mpd/src/output/plugins/OSXOutputPlugin.cxx

/*
 * Copyright 2003-2017 The Music Player Daemon Project
 * http://www.musicpd.org
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */

#include "config.h"
#include "OSXOutputPlugin.hxx"
#include "../OutputAPI.hxx"
#include "mixer/MixerList.hxx"
#include "util/ScopeExit.hxx"
#include "util/RuntimeError.hxx"
#include "util/Domain.hxx"
#include "util/Manual.hxx"
#include "util/ConstBuffer.hxx"
#include "pcm/PcmExport.hxx"
#include "thread/Mutex.hxx"
#include "thread/Cond.hxx"
#include "system/ByteOrder.hxx"
#include "util/StringBuffer.hxx"
#include "util/StringFormat.hxx"
#include "Log.hxx"

#include <CoreAudio/CoreAudio.h>
#include <AudioUnit/AudioUnit.h>
#include <CoreServices/CoreServices.h>
#include <boost/lockfree/spsc_queue.hpp>

#include <memory>

static constexpr unsigned MPD_OSX_BUFFER_TIME_MS = 100;

static StringBuffer<64>
StreamDescriptionToString(const AudioStreamBasicDescription desc) {
	// Only convert the lpcm formats (nothing else supported / used by MPD)
	assert(desc.mFormatID == kAudioFormatLinearPCM);
	
	return StringFormat<64>("%u channel %s %sinterleaved %u-bit %s %s (%uHz)",
							desc.mChannelsPerFrame,
							(desc.mFormatFlags & kAudioFormatFlagIsNonMixable) ? "" : "mixable",
							(desc.mFormatFlags & kAudioFormatFlagIsNonInterleaved) ? "non-" : "",
							desc.mBitsPerChannel,
							(desc.mFormatFlags & kAudioFormatFlagIsFloat) ? "Float" : "SInt",
							(desc.mFormatFlags & kAudioFormatFlagIsBigEndian) ? "BE" : "LE",
							(UInt32)desc.mSampleRate);
}


struct OSXOutput final : AudioOutput {
	/* configuration settings */
	OSType component_subtype;
	/* only applicable with kAudioUnitSubType_HALOutput */
	const char *device_name;
	const char *channel_map;
	bool hog_device;
	bool pause;
#ifdef ENABLE_DSD
	/**
	 * Enable DSD over PCM according to the DoP standard?
	 *
	 * @see http://dsd-guide.com/dop-open-standard
	 */
	bool dop_setting;
	Manual<PcmExport> pcm_export;
#endif

	AudioDeviceID dev_id;
	AudioComponentInstance au;
	AudioStreamBasicDescription asbd;

	boost::lockfree::spsc_queue<uint8_t> *ring_buffer;

	OSXOutput(const ConfigBlock &block);

	static AudioOutput *Create(EventLoop &, const ConfigBlock &block);
	int GetVolume();
	void SetVolume(unsigned new_volume);

private:
	void Enable() override;
	void Disable() noexcept override;

	void Open(AudioFormat &audio_format) override;
	void Close() noexcept override;

	std::chrono::steady_clock::duration Delay() const noexcept override;
	size_t Play(const void *chunk, size_t size) override;
	bool Pause() override;
};

static constexpr Domain osx_output_domain("osx_output");

static void
osx_os_status_to_cstring(OSStatus status, char *str, size_t size) {
	CFErrorRef cferr = CFErrorCreate(nullptr, kCFErrorDomainOSStatus, status, nullptr);
	CFStringRef cfstr = CFErrorCopyDescription(cferr);
	if (!CFStringGetCString(cfstr, str, size, kCFStringEncodingUTF8)) {
		/* conversion failed, return empty string */
		*str = '\0';
	}
	if (cferr)
		CFRelease(cferr);
	if (cfstr)
		CFRelease(cfstr);
}

static bool
osx_output_test_default_device(void)
{
	/* on a Mac, this is always the default plugin, if nothing
	   else is configured */
	return true;
}

OSXOutput::OSXOutput(const ConfigBlock &block)
	:AudioOutput(FLAG_ENABLE_DISABLE|FLAG_PAUSE)
{
	const char *device = block.GetBlockValue("device");

	if (device == nullptr || 0 == strcmp(device, "default")) {
		component_subtype = kAudioUnitSubType_DefaultOutput;
		device_name = nullptr;
	}
	else if (0 == strcmp(device, "system")) {
		component_subtype = kAudioUnitSubType_SystemOutput;
		device_name = nullptr;
	}
	else {
		component_subtype = kAudioUnitSubType_HALOutput;
		/* XXX am I supposed to strdup() this? */
		device_name = device;
	}

	channel_map = block.GetBlockValue("channel_map");
	hog_device = block.GetBlockValue("hog_device", false);
#ifdef ENABLE_DSD
	dop_setting = block.GetBlockValue("dop", false);
#endif
}

AudioOutput *
OSXOutput::Create(EventLoop &, const ConfigBlock &block)
{
	OSXOutput *oo = new OSXOutput(block);
	AudioObjectPropertyAddress aopa;
	AudioDeviceID dev_id = kAudioDeviceUnknown;
	UInt32 dev_id_size = sizeof(dev_id);
	
	if (oo->component_subtype == kAudioUnitSubType_SystemOutput)
		// get system output dev_id if configured
		aopa = {
			kAudioHardwarePropertyDefaultSystemOutputDevice,
			kAudioObjectPropertyScopeOutput,
			kAudioObjectPropertyElementMaster
		};
	else
		// fallback to default device initially (can still be changed by osx_output_set_device)
		aopa = {
			kAudioHardwarePropertyDefaultOutputDevice,
			kAudioObjectPropertyScopeOutput,
			kAudioObjectPropertyElementMaster
		};

	AudioObjectGetPropertyData(kAudioObjectSystemObject,
				   &aopa,
				   0,
				   NULL,
				   &dev_id_size,
				   &dev_id);
	oo->dev_id = dev_id;

	return oo;
}


int
OSXOutput::GetVolume()
{
	Float32 vol;
	AudioObjectPropertyAddress aopa = {
		.mSelector	= kAudioDevicePropertyVolumeScalar,
		.mScope		= kAudioObjectPropertyScopeOutput,
		.mElement	= kAudioObjectPropertyElementMaster,
	};
	UInt32 size = sizeof(vol);
	OSStatus status = AudioObjectGetPropertyData(dev_id,
						     &aopa,
						     0,
						     NULL,
						     &size,
						     &vol);

	if (status != noErr) {
		char errormsg[1024];
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("unable to get volume: %s", errormsg);
	}

	return static_cast<int>(vol * 100.0);
}

void
OSXOutput::SetVolume(unsigned new_volume) {
	Float32 vol = new_volume / 100.0;
	AudioObjectPropertyAddress aopa = {
		.mSelector	= kAudioDevicePropertyVolumeScalar,
		.mScope		= kAudioObjectPropertyScopeOutput,
		.mElement	= kAudioObjectPropertyElementMaster
	};
	UInt32 size = sizeof(vol);
	OSStatus status = AudioObjectSetPropertyData(dev_id,
						     &aopa,
						     0,
						     NULL,
						     size,
						     &vol);

	if (status != noErr) {
		char errormsg[1024];
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError( "unable to set new volume %u: %s",
				new_volume, errormsg);
	}
}

static void
osx_output_parse_channel_map(
	const char *device_name,
	const char *channel_map_str,
	SInt32 channel_map[],
	UInt32 num_channels)
{
	char *endptr;
	unsigned int inserted_channels = 0;
	bool want_number = true;

	while (*channel_map_str) {
		if (inserted_channels >= num_channels)
			throw FormatRuntimeError("%s: channel map contains more than %u entries or trailing garbage",
						 device_name, num_channels);

		if (!want_number && *channel_map_str == ',') {
			++channel_map_str;
			want_number = true;
			continue;
		}

		if (want_number &&
			(isdigit(*channel_map_str) || *channel_map_str == '-')
		) {
			channel_map[inserted_channels] = strtol(channel_map_str, &endptr, 10);
			if (channel_map[inserted_channels] < -1)
				throw FormatRuntimeError("%s: channel map value %d not allowed (must be -1 or greater)",
							 device_name, channel_map[inserted_channels]);

			channel_map_str = endptr;
			want_number = false;
			FormatDebug(osx_output_domain,
				"%s: channel_map[%u] = %d",
				device_name, inserted_channels, channel_map[inserted_channels]);
			++inserted_channels;
			continue;
		}

		throw FormatRuntimeError("%s: invalid character '%c' in channel map",
					 device_name, *channel_map_str);
	}

	if (inserted_channels < num_channels)
		throw FormatRuntimeError("%s: channel map contains less than %u entries",
					 device_name, num_channels);
}

static void
osx_output_set_channel_map(OSXOutput *oo)
{
	AudioStreamBasicDescription desc;
	OSStatus status;
	UInt32 size, num_channels;
	char errormsg[1024];

	size = sizeof(desc);
	memset(&desc, 0, size);
	status = AudioUnitGetProperty(oo->au,
		kAudioUnitProperty_StreamFormat,
		kAudioUnitScope_Output,
		0,
		&desc,
		&size);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("%s: unable to get number of output device channels: %s",
					 oo->device_name, errormsg);
	}

	num_channels = desc.mChannelsPerFrame;
	std::unique_ptr<SInt32[]> channel_map(new SInt32[num_channels]);
	osx_output_parse_channel_map(oo->device_name,
				     oo->channel_map,
				     channel_map.get(),
				     num_channels);

	size = num_channels * sizeof(SInt32);
	status = AudioUnitSetProperty(oo->au,
		kAudioOutputUnitProperty_ChannelMap,
		kAudioUnitScope_Input,
		0,
		channel_map.get(),
		size);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("%s: unable to set channel map: %s", oo->device_name, errormsg);
	}
}


static float
osx_output_score_sample_rate(Float64 destination_rate, unsigned int source_rate) {
	float score = 0;
	double int_portion;
	double frac_portion = modf(source_rate / destination_rate, &int_portion);
	// prefer sample rates that are multiples of the source sample rate
	score += (1 - frac_portion) * 1000;
	// prefer exact matches over other multiples
	score += (int_portion == 1.0) ? 500 : 0;
	// prefer higher multiples if source rate higher than dest rate
	if(source_rate >= destination_rate)
		score += (int_portion > 1 && int_portion < 100) ? (100 - int_portion) / 100 * 100 : 0;
	else
		score += (int_portion > 1 && int_portion < 100) ? (100 + int_portion) / 100 * 100 : 0;
	
	return score;
}

static float
osx_output_score_format(const AudioStreamBasicDescription &format_desc, const AudioFormat &format) {
	float score = 0;
	// Score only linear PCM formats (everything else MPD cannot use)
	if (format_desc.mFormatID == kAudioFormatLinearPCM) {
		score += osx_output_score_sample_rate(format_desc.mSampleRate, format.sample_rate);
		
		// Just choose the stream / format with the highest number of output channels
		score += format_desc.mChannelsPerFrame * 5;
		
		if (format.format == SampleFormat::FLOAT) {
			// for float, prefer the highest bitdepth we have
			if (format_desc.mBitsPerChannel >= 16)
				score += (format_desc.mBitsPerChannel / 8);
		} else {
			if (format_desc.mBitsPerChannel == ((format.format == SampleFormat::S24_P32) ? 24 : format.GetSampleSize() * 8))
				score += 5;
			else if (format_desc.mBitsPerChannel > format.GetSampleSize() * 8)
				score += 1;
			
		}
	}
	return score;
}

static Float64
osx_output_set_device_format(AudioDeviceID dev_id, const AudioFormat &audio_format)
{
	AudioObjectPropertyAddress aopa = {
		kAudioDevicePropertyStreams,
		kAudioObjectPropertyScopeOutput,
		kAudioObjectPropertyElementMaster
	};

	UInt32 property_size;
	OSStatus err = AudioObjectGetPropertyDataSize(dev_id, &aopa, 0, NULL, &property_size);	
	if (err != noErr) {	
		throw FormatRuntimeError("Cannot get number of streams: %d\n", err);	
	}	
	
	int n_streams = property_size / sizeof(AudioStreamID);	
	AudioStreamID streams[n_streams];	
	err = AudioObjectGetPropertyData(dev_id, &aopa, 0, NULL, &property_size, streams);	
	if (err != noErr) {	
		throw FormatRuntimeError("Cannot get streams: %d\n", err);	
	}	
	
	bool format_found = false;
	int output_stream;
	AudioStreamBasicDescription output_format;

	for (int i = 0; i < n_streams; i++) {	
		UInt32 direction;	
		AudioStreamID stream = streams[i];
		aopa.mSelector = kAudioStreamPropertyDirection;	
		property_size = sizeof(direction);	
		err = AudioObjectGetPropertyData(stream,	
						 &aopa,	
						 0,	
						 NULL,	
						 &property_size,	
						 &direction);	
		if (err != noErr) {
			throw FormatRuntimeError("Cannot get streams direction: %d\n", err);	
		}
		if (direction != 0) {	
			continue;
		}

		aopa.mSelector = kAudioStreamPropertyAvailablePhysicalFormats;
		err = AudioObjectGetPropertyDataSize(stream, &aopa, 0, NULL, &property_size);
		if (err != noErr)
			throw FormatRuntimeError("Unable to get format size s for stream %d. Error = %s", streams[i], err);

		int format_count = property_size / sizeof(AudioStreamRangedDescription);
		AudioStreamRangedDescription format_list[format_count];
		err = AudioObjectGetPropertyData(stream, &aopa, 0, NULL, &property_size, format_list);
		if (err != noErr)
			throw FormatRuntimeError("Unable to get available formats for stream %d. Error = %s", streams[i], err);

		float output_score = 0;
		for (int j = 0; j < format_count; j++) {
			AudioStreamBasicDescription format_desc = format_list[j].mFormat;
			std::string format_string;
			
			// for devices with kAudioStreamAnyRate
			// we use the requested samplerate here
			if (format_desc.mSampleRate == kAudioStreamAnyRate)
				format_desc.mSampleRate = audio_format.sample_rate;
			float score = osx_output_score_format(format_desc, audio_format);
			
			// print all (linear pcm) formats and their rating
			if(score > 0.0)
				FormatDebug(osx_output_domain, "Format: %s rated %f", StreamDescriptionToString(format_desc).c_str(), score);
			
			if (score > output_score) {
				output_score  = score;
				output_format = format_desc;
				output_stream = stream; // set the idx of the stream in the device
				format_found = true;
			}
		}
	}

	if (format_found) {
		aopa.mSelector = kAudioStreamPropertyPhysicalFormat;	
		err = AudioObjectSetPropertyData(output_stream,
						 &aopa,	
						 0,	
						 NULL,	
						 sizeof(output_format),	
						 &output_format);	
		if (err != noErr) {	
			throw FormatRuntimeError("Failed to change the stream format: %d\n", err);	
		}	
	}	

	return output_format.mSampleRate;
}

static OSStatus
osx_output_set_buffer_size(AudioUnit au, AudioStreamBasicDescription desc, UInt32 *frame_size)
{
	AudioValueRange value_range = {0, 0};
	UInt32 property_size = sizeof(AudioValueRange);
	OSStatus err = AudioUnitGetProperty(au,
					    kAudioDevicePropertyBufferFrameSizeRange,
					    kAudioUnitScope_Global,
					    0,
					    &value_range,
					    &property_size);
	if (err != noErr)
		return err;

	UInt32 buffer_frame_size = value_range.mMaximum;
	err = AudioUnitSetProperty(au,
				   kAudioDevicePropertyBufferFrameSize,
				   kAudioUnitScope_Global,
				   0,
				   &buffer_frame_size,
				   sizeof(buffer_frame_size));
	if (err != noErr)
                FormatWarning(osx_output_domain,
			      "Failed to set maximum buffer size: %d",
			      err);

	property_size = sizeof(buffer_frame_size);
	err = AudioUnitGetProperty(au,
				   kAudioDevicePropertyBufferFrameSize,
				   kAudioUnitScope_Global,
				   0,
				   &buffer_frame_size,
				   &property_size);
	if (err != noErr) {
                FormatWarning(osx_output_domain,
			      "Cannot get the buffer frame size: %d",
			      err);
		return err;
	}

	buffer_frame_size *= desc.mBytesPerFrame;

	// We set the frame size to a power of two integer that
	// is larger than buffer_frame_size.
	while (*frame_size < buffer_frame_size + 1) {
		*frame_size <<= 1;
	}

	return noErr;
}

static void
osx_output_hog_device(AudioDeviceID dev_id, bool hog)
{
	pid_t hog_pid;
	AudioObjectPropertyAddress aopa = {
		kAudioDevicePropertyHogMode,
		kAudioObjectPropertyScopeOutput,
		kAudioObjectPropertyElementMaster
	};
	UInt32 size = sizeof(hog_pid);
	OSStatus err = AudioObjectGetPropertyData(dev_id,
						  &aopa,
						  0,
						  NULL,
						  &size,
						  &hog_pid);
	if (err != noErr) {
		FormatDebug(osx_output_domain,
			    "Cannot get hog information: %d",
			    err);
		return;
	}
	if (hog) {
		if (hog_pid != -1) {
		        FormatDebug(osx_output_domain,
				    "Device is already hogged.");
			return;
		}
	} else {
		if (hog_pid != getpid()) {
		        FormatDebug(osx_output_domain,
				    "Device is not owned by this process.");
			return;
		}
	}
	hog_pid = hog ? getpid() : -1;
	size = sizeof(hog_pid);
	err = AudioObjectSetPropertyData(dev_id,
					 &aopa,
					 0,
					 NULL,
					 size,
					 &hog_pid);
	if (err != noErr) {
		FormatDebug(osx_output_domain,
			    "Cannot hog the device: %d",
			    err);
	} else {
		FormatDebug(osx_output_domain,
			    hog_pid == -1 ? "Device is unhogged" 
					  : "Device is hogged");
	}
}


static void
osx_output_set_device(OSXOutput *oo)
{
	OSStatus status;
	UInt32 size, numdevices;
	AudioObjectPropertyAddress propaddr;
	CFStringRef cfname = nullptr;
	char errormsg[1024];
	char name[256];
	unsigned int i;

	AtScopeExit(&cfname) {
		if (cfname)
			CFRelease(cfname);
	};

	if (oo->component_subtype != kAudioUnitSubType_HALOutput)
		return;

	/* how many audio devices are there? */
	propaddr = { kAudioHardwarePropertyDevices, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMaster };
	status = AudioObjectGetPropertyDataSize(kAudioObjectSystemObject, &propaddr, 0, nullptr, &size);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to determine number of OS X audio devices: %s",
					 errormsg);
	}

	/* what are the available audio device IDs? */
	numdevices = size / sizeof(AudioDeviceID);
	std::unique_ptr<AudioDeviceID[]> deviceids(new AudioDeviceID[numdevices]);
	status = AudioObjectGetPropertyData(kAudioObjectSystemObject, &propaddr, 0, nullptr, &size, deviceids.get());
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to determine OS X audio device IDs: %s",
					 errormsg);
	}

	/* which audio device matches oo->device_name? */
	propaddr = { kAudioObjectPropertyName, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMaster };
	size = sizeof(CFStringRef);
	for (i = 0; i < numdevices; i++) {
		status = AudioObjectGetPropertyData(deviceids[i], &propaddr, 0, nullptr, &size, &cfname);
		if (status != noErr) {
			osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
			throw FormatRuntimeError("Unable to determine OS X device name "
						 "(device %u): %s",
						 (unsigned int) deviceids[i],
						 errormsg);
		}

		if (!CFStringGetCString(cfname, name, sizeof(name), kCFStringEncodingUTF8))
			throw std::runtime_error("Unable to convert device name from CFStringRef to char*");

		if (strcmp(oo->device_name, name) == 0) {
			FormatDebug(osx_output_domain,
				    "found matching device: ID=%u, name=%s",
				    (unsigned)deviceids[i], name);
			break;
		}
	}
	if (i == numdevices) {
		FormatWarning(osx_output_domain,
			      "Found no audio device with name '%s' "
			      "(will use default audio device)",
			      oo->device_name);
		return;
	}

	status = AudioUnitSetProperty(oo->au,
				      kAudioOutputUnitProperty_CurrentDevice,
				      kAudioUnitScope_Global,
				      0,
				      &(deviceids[i]),
				      sizeof(AudioDeviceID));
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to set OS X audio output device: %s",
					 errormsg);
	}

	oo->dev_id = deviceids[i];
	FormatDebug(osx_output_domain,
		    "set OS X audio output device ID=%u, name=%s",
		    (unsigned)deviceids[i], name);

	if (oo->channel_map)
		osx_output_set_channel_map(oo);
}


/*
	This function (the 'render callback' osx_render) is called by the
	OS X audio subsystem (CoreAudio) to request audio data that will be
	played by the audio hardware. This function has hard time constraints
	so it cannot do IO (debug statements) or memory allocations.
*/

static OSStatus
osx_render(void *vdata,
	   gcc_unused AudioUnitRenderActionFlags *io_action_flags,
	   gcc_unused const AudioTimeStamp *in_timestamp,
	   gcc_unused UInt32 in_bus_number,
	   UInt32 in_number_frames,
	   AudioBufferList *buffer_list)
{
	OSXOutput *od = (OSXOutput *) vdata;

	int count = in_number_frames * od->asbd.mBytesPerFrame;
	buffer_list->mBuffers[0].mDataByteSize =
		od->ring_buffer->pop((uint8_t *)buffer_list->mBuffers[0].mData,
				     count);
 	return noErr;
}

void
OSXOutput::Enable()
{
	char errormsg[1024];

	AudioComponentDescription desc;
	desc.componentType = kAudioUnitType_Output;
	desc.componentSubType = component_subtype;
	desc.componentManufacturer = kAudioUnitManufacturer_Apple;
	desc.componentFlags = 0;
	desc.componentFlagsMask = 0;

	AudioComponent comp = AudioComponentFindNext(nullptr, &desc);
	if (comp == 0)
		throw std::runtime_error("Error finding OS X component");

	OSStatus status = AudioComponentInstanceNew(comp, &au);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to open OS X component: %s",
					 errormsg);
	}
#ifdef ENABLE_DSD
	pcm_export.Construct();
#endif

	try {
		osx_output_set_device(this);
	} catch (...) {
		AudioComponentInstanceDispose(au);
#ifdef ENABLE_DSD
		pcm_export.Destruct();
#endif
		throw;
	}
	
	if (hog_device)
		osx_output_hog_device(dev_id, true);
}

void
OSXOutput::Disable() noexcept
{
	AudioComponentInstanceDispose(au);
#ifdef ENABLE_DSD
	pcm_export.Destruct();
#endif

	if (hog_device)
		osx_output_hog_device(dev_id, false);
}

void
OSXOutput::Close() noexcept
{
	AudioOutputUnitStop(au);
	AudioUnitUninitialize(au);
	delete ring_buffer;
}

void
OSXOutput::Open(AudioFormat &audio_format)
{
	char errormsg[1024];
#ifdef ENABLE_DSD
	PcmExport::Params params;
	params.alsa_channel_order = true;
	bool dop = dop_setting;
	params.dop = false;
#endif

	memset(&asbd, 0, sizeof(asbd));
	asbd.mFormatID = kAudioFormatLinearPCM;
	asbd.mFormatFlags = kLinearPCMFormatFlagIsSignedInteger;

	switch (audio_format.format) {
	case SampleFormat::S8:
		asbd.mBitsPerChannel = 8;
		break;

	case SampleFormat::S16:
		asbd.mBitsPerChannel = 16;
		break;

	case SampleFormat::S24_P32:
		asbd.mBitsPerChannel = 24;
		break;

	case SampleFormat::S32:
		asbd.mBitsPerChannel = 32;
		break;

#ifdef ENABLE_DSD
	case SampleFormat::DSD:
		if(dop) {
			asbd.mBitsPerChannel = 24;
			params.dop = true;
			break;
		}
#endif

	default:
		audio_format.format = SampleFormat::S32;
		asbd.mBitsPerChannel = 32;
		break;
	}
#ifdef ENABLE_DSD
	asbd.mSampleRate = params.CalcOutputSampleRate(audio_format.sample_rate);
#endif

	if (IsBigEndian())
		asbd.mFormatFlags |= kLinearPCMFormatFlagIsBigEndian;

	asbd.mBytesPerPacket = audio_format.GetFrameSize();
#ifdef ENABLE_DSD
	if (audio_format.format == SampleFormat::DSD)
		asbd.mBytesPerPacket = 4 * audio_format.channels;
#endif
	asbd.mFramesPerPacket = 1;
	asbd.mBytesPerFrame = asbd.mBytesPerPacket;
	asbd.mChannelsPerFrame = audio_format.channels;

	Float64 sample_rate = osx_output_set_device_format(dev_id, audio_format);

#ifdef ENABLE_DSD
	if(params.dop && (sample_rate != asbd.mSampleRate)) { // fall back to PCM in case sample_rate cannot be synchronized
		params.dop = false;
		audio_format.format = SampleFormat::S32;
		asbd.mBitsPerChannel = 32;
		asbd.mBytesPerPacket = audio_format.GetFrameSize();
		asbd.mSampleRate = params.CalcOutputSampleRate(audio_format.sample_rate);
		asbd.mBytesPerFrame = asbd.mBytesPerPacket;
	}
#endif

	OSStatus status =
		AudioUnitSetProperty(au, kAudioUnitProperty_StreamFormat,
				     kAudioUnitScope_Input, 0,
				     &asbd,
				     sizeof(asbd));
	if (status != noErr)
		throw std::runtime_error("Unable to set format on OS X device");

	AURenderCallbackStruct callback;
	callback.inputProc = osx_render;
	callback.inputProcRefCon = this;

	status =
		AudioUnitSetProperty(au,
				     kAudioUnitProperty_SetRenderCallback,
				     kAudioUnitScope_Input, 0,
				     &callback, sizeof(callback));
	if (status != noErr) {
		AudioComponentInstanceDispose(au);
		throw std::runtime_error("Unable to set callback for OS X audio unit");
	}

	status = AudioUnitInitialize(au);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to initialize OS X audio unit: %s",
					 errormsg);
	}

	UInt32 buffer_frame_size = 1;
	status = osx_output_set_buffer_size(au, asbd, &buffer_frame_size);
	if (status != noErr) {
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to set frame size: %s",
					 errormsg);
	}
	pcm_export->Open(audio_format.format, audio_format.channels, params);

#ifdef ENABLE_DSD
	size_t ring_buffer_size = std::max<size_t>(buffer_frame_size,
						   MPD_OSX_BUFFER_TIME_MS * pcm_export->GetFrameSize(audio_format) * asbd.mSampleRate / 1000);
#else
	size_t ring_buffer_size = std::max<size_t>(buffer_frame_size,
-						   MPD_OSX_BUFFER_TIME_MS * audio_format.GetFrameSize() * audio_format.sample_rate / 1000);
#endif
	ring_buffer = new boost::lockfree::spsc_queue<uint8_t>(ring_buffer_size);

	status = AudioOutputUnitStart(au);
	if (status != 0) {
		AudioUnitUninitialize(au);
		osx_os_status_to_cstring(status, errormsg, sizeof(errormsg));
		throw FormatRuntimeError("Unable to start audio output: %s",
					 errormsg);
	}
	pause = false;
}

size_t
OSXOutput::Play(const void *chunk, size_t size)
{
	assert(size > 0);
	if(pause) {
		pause = false;
		OSStatus status = AudioOutputUnitStart(au);
		if (status != 0) {
			AudioUnitUninitialize(au);
			throw std::runtime_error("Unable to restart audio output after pause");
		}
	}
#ifdef ENABLE_DSD
	const auto e = pcm_export->Export({chunk, size});
	if (e.size == 0)
		/* the DoP (DSD over PCM) filter converts two frames
		   at a time and ignores the last odd frame; if there
		   was only one frame (e.g. the last frame in the
		   file), the result is empty; to avoid an endless
		   loop, bail out here, and pretend the one frame has
		   been played */
		return size;

	size_t bytes_written = ring_buffer->push((const uint8_t *)e.data,
											 e.size);
	return pcm_export->CalcSourceSize(bytes_written);
#endif
	return ring_buffer->push((const uint8_t *)chunk, size);
}

std::chrono::steady_clock::duration
OSXOutput::Delay() const noexcept
{
	// Idle if paused
	if(pause)
		return std::chrono::seconds(1);
	
	return ring_buffer->write_available()
		? std::chrono::steady_clock::duration::zero()
		: std::chrono::milliseconds(MPD_OSX_BUFFER_TIME_MS / 4);
}
	
bool OSXOutput::Pause() {
	if(!pause) {
		pause = true;
		AudioOutputUnitStop(au);
	}
	return true;
}
	
int
osx_output_get_volume(OSXOutput &output)
{
	return output.GetVolume();
}

void
osx_output_set_volume(OSXOutput &output, unsigned new_volume)
{
	return output.SetVolume(new_volume);
}

const struct AudioOutputPlugin osx_output_plugin = {
	"osx",
	osx_output_test_default_device,
	&OSXOutput::Create,
	&osx_mixer_plugin,
};