TeleportFling/cmd/teleportfling/main.go

// Command teleportfling is a standalone sender for the Teleport protocol.
//
// It captures a Wayland screen (PipeWire via xdg-desktop-portal) plus the
// system's default audio output and streams them over TCP as the Teleport
// protocol, announcing itself on the LAN multicast group so an OBS instance
// with the obs-teleport plugin can discover and decode the stream.
//
// Usage:
//
//	teleportfling [--name NAME] [--port PORT] [--quality 1..100]
//	             [--fps N] [--source screen|pattern] [--audio]
//	             [--stream-index N] [--duration SECONDS]
//
// --source pattern selects the M1 synthetic test pattern (colour bars with a
// moving box) instead of real screen capture, which is useful for testing
// without granting screen-share permission.
package main

import (
	"errors"
	"flag"
	"image"
	"image/color"
	"io"
	"log"
	"os"
	"os/signal"
	"strconv"
	"sync/atomic"
	"syscall"
	"time"

	"teleportfling/internal/capture"
	"teleportfling/internal/discovery"
	"teleportfling/internal/output"
	"teleportfling/internal/protocol"
)

const (
	// sampleRate and speakers describe the captured/encoded audio stream.
	sampleRate = 48000
	speakers   = 2
	// audioChunk sets how much audio we packetize per WAVE message (~10 ms).
	audioChunk = 10 * time.Millisecond
)

func main() {
	var (
		name        = flag.String("name", "", "announce name (default: hostname)")
		port        = flag.Int("port", 9756, "TCP listening port")
		quality     = flag.Int("quality", 80, "JPEG quality 1..100")
		fps         = flag.Int("fps", 30, "video frames per second")
		source      = flag.String("source", "screen", "capture source: screen or pattern")
		withAudio   = flag.Bool("audio", true, "capture and stream system audio")
		streamIndex = flag.Int("stream-index", 0, "monitor index to capture (screen source)")
		duration    = flag.Duration("duration", 0, "stream duration (0 = run until interrupted)")
	)
	flag.Parse()

	// Build the sender: TCP listener + multicast announcer.
	sender := output.New()
	p, err := sender.Listen(addr(*port))
	if err != nil {
		log.Fatalf("output: listen: %v", err)
	}
	announcer := discovery.Start(*name, p)

	var (
		totalFrames atomic.Int64
		encoder     = mustNewEncoder()
		start       = time.Now()
		stop        = make(chan struct{})
	)

	var (
		cam  capture.Capture
		loop frameSource
	)

	switch *source {
	case "screen":
		cam, err = capture.OpenPipeWire(*streamIndex, *withAudio)
		if err != nil {
			log.Fatalf("capture: %v", err)
		}
		loop = captureLoop{cam}
		log.Printf("teleportfling: advertising on %d, capturing screen via PipeWire", p)
	case "pattern":
		loop = &patternLoop{
			w:   1920,
			h:   1080,
			fps: *fps,
		}
		log.Printf("teleportfling: advertising on %d, streaming test pattern", p)
	default:
		log.Fatalf("teleportfling: unknown source %q (want screen or pattern)", *source)
	}

	// Interrupt / SIGTERM handling.
	sigc := make(chan os.Signal, 1)
	signal.Notify(sigc, syscall.SIGINT, syscall.SIGTERM)

	// — Audio loop —
	audioDone := make(chan struct{})
	go func() {
		defer close(audioDone)

		var src io.ReadCloser
		switch {
		case cam == nil:
			// Pattern source: synthesize silence to keep the audio pipeline alive.
			src = capture.NewSilenceSource()
		case cam.Audio() != nil:
			src = cam.Audio()
		default:
			log.Printf("teleportfling: system audio unavailable, streaming silence")
			src = capture.NewSilenceSource()
		}
		defer func() { _ = src.Close() }()

		audioLoop(sender, src, start, stop)
	}()

	// — Video loop —
	videoDone := make(chan struct{})
	go func() {
		defer close(videoDone)
		videoLoop(sender, encoder, loop, *fps, *quality, start, stop, &totalFrames)
	}()

	// — Stats ticker —
	statsDone := make(chan struct{})
	go func() {
		defer close(statsDone)
		tick := time.NewTicker(5 * time.Second)
		defer tick.Stop()
		for {
			select {
			case <-tick.C:
				log.Printf("stats: %d frames, %d conns",
					totalFrames.Load(), sender.NumConns())
			case <-stop:
				return
			}
		}
	}()

	// — Wait for interrupt/duration —
	select {
	case <-sigc:
		log.Printf("teleportfling: stopping…")
	case <-func() <-chan struct{} {
		if *duration > 0 {
			ch := make(chan struct{})
			time.AfterFunc(*duration, func() { close(ch) })
			return ch
		}
		return nil
	}():
		log.Printf("teleportfling: duration reached")
	}

	close(stop)
	<-audioDone
	<-videoDone
	<-statsDone

	announcer.Stop()
	sender.Close()
	encoder.Close()
	if cam != nil {
		if err := cam.Close(); err != nil {
			log.Printf("teleportfling: capture close: %v", err)
		}
	}
	log.Printf("teleportfling: stopped after %s", time.Since(start).Round(time.Millisecond))
}

// audioLoop reads raw PCM from src and emits WAVE packets in audioChunk-sized
// pieces. PCM is assumed interleaved signed-16-bit at sampleRate/speakers.
//
// start is the shared reference clock used by the video loop: audio and video
// timestamps must share one time base, otherwise a constant skew between them
// makes OBS buffer one stream to re-sync the other, adding latency.
func audioLoop(sender *output.Sender, src io.Reader, start time.Time, stop <-chan struct{}) {
	framesPerChunk := int(sampleRate) * int(audioChunk) / int(time.Second)
	chunkBytes := framesPerChunk * speakers * 2 // S16
	buf := make([]byte, chunkBytes)

	for {
		n, err := io.ReadFull(src, buf)
		if n > 0 {
			frames := n / (speakers * 2)
			ts := uint64(time.Since(start))
			packet, perr := protocol.BuildWavePacket(ts, protocol.AudioFormatS16, sampleRate, speakers, int32(frames), buf[:n])
			if perr != nil {
				log.Printf("teleportfling: wave: %v", perr)
			} else {
				sender.Send(packet)
			}
		}
		if err != nil {
			if !errors.Is(err, io.EOF) && !errors.Is(err, io.ErrClosedPipe) {
				log.Printf("teleportfling: audio: %v", err)
			}
			select {
			case <-stop:
				return
			default:
			}
		}
	}
}

// videoLoop pulls frames from loop and sends them at fps, encoding each to
// JPEG with the given quality.
func videoLoop(sender *output.Sender, encoder *protocol.JPEGEncoder, loop frameSource, fps, quality int, start time.Time, stop <-chan struct{}, total *atomic.Int64) {
	frameInterval := time.Second / time.Duration(fps)
	next := start

	for {
		frame, err := loop.Next()
		if err != nil {
			if !errors.Is(err, io.EOF) && !errors.Is(err, io.ErrClosedPipe) {
				log.Printf("teleportfling: capture: %v", err)
			}
			select {
			case <-stop:
				return
			default:
			}
			continue
		}

		// Drop frames if we're running ahead of the target fps (e.g. a 60 Hz
		// monitor captured at 30 fps) to keep timestamps monotonic.
		now := time.Now()
		if now.Before(next) {
			continue
		}
		next = now.Add(frameInterval)

		ts := uint64(now.Sub(start))
		buf, err := encodeFrame(encoder, frame, quality)
		if err != nil {
			log.Printf("teleportfling: jpeg: %v", err)
			continue
		}

		packet, err := protocol.WritePacket(
			protocol.Header{Type: protocol.VideoType, Timestamp: ts, Size: int32(len(buf))},
			ptr(protocol.DefaultBT709Full()),
			nil,
			buf,
		)
		if err != nil {
			log.Printf("teleportfling: packet: %v", err)
			continue
		}
		sender.Send(packet)
		total.Add(1)
	}
}

// encodeFrame compresses a captured frame based on its concrete type.
func encodeFrame(encoder *protocol.JPEGEncoder, frame *capture.VideoFrame, quality int) ([]byte, error) {
	if frame.Pix != nil {
		// BGRA from the PipeWire backend.
		return encoder.EncodeBGRA(frame.Pix, frame.Width, frame.Height, quality)
	}
	return nil, errors.New("capture: unsupported frame type")
}

// frameSource abstracts the frame source: real capture or the test pattern.
type frameSource interface {
	Next() (*capture.VideoFrame, error)
}

// captureLoop wraps the PipeWire capture backend.
type captureLoop struct {
	cam capture.Capture
}

func (c captureLoop) Next() (*capture.VideoFrame, error) {
	return c.cam.Video().NextFrame()
}

// patternLoop synthesizes the M1 test pattern (colour bars + moving box).
type patternLoop struct {
	w, h int
	fps  int
	seq  int64
}

func (p *patternLoop) Next() (*capture.VideoFrame, error) {
	img := testPattern(p.w, p.h, int(p.seq))
	p.seq++
	return ycrcbToBGRA(img), nil
}

// ycrcbToBGRA converts a YCbCr image to a BGRA VideoFrame so both sources
// share the encode path (EncodeBGRA).
func ycrcbToBGRA(img *image.YCbCr) *capture.VideoFrame {
	w, h := img.Rect.Dx(), img.Rect.Dy()
	frame := &capture.VideoFrame{
		Pix:    make([]byte, w*h*4),
		Width:  w,
		Height: h,
		Stride: w * 4,
	}
	for y := 0; y < h; y++ {
		for x := 0; x < w; x++ {
			yi := y*img.YStride + x
			ci := (y/2)*img.CStride + x/2
			r, g, b := color.YCbCrToRGB(img.Y[yi], img.Cb[ci], img.Cr[ci])
			off := (y*w + x) * 4
			frame.Pix[off], frame.Pix[off+1], frame.Pix[off+2], frame.Pix[off+3] = b, g, r, 255
		}
	}
	return frame
}

// mustNewEncoder creates a JPEG encoder or panics.
func mustNewEncoder() *protocol.JPEGEncoder {
	enc, err := protocol.NewJPEGEncoder()
	if err != nil {
		log.Fatal(err)
	}
	return enc
}

// ptr returns a pointer to v, for passing headers to WritePacket.
func ptr[T any](v T) *T { return &v }

// addr formats a port as a listen address.
func addr(port int) string {
	return ":" + strconv.Itoa(port)
}

// testPattern renders a standard SMPTE colour bar with a moving white box at
// the given frame index. The result is a *image.YCbCr 4:2:0 image so the
// encoder uses the YUV path — closest to what real PipeWire capture will
// produce.
func testPattern(w, h, frame int) *image.YCbCr {
	img := image.NewYCbCr(image.Rect(0, 0, w, h), image.YCbCrSubsampleRatio420)

	// 7 vertical colour bars (grey, yellow, cyan, green, magenta, red, blue).
	bars := []color.RGBA{
		{R: 191, G: 191, B: 191}, // 75% grey
		{R: 191, G: 191, B: 0},   // yellow
		{R: 0, G: 191, B: 191},   // cyan
		{R: 0, G: 191, B: 0},     // green
		{R: 191, G: 0, B: 191},   // magenta
		{R: 191, G: 0, B: 0},     // red
		{R: 0, G: 0, B: 191},     // blue
	}

	const barCount = 7
	barW := w / barCount
	const boxSize = 80

	// Moving white box sweeps left→right across the lower black block.
	boxMinX := (frame*(w+boxSize)/120)%(w+boxSize) - boxSize/2

	buf := make([]color.RGBA, w*h)
	for by := 0; by < h; by++ {
		rowIsBars := by < h*2/3
		for bx := 0; bx < w; bx++ {
			var c color.RGBA
			switch {
			case rowIsBars:
				idx := bx / barW
				if idx >= barCount {
					idx = barCount - 1
				}
				c = bars[idx]
			case by%8 < 4 && bx > w/3 && bx < w*2/3:
				// Periodic white band across the lower black block for motion.
				c = color.RGBA{R: 255, G: 255, B: 255, A: 255}
			default:
				c = color.RGBA{}
			}
			// Overlay the moving box on the bottom band.
			if bx >= boxMinX && bx < boxMinX+boxSize && by >= h*2/3 {
				c = color.RGBA{R: 255, G: 255, B: 255, A: 255}
			}
			buf[by*w+bx] = c
		}
	}

	// Chroma planes: average each 2x2 RGB block, then convert to Cb/Cr.
	for by := 0; by < h; by += 2 {
		for bx := 0; bx < w; bx += 2 {
			var rSum, gSum, bSum uint32
			n := uint32(0)
			for dy := 0; dy < 2; dy++ {
				for dx := 0; dx < 2; dx++ {
					xx, yy := bx+dx, by+dy
					if xx >= w || yy >= h {
						continue
					}
					px := buf[yy*w+xx]
					rSum += uint32(px.R)
					gSum += uint32(px.G)
					bSum += uint32(px.B)
					n++
				}
			}
			_, cb, cr := color.RGBToYCbCr(uint8(rSum/n), uint8(gSum/n), uint8(bSum/n))
			img.Cb[(by/2)*img.CStride+bx/2] = cb
			img.Cr[(by/2)*img.CStride+bx/2] = cr
		}
	}

	// Luma plane: Y = YCbCr luma of every pixel.
	for by := 0; by < h; by++ {
		for bx := 0; bx < w; bx++ {
			px := buf[by*w+bx]
			y, _, _ := color.RGBToYCbCr(px.R, px.G, px.B)
			img.Y[by*img.YStride+bx] = y
		}
	}

	return img
}