visionA/local-agent/server/internal/camera/ffmpeg_camera.go

package camera

import (
	"bufio"
	"bytes"
	"fmt"
	"io"
	"os/exec"
	"runtime"
	"sync"
	"time"
)

// stderrTailBytes 是保留的 ffmpeg stderr 尾端大小上限。
// ffmpeg 開攝影機失敗（avfoundation not authorized / device busy 等）的關鍵訊息
// 都在 stderr 末尾，保留尾端即可診斷；限制大小避免長時間執行累積無界記憶體。
const stderrTailBytes = 8 * 1024

// FFmpegCamera captures webcam frames using ffmpeg subprocess.
// Supports macOS (AVFoundation) and Windows (DirectShow).
// ffmpeg outputs a continuous MJPEG stream to stdout which is parsed
// by scanning for JPEG SOI (0xFFD8) and EOI (0xFFD9) markers.
type FFmpegCamera struct {
	cmd         *exec.Cmd
	stdout      io.ReadCloser
	stderrBuf   *ringBuffer // 保留 ffmpeg stderr 尾端，供失敗診斷（原本 =nil 直接丟棄）
	latestFrame []byte
	mu          sync.Mutex
	done        chan struct{}
	err         error

	// firstFrame 在第一張完整 JPEG frame 抵達時 close 一次，讓 WaitForFirstFrame 得知
	// 攝影機真的開起來了（cmd.Start() 成功只代表 ffmpeg 進程 fork 成功，不代表拿到 camera）。
	firstFrame     chan struct{}
	firstFrameOnce sync.Once
}

// ringBuffer 保留寫入資料的最後 max bytes（thread-safe），用來留住 ffmpeg stderr 尾端。
type ringBuffer struct {
	mu  sync.Mutex
	buf []byte
	max int
}

func newRingBuffer(max int) *ringBuffer {
	return &ringBuffer{max: max}
}

func (r *ringBuffer) Write(p []byte) (int, error) {
	r.mu.Lock()
	defer r.mu.Unlock()
	r.buf = append(r.buf, p...)
	if len(r.buf) > r.max {
		r.buf = r.buf[len(r.buf)-r.max:]
	}
	return len(p), nil
}

func (r *ringBuffer) String() string {
	r.mu.Lock()
	defer r.mu.Unlock()
	return string(bytes.TrimSpace(append([]byte(nil), r.buf...)))
}

// NewFFmpegCamera starts an ffmpeg process to capture from the given camera.
// On macOS, cameraIndex is used (e.g. 0 for first camera).
// On Windows, cameraName from device detection is used; cameraIndex is ignored
// unless no name is provided.
func NewFFmpegCamera(cameraIndex, width, height, framerate int) (*FFmpegCamera, error) {
	return NewFFmpegCameraWithName(cameraIndex, "", width, height, framerate)
}

// NewFFmpegCameraWithName starts ffmpeg with explicit camera name (needed for Windows dshow).
func NewFFmpegCameraWithName(cameraIndex int, cameraName string, width, height, framerate int) (*FFmpegCamera, error) {
	args := buildCaptureArgs(cameraIndex, cameraName, width, height, framerate)
	cmd := exec.Command("ffmpeg", args...)
	return newFFmpegCameraFromCmd(cmd)
}

// newFFmpegCameraFromCmd wires up stdout/stderr and starts the given ffmpeg-like
// command. Extracted so tests can substitute a fake command (e.g. a shell script
// that simulates early-exit or no-frame) to exercise WaitForFirstFrame.
func newFFmpegCameraFromCmd(cmd *exec.Cmd) (*FFmpegCamera, error) {
	stdout, err := cmd.StdoutPipe()
	if err != nil {
		return nil, fmt.Errorf("failed to get stdout pipe: %w", err)
	}

	// 保留 ffmpeg stderr 尾端而不是丟棄（原本 cmd.Stderr = nil 讓 avfoundation
	// 權限被拒 / 裝置忙碌等錯誤全數消失、極難 debug）。ringBuffer 只留末端、有界。
	stderrBuf := newRingBuffer(stderrTailBytes)
	cmd.Stderr = stderrBuf

	if err := cmd.Start(); err != nil {
		return nil, fmt.Errorf("failed to start ffmpeg: %w", err)
	}

	cam := &FFmpegCamera{
		cmd:        cmd,
		stdout:     stdout,
		stderrBuf:  stderrBuf,
		done:       make(chan struct{}),
		firstFrame: make(chan struct{}),
	}

	go cam.readLoop()

	return cam, nil
}

// WaitForFirstFrame 等到攝影機真的產出第一張 frame 才回 nil；否則回明確錯誤。
//
// 存在原因：cmd.Start() 只代表 ffmpeg 進程 fork 成功，avfoundation 抓不到 camera
// （權限被 TCC 靜默拒絕 / 裝置忙碌）是在進程啟動「之後」才失敗、ffmpeg 隨即 exit。
// 呼叫端（manager.Open → handler）改在 Start 後呼叫此函式，把「真的拿到攝影機」
// 這件事納入成功判斷，避免 HTTP 200 假成功。
//
// 三種結束情況：
//  1. 收到第一張 frame → nil
//  2. ffmpeg 提早結束（readLoop 讀到 EOF，設 c.err）→ 回含 stderr 尾端的錯誤
//  3. 逾時 → 回逾時錯誤（附 stderr 尾端，可能含權限 / 裝置訊息）
func (c *FFmpegCamera) WaitForFirstFrame(timeout time.Duration) error {
	timer := time.NewTimer(timeout)
	defer timer.Stop()

	select {
	case <-c.firstFrame:
		return nil
	case <-c.done:
		// ffmpeg 已結束卻沒送出任何 frame → 開攝影機失敗。
		c.mu.Lock()
		streamErr := c.err
		c.mu.Unlock()
		// 有可能 done 與 firstFrame 幾乎同時（極少見）：done 後再確認一次是否其實已有 frame。
		select {
		case <-c.firstFrame:
			return nil
		default:
		}
		return fmt.Errorf("camera did not start: ffmpeg exited before producing a frame: %w%s",
			streamErr, c.stderrTailSuffix())
	case <-timer.C:
		return fmt.Errorf("camera did not start: timed out after %s waiting for first frame%s",
			timeout, c.stderrTailSuffix())
	}
}

// stderrTailSuffix 回傳可附加到錯誤訊息的 ffmpeg stderr 尾端（若有）。
func (c *FFmpegCamera) stderrTailSuffix() string {
	if c.stderrBuf == nil {
		return ""
	}
	tail := c.stderrBuf.String()
	if tail == "" {
		return ""
	}
	return " (ffmpeg: " + tail + ")"
}

// StderrTail 回傳目前保留的 ffmpeg stderr 尾端（供上層 log 診斷）。
func (c *FFmpegCamera) StderrTail() string {
	if c.stderrBuf == nil {
		return ""
	}
	return c.stderrBuf.String()
}

// buildCaptureArgs returns the ffmpeg arguments for the current OS.
func buildCaptureArgs(cameraIndex int, cameraName string, width, height, framerate int) []string {
	videoSize := fmt.Sprintf("%dx%d", width, height)
	fps := fmt.Sprintf("%d", framerate)

	switch runtime.GOOS {
	case "windows":
		// DirectShow on Windows: -f dshow -i video="Camera Name"
		inputName := cameraName
		if inputName == "" {
			// Fallback: try to detect first camera
			devices := ListFFmpegDevices()
			if len(devices) > 0 {
				inputName = devices[0].Name
			} else {
				inputName = "Integrated Camera"
			}
		}
		return []string{
			"-f", "dshow",
			"-framerate", fps,
			"-video_size", videoSize,
			"-i", fmt.Sprintf("video=%s", inputName),
			"-f", "image2pipe",
			"-vcodec", "mjpeg",
			"-q:v", "5",
			"-an",
			"-",
		}
	default:
		// AVFoundation on macOS: -f avfoundation -i "index:none"
		return []string{
			"-f", "avfoundation",
			"-framerate", fps,
			"-video_size", videoSize,
			"-i", fmt.Sprintf("%d:none", cameraIndex),
			"-f", "image2pipe",
			"-vcodec", "mjpeg",
			"-q:v", "5",
			"-an",
			"-",
		}
	}
}

// readLoop continuously reads ffmpeg's stdout and extracts JPEG frames.
func (c *FFmpegCamera) readLoop() {
	defer close(c.done)

	reader := bufio.NewReaderSize(c.stdout, 1024*1024) // 1MB buffer
	buf := make([]byte, 0, 512*1024)                   // 512KB initial frame buffer
	inFrame := false

	for {
		b, err := reader.ReadByte()
		if err != nil {
			c.mu.Lock()
			c.err = fmt.Errorf("ffmpeg stream ended: %w", err)
			c.mu.Unlock()
			return
		}

		if !inFrame {
			// Look for SOI marker: 0xFF 0xD8
			if b == 0xFF {
				next, err := reader.ReadByte()
				if err != nil {
					c.mu.Lock()
					c.err = fmt.Errorf("ffmpeg stream ended: %w", err)
					c.mu.Unlock()
					return
				}
				if next == 0xD8 {
					// Start of JPEG
					buf = buf[:0]
					buf = append(buf, 0xFF, 0xD8)
					inFrame = true
				}
			}
			continue
		}

		// Inside a frame, collect bytes
		buf = append(buf, b)

		// Look for EOI marker: 0xFF 0xD9
		if b == 0xD9 && len(buf) >= 2 && buf[len(buf)-2] == 0xFF {
			// Complete JPEG frame
			frame := make([]byte, len(buf))
			copy(frame, buf)

			c.mu.Lock()
			c.latestFrame = frame
			c.mu.Unlock()

			// 通知 WaitForFirstFrame：攝影機真的產出 frame 了（只觸發一次）。
			c.firstFrameOnce.Do(func() { close(c.firstFrame) })

			inFrame = false
		}
	}
}

// ReadFrame returns the most recently captured JPEG frame.
func (c *FFmpegCamera) ReadFrame() ([]byte, error) {
	c.mu.Lock()
	defer c.mu.Unlock()

	if c.err != nil {
		return nil, c.err
	}
	if c.latestFrame == nil {
		return nil, fmt.Errorf("no frame available yet")
	}

	// Return a copy to avoid data races
	frame := make([]byte, len(c.latestFrame))
	copy(frame, c.latestFrame)
	return frame, nil
}

// Close stops the ffmpeg process and cleans up resources.
func (c *FFmpegCamera) Close() error {
	if c.cmd != nil && c.cmd.Process != nil {
		_ = c.cmd.Process.Kill()
		_ = c.cmd.Wait()
	}
	<-c.done
	return nil
}