TeleportFling/internal/flinger/flinger.go

// Package flinger implements the TeleportFling streaming engine: it captures
// a screen (and optionally system audio), encodes video to JPEG, packetizes
// both into the Teleport protocol, and broadcasts them to connected OBS
// receivers.
//
// The engine is deliberately GUI-free so it can run headless (CLI/daemon) or
// be driven by a desktop app. Callers create an Engine with a Config, call
// Start, poll Status, and call Stop when done.
package flinger

import (
	"errors"
	"image"
	"image/color"
	"io"
	"log"
	"strconv"
	"sync/atomic"
	"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
)

// Config configures the streaming engine.
type Config struct {
	// Name is the announce name advertised to receivers (empty → hostname).
	Name string
	// Port is the TCP listening port for receivers.
	Port int
	// Quality is the JPEG quality, 1–100.
	Quality int
	// FPS is the target video frame rate.
	FPS int
	// Source is "screen" (PipeWire) or "pattern" (synthetic test signal).
	Source string
	// Audio enables system audio capture and streaming.
	Audio bool
	// StreamIndex selects which monitor to capture (screen source only).
	StreamIndex int
}

// DefaultConfig returns the recommended defaults.
func DefaultConfig() Config {
	return Config{
		Port:    9756,
		Quality: 80,
		FPS:     30,
		Source:  "screen",
		Audio:   true,
	}
}

// Status is a point-in-time snapshot of the running engine.
type Status struct {
	Running bool
	Frames  int64
	Conns   int
}

// Engine owns the capture, encode and send pipeline.
type Engine struct {
	cfg Config

	sender    *output.Sender
	announcer *discovery.Announcer
	encoder   *protocol.JPEGEncoder
	cam       capture.Capture
	loop      frameSource

	start time.Time
	stop  chan struct{}

	frames atomic.Int64
}

// New creates an engine from cfg. Capture is opened eagerly so that
// misconfiguration (e.g. no screen-share permission) surfaces before Start.
func New(cfg Config) (*Engine, error) {
	e := &Engine{cfg: cfg}

	sender := output.New()
	if _, err := sender.Listen(addr(cfg.Port)); err != nil {
		return nil, err
	}

	enc, err := protocol.NewJPEGEncoder()
	if err != nil {
		sender.Close()
		return nil, err
	}

	e.sender = sender
	e.encoder = enc

	switch cfg.Source {
	case "screen":
		cam, err := capture.OpenPipeWire(cfg.StreamIndex, cfg.Audio)
		if err != nil {
			e.encoder.Close()
			sender.Close()
			return nil, err
		}
		e.cam = cam
		e.loop = captureLoop{cam}
	case "pattern":
		e.loop = &patternLoop{w: 1920, h: 1080}
	default:
		e.encoder.Close()
		sender.Close()
		return nil, errors.New("unknown source " + cfg.Source)
	}

	return e, nil
}

// Start begins the audio, video and stats loops and starts announcing the
// stream. It is idempotent.
func (e *Engine) Start() {
	if e.stop != nil {
		return
	}

	e.start = time.Now()
	e.stop = make(chan struct{})

	e.announcer = discovery.Start(e.cfg.Name, e.sender.Port())

	var src io.ReadCloser
	switch {
	case e.cam == nil:
		src = capture.NewSilenceSource()
	case e.cam.Audio() != nil:
		src = e.cam.Audio()
	default:
		log.Printf("flinger: system audio unavailable, streaming silence")
		src = capture.NewSilenceSource()
	}

	go e.audioLoop(src)
	go e.videoLoop()
	go e.statsLoop()
}

// Stop halts the loops, stops announcing and closes all resources. It is
// idempotent. After Stop the engine must not be restarted.
func (e *Engine) Stop() {
	if e.stop == nil {
		return
	}

	close(e.stop)
	// Give the loops a moment to observe the stop signal.
	time.Sleep(50 * time.Millisecond)

	if e.announcer != nil {
		e.announcer.Stop()
	}
	e.sender.Close()
	e.encoder.Close()
	if e.cam != nil {
		if err := e.cam.Close(); err != nil {
			log.Printf("flinger: capture close: %v", err)
		}
	}
}

// Status returns a snapshot of the running engine.
func (e *Engine) Status() Status {
	return Status{
		Running: e.stop != nil,
		Frames:  e.frames.Load(),
		Conns:   e.sender.NumConns(),
	}
}

// audioLoop reads raw PCM and emits WAVE packets. start is the shared
// reference clock used by the video loop so audio and video timestamps stay
// aligned on the receiver.
func (e *Engine) audioLoop(src io.ReadCloser) {
	defer func() { _ = src.Close() }()

	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(e.start))
			packet, perr := protocol.BuildWavePacket(ts, protocol.AudioFormatS16, sampleRate, speakers, int32(frames), buf[:n])
			if perr != nil {
				log.Printf("flinger: wave: %v", perr)
			} else {
				e.sender.Send(packet)
			}
		}
		if err != nil {
			select {
			case <-e.stop:
				return
			default:
			}
			if !errors.Is(err, io.EOF) && !errors.Is(err, io.ErrClosedPipe) {
				log.Printf("flinger: audio: %v", err)
			}
		}
	}
}

// videoLoop pulls frames and sends them at the configured fps.
func (e *Engine) videoLoop() {
	frameInterval := time.Second / time.Duration(e.cfg.FPS)
	next := e.start

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

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

		ts := uint64(now.Sub(e.start))
		buf, err := e.encoder.EncodeBGRA(frame.Pix, frame.Width, frame.Height, e.cfg.Quality)
		if err != nil {
			log.Printf("flinger: 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("flinger: packet: %v", err)
			continue
		}
		e.sender.Send(packet)
		e.frames.Add(1)
	}
}

// statsLoop logs a periodic summary.
func (e *Engine) statsLoop() {
	tick := time.NewTicker(5 * time.Second)
	defer tick.Stop()
	for {
		select {
		case <-tick.C:
			st := e.Status()
			log.Printf("flinger: %d frames, %d conns", st.Frames, st.Conns)
		case <-e.stop:
			return
		}
	}
}

// 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
	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
}

// 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.
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:
				c = color.RGBA{R: 255, G: 255, B: 255, A: 255}
			default:
				c = color.RGBA{}
			}
			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
}