGeneric, func(u unit.Unit) error {
+		stream.AddReader(m.writer, audioMedia, audioFormatMPEG4Audio, func(u unit.Unit) error {
-			tunit := u.(*unit.MPEG4AudioGeneric)
+			tunit := u.(*unit.MPEG4Audio)
 
 			if tunit.AUs == nil {
 				return nil
@@ -496,39 +496,7 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*description.Media,
 
 		return audioMedia, &gohlslib.Track{
 			Codec: &codecs.MPEG4Audio{
-				Config: *audioFormatMPEG4AudioGeneric.Config,
+				Config: *audioFormatMPEG4Audio.GetConfig(),
-			},
-		}
-	}
-
-	var audioFormatMPEG4AudioLATM *format.MPEG4AudioLATM
-	audioMedia = stream.Desc().FindFormat(&audioFormatMPEG4AudioLATM)
-
-	if audioMedia != nil &&
-		audioFormatMPEG4AudioLATM.Config != nil &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs) == 1 &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs[0].Layers) == 1 {
-		stream.AddReader(m.writer, audioMedia, audioFormatMPEG4AudioLATM, func(u unit.Unit) error {
-			tunit := u.(*unit.MPEG4AudioLATM)
-
-			if tunit.AU == nil {
-				return nil
-			}
-
-			err := m.muxer.WriteMPEG4Audio(
-				tunit.NTP,
-				tunit.PTS,
-				[][]byte{tunit.AU})
-			if err != nil {
-				return fmt.Errorf("muxer error: %v", err)
-			}
-
-			return nil
-		})
-
-		return audioMedia, &gohlslib.Track{
-			Codec: &codecs.MPEG4Audio{
-				Config: *audioFormatMPEG4AudioLATM.Config.Programs[0].Layers[0].AudioSpecificConfig,
 			},
 		}
 	}

internal/core/hls_source.go

@@ -168,7 +168,7 @@ func (s *hlsSource) run(ctx context.Context, cnf *conf.PathConf, reloadConf chan
 					}
 
 					c.OnDataMPEG4Audio(track, func(pts time.Duration, aus [][]byte) {
-						stream.WriteUnit(medi, medi.Formats[0], &unit.MPEG4AudioGeneric{
+						stream.WriteUnit(medi, medi.Formats[0], &unit.MPEG4Audio{
 							Base: unit.Base{
 								NTP: time.Now(),
 								PTS: pts,

internal/core/mpegts.go

@@ -135,7 +135,7 @@ func mpegtsSetupTracks(r *mpegts.Reader, stream **stream.Stream) ([]*description
 			}
 
 			r.OnDataMPEG4Audio(track, func(pts int64, aus [][]byte) error {
-				(*stream).WriteUnit(medi, medi.Formats[0], &unit.MPEG4AudioGeneric{
+				(*stream).WriteUnit(medi, medi.Formats[0], &unit.MPEG4Audio{
 					Base: unit.Base{
 						NTP: time.Now(),
 						PTS: decodeTime(pts),

internal/core/rtmp_conn.go

@@ -395,12 +395,12 @@ func (c *rtmpConn) setupAudio(
 	stream *stream.Stream,
 	writer *asyncwriter.Writer,
 ) (*description.Media, format.Format) {
-	var audioFormatMPEG4Generic *format.MPEG4AudioGeneric
+	var audioFormatMPEG4Audio *format.MPEG4Audio
-	audioMedia := stream.Desc().FindFormat(&audioFormatMPEG4Generic)
+	audioMedia := stream.Desc().FindFormat(&audioFormatMPEG4Audio)
 
 	if audioMedia != nil {
-		stream.AddReader(writer, audioMedia, audioFormatMPEG4Generic, func(u unit.Unit) error {
+		stream.AddReader(writer, audioMedia, audioFormatMPEG4Audio, func(u unit.Unit) error {
-			tunit := u.(*unit.MPEG4AudioGeneric)
+			tunit := u.(*unit.MPEG4Audio)
 
 			if tunit.AUs == nil {
 				return nil
@@ -410,7 +410,7 @@ func (c *rtmpConn) setupAudio(
 				c.nconn.SetWriteDeadline(time.Now().Add(time.Duration(c.writeTimeout)))
 				err := (*w).WriteMPEG4Audio(
 					tunit.PTS+time.Duration(i)*mpeg4audio.SamplesPerAccessUnit*
-						time.Second/time.Duration(audioFormatMPEG4Generic.ClockRate()),
+						time.Second/time.Duration(audioFormatMPEG4Audio.ClockRate()),
 					au,
 				)
 				if err != nil {
@@ -421,28 +421,7 @@ func (c *rtmpConn) setupAudio(
 			return nil
 		})
 
-		return audioMedia, audioFormatMPEG4Generic
+		return audioMedia, audioFormatMPEG4Audio
-	}
-
-	var audioFormatMPEG4AudioLATM *format.MPEG4AudioLATM
-	audioMedia = stream.Desc().FindFormat(&audioFormatMPEG4AudioLATM)
-
-	if audioMedia != nil &&
-		audioFormatMPEG4AudioLATM.Config != nil &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs) == 1 &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs[0].Layers) == 1 {
-		stream.AddReader(writer, audioMedia, audioFormatMPEG4AudioLATM, func(u unit.Unit) error {
-			tunit := u.(*unit.MPEG4AudioLATM)
-
-			if tunit.AU == nil {
-				return nil
-			}
-
-			c.nconn.SetWriteDeadline(time.Now().Add(time.Duration(c.writeTimeout)))
-			return (*w).WriteMPEG4Audio(tunit.PTS, tunit.AU)
-		})
-
-		return audioMedia, audioFormatMPEG4AudioLATM
 	}
 
 	var audioFormatMPEG1 *format.MPEG1Audio
@@ -590,9 +569,9 @@ func (c *rtmpConn) runPublish(conn *rtmp.Conn, u *url.URL) error {
 		medias = append(medias, audioMedia)
 
 		switch audioFormat.(type) {
-		case *format.MPEG4AudioGeneric:
+		case *format.MPEG4Audio:
 			r.OnDataMPEG4Audio(func(pts time.Duration, au []byte) {
-				stream.WriteUnit(audioMedia, audioFormat, &unit.MPEG4AudioGeneric{
+				stream.WriteUnit(audioMedia, audioFormat, &unit.MPEG4Audio{
 					Base: unit.Base{
 						NTP: time.Now(),
 						PTS: pts,

internal/core/rtmp_source.go

@@ -148,9 +148,9 @@ func (s *rtmpSource) runReader(u *url.URL, nconn net.Conn) error {
 		medias = append(medias, audioMedia)
 
 		switch audioFormat.(type) {
-		case *format.MPEG4AudioGeneric:
+		case *format.MPEG4Audio:
 			mc.OnDataMPEG4Audio(func(pts time.Duration, au []byte) {
-				stream.WriteUnit(audioMedia, audioFormat, &unit.MPEG4AudioGeneric{
+				stream.WriteUnit(audioMedia, audioFormat, &unit.MPEG4Audio{
 					Base: unit.Base{
 						NTP: time.Now(),
 						PTS: pts,

internal/core/srt_conn.go

@@ -467,13 +467,13 @@ func (c *srtConn) runRead(req srtNewConnReq, pathName string, user string, pass
 					return bw.Flush()
 				})
 
-			case *format.MPEG4AudioGeneric:
+			case *format.MPEG4Audio:
 				track := addTrack(medi, &mpegts.CodecMPEG4Audio{
-					Config: *forma.Config,
+					Config: *forma.GetConfig(),
 				})
 
 				res.stream.AddReader(writer, medi, forma, func(u unit.Unit) error {
-					tunit := u.(*unit.MPEG4AudioGeneric)
+					tunit := u.(*unit.MPEG4Audio)
 					if tunit.AUs == nil {
 						return nil
 					}
@@ -486,29 +486,6 @@ func (c *srtConn) runRead(req srtNewConnReq, pathName string, user string, pass
 					return bw.Flush()
 				})
 
-			case *format.MPEG4AudioLATM:
-				if forma.Config != nil &&
-					len(forma.Config.Programs) == 1 &&
-					len(forma.Config.Programs[0].Layers) == 1 {
-					track := addTrack(medi, &mpegts.CodecMPEG4Audio{
-						Config: *forma.Config.Programs[0].Layers[0].AudioSpecificConfig,
-					})
-
-					res.stream.AddReader(writer, medi, forma, func(u unit.Unit) error {
-						tunit := u.(*unit.MPEG4AudioLATM)
-						if tunit.AU == nil {
-							return nil
-						}
-
-						sconn.SetWriteDeadline(time.Now().Add(time.Duration(c.writeTimeout)))
-						err = w.WriteMPEG4Audio(track, durationGoToMPEGTS(tunit.PTS), [][]byte{tunit.AU})
-						if err != nil {
-							return err
-						}
-						return bw.Flush()
-					})
-				}
-
 			case *format.Opus:
 				track := addTrack(medi, &mpegts.CodecOpus{
 					ChannelCount: func() int {

internal/formatprocessor/mpeg4_audio.go

@@ -11,19 +11,19 @@ import (
 	"github.com/bluenviron/mediamtx/internal/unit"
 )
 
-type formatProcessorMPEG4AudioGeneric struct {
+type formatProcessorMPEG4Audio struct {
 	udpMaxPayloadSize int
 	format            *format.MPEG4Audio
 	encoder           *rtpmpeg4audio.Encoder
 	decoder           *rtpmpeg4audio.Decoder
 }
 
-func newMPEG4AudioGeneric(
+func newMPEG4Audio(
 	udpMaxPayloadSize int,
 	forma *format.MPEG4Audio,
 	generateRTPPackets bool,
-) (*formatProcessorMPEG4AudioGeneric, error) {
+) (*formatProcessorMPEG4Audio, error) {
-	t := &formatProcessorMPEG4AudioGeneric{
+	t := &formatProcessorMPEG4Audio{
 		udpMaxPayloadSize: udpMaxPayloadSize,
 		format:            forma,
 	}
@@ -38,7 +38,7 @@ func newMPEG4AudioGeneric(
 	return t, nil
 }
 
-func (t *formatProcessorMPEG4AudioGeneric) createEncoder() error {
+func (t *formatProcessorMPEG4Audio) createEncoder() error {
 	t.encoder = &rtpmpeg4audio.Encoder{
 		PayloadMaxSize:   t.udpMaxPayloadSize - 12,
 		PayloadType:      t.format.PayloadTyp,
@@ -49,8 +49,8 @@ func (t *formatProcessorMPEG4AudioGeneric) createEncoder() error {
 	return t.encoder.Init()
 }
 
-func (t *formatProcessorMPEG4AudioGeneric) ProcessUnit(uu unit.Unit) error { //nolint:dupl
+func (t *formatProcessorMPEG4Audio) ProcessUnit(uu unit.Unit) error { //nolint:dupl
-	u := uu.(*unit.MPEG4AudioGeneric)
+	u := uu.(*unit.MPEG4Audio)
 
 	pkts, err := t.encoder.Encode(u.AUs)
 	if err != nil {
@@ -67,13 +67,13 @@ func (t *formatProcessorMPEG4AudioGeneric) ProcessUnit(uu unit.Unit) error { //n
 	return nil
 }
 
-func (t *formatProcessorMPEG4AudioGeneric) ProcessRTPPacket( //nolint:dupl
+func (t *formatProcessorMPEG4Audio) ProcessRTPPacket( //nolint:dupl
 	pkt *rtp.Packet,
 	ntp time.Time,
 	pts time.Duration,
 	hasNonRTSPReaders bool,
 ) (Unit, error) {
-	u := &unit.MPEG4AudioGeneric{
+	u := &unit.MPEG4Audio{
 		Base: unit.Base{
 			RTPPackets: []*rtp.Packet{pkt},
 			NTP:        ntp,

internal/formatprocessor/mpeg4_audio_latm.go

@@ -1,112 +0,0 @@
-package formatprocessor
-
-import (
-	"fmt"
-	"time"
-
-	"github.com/bluenviron/gortsplib/v4/pkg/format"
-	"github.com/bluenviron/gortsplib/v4/pkg/format/rtpmpeg4audiolatm"
-	"github.com/pion/rtp"
-
-	"github.com/bluenviron/mediamtx/internal/unit"
-)
-
-type formatProcessorMPEG4AudioLATM struct {
-	udpMaxPayloadSize int
-	format            *format.MPEG4AudioLATM
-	encoder           *rtpmpeg4audiolatm.Encoder
-	decoder           *rtpmpeg4audiolatm.Decoder
-}
-
-func newMPEG4AudioLATM(
-	udpMaxPayloadSize int,
-	forma *format.MPEG4AudioLATM,
-	generateRTPPackets bool,
-) (*formatProcessorMPEG4AudioLATM, error) {
-	t := &formatProcessorMPEG4AudioLATM{
-		udpMaxPayloadSize: udpMaxPayloadSize,
-		format:            forma,
-	}
-
-	if generateRTPPackets {
-		err := t.createEncoder()
-		if err != nil {
-			return nil, err
-		}
-	}
-
-	return t, nil
-}
-
-func (t *formatProcessorMPEG4AudioLATM) createEncoder() error {
-	t.encoder = &rtpmpeg4audiolatm.Encoder{
-		PayloadType: t.format.PayloadTyp,
-	}
-	return t.encoder.Init()
-}
-
-func (t *formatProcessorMPEG4AudioLATM) ProcessUnit(uu unit.Unit) error { //nolint:dupl
-	u := uu.(*unit.MPEG4AudioLATM)
-
-	pkts, err := t.encoder.Encode(u.AU)
-	if err != nil {
-		return err
-	}
-
-	ts := uint32(multiplyAndDivide(u.PTS, time.Duration(t.format.ClockRate()), time.Second))
-	for _, pkt := range pkts {
-		pkt.Timestamp += ts
-	}
-
-	u.RTPPackets = pkts
-
-	return nil
-}
-
-func (t *formatProcessorMPEG4AudioLATM) ProcessRTPPacket( //nolint:dupl
-	pkt *rtp.Packet,
-	ntp time.Time,
-	pts time.Duration,
-	hasNonRTSPReaders bool,
-) (Unit, error) {
-	u := &unit.MPEG4AudioLATM{
-		Base: unit.Base{
-			RTPPackets: []*rtp.Packet{pkt},
-			NTP:        ntp,
-			PTS:        pts,
-		},
-	}
-
-	// remove padding
-	pkt.Header.Padding = false
-	pkt.PaddingSize = 0
-
-	if pkt.MarshalSize() > t.udpMaxPayloadSize {
-		return nil, fmt.Errorf("payload size (%d) is greater than maximum allowed (%d)",
-			pkt.MarshalSize(), t.udpMaxPayloadSize)
-	}
-
-	// decode from RTP
-	if hasNonRTSPReaders || t.decoder != nil {
-		if t.decoder == nil {
-			var err error
-			t.decoder, err = t.format.CreateDecoder()
-			if err != nil {
-				return nil, err
-			}
-		}
-
-		au, err := t.decoder.Decode(pkt)
-		if err != nil {
-			if err == rtpmpeg4audiolatm.ErrMorePacketsNeeded {
-				return u, nil
-			}
-			return nil, err
-		}
-
-		u.AU = au
-	}
-
-	// route packet as is
-	return u, nil
-}

internal/formatprocessor/processor.go

@@ -63,11 +63,8 @@ func New(
 	case *format.Opus:
 		return newOpus(udpMaxPayloadSize, forma, generateRTPPackets)
 
-	case *format.MPEG4AudioGeneric:
+	case *format.MPEG4Audio:
-		return newMPEG4AudioGeneric(udpMaxPayloadSize, forma, generateRTPPackets)
+		return newMPEG4Audio(udpMaxPayloadSize, forma, generateRTPPackets)
-
-	case *format.MPEG4AudioLATM:
-		return newMPEG4AudioLATM(udpMaxPayloadSize, forma, generateRTPPackets)
 
 	case *format.MPEG1Audio:
 		return newMPEG1Audio(udpMaxPayloadSize, forma, generateRTPPackets)

internal/record/agent.go

@@ -571,16 +571,16 @@ func NewAgent(
 					return nil
 				})
 
-			case *format.MPEG4AudioGeneric:
+			case *format.MPEG4Audio:
 				codec := &fmp4.CodecMPEG4Audio{
-					Config: *forma.Config,
+					Config: *forma.GetConfig(),
 				}
 				track := addTrack(codec)
 
-				sampleRate := time.Duration(forma.Config.SampleRate)
+				sampleRate := time.Duration(forma.ClockRate())
 
 				stream.AddReader(r.writer, media, forma, func(u unit.Unit) error {
-					tunit := u.(*unit.MPEG4AudioGeneric)
+					tunit := u.(*unit.MPEG4Audio)
 					if tunit.AUs == nil {
 						return nil
 					}
@@ -603,26 +603,6 @@ func NewAgent(
 					return nil
 				})
 
-			case *format.MPEG4AudioLATM:
-				codec := &fmp4.CodecMPEG4Audio{
-					Config: *forma.Config.Programs[0].Layers[0].AudioSpecificConfig,
-				}
-				track := addTrack(codec)
-
-				stream.AddReader(r.writer, media, forma, func(u unit.Unit) error {
-					tunit := u.(*unit.MPEG4AudioLATM)
-					if tunit.AU == nil {
-						return nil
-					}
-
-					return track.record(&sample{
-						PartSample: &fmp4.PartSample{
-							Payload: tunit.AU,
-						},
-						dts: tunit.PTS,
-					})
-				})
-
 			case *format.MPEG1Audio:
 				codec := &fmp4.CodecMPEG1Audio{
 					SampleRate:   32000,

internal/record/agent_test.go

@@ -132,7 +132,7 @@ func TestAgent(t *testing.T) {
 			},
 		})
 
-		stream.WriteUnit(desc.Medias[2], desc.Medias[2].Formats[0], &unit.MPEG4AudioGeneric{
+		stream.WriteUnit(desc.Medias[2], desc.Medias[2].Formats[0], &unit.MPEG4Audio{
 			Base: unit.Base{
 				PTS: (50 + time.Duration(i)) * time.Second,
 			},

internal/rtmp/writer.go

@@ -87,7 +87,7 @@ func (w *Writer) writeTracks(videoTrack format.Format, audioTrack format.Format)
 						case *format.MPEG1Audio:
 							return message.CodecMPEG1Audio
 
-						case *format.MPEG4AudioGeneric, *format.MPEG4AudioLATM:
+						case *format.MPEG4Audio:
 							return message.CodecMPEG4Audio
 
 						default:
@@ -127,12 +127,8 @@ func (w *Writer) writeTracks(videoTrack format.Format, audioTrack format.Format)
 
 	var audioConfig *mpeg4audio.AudioSpecificConfig
 
-	switch track := audioTrack.(type) {
+	if track, ok := audioTrack.(*format.MPEG4Audio); ok {
-	case *format.MPEG4Audio:
+		audioConfig = track.GetConfig()
-		audioConfig = track.Config
-
-	case *format.MPEG4AudioLATM:
-		audioConfig = track.Config.Programs[0].Layers[0].AudioSpecificConfig
 	}
 
 	if audioConfig != nil {

internal/unit/mpeg4_audio.go

@@ -0,0 +1,7 @@
+package unit
+
+// MPEG4Audio is a MPEG-4 Audio data unit.
+type MPEG4Audio struct {
+	Base
+	AUs [][]byte
+}

internal/unit/mpeg4_audio_generic.go

@@ -1,7 +0,0 @@
-package unit
-
-// MPEG4AudioGeneric is a MPEG-4 Audio data unit.
-type MPEG4AudioGeneric struct {
-	Base
-	AUs [][]byte
-}

internal/unit/mpeg4_audio_latm.go

@@ -1,7 +0,0 @@
-package unit
-
-// MPEG4AudioLATM is a MPEG-4 Audio data unit.
-type MPEG4AudioLATM struct {
-	Base
-	AU []byte
-}