talk/internal/audio/audio_test.go

package audio

import (
	"bytes"
	"context"
	"encoding/binary"
	"math"
	"os"
	"strings"
	"testing"
)

// buildWAV builds a WAV file per the RIFF specification (implemented
// independently of the parser so parser and test fixture cannot share the
// same bug).
func buildWAV(audioFormat uint16, channels, rate, bits int, samples []byte) []byte {
	var buf bytes.Buffer
	buf.WriteString("RIFF")
	payload := 28 + len(samples) // fmt (8+16) + data (8+n) + 4 bytes WAVE
	binary.Write(&buf, binary.LittleEndian, uint32(payload))
	buf.WriteString("WAVE")

	buf.WriteString("fmt ")
	binary.Write(&buf, binary.LittleEndian, uint32(16))
	binary.Write(&buf, binary.LittleEndian, audioFormat)
	binary.Write(&buf, binary.LittleEndian, uint16(channels))
	binary.Write(&buf, binary.LittleEndian, uint32(rate))
	blockAlign := channels * bits / 8
	binary.Write(&buf, binary.LittleEndian, uint32(rate*blockAlign))
	binary.Write(&buf, binary.LittleEndian, uint16(blockAlign))
	binary.Write(&buf, binary.LittleEndian, uint16(bits))

	buf.WriteString("data")
	binary.Write(&buf, binary.LittleEndian, uint32(len(samples)))
	buf.Write(samples)
	return buf.Bytes()
}

func TestDecodeWAV(t *testing.T) {
	t.Run("16-bit mono", func(t *testing.T) {
		wav := buildWAV(1, 1, 22050, 16, []byte{1, 2, 3, 4})
		f, pcm, err := decodeWAV(wav)
		if err != nil {
			t.Fatalf("decodeWAV: %v", err)
		}
		if f.SampleRate != 22050 || f.Channels != 1 || f.BitsPerCoded != 16 || f.Float {
			t.Errorf("format = %+v", f)
		}
		if string(pcm) != "\x01\x02\x03\x04" {
			t.Errorf("PCM = %x", pcm)
		}
	})
	t.Run("8-bit stereo with extra chunk", func(t *testing.T) {
		wav := buildWAV(1, 2, 8000, 8, []byte{7, 8, 9})
		// insert a LIST chunk (word-aligned, odd length) between fmt and data
		withList := append(wav[:36], append([]byte("LIST\x03\x00\x00\x00abc\x00"), wav[36:]...)...)
		f, pcm, err := decodeWAV(withList)
		if err != nil {
			t.Fatalf("decodeWAV: %v", err)
		}
		if f.BitsPerCoded != 8 || f.Channels != 2 {
			t.Errorf("format = %+v", f)
		}
		if string(pcm) != "\x07\x08\x09" {
			t.Errorf("PCM = %x", pcm)
		}
	})
	t.Run("float32", func(t *testing.T) {
		wav := buildWAV(3, 1, 44100, 32, []byte{0, 0, 0, 0})
		f, _, err := decodeWAV(wav)
		if err != nil {
			t.Fatalf("decodeWAV: %v", err)
		}
		if !f.Float || f.BitsPerCoded != 32 {
			t.Errorf("format = %+v, want Float32", f)
		}
	})
	t.Run("empty data chunk is allowed", func(t *testing.T) {
		wav := buildWAV(1, 1, 22050, 16, nil)
		_, pcm, err := decodeWAV(wav)
		if err != nil {
			t.Fatalf("decodeWAV: %v", err)
		}
		if len(pcm) != 0 {
			t.Errorf("PCM = %x, want empty", pcm)
		}
	})
}

func TestDecodeWAVErrors(t *testing.T) {
	cases := []struct {
		name string
		data []byte
		want string
	}{
		{"too short", []byte("RIFF"), "too short"},
		{"no riff", []byte("RIFX" + "\x00\x00\x00\x00" + "WAVE"), "RIFF"},
		{"24-bit pcm", buildWAV(1, 1, 44100, 24, []byte{1, 2, 3}), "unsupported format"},
		{"3 channels", buildWAV(1, 3, 44100, 16, []byte{1, 2, 3, 4, 5, 6}), "channels"},
		{"without data", func() []byte {
			wav := buildWAV(1, 1, 22050, 16, []byte{1, 2})
			return wav[:36] // keep only the fmt chunk
		}(), "data"},
		{"broken chunk", append(buildWAV(1, 1, 22050, 16, nil), "data\xff\xff\xff\xff"...), "exceeds"},
	}
	for _, tc := range cases {
		t.Run(tc.name, func(t *testing.T) {
			_, _, err := decodeWAV(tc.data)
			if err == nil || !strings.Contains(err.Error(), tc.want) {
				t.Errorf("decodeWAV error = %v, want message containing %q", err, tc.want)
			}
		})
	}
}

func TestNewSink(t *testing.T) {
	for _, name := range []string{NameOto, NameNull} {
		s, err := NewSink(name, nil)
		if err != nil {
			t.Fatalf("NewSink(%q): %v", name, err)
		}
		if s.Name() != name {
			t.Errorf("NewSink(%q).Name() = %q", name, s.Name())
		}
	}
	if _, err := NewSink("mpv", nil); err == nil {
		t.Error("NewSink(\"mpv\") should fail")
	}
	t.Run("null sink logs via the factory writer", func(t *testing.T) {
		var buf bytes.Buffer
		s, err := NewSink(NameNull, &buf)
		if err != nil {
			t.Fatalf("NewSink(%q): %v", NameNull, err)
		}
		if err := s.Play(context.Background(), []byte("audio")); err != nil {
			t.Fatalf("Play: %v", err)
		}
		if !strings.Contains(buf.String(), "discarded (null sink)") {
			t.Errorf("log = %q, want discard line via the factory writer", buf.String())
		}
	})
}

func TestNullSinkDiscardsAndLogs(t *testing.T) {
	var buf bytes.Buffer
	sink := NullSink{W: &buf}
	if err := sink.Play(context.Background(), []byte("audio")); err != nil {
		t.Fatalf("NullSink.Play: %v", err)
	}
	if !strings.Contains(buf.String(), "discarded") {
		t.Errorf("NullSink log = %q, want discard line", buf.String())
	}
	if sink.Name() != NameNull {
		t.Errorf("Name() = %q", sink.Name())
	}
}

// TestOtoSinkInvalidData: OtoSink.Play must fail at WAV decoding before any
// audio device is touched; this is the CI/server-safe error path of the oto
// sink.
func TestOtoSinkInvalidData(t *testing.T) {
	err := (OtoSink{}).Play(context.Background(), []byte("not audio"))
	if err == nil {
		t.Fatal("OtoSink.Play with garbage: no error")
	}
	if !strings.Contains(err.Error(), "wav:") {
		t.Errorf("error should come from the WAV parser: %v", err)
	}
}

// TestOtoSinkRealPlayback runs only with TALK_TEST_REAL_AUDIO=1 (produces
// real sound); default test runs stay silent.
func TestOtoSinkRealPlayback(t *testing.T) {
	if testing.Short() {
		t.Skip("skipped (-short)")
	}
	if os.Getenv("TALK_TEST_REAL_AUDIO") != "1" {
		t.Skip("skipped (audible); run with TALK_TEST_REAL_AUDIO=1")
	}
	// 0.1 s 440 Hz sine, 16 bit mono 22050 Hz.
	const rate = 22050
	samples := make([]byte, rate/10*2)
	for i := 0; i < rate/10; i++ {
		v := int16(8000 * math.Sin(float64(i)/float64(rate)*2*math.Pi*440))
		binary.LittleEndian.PutUint16(samples[i*2:], uint16(v))
	}
	if err := (OtoSink{}).Play(context.Background(), buildWAV(1, 1, rate, 16, samples)); err != nil {
		t.Fatalf("OtoSink.Play: %v", err)
	}
}