Replace internal decoder with ffplay subprocess

This commit is contained in:
Stu 2025-12-10 05:27:24 -05:00
parent 3d43123840
commit 4929918d4b

422
main.go
View File

@ -3704,24 +3704,16 @@ func buildVideoPane(state *appState, min fyne.Size, src *videoSource, onCover fu
}
type playSession struct {
path string
fps float64
width int
height int
targetW int
targetH int
volume float64
muted bool
paused bool
current float64
stop chan struct{}
done chan struct{}
prog func(float64)
img *canvas.Image
mu sync.Mutex
videoCmd *exec.Cmd
audioCmd *exec.Cmd
frameN int
path string
fps float64
volume float64
muted bool
paused bool
current float64
mu sync.Mutex
cmd *exec.Cmd
stdin io.WriteCloser
done chan struct{}
}
var audioCtxGlobal struct {
@ -3743,42 +3735,26 @@ func newPlaySession(path string, w, h int, fps float64, targetW, targetH int, pr
fmt.Printf("═══════════════════════════════════════════════════════\n")
fmt.Printf("📁 Video: %s\n", filepath.Base(path))
fmt.Printf("📐 Source: %dx%d @ %.2f fps\n", w, h, fps)
fmt.Printf("🎯 Target: %dx%d\n", targetW, targetH)
fmt.Printf("🎯 Playback via ffplay subprocess\n")
fmt.Printf("═══════════════════════════════════════════════════════\n\n")
// Validate input parameters
if fps <= 0 {
fps = 24
fmt.Printf("⚠️ Invalid FPS (%.2f), defaulting to 24\n", fps)
}
if targetW <= 0 {
targetW = 640
fmt.Printf("⚠️ Invalid target width (%d), defaulting to 640\n", targetW)
}
if targetH <= 0 {
targetH = int(float64(targetW) * (float64(h) / float64(utils.MaxInt(w, 1))))
fmt.Printf("⚠️ Invalid target height (%d), calculating to %d\n", targetH, targetH)
}
// Check if video file exists
if _, err := os.Stat(path); os.IsNotExist(err) {
fmt.Printf("❌ ERROR: Video file does not exist: %s\n", path)
return nil
}
if _, err := exec.LookPath("ffplay"); err != nil {
fmt.Printf("❌ ERROR: ffplay not found in PATH: %v\n", err)
return nil
}
fmt.Printf("✅ Play session created successfully\n")
return &playSession{
path: path,
fps: fps,
width: w,
height: h,
targetW: targetW,
targetH: targetH,
volume: 100,
stop: make(chan struct{}),
done: make(chan struct{}),
prog: prog,
img: img,
path: path,
fps: fps,
volume: 100,
done: make(chan struct{}),
}
}
@ -3786,19 +3762,27 @@ func (p *playSession) Play() {
fmt.Printf("▶️ PLAY called (current position: %.2fs)\n", p.current)
p.mu.Lock()
defer p.mu.Unlock()
if p.videoCmd == nil && p.audioCmd == nil {
fmt.Printf("⚡ Starting playback from scratch...\n")
p.startLocked(p.current)
if p.cmd == nil {
fmt.Printf("⚡ Starting ffplay subprocess...\n")
p.startFFplayLocked(p.current)
return
}
fmt.Printf("▶️ Resuming playback\n")
p.paused = false
// Toggle pause/resume by sending "p" to stdin
if p.stdin != nil {
_, _ = p.stdin.Write([]byte("p"))
p.paused = !p.paused
fmt.Printf("⏯️ Toggled pause -> %v\n", p.paused)
}
}
func (p *playSession) Pause() {
p.mu.Lock()
defer p.mu.Unlock()
p.paused = true
if p.stdin != nil {
_, _ = p.stdin.Write([]byte("p"))
p.paused = !p.paused
fmt.Printf("⏸️ Pause toggle -> %v\n", p.paused)
}
}
func (p *playSession) Seek(offset float64) {
@ -3807,22 +3791,9 @@ func (p *playSession) Seek(offset float64) {
if offset < 0 {
offset = 0
}
paused := p.paused
p.current = offset
p.stopLocked()
p.startLocked(p.current)
p.paused = paused
if p.paused {
// Ensure loops honor paused right after restart.
time.AfterFunc(30*time.Millisecond, func() {
p.mu.Lock()
defer p.mu.Unlock()
p.paused = true
})
}
if p.prog != nil {
p.prog(p.current)
}
p.startFFplayLocked(p.current)
}
func (p *playSession) SetVolume(v float64) {
@ -3840,41 +3811,15 @@ func (p *playSession) SetVolume(v float64) {
} else {
p.muted = true
}
// Runtime volume changes are not pushed to ffplay; a restart or pause toggle will pick it up.
}
// StepFrame steps forward or backward by one frame
// direction: 1 for forward, -1 for backward
func (p *playSession) StepFrame(direction int) {
p.mu.Lock()
defer p.mu.Unlock()
// Ensure we're paused for frame stepping
if !p.paused {
p.paused = true
}
// Calculate new position (1 frame = 1/fps seconds)
frameDuration := 1.0 / p.fps
newPos := p.current + (float64(direction) * frameDuration)
if newPos < 0 {
newPos = 0
}
p.current = newPos
p.stopLocked()
p.startLocked(p.current)
p.paused = true
// Ensure paused state sticks
time.AfterFunc(30*time.Millisecond, func() {
p.mu.Lock()
defer p.mu.Unlock()
p.paused = true
})
if p.prog != nil {
p.prog(p.current)
}
// Approximate by seeking one frame forward/backward
step := 1.0 / math.Max(p.fps, 24)
p.Seek(p.current + float64(direction)*step)
}
// GetCurrentPosition returns the current playback position
@ -3891,288 +3836,49 @@ func (p *playSession) Stop() {
}
func (p *playSession) stopLocked() {
select {
case <-p.stop:
default:
close(p.stop)
if p.cmd != nil && p.cmd.Process != nil {
_ = p.cmd.Process.Kill()
_, _ = p.cmd.Process.Wait()
}
if p.videoCmd != nil && p.videoCmd.Process != nil {
_ = p.videoCmd.Process.Kill()
_ = p.videoCmd.Wait()
}
if p.audioCmd != nil && p.audioCmd.Process != nil {
_ = p.audioCmd.Process.Kill()
_ = p.audioCmd.Wait()
}
p.videoCmd = nil
p.audioCmd = nil
p.stop = make(chan struct{})
p.cmd = nil
p.stdin = nil
p.done = make(chan struct{})
}
func (p *playSession) startLocked(offset float64) {
p.paused = false
p.current = offset
p.frameN = 0
logging.Debug(logging.CatFFMPEG, "playSession start path=%s offset=%.3f fps=%.3f target=%dx%d", p.path, offset, p.fps, p.targetW, p.targetH)
p.runVideo(offset)
p.runAudio(offset)
}
func (p *playSession) runVideo(offset float64) {
fmt.Printf("📹 Starting video decode pipeline...\n")
fmt.Printf(" - Resolution: %dx%d\n", p.targetW, p.targetH)
fmt.Printf(" - Frame rate: %.2f fps\n", p.fps)
fmt.Printf(" - Offset: %.2fs\n", offset)
func (p *playSession) startFFplayLocked(offset float64) {
var stderr bytes.Buffer
args := []string{
"-hide_banner", "-loglevel", "error",
"-hide_banner",
"-loglevel", "warning",
"-autoexit",
"-ss", fmt.Sprintf("%.3f", offset),
"-i", p.path,
"-vf", fmt.Sprintf("scale=%d:%d:flags=bilinear", p.targetW, p.targetH),
"-f", "rawvideo",
"-pix_fmt", "rgb24",
"-r", fmt.Sprintf("%.3f", p.fps),
"-vsync", "0", // Avoid frame duplication
"-",
"-window_title", "VT Player",
}
fmt.Printf("🔧 FFmpeg command: ffmpeg %s\n", strings.Join(args, " "))
cmd := exec.Command("ffmpeg", args...)
if p.volume > 0 {
args = append([]string{"-volume", fmt.Sprintf("%d", int(p.volume))}, args...)
}
cmd := exec.Command("ffplay", args...)
cmd.Stderr = &stderr
stdout, err := cmd.StdoutPipe()
stdin, err := cmd.StdinPipe()
if err != nil {
fmt.Printf("❌ ERROR: Failed to create video pipe: %v\n", err)
logging.Debug(logging.CatFFMPEG, "video pipe error: %v", err)
fmt.Printf("❌ ERROR: failed to create ffplay stdin: %v\n", err)
return
}
fmt.Printf("⚡ Starting FFmpeg process...\n")
startTime := time.Now()
cmd.Stdout = os.Stdout
if err := cmd.Start(); err != nil {
errMsg := strings.TrimSpace(stderr.String())
fmt.Printf("❌ ERROR: FFmpeg failed to start: %v\n", err)
if errMsg != "" {
fmt.Printf(" FFmpeg error: %s\n", errMsg)
fmt.Printf("❌ ERROR: ffplay failed to start: %v\n", err)
if msg := strings.TrimSpace(stderr.String()); msg != "" {
fmt.Printf(" ffplay stderr: %s\n", msg)
}
// Check if ffmpeg is available
if _, pathErr := exec.LookPath("ffmpeg"); pathErr != nil {
fmt.Printf("❌ FATAL: ffmpeg not found in PATH: %v\n", pathErr)
}
logging.Debug(logging.CatFFMPEG, "video start failed: %v (%s)", err, errMsg)
return
}
fmt.Printf("✅ FFmpeg started (PID: %d) in %.3fs\n", cmd.Process.Pid, time.Since(startTime).Seconds())
// Pace frames to the source frame rate instead of hammering refreshes as fast as possible.
frameDur := time.Second
if p.fps > 0 {
frameDur = time.Duration(float64(time.Second) / math.Max(p.fps, 0.1))
}
nextFrameAt := time.Now()
p.videoCmd = cmd
frameSize := p.targetW * p.targetH * 3
buf := make([]byte, frameSize)
fmt.Printf("📦 Frame buffer allocated: %d bytes (%.2f MB)\n", frameSize, float64(frameSize)/(1024*1024))
p.cmd = cmd
p.stdin = stdin
p.paused = false
go func() {
defer cmd.Process.Kill()
// Performance monitoring variables
var lastFPSReport time.Time
var fpsCounter int
loopStart := time.Now()
fmt.Printf("🔄 Frame decode loop started\n")
for {
select {
case <-p.stop:
fmt.Printf("⏹️ Video decode loop stopped\n")
logging.Debug(logging.CatFFMPEG, "video loop stop")
return
default:
}
if p.paused {
time.Sleep(30 * time.Millisecond)
nextFrameAt = time.Now().Add(frameDur)
continue
}
readStart := time.Now()
_, err := io.ReadFull(stdout, buf)
readDuration := time.Since(readStart)
if err != nil {
if errors.Is(err, io.EOF) {
fmt.Printf("📺 Video playback completed (reached end)\n")
return
}
msg := strings.TrimSpace(stderr.String())
fmt.Printf("❌ ERROR: Frame read failed: %v\n", err)
if msg != "" {
fmt.Printf(" FFmpeg error: %s\n", msg)
}
logging.Debug(logging.CatFFMPEG, "video read failed: %v (%s)", err, msg)
return
}
// Track first frame timing
if p.frameN == 0 {
elapsed := time.Since(loopStart)
fmt.Printf("🎞️ FIRST FRAME decoded in %.3fs (read: %.3fs)\n", elapsed.Seconds(), readDuration.Seconds())
}
// Improved frame pacing - use a more stable timing approach
now := time.Now()
if now.Before(nextFrameAt) {
time.Sleep(nextFrameAt.Sub(now))
}
nextFrameAt = nextFrameAt.Add(frameDur)
// Allocate a fresh frame to avoid concurrent texture reuse issues.
frame := image.NewRGBA(image.Rect(0, 0, p.targetW, p.targetH))
utils.CopyRGBToRGBA(frame.Pix, buf)
renderStart := time.Now()
fyne.CurrentApp().Driver().DoFromGoroutine(func() {
if p.img != nil {
// Ensure we render the live frame, not a stale resource preview.
p.img.Resource = nil
p.img.File = ""
p.img.Image = frame
p.img.Refresh()
}
}, false)
renderDuration := time.Since(renderStart)
// Log first few frames in detail
if p.frameN < 3 {
fmt.Printf("🖼️ Frame %d: decode=%.3fs render=%.3fs\n", p.frameN+1, readDuration.Seconds(), renderDuration.Seconds())
logging.Debug(logging.CatFFMPEG, "video frame %d drawn (%.2fs)", p.frameN+1, p.current)
}
p.frameN++
fpsCounter++
// FPS report every 2 seconds
if time.Since(lastFPSReport) >= 2*time.Second {
fps := float64(fpsCounter) / time.Since(lastFPSReport).Seconds()
fmt.Printf("📊 Performance: %.1f FPS (target: %.1f FPS) | Frames: %d\n", fps, p.fps, p.frameN)
fpsCounter = 0
lastFPSReport = time.Now()
}
if p.fps > 0 {
p.current = offset + (float64(p.frameN) / p.fps)
}
if p.prog != nil {
p.prog(p.current)
}
}
}()
}
func (p *playSession) runAudio(offset float64) {
const sampleRate = 48000
const channels = 2
const bytesPerSample = 2
var stderr bytes.Buffer
cmd := exec.Command("ffmpeg",
"-hide_banner", "-loglevel", "error",
"-ss", fmt.Sprintf("%.3f", offset),
"-i", p.path,
"-vn",
"-ac", fmt.Sprintf("%d", channels),
"-ar", fmt.Sprintf("%d", sampleRate),
"-f", "s16le",
"-",
)
cmd.Stderr = &stderr
stdout, err := cmd.StdoutPipe()
if err != nil {
logging.Debug(logging.CatFFMPEG, "audio pipe error: %v", err)
return
}
if err := cmd.Start(); err != nil {
errMsg := strings.TrimSpace(stderr.String())
fmt.Printf("❌ ERROR: Audio FFmpeg failed to start: %v\n", err)
if errMsg != "" {
fmt.Printf(" Audio FFmpeg error: %s\n", errMsg)
}
logging.Debug(logging.CatFFMPEG, "audio start failed: %v (%s)", err, errMsg)
return
}
fmt.Printf("✅ Audio FFmpeg started (PID: %d)\n", cmd.Process.Pid)
p.audioCmd = cmd
ctx, err := getAudioContext(sampleRate, channels, bytesPerSample)
if err != nil {
logging.Debug(logging.CatFFMPEG, "audio context error: %v", err)
return
}
player := ctx.NewPlayer()
if player == nil {
logging.Debug(logging.CatFFMPEG, "audio player creation failed")
return
}
localPlayer := player
go func() {
defer cmd.Process.Kill()
defer localPlayer.Close()
chunk := make([]byte, 4096)
tmp := make([]byte, 4096)
loggedFirst := false
for {
select {
case <-p.stop:
logging.Debug(logging.CatFFMPEG, "audio loop stop")
return
default:
}
if p.paused {
time.Sleep(30 * time.Millisecond)
continue
}
n, err := stdout.Read(chunk)
if n > 0 {
if !loggedFirst {
logging.Debug(logging.CatFFMPEG, "audio stream delivering bytes")
loggedFirst = true
}
gain := p.volume / 100.0
if gain < 0 {
gain = 0
}
if gain > 2 {
gain = 2
}
copy(tmp, chunk[:n])
if p.muted || gain <= 0 {
for i := 0; i < n; i++ {
tmp[i] = 0
}
} else if math.Abs(1-gain) > 0.001 {
for i := 0; i+1 < n; i += 2 {
sample := int16(binary.LittleEndian.Uint16(tmp[i:]))
amp := int(float64(sample) * gain)
if amp > math.MaxInt16 {
amp = math.MaxInt16
}
if amp < math.MinInt16 {
amp = math.MinInt16
}
binary.LittleEndian.PutUint16(tmp[i:], uint16(int16(amp)))
}
}
localPlayer.Write(tmp[:n])
}
if err != nil {
if !errors.Is(err, io.EOF) {
logging.Debug(logging.CatFFMPEG, "audio read failed: %v (%s)", err, strings.TrimSpace(stderr.String()))
}
return
}
}
_ = cmd.Wait()
close(p.done)
}()
}