add SDL audio sink

2026-08-28 21:58:43 +03:00
parent d57e403c2e
commit bcb0ae6a6d
2 changed files with 417 additions and 0 deletions
@@ -0,0 +1,204 @@
+package output
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"math"
+	"time"
+
+	"mxl-player/internal/playback"
+	"mxl-player/internal/sdl"
+)
+
+const (
+	defaultMaxQueuedBatches = int32(20)
+	defaultAudioQueuePoll   = 10 * time.Millisecond
+)
+
+var (
+	ErrInvalidAudioFormat = errors.New("invalid audio output format")
+	ErrOpenAudioStream    = errors.New("open SDL audio stream")
+	ErrResumeAudioStream  = errors.New("resume SDL audio stream")
+	ErrQueueAudioData     = errors.New("queue SDL audio data")
+)
+
+type audioBackend interface {
+	Open(uint32, sdl.AudioSpec) uintptr
+	Resume(uintptr) bool
+	Put(uintptr, []byte) bool
+	Queued(uintptr) int32
+	Destroy(uintptr)
+	Error() string
+}
+
+type systemAudioBackend struct{}
+
+func (systemAudioBackend) Open(device uint32, spec sdl.AudioSpec) uintptr {
+	return sdl.OpenAudioDeviceStream(device, spec)
+}
+func (systemAudioBackend) Resume(stream uintptr) bool { return sdl.ResumeAudioStreamDevice(stream) }
+func (systemAudioBackend) Put(stream uintptr, data []byte) bool {
+	return sdl.PutAudioStreamData(stream, data)
+}
+func (systemAudioBackend) Queued(stream uintptr) int32 { return sdl.GetAudioStreamQueued(stream) }
+func (systemAudioBackend) Destroy(stream uintptr)      { sdl.DestroyAudioStream(stream) }
+func (systemAudioBackend) Error() string               { return sdl.GetError() }
+
+type audioFormat struct {
+	channels  int32
+	frequency int32
+}
+
+type audioQueueWaitFunc func(context.Context, time.Duration) error
+
+type SDLAudioSink struct {
+	deviceID         uint32
+	backend          audioBackend
+	stream           uintptr
+	format           audioFormat
+	maxQueuedBatches int32
+	queuePoll        time.Duration
+	wait             audioQueueWaitFunc
+}
+
+var _ playback.AudioSink = (*SDLAudioSink)(nil)
+
+func NewSDLAudioSink(deviceID uint32) *SDLAudioSink {
+	return &SDLAudioSink{
+		deviceID:         deviceID,
+		backend:          systemAudioBackend{},
+		maxQueuedBatches: defaultMaxQueuedBatches,
+		queuePoll:        defaultAudioQueuePoll,
+		wait:             waitForAudioQueue,
+	}
+}
+
+func waitForAudioQueue(ctx context.Context, delay time.Duration) error {
+	timer := time.NewTimer(delay)
+	defer timer.Stop()
+
+	select {
+	case <-timer.C:
+		return nil
+	case <-ctx.Done():
+		return ctx.Err()
+	}
+}
+
+func audioOutputFormat(frame playback.AudioFrame) (audioFormat, error) {
+	if frame.Channels == 0 || frame.Channels > math.MaxInt32 ||
+		frame.SampleRateNumerator <= 0 || frame.SampleRateDenominator <= 0 ||
+		frame.SampleRateNumerator%frame.SampleRateDenominator != 0 {
+		return audioFormat{}, fmt.Errorf(
+			"%w: channels=%d rate=%d/%d",
+			ErrInvalidAudioFormat,
+			frame.Channels,
+			frame.SampleRateNumerator,
+			frame.SampleRateDenominator,
+		)
+	}
+
+	frequency := frame.SampleRateNumerator / frame.SampleRateDenominator
+	if frequency <= 0 || frequency > math.MaxInt32 {
+		return audioFormat{}, fmt.Errorf(
+			"%w: frequency=%d",
+			ErrInvalidAudioFormat,
+			frequency,
+		)
+	}
+
+	return audioFormat{
+		channels:  int32(frame.Channels),
+		frequency: int32(frequency),
+	}, nil
+}
+
+func (s *SDLAudioSink) ConsumeAudio(
+	ctx context.Context,
+	frame playback.AudioFrame,
+) error {
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+
+	format, err := audioOutputFormat(frame)
+	if err != nil {
+		return err
+	}
+	interleaved, err := InterleaveF32(frame)
+	if err != nil {
+		return err
+	}
+
+	if s.stream == 0 || s.format != format {
+		if err := s.recreateStream(format); err != nil {
+			return err
+		}
+	}
+
+	if s.maxQueuedBatches <= 0 || len(interleaved) > math.MaxInt32/int(s.maxQueuedBatches) {
+		return fmt.Errorf(
+			"%w: batch bytes=%d queue batches=%d",
+			ErrInvalidAudioFormat,
+			len(interleaved),
+			s.maxQueuedBatches,
+		)
+	}
+	maxQueuedBytes := int32(len(interleaved)) * s.maxQueuedBatches
+
+	for {
+		queued := s.backend.Queued(s.stream)
+		if queued < 0 {
+			return fmt.Errorf("%w: query queued bytes: %s", ErrQueueAudioData, s.backend.Error())
+		}
+		if queued <= maxQueuedBytes {
+			break
+		}
+		if err := s.wait(ctx, s.queuePoll); err != nil {
+			if ctx.Err() != nil {
+				return ctx.Err()
+			}
+			return err
+		}
+	}
+
+	if !s.backend.Put(s.stream, interleaved) {
+		return fmt.Errorf("%w: %s", ErrQueueAudioData, s.backend.Error())
+	}
+	return nil
+}
+
+func (s *SDLAudioSink) recreateStream(format audioFormat) error {
+	if s.stream != 0 {
+		s.backend.Destroy(s.stream)
+		s.stream = 0
+		s.format = audioFormat{}
+	}
+
+	stream := s.backend.Open(s.deviceID, sdl.AudioSpec{
+		Format:   sdl.AudioF32,
+		Channels: format.channels,
+		Freq:     format.frequency,
+	})
+	if stream == 0 {
+		return fmt.Errorf("%w: %s", ErrOpenAudioStream, s.backend.Error())
+	}
+	if !s.backend.Resume(stream) {
+		s.backend.Destroy(stream)
+		return fmt.Errorf("%w: %s", ErrResumeAudioStream, s.backend.Error())
+	}
+
+	s.stream = stream
+	s.format = format
+	return nil
+}
+
+func (s *SDLAudioSink) Close() error {
+	if s.stream != 0 {
+		s.backend.Destroy(s.stream)
+		s.stream = 0
+	}
+	s.format = audioFormat{}
+	return nil
+}
@@ -0,0 +1,213 @@
+package output
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"mxl-player/internal/playback"
+	"mxl-player/internal/sdl"
+)
+
+type fakeAudioBackend struct {
+	nextStream uintptr
+	openSpecs  []sdl.AudioSpec
+	openDevice []uint32
+	resumeOK   bool
+	putOK      bool
+	puts       [][]byte
+	queued     []int32
+	queueCalls int
+	destroyed  []uintptr
+	errText    string
+}
+
+func (b *fakeAudioBackend) Open(device uint32, spec sdl.AudioSpec) uintptr {
+	b.openDevice = append(b.openDevice, device)
+	b.openSpecs = append(b.openSpecs, spec)
+	return b.nextStream
+}
+func (b *fakeAudioBackend) Resume(uintptr) bool { return b.resumeOK }
+func (b *fakeAudioBackend) Put(_ uintptr, data []byte) bool {
+	b.puts = append(b.puts, append([]byte(nil), data...))
+	return b.putOK
+}
+func (b *fakeAudioBackend) Queued(uintptr) int32 {
+	if len(b.queued) == 0 {
+		return 0
+	}
+	index := b.queueCalls
+	if index >= len(b.queued) {
+		index = len(b.queued) - 1
+	}
+	b.queueCalls++
+	return b.queued[index]
+}
+func (b *fakeAudioBackend) Destroy(stream uintptr) {
+	b.destroyed = append(b.destroyed, stream)
+}
+func (b *fakeAudioBackend) Error() string { return b.errText }
+
+func newTestSDLAudioSink(backend audioBackend) *SDLAudioSink {
+	return &SDLAudioSink{
+		deviceID:         7,
+		backend:          backend,
+		maxQueuedBatches: 2,
+		queuePoll:        time.Millisecond,
+		wait:             func(context.Context, time.Duration) error { return nil },
+	}
+}
+
+func TestAudioOutputFormat(t *testing.T) {
+	tests := []struct {
+		name    string
+		frame   playback.AudioFrame
+		want    audioFormat
+		wantErr bool
+	}{
+		{
+			name:  "valid",
+			frame: playback.AudioFrame{Channels: 2, SampleRateNumerator: 48000, SampleRateDenominator: 1},
+			want:  audioFormat{channels: 2, frequency: 48000},
+		},
+		{name: "zero channels", frame: playback.AudioFrame{SampleRateNumerator: 48000, SampleRateDenominator: 1}, wantErr: true},
+		{name: "zero numerator", frame: playback.AudioFrame{Channels: 2, SampleRateDenominator: 1}, wantErr: true},
+		{name: "zero denominator", frame: playback.AudioFrame{Channels: 2, SampleRateNumerator: 48000}, wantErr: true},
+		{name: "fractional frequency", frame: playback.AudioFrame{Channels: 2, SampleRateNumerator: 30000, SampleRateDenominator: 1001}, wantErr: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := audioOutputFormat(tt.frame)
+			if tt.wantErr {
+				if !errors.Is(err, ErrInvalidAudioFormat) {
+					t.Fatalf("audioOutputFormat() error = %v, want %v", err, ErrInvalidAudioFormat)
+				}
+				return
+			}
+			if err != nil || got != tt.want {
+				t.Fatalf("audioOutputFormat() = %#v, %v; want %#v, nil", got, err, tt.want)
+			}
+		})
+	}
+}
+
+func TestSDLAudioSinkOpensInterleavesAndReusesStream(t *testing.T) {
+	backend := &fakeAudioBackend{nextStream: 11, resumeOK: true, putOK: true}
+	sink := newTestSDLAudioSink(backend)
+	frame := validAudioFrame()
+
+	if err := sink.ConsumeAudio(context.Background(), frame); err != nil {
+		t.Fatalf("first ConsumeAudio() error = %v", err)
+	}
+	if err := sink.ConsumeAudio(context.Background(), frame); err != nil {
+		t.Fatalf("second ConsumeAudio() error = %v", err)
+	}
+	if len(backend.openSpecs) != 1 {
+		t.Fatalf("open calls = %d, want 1", len(backend.openSpecs))
+	}
+	wantSpec := sdl.AudioSpec{Format: sdl.AudioF32, Channels: 2, Freq: 48000}
+	if backend.openDevice[0] != 7 || backend.openSpecs[0] != wantSpec {
+		t.Fatalf("open = device %d spec %#v, want 7 %#v", backend.openDevice[0], backend.openSpecs[0], wantSpec)
+	}
+	wantData := []byte{1, 2, 3, 4, 9, 10, 11, 12, 5, 6, 7, 8, 13, 14, 15, 16}
+	if len(backend.puts) != 2 || !bytes.Equal(backend.puts[0], wantData) {
+		t.Fatalf("queued data = %v, want %v twice", backend.puts, wantData)
+	}
+}
+
+func TestSDLAudioSinkRecreatesStreamOnFormatChange(t *testing.T) {
+	backend := &fakeAudioBackend{nextStream: 11, resumeOK: true, putOK: true}
+	sink := newTestSDLAudioSink(backend)
+	first := validAudioFrame()
+	if err := sink.ConsumeAudio(context.Background(), first); err != nil {
+		t.Fatalf("first ConsumeAudio() error = %v", err)
+	}
+
+	backend.nextStream = 12
+	second := first
+	second.SampleRateNumerator = 96000
+	if err := sink.ConsumeAudio(context.Background(), second); err != nil {
+		t.Fatalf("second ConsumeAudio() error = %v", err)
+	}
+	if len(backend.openSpecs) != 2 || len(backend.destroyed) != 1 || backend.destroyed[0] != 11 {
+		t.Fatalf("opens=%d destroyed=%v, want 2 and [11]", len(backend.openSpecs), backend.destroyed)
+	}
+}
+
+func TestSDLAudioSinkReportsOpenResumeAndPutFailures(t *testing.T) {
+	frame := validAudioFrame()
+
+	t.Run("open", func(t *testing.T) {
+		backend := &fakeAudioBackend{errText: "open failed"}
+		err := newTestSDLAudioSink(backend).ConsumeAudio(context.Background(), frame)
+		if !errors.Is(err, ErrOpenAudioStream) {
+			t.Fatalf("ConsumeAudio() error = %v, want %v", err, ErrOpenAudioStream)
+		}
+	})
+
+	t.Run("resume", func(t *testing.T) {
+		backend := &fakeAudioBackend{nextStream: 11, errText: "resume failed"}
+		sink := newTestSDLAudioSink(backend)
+		err := sink.ConsumeAudio(context.Background(), frame)
+		if !errors.Is(err, ErrResumeAudioStream) || sink.stream != 0 {
+			t.Fatalf("ConsumeAudio() error=%v stream=%d, want resume error and zero stream", err, sink.stream)
+		}
+		if len(backend.destroyed) != 1 || backend.destroyed[0] != 11 {
+			t.Fatalf("destroyed = %v, want [11]", backend.destroyed)
+		}
+	})
+
+	t.Run("put", func(t *testing.T) {
+		backend := &fakeAudioBackend{nextStream: 11, resumeOK: true, errText: "put failed"}
+		err := newTestSDLAudioSink(backend).ConsumeAudio(context.Background(), frame)
+		if !errors.Is(err, ErrQueueAudioData) {
+			t.Fatalf("ConsumeAudio() error = %v, want %v", err, ErrQueueAudioData)
+		}
+	})
+}
+
+func TestSDLAudioSinkWaitsForQueueAndHonorsCancellation(t *testing.T) {
+	backend := &fakeAudioBackend{
+		nextStream: 11,
+		resumeOK:   true,
+		putOK:      true,
+		queued:     []int32{100, 100},
+	}
+	sink := newTestSDLAudioSink(backend)
+	ctx, cancel := context.WithCancel(context.Background())
+	waits := 0
+	sink.wait = func(ctx context.Context, _ time.Duration) error {
+		waits++
+		cancel()
+		return ctx.Err()
+	}
+
+	err := sink.ConsumeAudio(ctx, validAudioFrame())
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("ConsumeAudio() error = %v, want context.Canceled", err)
+	}
+	if waits != 1 || len(backend.puts) != 0 {
+		t.Fatalf("waits=%d puts=%d, want 1 and 0", waits, len(backend.puts))
+	}
+}
+
+func TestSDLAudioSinkCloseIsIdempotent(t *testing.T) {
+	backend := &fakeAudioBackend{nextStream: 11, resumeOK: true, putOK: true}
+	sink := newTestSDLAudioSink(backend)
+	if err := sink.ConsumeAudio(context.Background(), validAudioFrame()); err != nil {
+		t.Fatalf("ConsumeAudio() error = %v", err)
+	}
+
+	if err := sink.Close(); err != nil {
+		t.Fatalf("Close() error = %v", err)
+	}
+	if err := sink.Close(); err != nil {
+		t.Fatalf("second Close() error = %v", err)
+	}
+	if len(backend.destroyed) != 1 || sink.stream != 0 || sink.format != (audioFormat{}) {
+		t.Fatalf("destroyed=%v stream=%d format=%#v", backend.destroyed, sink.stream, sink.format)
+	}
+}