Talon/voice-cat
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Complete managed macOS platform bring-up
.NET port / test (macos-latest) (push) Canceled after 0s
Details
.NET port / test (ubuntu-24.04) (push) Canceled after 0s
Details
.NET port / test (windows-latest) (push) Canceled after 0s
Details
.NET port / apple-client (push) Canceled after 0s
Details
.NET port / cpp-conformance (push) Canceled after 0s
Details

Browse Source
This commit is contained in:
Talon
2026-09-18 18:51:53 +02:00
parent ab460ae12b
commit 310c0f09dd
21 changed files with 421 additions and 83 deletions
Download Patch File Download Diff File Expand all files Collapse all files
clients/apple/dotnet/VoiceCat.Mac/MacAudioBackend.cs
+65 -21
View File
@@ -1,4 +1,5 @@
using System.Runtime.InteropServices;
using AudioUnit;
using AVFoundation;
using Foundation;
using VoiceCat.Audio;
@@ -7,25 +8,21 @@ namespace VoiceCat.Mac;
internal sealed class MacAudioBackend : IAudioDeviceBackend
{
private static readonly AudioDeviceInfo[] DefaultInput = [new("", "System default microphone", true)];
private static readonly AudioDeviceInfo[] DefaultOutput = [new("", "System default output", true)];
private Exception? failure;
internal Exception? Failure => Volatile.Read(ref failure);
public IReadOnlyList<AudioDeviceInfo> Enumerate(bool input) => input ? DefaultInput : DefaultOutput;
public IReadOnlyList<AudioDeviceInfo> Enumerate(bool input) => CoreAudioDevices.List(input);
public IAudioCapture OpenCapture(string? deviceId, bool loopback, CapturePcmHandler pcm)
{
if (loopback) throw new NotSupportedException("Screen audio uses ScreenCaptureKit, not microphone capture.");
if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio input is not implemented yet.");
return new Capture(pcm, ReportFailure);
return new Capture(deviceId, pcm, ReportFailure);
}
public IAudioPlayback OpenPlayback(string? deviceId = null)
{
if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio output is not implemented yet.");
return new Playback(ReportFailure);
return new Playback(deviceId, ReportFailure);
}
private void ReportFailure(Exception exception) => Interlocked.CompareExchange(ref failure, exception, null);
@@ -43,11 +40,16 @@ internal sealed class MacAudioBackend : IAudioDeviceBackend
private bool inputProvided;
private int disposed;
internal Capture(CapturePcmHandler handler, Action<Exception> reportFailure)
internal Capture(string? deviceId, CapturePcmHandler handler, Action<Exception> reportFailure)
{
this.handler = handler;
this.reportFailure = reportFailure;
AVAudioInputNode input = engine.InputNode;
if (!string.IsNullOrEmpty(deviceId))
{
AudioUnitStatus status = input.AudioUnit!.SetCurrentDevice(CoreAudioDevices.ParseId(deviceId), AudioUnitScopeType.Global, 0);
if (status != AudioUnitStatus.NoError) throw new InvalidOperationException($"Core Audio input selection failed ({status}).");
}
AVAudioFormat inputFormat = input.GetBusOutputFormat(0);
uint channels = Math.Clamp(inputFormat.ChannelCount, 1u, 2u);
outputFormat = new(AVAudioCommonFormat.PCMInt16, 48_000, channels, true);
@@ -55,7 +57,19 @@ internal sealed class MacAudioBackend : IAudioDeviceBackend
uint outputCapacity = checked((uint)Math.Ceiling(4_096 * 48_000 / inputFormat.SampleRate) + 64);
converted = new(outputFormat, outputCapacity);
inputProvider = ProvideInput;
input.InstallTapOnBus(0, 960, inputFormat, Convert);
NSError? tapError = null;
if (OperatingSystem.IsMacOSVersionAtLeast(27))
input.InstallTapOnBus(0, 960, inputFormat, out tapError, Convert);
else
input.InstallTapOnBus(0, 960, inputFormat, Convert);
if (tapError is not null)
{
converted.Dispose();
converter.Dispose();
outputFormat.Dispose();
engine.Dispose();
throw new InvalidOperationException("Core Audio capture tap could not be installed: " + tapError.LocalizedDescription);
}
engine.Prepare();
if (!engine.StartAndReturnError(out var error))
{
@@ -126,6 +140,7 @@ internal sealed class MacAudioBackend : IAudioDeviceBackend
private sealed class Playback : IAudioPlayback
{
private readonly PcmRing pcm = new(32_768);
private readonly short[] renderScratch = new short[8_192];
private readonly AVAudioEngine engine = new();
private readonly AVAudioFormat format;
private readonly AVAudioSourceNode source;
@@ -133,13 +148,33 @@ internal sealed class MacAudioBackend : IAudioDeviceBackend
private int callbackFailed;
private int disposed;
internal Playback(Action<Exception> reportFailure)
internal Playback(string? deviceId, Action<Exception> reportFailure)
{
this.reportFailure = reportFailure;
format = new(AVAudioCommonFormat.PCMInt16, 48_000, 2, true);
if (!string.IsNullOrEmpty(deviceId))
{
AudioUnitStatus status = engine.OutputNode.AudioUnit!.SetCurrentDevice(CoreAudioDevices.ParseId(deviceId), AudioUnitScopeType.Global, 0);
if (status != AudioUnitStatus.NoError) throw new InvalidOperationException($"Core Audio output selection failed ({status}).");
}
// AVAudioEngine's native mixer path is planar Float32. Convert from the managed
// Int16 ring directly in this allocation-free callback, avoiding an extra graph
// converter and matching the established Swift/VPIO renderer.
format = new(AVAudioCommonFormat.PCMFloat32, 48_000, 2, false);
source = new(format, Render);
engine.AttachNode(source);
engine.Connect(source, engine.MainMixerNode, format);
NSError? connectionError = null;
if (OperatingSystem.IsMacOSVersionAtLeast(27))
engine.Connect(source, engine.MainMixerNode, format, out connectionError);
else
engine.Connect(source, engine.MainMixerNode, format);
if (connectionError is not null)
{
engine.DetachNode(source);
source.Dispose();
format.Dispose();
engine.Dispose();
throw new InvalidOperationException("Core Audio playback could not be connected: " + connectionError.LocalizedDescription);
}
engine.Prepare();
if (!engine.StartAndReturnError(out var error))
{
@@ -159,17 +194,26 @@ internal sealed class MacAudioBackend : IAudioDeviceBackend
{
int bufferCount = Marshal.ReadInt32(outputData);
int firstBuffer = IntPtr.Size == 8 ? 8 : 4;
if (bufferCount != 1) return Fail("Core Audio returned a non-interleaved playback buffer.");
int channels = Marshal.ReadInt32(outputData, firstBuffer);
int byteCount = Marshal.ReadInt32(outputData, firstBuffer + 4);
nint data = Marshal.ReadIntPtr(outputData, firstBuffer + 8);
int requested = checked((int)frameCount * 2);
if (channels != 2 || data == 0 || byteCount < requested * sizeof(short)) return Fail("Core Audio returned an invalid stereo playback buffer.");
var output = new Span<short>((void*)data, requested);
const int bufferStride64 = 16;
int stride = IntPtr.Size == 8 ? bufferStride64 : 12;
int frames = checked((int)frameCount);
int requested = checked(frames * 2);
if (bufferCount != 2 || requested > renderScratch.Length) return Fail("Core Audio returned an invalid planar stereo playback buffer.");
Span<short> discard = stackalloc short[1_920];
while (pcm.Count > 11_520) pcm.Read(discard[..Math.Min(discard.Length, pcm.Count - 11_520)]);
int read = pcm.Read(output);
output[read..].Clear();
Span<short> source = renderScratch.AsSpan(0, requested);
int read = pcm.Read(source);
source[read..].Clear();
for (int channel = 0; channel < 2; channel++)
{
int offset = firstBuffer + channel * stride;
int channels = Marshal.ReadInt32(outputData, offset);
int byteCount = Marshal.ReadInt32(outputData, offset + 4);
nint data = Marshal.ReadIntPtr(outputData, offset + 8);
if (channels != 1 || data == 0 || byteCount < frames * sizeof(float)) return Fail("Core Audio returned an invalid Float32 playback channel.");
var output = new Span<float>((void*)data, frames);
for (int frame = 0; frame < frames; frame++) output[frame] = source[frame * 2 + channel] * (1.0f / 32_768.0f);
}
if (isSilence != IntPtr.Zero) Marshal.WriteByte(isSilence, read == 0 ? (byte)1 : (byte)0);
return 0;
}