xref: /dports/audio/sbagen/sbagen-1.4.5/sbagen.c

//
//	SBaGen - Sequenced Binaural Beat Generator
//
//	(c) 1999-2011 Jim Peters <jim@uazu.net>.  All Rights Reserved.
//	For latest version see http://sbagen.sf.net/ or
//	http://uazu.net/sbagen/.  Released under the GNU GPL version 2.
//	Use at your own risk.
//
//	" 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, version 2.
//
//	  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. "
//
//	See the file COPYING for details of this license.
//
//	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
//
//	Some code fragments in the Win32 audio handling are based on
//	code from PLIB (c) 2001 by Steve Baker, originally released
//	under the LGPL (slDSP.cxx and sl.h).  For the original source,
//	see the PLIB project: http://plib.sf.net
//
//	The code for the Mac audio output was based on code from the
//	FINK project's patches to ESounD, by Shawn Hsiao and Masanori
//	Sekino.  See: http://fink.sf.net

#define VERSION "1.4.5"

// This should be built with one of the following target macros
// defined, which selects options for that platform, or else with some
// of the individual named flags #defined as listed later.
//
//  T_LINUX	To build the LINUX version with /dev/dsp support
//  T_MINGW	To build for Windows using MinGW and Win32 calls
//  T_MSVC	To build for Windows using MSVC and Win32 calls
//  T_MACOSX	To build for MacOSX using CoreAudio
//  T_POSIX	To build for simple file output on any Posix-compliant OS
//
// Ogg and MP3 support is handled separately from the T_* macros.

// Define OSS_AUDIO to use /dev/dsp for audio output
// Define WIN_AUDIO to use Win32 calls
// Define MAC_AUDIO to use Mac CoreAudio calls
// Define NO_AUDIO if no audio output device is usable
// Define UNIX_TIME to use UNIX calls for getting time
// Define WIN_TIME to use Win32 calls for getting time
// Define ANSI_TTY to use ANSI sequences to clear/redraw lines
// Define UNIX_MISC to use UNIX calls for various miscellaneous things
// Define WIN_MISC to use Windows calls for various miscellaneous things
// Define EXIT_KEY to require the user to hit RETURN before exiting after error

// Define OGG_DECODE to include OGG support code
// Define MP3_DECODE to include MP3 support code

#ifdef T_LINUX
#define OSS_AUDIO
#define UNIX_TIME
#define UNIX_MISC
#define ANSI_TTY
#endif

#ifdef T_MINGW
#define WIN_AUDIO
#define WIN_TIME
#define WIN_MISC
#define EXIT_KEY
#endif

#ifdef T_MSVC
#define WIN_AUDIO
#define WIN_TIME
#define WIN_MISC
#define EXIT_KEY
#endif

#ifdef T_MACOSX
#define MAC_AUDIO
#define UNIX_TIME
#define UNIX_MISC
#define ANSI_TTY
#endif

#ifdef T_POSIX
#define NO_AUDIO
#define UNIX_TIME
#define UNIX_MISC
#endif

// Make sure NO_AUDIO is set if necessary
#ifndef OSS_AUDIO
#ifndef MAC_AUDIO
#ifndef WIN_AUDIO
#define NO_AUDIO
#endif
#endif
#endif

// Make sure one of the _TIME macros is set
#ifndef UNIX_TIME
#ifndef WIN_TIME
#error UNIX_TIME or WIN_TIME not defined.  Maybe you did not define one of T_LINUX/T_MINGW/T_MACOSX/etc ?
#endif
#endif

// Make sure one of the _MISC macros is set
#ifndef UNIX_MISC
#ifndef WIN_MISC
#error UNIX_MISC or WIN_MISC not defined.  Maybe you did not define one of T_LINUX/T_MINGW/T_MACOSX/etc ?
#endif
#endif

#include <stdio.h>
#include <math.h>
#include <stdarg.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <ctype.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <time.h>

#ifdef T_MSVC
 #include <io.h>
 #define write _write
 #define vsnprintf _vsnprintf
 typedef long long S64;		// I have no idea if this is correct for MSVC
#else
 #include <unistd.h>
 #include <sys/time.h>
 typedef long long S64;
#endif

#ifdef T_MINGW
 #define vsnprintf _vsnprintf
#endif

#ifdef OSS_AUDIO
 #include <sys/soundcard.h>
 //WAS: #include <sys/soundcard.h>
#endif
#ifdef WIN_AUDIO
 #include <windows.h>
 #include <mmsystem.h>
#endif
#ifdef MAC_AUDIO
 #include <Carbon.h>
 #include <CoreAudio/CoreAudio.h>
#endif
#ifdef UNIX_TIME
 #include <sys/ioctl.h>
 #include <sys/times.h>
#endif
#ifdef UNIX_MISC
 #include <pthread.h>
#endif

typedef struct Channel Channel;
typedef struct Voice Voice;
typedef struct Period Period;
typedef struct NameDef NameDef;
typedef struct BlockDef BlockDef;
typedef unsigned char uchar;

int inbuf_loop(void *vp) ;
int inbuf_read(int *dst, int dlen) ;
void inbuf_start(int(*rout)(int*,int), int len) ;
inline int t_per24(int t0, int t1) ;
inline int t_per0(int t0, int t1) ;
inline int t_mid(int t0, int t1) ;
int main(int argc, char **argv) ;
void status(char *) ;
void dispCurrPer( FILE* ) ;
void init_sin_table() ;
void debug(char *fmt, ...) ;
void warn(char *fmt, ...) ;
void * Alloc(size_t len) ;
char * StrDup(char *str) ;
static inline int calcNow() ;
void loop() ;
void outChunk() ;
void corrVal(int ) ;
int readLine() ;
char * getWord() ;
void badSeq() ;
void readSeqImm(int ac, char **av) ;
void readSeq(int ac, char **av) ;
void readPreProg(int ac, char **av) ;
void correctPeriods();
void setup_device(void) ;
void readNameDef();
void readTimeLine();
int voicesEq(Voice *, Voice *);
void error(char *fmt, ...) ;
int sprintTime(char *, int);
int sprintVoice(char *, Voice *, Voice *);
int readTime(char *, int *);
void writeWAV();
void writeOut(char *, int);
void sinc_interpolate(double *, int, int *);
inline int userTime();
void find_wav_data_start(FILE *in);
int raw_mix_in(int *dst, int dlen);
int scanOptions(int *acp, char ***avp);
void handleOptions(char *p);
void setupOptC(char *spec) ;
extern int out_rate, out_rate_def;
void create_drop(int ac, char **av);
void create_slide(int ac, char **av);

#define ALLOC_ARR(cnt, type) ((type*)Alloc((cnt) * sizeof(type)))
#define uint unsigned int

#ifdef OGG_DECODE
#include "oggdec.c"
#endif
#ifdef MP3_DECODE
#include "mp3dec.c"
#endif

#ifdef WIN_AUDIO
void CALLBACK win32_audio_callback(HWAVEOUT, UINT, DWORD, DWORD, DWORD);
#endif
#ifdef MAC_AUDIO
OSStatus mac_callback(AudioDeviceID, const AudioTimeStamp *, const AudioBufferList *,
		      const AudioTimeStamp *, AudioBufferList *, const AudioTimeStamp *,
		      void *inClientData);
#endif

#define NL "\n"

void
help() {
   printf("SBaGen - Sequenced Binaural Beat Generator, version " VERSION
	  NL "Copyright (c) 1999-2011 Jim Peters, http://uazu.net/, all rights "
	  NL "  reserved, released under the GNU GPL v2.  See file COPYING."
	  NL
	  NL "Usage: sbagen [options] seq-file ..."
	  NL "       sbagen [options] -i tone-specs ..."
	  NL "       sbagen [options] -p pre-programmed-sequence-specs ..."
	  NL
	  NL "Options:  -h        Display this help-text"
	  NL "          -Q        Quiet - don't display running status"
	  NL "          -D        Display the full interpreted sequence instead of playing it"
	  NL "          -i        Immediate.  Take the remainder of the command line to be"
	  NL "                     tone-specifications, and play them continuously"
	  NL "          -p        Pre-programmed sequence.  Take the remainder of the command"
	  NL "                     line to be a type and arguments, e.g. \"drop 00ds+\""
	  NL "          -q mult   Quick.  Run through quickly (real time x 'mult') from the"
	  NL "                     start time, rather than wait for real time to pass"
	  NL
	  NL "          -r rate   Select the output rate (default is 44100 Hz, or from -m)"
#ifndef MAC_AUDIO
	  NL "          -b bits   Select the number bits for output (8 or 16, default 16)"
#endif
	  NL "          -L time   Select the length of time (hh:mm or hh:mm:ss) to output"
	  NL "                     for.  Default is to output forever."
	  NL "          -S        Output from the first tone-set in the sequence (Start),"
	  NL "                     instead of working in real-time.  Equivalent to '-q 1'."
	  NL "          -E        Output until the last tone-set in the sequence (End),"
	  NL "                     instead of outputting forever."
	  NL "          -T time   Start at the given clock-time (hh:mm)"
	  NL
	  NL "          -o file   Output raw data to the given file instead of /dev/dsp"
	  NL "          -O        Output raw data to the standard output"
	  NL "          -W        Output a WAV-format file instead of raw data"
	  NL "          -m file   Read audio data from the given file and mix it with the"
	  NL "                      generated binaural beats; may be "
#ifdef OGG_DECODE
	  "ogg/"
#endif
#ifdef MP3_DECODE
	  "mp3/"
#endif
	  "wav/raw format"
	  NL "          -M        Read raw audio data from the standard input and mix it"
	  NL "                      with the generated binaural beats (raw only)"
	  NL
	  NL "          -R rate   Select rate in Hz that frequency changes are recalculated"
	  NL "                     (for file/pipe output only, default is 10Hz)"
	  NL "          -F fms    Fade in/out time in ms (default 60000ms, or 1min)"
#ifdef OSS_AUDIO
	  NL "          -d dev    Select a different output device instead of /dev/dsp"
#endif
	  NL "          -c spec   Compensate for low-frequency headphone roll-off; see docs"
	  NL
	  );
   exit(0);
}

void
usage() {
  error("SBaGen - Sequenced Binaural Beat Generator, version " VERSION
	NL "Copyright (c) 1999-2011 Jim Peters, http://uazu.net/, all rights "
	NL "  reserved, released under the GNU GPL v2.  See file COPYING."
	NL
	NL "Usage: sbagen [options] seq-file ..."
	NL "       sbagen [options] -i tone-specs ..."
	NL "       sbagen [options] -p pre-programmed-sequence-specs ..."
	NL
	NL "For full usage help, type 'sbagen -h'.  For latest version see"
	NL "http://uazu.net/sbagen/ or http://sbagen.sf.net/"
#ifdef EXIT_KEY
	NL
	NL "Windows users please note that this utility is designed to be run as the"
	NL "associated application for SBG files.  This should have been set up for you by"
	NL "the installer.  You can run all the SBG files directly from the desktop by"
	NL "double-clicking on them, and edit them using NotePad from the right-click menu."
	NL "Alternatively, SBaGen may be run from the MS-DOS prompt (CMD on WinXP), or from"
	NL "BAT files.  SBaGen is powerful software -- it is worth the effort of figuring"
	NL "all this out.  See SBAGEN.TXT for the full documentation."
	NL
	NL "Editing the SBG files gives you access to the full tweakable power of SBaGen, "
	NL "but if you want a simple GUI interface to the most basic features, you could "
	NL "look at a user-contributed tool called SBaGUI:"
	NL
	NL "  http://sbagen.opensrc.org/wiki.php?page=SBaGUI"
#endif
	NL);
}


#define DEBUG_CHK_UTIME 0	// Check how much user time is being consumed
#define DEBUG_DUMP_WAVES 0	// Dump out wave tables (to plot with gnuplot)
#define DEBUG_DUMP_AMP 0	// Dump output amplitude to stdout per chunk
#define N_CH 16			// Number of channels

struct Voice {
  int typ;			// Voice type: 0 off, 1 binaural, 2 pink noise, 3 bell, 4 spin,
   				//   5 mix, 6 mixspin, 7 mixbeat, -1 to -100 wave00 to wave99
  double amp;			// Amplitude level (0-4096 for 0-100%)
  double carr;			// Carrier freq (for binaural/bell), width (for spin)
  double res;			// Resonance freq (-ve or +ve) (for binaural/spin)
};

struct Channel {
  Voice v;			// Current voice setting (updated from current period)
  int typ;			// Current type: 0 off, 1 binaural, 2 pink noise, 3 bell, 4 spin,
   				//   5 mix, 6 mixspin, 7 mixbeat, -1 to -100 wave00 to wave99
  int amp, amp2;		// Current state, according to current type
  int inc1, off1;		//  ::  (for binaural tones, offset + increment into sine
  int inc2, off2;		//  ::   table * 65536)
};

struct Period {
  Period *nxt, *prv;		// Next/prev in chain
  int tim;			// Start time (end time is ->nxt->tim)
  Voice v0[N_CH], v1[N_CH];	// Start and end voices
  int fi, fo;			// Temporary: Fade-in, fade-out modes
};

struct NameDef {
  NameDef *nxt;
  char *name;			// Name of definition
  BlockDef *blk;		// Non-zero for block definition
  Voice vv[N_CH];		// Voice-set for it (unless a block definition)
};

struct BlockDef {
  BlockDef *nxt;		// Next in chain
  char *lin;			// StrDup'd line
};

#define ST_AMP 0x7FFFF		// Amplitude of wave in sine-table
#define NS_ADJ 12		// Noise is generated internally with amplitude ST_AMP<<NS_ADJ
#define NS_DITHER 16		// How many bits right to shift the noise for dithering
#define NS_AMP (ST_AMP<<NS_ADJ)
#define ST_SIZ 16384		// Number of elements in sine-table (power of 2)
int *sin_table;
#define AMP_DA(pc) (40.96 * (pc))	// Display value (%age) to ->amp value
#define AMP_AD(amp) ((amp) / 40.96)	// Amplitude value to display %age
int *waves[100];		// Pointers are either 0 or point to a sin_table[]-style array of int

Channel chan[N_CH];		// Current channel states
int now;			// Current time (milliseconds from midnight)
Period *per= 0;			// Current period
NameDef *nlist;			// Full list of name definitions

int *tmp_buf;			// Temporary buffer for 20-bit mix values
short *out_buf;			// Output buffer
int out_bsiz;			// Output buffer size (bytes)
int out_blen;			// Output buffer length (samples) (1.0* or 0.5* out_bsiz)
int out_bps;			// Output bytes per sample (2 or 4)
int out_buf_ms;			// Time to output a buffer-ful in ms
int out_buf_lo;			// Time to output a buffer-ful, fine-tuning in ms/0x10000
int out_fd;			// Output file descriptor
int out_rate= 44100;		// Sample rate
int out_rate_def= 1;		// Sample rate is default value, not set by user
int out_mode= 1;		// Output mode: 0 unsigned char[2], 1 short[2], 2 swapped short[2]
int out_prate= 10;		// Rate of parameter change (for file and pipe output only)
int fade_int= 60000;		// Fade interval (ms)
FILE *in;			// Input sequence file
int in_lin;			// Current input line
char buf[4096];			// Buffer for current line
char buf_copy[4096];		// Used to keep unmodified copy of line
char *lin;			// Input line (uses buf[])
char *lin_copy;			// Copy of input line
double spin_carr_max;		// Maximum 'carrier' value for spin (really max width in us)

#define NS_BIT 10
int ns_tbl[1<<NS_BIT];
int ns_off= 0;

int fast_tim0= -1;		// First time mentioned in the sequence file (for -q and -S option)
int fast_tim1= -1;		// Last time mentioned in the sequence file (for -E option)
int fast_mult= 0;		// 0 to sync to clock (adjusting as necessary), or else sync to
				//  output rate, with the multiplier indicated
S64 byte_count= -1;		// Number of bytes left to output, or -1 if unlimited
int tty_erase;			// Chars to erase from current line (for ESC[K emulation)

int opt_D;
int opt_M;
int opt_Q;
int opt_S;
int opt_E;
int opt_W;
int opt_O;
int opt_L= -1;			// Length in ms, or -1
int opt_T= -1;			// Start time in ms, or -1
char *opt_o;			// File name to output to, or 0
char *opt_m;			// File name to read mix data from, or 0
char *opt_d= "/dev/dsp";	// Output device

FILE *mix_in;			// Input stream for mix sound data, or 0
int mix_cnt;			// Version number from mix filename (#<digits>), or -1
int bigendian;			// Is this platform Big-endian?
int mix_flag= 0;		// Has 'mix/*' been used in the sequence?

int opt_c;			// Number of -c option points provided (max 16)
struct AmpAdj {
   double freq, adj;
} ampadj[16];			// List of maximum 16 (freq,adj) pairs, freq-increasing order

char *pdir;			// Program directory (used as second place to look for -m files)

#ifdef WIN_AUDIO
 #define BUFFER_COUNT 8
 #define BUFFER_SIZE 8192*4
 HWAVEOUT aud_handle;
 WAVEHDR *aud_head[BUFFER_COUNT];
 int aud_current;		// Current header
 int aud_cnt;			// Number of headers in use
#endif

#ifdef MAC_AUDIO
 #define BUFFER_COUNT 8
 #define BUFFER_SIZE 4096*4
 char aud_buf[BUFFER_COUNT][BUFFER_SIZE];
 int aud_rd;	// Next buffer to read out of list (to send to device)
 int aud_wr;	// Next buffer to write.  aud_rd==aud_wr means empty buffer list
 static AudioDeviceID aud_dev;
#endif

//
//	Delay for a short period of time (in ms)
//

#ifdef UNIX_MISC
void
delay(int ms) {
   struct timespec ts;
   ts.tv_sec= ms / 1000;
   ts.tv_nsec= (ms % 1000) * 1000000;
   nanosleep(&ts, 0);
}
#endif
#ifdef WIN_MISC
void
delay(int ms) {
   Sleep(ms);
}
#endif


//
//	WAV/OGG/MP3 input data buffering
//

int *inbuf;		// Buffer for input data (as 20-bit samples)
int ib_len;		// Length of input buffer (in ints)
volatile int ib_rd;	// Read-offset in inbuf
volatile int ib_wr;	// Write-offset in inbuf
volatile int ib_eof;	// End of file flag
int ib_cycle= 100;	// Time in ms for a complete loop through the buffer
int (*ib_read)(int*,int);  // Routine to refill buffer

int
inbuf_loop(void *vp) {
   int now= -1;
   int waited= 0;	// Used to bail out if the main thread dies for some reason
   int a;

   while (1) {
      int rv;
      int rd= ib_rd;
      int wr= ib_wr;
      int cnt= (rd-1-wr) & (ib_len-1);
      if (cnt > ib_len-wr) cnt= ib_len-wr;
      if (cnt > ib_len/8) cnt= ib_len/8;

      // Choose to only work in ib_len/8 units, although this is not
      // 100% necessary
      if (cnt < ib_len/8) {
	 // Wait a little while for the buffer to empty (minimum 1ms)
	 if (waited > 10000 + ib_cycle)
	    error("Mix stream halted for more than 10 seconds; aborting");
	 delay(a= 1+ib_cycle/4);
	 waited += a;
	 continue;
      }
      waited= 0;

      rv= ib_read(inbuf+wr, cnt);
      //debug("ib_read %d-%d (%d) -> %d", wr, wr+cnt-1, cnt, rv);
      if (rv != cnt) {
	 ib_eof= 1;
	 return 0;
      }

      ib_wr= (wr + rv) & (ib_len-1);

      // Whenever we roll over, recalculate 'ib_cycle'
      if (ib_wr < wr) {
	 int prev= now;
	 now= calcNow();
	 if (prev >= 0 && now > prev)
	    ib_cycle= now - prev;
	 //debug("Input buffer cycle duration is now %dms", ib_cycle);
      }
   }
   return 0;
}

//
//	Read a chunk of int data from the input buffer.  This will
//	always return enough data unless we have hit the end of the
//	file, in which case it returns a lower number or 0.  If not
//	enough data has been read by the input thread, then this
//	thread pauses until data is ready -- but this should hopefully
//	never happen.
//

int
inbuf_read(int *dst, int dlen) {
   int rv= 0;
   int waited= 0;	// As a precaution, bail out if other thread hangs for some reason
   int a;

   while (dlen > 0) {
      int rd= ib_rd;
      int wr= ib_wr;
      int avail= (wr-rd) & (ib_len-1);
      int toend= ib_len-rd;
      if (avail > toend) avail= toend;
      if (avail > dlen) avail= dlen;

      if (avail == 0) {
	 if (ib_eof) return rv;

	 // Necessary to wait for incoming mix data.  This should
	 // never happen in normal running, though, unless we are
	 // outputting to a file
	 if (waited > 10000)
	    error("Mix stream problem; waited more than 10 seconds for data; aborting");
	 //debug("Waiting for input thread (%d)", ib_eof);
	 delay(a= ib_cycle/4 > 100 ? 100 : 1+ib_cycle/4);
	 waited += a;
	 continue;
      }
      waited= 0;

      memcpy(dst, inbuf+rd, avail * sizeof(int));
      dst += avail;
      dlen -= avail;
      rv += avail;
      ib_rd= (rd + avail) & (ib_len-1);
   }
   return rv;
}

//
//	Start off the thread that fills the buffer
//

void
inbuf_start(int(*rout)(int*,int), int len) {
   if (0 != (len & (len-1)))
      error("inbuf_start() called with length not a power of two");

   ib_read= rout;
   ib_len= len;
   inbuf= ALLOC_ARR(ib_len, int);
   ib_rd= 0;
   ib_wr= 0;
   ib_eof= 0;
   if (!opt_Q) warn("Initialising %d-sample buffer for mix stream", ib_len/2);

   // Preload 75% of the buffer -- or at least attempt to do so;
   // errors/eof/etc will be picked up in the inbuf_loop() routine
   ib_wr= ib_read(inbuf, ib_len*3/4);

   // Start the thread off
#ifdef UNIX_MISC
   {
      pthread_t thread;
      if (0 != pthread_create(&thread, NULL, (void*)&inbuf_loop, NULL))
	 error("Failed to start input buffering thread");
   }
#endif
#ifdef WIN_MISC
   {
      DWORD tmp;
      if (0 == CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE)&inbuf_loop, 0, 0, &tmp))
	 error("Failed to start input buffering thread");
   }
#endif
}


//
//	Time-keeping functions
//

#define H24 (86400000)			// 24 hours
#define H12 (43200000)			// 12 hours

inline int t_per24(int t0, int t1) {		// Length of period starting at t0, ending at t1.
  int td= t1 - t0;				// NB for t0==t1 this gives 24 hours, *NOT 0*
  return td > 0 ? td : td + H24;
}
inline int t_per0(int t0, int t1) {		// Length of period starting at t0, ending at t1.
  int td= t1 - t0;				// NB for t0==t1 this gives 0 hours
  return td >= 0 ? td : td + H24;
}
inline int t_mid(int t0, int t1) {		// Midpoint of period from t0 to t1
  return ((t1 < t0) ? (H24 + t0 + t1) / 2 : (t0 + t1) / 2) % H24;
}

//
//	M A I N
//

int
main(int argc, char **argv) {
   short test= 0x1100;
   int rv;
   char *p;

   pdir= StrDup(argv[0]);
   p= strchr(pdir, 0);
   while (p > pdir && p[-1] != '/' && p[-1] != '\\') *--p= 0;

   argc--; argv++;
   init_sin_table();
   bigendian= ((char*)&test)[0] != 0;

   // Process all the options
   rv= scanOptions(&argc, &argv);

   if (argc < 1) usage();

   if (rv == 'i') {
      // Immediate mode
      readSeqImm(argc, argv);
   } else if (rv == 'p') {
      // Pre-programmed sequence
      readPreProg(argc, argv);
   } else {
      // Sequenced mode -- sequence may include options, so options
      // are not settled until below this point
      if (argc < 1) usage();
      readSeq(argc, argv);
   }

   if (opt_W && !opt_o && !opt_O)
      error("Use -o or -O with the -W option");
   if (opt_W && opt_L < 0 && !opt_E) {
      fprintf(stderr, "*** The length has not been specified for the -W option; assuming 1 hour ***\n");
      fprintf(stderr, "(Use -L or -E with the -W option to control the length of the WAV file)\n\n");
      opt_L= 60*60*1000;
   }

   mix_in= 0;
   if (opt_M || opt_m) {
      char *p;
      char tmp[4];
      int raw= 1;
      if (opt_M) {
	 mix_in= stdin;
	 tmp[0]= 0;
      }
      if (opt_m) {
	 // Pick up #<digits> on end of filename
	 p= strchr(opt_m, 0);
	 mix_cnt= -1;
	 if (p > opt_m && isdigit(p[-1])) {
	    mix_cnt= 0;
	    while (p > opt_m && isdigit(p[-1]))
	       mix_cnt= mix_cnt * 10 + *--p - '0';
	    if (p > opt_m && p[-1] == '#')
	       *--p= 0;
	    else {
	       p= strchr(opt_m, 0);
	       mix_cnt= -1;
	    }
	 }
	 // p points to end of filename (NUL)

	 // Open file
	 mix_in= fopen(opt_m, "rb");
	 if (!mix_in && opt_m[0] != '/') {
	    int len= strlen(opt_m) + strlen(pdir) + 1;
	    char *tmp= ALLOC_ARR(len, char);
	    strcpy(tmp, pdir);
	    strcat(tmp, opt_m);
	    mix_in= fopen(tmp, "rb");
	    free(tmp);
	 }
	 if (!mix_in)
	    error("Can't open -m option mix input file: %s", opt_m);

	 // Pick up extension
	 if (p-opt_m >= 4 && p[-4] == '.') {
	    tmp[0]= tolower(p[-3]);
	    tmp[1]= tolower(p[-2]);
	    tmp[2]= tolower(p[-1]);
	    tmp[3]= 0;
	 }
      }
      if (0 == strcmp(tmp, "wav"))	// Skip header on WAV files
	 find_wav_data_start(mix_in);
      if (0 == strcmp(tmp, "ogg")) {
#ifdef OGG_DECODE
	 ogg_init(); raw= 0;
#else
	 error("Sorry: Ogg support wasn't compiled into this executable");
#endif
      }
      if (0 == strcmp(tmp, "mp3")) {
#ifdef MP3_DECODE
	 mp3_init(); raw= 0;
#else
	 error("Sorry: MP3 support wasn't compiled into this executable");
#endif
      }
      // If this is a raw/wav data stream, setup a 256*1024-int
      // buffer (3s@44.1kHz)
      if (raw) inbuf_start(raw_mix_in, 256*1024);
   }

   loop();
   return 0;
}

//
//	Scan options.  Returns a flag indicating what is expected to
//	interpret the rest of the arguments: 0 normal, 'i' immediate
//	(-i option), 'p' -p option.
//

int
scanOptions(int *acp, char ***avp) {
   int argc= *acp;
   char **argv= *avp;
   int val;
   char dmy;
   int rv= 0;

   // Scan options
   while (argc > 0 && argv[0][0] == '-' && argv[0][1]) {
      char opt, *p= 1 + *argv++; argc--;
      while ((opt= *p++)) {
	 // Check options that are available on both
	 switch (opt) {
	  case 'Q': opt_Q= 1; break;
	  case 'E': opt_E= 1; break;
	  case 'm':
	     if (argc-- < 1) error("-m option expects filename");
	     // Earliest takes precedence, so command-line overrides sequence file
	     if (!opt_m) opt_m= *argv++;
	     break;
	  case 'S': opt_S= 1;
	     if (!fast_mult) fast_mult= 1; 		// Don't try to sync with real time
	     break;
	  case 'L':
	     if (argc-- < 1 || 0 == (val= readTime(*argv, &opt_L)) ||
		 1 == sscanf(*argv++ + val, " %c", &dmy))
		error("-L expects hh:mm or hh:mm:ss time");
	     break;
	  case 'T':
	     if (argc-- < 1 || 0 == (val= readTime(*argv, &opt_T)) ||
		 1 == sscanf(*argv++ + val, " %c", &dmy))
		error("-T expects hh:mm or hh:mm:ss time");
	     if (!fast_mult) fast_mult= 1;		// Don't try to sync with real time
	     break;
	  case 'F':
	     if (argc-- < 1 || 1 != sscanf(*argv++, "%d %c", &fade_int, &dmy))
		error("-F expects fade-time in ms");
	     break;
	  case 'c':
	     if (argc-- < 1) error("-c expects argument");
	     setupOptC(*argv++);
	     break;
	  case 'i': rv= 'i'; break;
	  case 'p': rv= 'p'; break;
	  case 'h': help(); break;
	  case 'D': opt_D= 1; break;
	  case 'M': opt_M= 1; break;
	  case 'O': opt_O= 1;
	     if (!fast_mult) fast_mult= 1; 		// Don't try to sync with real time
	     break;
	  case 'W': opt_W= 1;
	     if (!fast_mult) fast_mult= 1; 		// Don't try to sync with real time
	     break;
	  case 'q':
	     opt_S= 1;
	     if (argc-- < 1 || 1 != sscanf(*argv++, "%d %c", &fast_mult, &dmy))
		error("Expecting an integer after -q");
	     if (fast_mult < 1) fast_mult= 1;
	     break;
	  case 'r':
	     if (argc-- < 1 || 1 != sscanf(*argv++, "%d %c", &out_rate, &dmy))
		error("Expecting an integer after -r");
	     out_rate_def= 0;
	     break;
#ifndef MAC_AUDIO
	  case 'b':
	     if (argc-- < 1 ||
		 1 != sscanf(*argv++, "%d %c", &val, &dmy) ||
		 !(val == 8 || val == 16))
		error("Expecting -b 8 or -b 16");
	     out_mode= (val == 8) ? 0 : 1;
	     break;
#endif
	  case 'o':
	     if (argc-- < 1) error("Expecting filename after -o");
	     opt_o= *argv++;
	     if (!fast_mult) fast_mult= 1;		// Don't try to sync with real time
	     break;
#ifdef OSS_AUDIO
	  case 'd':
	     if (argc-- < 1) error("Expecting device filename after -d");
	     opt_d= *argv++;
	     break;
#endif
	  case 'R':
	     if (argc-- < 1 || 1 != sscanf(*argv++, "%d %c", &out_prate, &dmy))
		error("Expecting integer after -R");
	     break;
	  default:
	     error("Option -%c not known; run 'sbagen -h' for help", opt);
	 }
      }
   }

   *acp= argc;
   *avp= argv;
   return rv;
}

//
//	Handle an option string, breaking it into an (argc/argv) list
//	for scanOptions.
//

void
handleOptions(char *str0) {
   // Always StrDup() string and don't bother to free(), as normal
   // argv[] strings stick around for the life of the program
   char *str= StrDup(str0);
   int const max_argc= 32;
   char *argv[max_argc+1];
   int argc= 0;

   while (*str) {
      if (argc >= max_argc)
	 error("Too many options at line: %d\n  %s", in_lin, lin_copy);
      argv[argc++]= str;
      while (*str && !isspace(*str)) str++;
      if (!*str) continue;
      *str++= 0;	// NUL-term this word
      while (isspace(*str)) str++;
   }
   argv[argc]= 0;	// Terminate argv list with a NULL

   // Process the options
   {
      char **av= argv;
      int ac= argc;
      int rv;

      rv= scanOptions(&ac, &av);

      if (rv == 'i') {
	 // Immediate mode
	 readSeqImm(ac, av);
      } else if (rv == 'p') {
	 // Pre-programmed sequence
	 readPreProg(ac, av);
      } else if (ac)
	 error("Trailing garbage after options at line: %d\n  %s", in_lin, lin_copy);
   }
}

//
//	Setup the ampadj[] array from the given -c spec-string
//

void
setupOptC(char *spec) {
   char *p= spec, *q;
   int a, b;

   while (1) {
      while (isspace(*p) || *p == ',') p++;
      if (!*p) break;

      if (opt_c >= sizeof(ampadj) / sizeof(ampadj[0]))
	 error("Too many -c option frequencies; maxmimum is %d",
	       sizeof(ampadj) / sizeof(ampadj[0]));

      ampadj[opt_c].freq= strtod(p, &q);
      if (p == q) goto bad;
      if (*q++ != '=') goto bad;
      ampadj[opt_c].adj= strtod(q, &p);
      if (p == q) goto bad;
      opt_c++;
   }

   // Sort the list
   for (a= 0; a<opt_c; a++)
      for (b= a+1; b<opt_c; b++)
	 if (ampadj[a].freq > ampadj[b].freq) {
	    double tmp;
	    tmp= ampadj[a].freq; ampadj[a].freq= ampadj[b].freq; ampadj[b].freq= tmp;
	    tmp= ampadj[a].adj; ampadj[a].adj= ampadj[b].adj; ampadj[b].adj= tmp;
	 }
   return;

 bad:
   error("Bad -c option spec; expecting <freq>=<amp>[,<freq>=<amp>]...:\n  %s", spec);
}


//
//	If this is a WAV file we've been given, skip forward to the
//	'data' section.  Don't bother checking any of the 'fmt '
//	stuff.  If they didn't give us a valid 16-bit stereo file at
//	the right rate, then tough!
//

void
find_wav_data_start(FILE *in) {
   unsigned char buf[16];

   if (1 != fread(buf, 12, 1, in)) goto bad;
   if (0 != memcmp(buf, "RIFF", 4)) goto bad;
   if (0 != memcmp(buf+8, "WAVE", 4)) goto bad;

   while (1) {
      int len;
      if (1 != fread(buf, 8, 1, in)) goto bad;
      if (0 == memcmp(buf, "data", 4)) return;		// We're in the right place!
      len= buf[4] + (buf[5]<<8) + (buf[6]<<16) + (buf[7]<<24);
      if (len & 1) len++;
      if (out_rate_def && 0 == memcmp(buf, "fmt ", 4)) {
	 // Grab the sample rate to use as the default if available
	 if (1 != fread(buf, 8, 1, in)) goto bad;
	 len -= 8;
	 out_rate= buf[4] + (buf[5]<<8) + (buf[6]<<16) + (buf[7]<<24);
	 out_rate_def= 0;
      }
      if (0 != fseek(in, len, SEEK_CUR)) goto bad;
   }

 bad:
   warn("WARNING: Not a valid WAV file, treating as RAW");
   rewind(in);
}

//
//	Input raw audio data from the 'mix_in' stream, and convert to
//	32-bit values (max 'dlen')
//

int
raw_mix_in(int *dst, int dlen) {
   short *tmp= (short*)(dst + dlen/2);
   int a, rv;

   rv= fread(tmp, 2, dlen, mix_in);
   if (rv == 0) {
      if (feof(mix_in))
	 return 0;
      error("Read error on mix input:\n  %s", strerror(errno));
   }

   // Now convert 16-bit little-endian input data into 20-bit native
   // int values
   if (bigendian) {
      char *rd= (char*)tmp;
      for (a= 0; a<rv; a++) {
	 *dst++= ((rd[0]&255) + (rd[1]<<8)) << 4;
	 rd += 2;
      }
   } else {
      for (a= 0; a<rv; a++)
	 *dst++= *tmp++ << 4;
   }

   return rv;
}



//
//	Update a status line
//

void
status(char *err) {
  int a;
  int nch= N_CH;
  char *p= buf, *p0, *p1;

  if (opt_Q) return;

#ifdef ANSI_TTY
  if (tty_erase) p += sprintf(p, "\033[K");
#endif

  p0= p;		// Start of line
  *p++= ' '; *p++= ' ';
  p += sprintTime(p, now);
  while (nch > 1 && chan[nch-1].v.typ == 0) nch--;
  for (a= 0; a<nch; a++)
    p += sprintVoice(p, &chan[a].v, 0);
  if (err) p += sprintf(p, " %s", err);
  p1= p;		// End of line

#ifndef ANSI_TTY
  // Truncate line to 79 characters on Windows
  if (p1-p0 > 79) {
    p1 = p0 + 76;
    p1 += sprintf(p1, "...");
  }
#endif

#ifndef ANSI_TTY
  while (tty_erase > p-p0) *p++= ' ';
#endif

  tty_erase= p1-p0;		// Characters that will need erasing
  fprintf(stderr, "%s\r", buf);
  fflush(stderr);
}

void 				// Display current period details
dispCurrPer(FILE *fp) {
  int a;
  Voice *v0, *v1;
  char *p0, *p1;
  int len0, len1;
  int nch= N_CH;

  if (opt_Q) return;

  p0= buf;
  p1= buf_copy;

  p0 += sprintf(p0, "* ");
  p0 += sprintTime(p0, per->tim);
  p1 += sprintf(p1, "  ");
  p1 += sprintTime(p1, per->nxt->tim);

  v0= per->v0; v1= per->v1;
  while (nch > 1 && v0[nch-1].typ == 0) nch--;
  for (a= 0; a<nch; a++, v0++, v1++) {
    p0 += len0= sprintVoice(p0, v0, 0);
    p1 += len1= sprintVoice(p1, v1, v0);
    while (len0 < len1) { *p0++= ' '; len0++; }
    while (len1 < len0) { *p1++= ' '; len1++; }
  }
  *p0= 0; *p1= 0;
  fprintf(fp, "%s\n%s\n", buf, buf_copy);
  fflush(fp);
}

int
sprintTime(char *p, int tim) {
  return sprintf(p, "%02d:%02d:%02d",
		 tim % 86400000 / 3600000,
		 tim % 3600000 / 60000,
		 tim % 60000 / 1000);
}

int
sprintVoice(char *p, Voice *vp, Voice *dup) {
   switch (vp->typ) {
    case 0:
       return sprintf(p, " -");
    case 1:
       if (dup && vp->carr == dup->carr && vp->res == dup->res && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " %.2f%+.2f/%.2f", vp->carr, vp->res, AMP_AD(vp->amp));
    case 2:
       if (dup && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " pink/%.2f", AMP_AD(vp->amp));
    case 3:
       if (dup && vp->carr == dup->carr && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " bell%+.2f/%.2f", vp->carr, AMP_AD(vp->amp));
    case 4:
       if (dup && vp->carr == dup->carr && vp->res == dup->res && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " spin:%.2f%+.2f/%.2f", vp->carr, vp->res, AMP_AD(vp->amp));
    case 5:
       if (dup && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " mix/%.2f", AMP_AD(vp->amp));
    default:
       if (vp->typ < -100 || vp->typ > -1)
	  return sprintf(p, " ERROR");
       if (dup && vp->typ == dup->typ &&
	   vp->carr == dup->carr && vp->res == dup->res && vp->amp == dup->amp)
	  return sprintf(p, "  ::");
       return sprintf(p, " wave%02d:%.2f%+.2f/%.2f", -1-vp->typ, vp->carr, vp->res, AMP_AD(vp->amp));
   }
}

void
init_sin_table() {
  int a;
  int *arr= (int*)Alloc(ST_SIZ * sizeof(int));

  for (a= 0; a<ST_SIZ; a++)
    arr[a]= (int)(ST_AMP * sin((a * 3.14159265358979323846 * 2) / ST_SIZ));

  sin_table= arr;
}

void
error(char *fmt, ...) {
  va_list ap; va_start(ap, fmt);
  vfprintf(stderr, fmt, ap);
  fprintf(stderr, "\n");
#ifdef EXIT_KEY
  fprintf(stderr, "Press <RETURN> to continue: ");
  fflush(stderr);
  getchar();
#endif
  exit(1);
}

void
debug(char *fmt, ...) {
  va_list ap; va_start(ap, fmt);
  vfprintf(stderr, fmt, ap);
  fprintf(stderr, "\n");
}

void
warn(char *fmt, ...) {
  va_list ap; va_start(ap, fmt);
  vfprintf(stderr, fmt, ap);
  fprintf(stderr, "\n");
}

void *
Alloc(size_t len) {
  void *p= calloc(1, len);
  if (!p) error("Out of memory");
  return p;
}

char *
StrDup(char *str) {
  char *rv= strdup(str);
  if (!rv) error("Out of memory");
  return rv;
}

#ifdef UNIX_TIME
// Precalculate a reference timestamp to accelerate calcNow().  This
// can be any recent time.  We recalculate it every 10 minutes.  The
// only reason for doing this is to cope with clocks going forwards or
// backwards when entering or leaving summer time so that people wake
// up on time on these two dates; an hour of the sequence will be
// repeated or skipped.  The 'time_ref*' variables will be initialised
// on the first call to calcNow().

static int time_ref_epoch= 0;	// Reference time compared to UNIX epoch
static int time_ref_ms;		// Reference time in sbagen 24-hour milliseconds

void
setupRefTime() {
  struct tm *tt;
  time_t tim= time(0);
  tt= localtime(&tim);
  time_ref_epoch= tim;
  time_ref_ms= 1000*tt->tm_sec + 60000*tt->tm_min + 3600000*tt->tm_hour;
}

static inline int
calcNow() {
  struct timeval tv;
  if (0 != gettimeofday(&tv, 0)) error("Can't get current time");
  if (tv.tv_sec - time_ref_epoch > 600) setupRefTime();
  return (time_ref_ms + (tv.tv_sec - time_ref_epoch) * 1000 + tv.tv_usec / 1000) % H24;
}
#endif

#ifdef WIN_TIME
inline int
calcNow() {
  SYSTEMTIME st;
  GetLocalTime(&st);
  return st.wMilliseconds + 1000*st.wSecond + 60000*st.wMinute + 3600000*st.wHour;
}
#endif

#if DEBUG_CHK_UTIME
inline int
userTime() {
  struct tms buf;
  times(&buf);
  return buf.tms_utime;
}
#else
// Dummy to avoid complaints on MSVC
int userTime() { return 0; }
#endif

//
//	Simple random number generator.  Generates a repeating
//	sequence of 65536 odd numbers in the range -65535->65535.
//
//	Based on ZX Spectrum random number generator:
//	  seed= (seed+1) * 75 % 65537 - 1
//

#define RAND_MULT 75

static int seed= 2;

//inline int qrand() {
//  return (seed= seed * 75 % 131074) - 65535;
//}

//
//	Generate next sample for simulated pink noise, with same
//	scaling as the sin_table[].  This version uses an inlined
//	random number generator, and smooths the lower frequency bands
//	as well.
//

#define NS_BANDS 9
typedef struct Noise Noise;
struct Noise {
  int val;		// Current output value
  int inc;		// Increment
};
Noise ntbl[NS_BANDS];
int nt_off;
int noise_buf[256];
uchar noise_off= 0;

static inline int
noise2() {
  int tot;
  int off= nt_off++;
  int cnt= 1;
  Noise *ns= ntbl;
  Noise *ns1= ntbl + NS_BANDS;

  tot= ((seed= seed * RAND_MULT % 131074) - 65535) * (NS_AMP / 65535 / (NS_BANDS + 1));

  while ((cnt & off) && ns < ns1) {
    int val= ((seed= seed * RAND_MULT % 131074) - 65535) * (NS_AMP / 65535 / (NS_BANDS + 1));
    tot += ns->val += ns->inc= (val - ns->val) / (cnt += cnt);
    ns++;
  }

  while (ns < ns1) {
    tot += (ns->val += ns->inc);
    ns++;
  }

  return noise_buf[noise_off++]= (tot >> NS_ADJ);
}

//	//
//	//	Generate next sample for simulated pink noise, scaled the same
//	//	as the sin_table[].  This version uses a library random number
//	//	generator, and no smoothing.
//	//
//
//	inline double
//	noise() {
//	  int tot= 0;
//	  int bit= ~0;
//	  int a;
//	  int off;
//
//	  ns_tbl[ns_off]= (rand() - (RAND_MAX / 2)) / (NS_BIT + 1);
//	  off= ns_off;
//	  for (a= 0; a<=NS_BIT; a++, bit <<= 1) {
//	    off &= bit;
//	    tot += ns_tbl[off];
//	  }
//	  ns_off= (ns_off + 1) & ((1<<NS_BIT) - 1);
//
//	  return tot * (ST_AMP / (RAND_MAX * 0.5));
//	}

//
//	Play loop
//

void
loop() {
  int c, cnt;
  int err;		// Error to add to 'now' until next cnt==0
  int fast= fast_mult != 0;
  int vfast= fast_mult > 20;		// Very fast - update status line often
  int utime= 0;
  int now_lo= 0;			// Low-order 16 bits of 'now' (fractional)
  int err_lo= 0;
  int ms_inc;

  setup_device();
  spin_carr_max= 127.0 / 1E-6 / out_rate;
  cnt= 1 + 1999 / out_buf_ms;	// Update every 2 seconds or so
  now= opt_S ? fast_tim0 : calcNow();
  if (opt_T != -1) now= opt_T;
  err= fast ? out_buf_ms * (fast_mult - 1) : 0;
  if (opt_L)
    byte_count= out_bps * (S64)(opt_L * 0.001 * out_rate);
  if (opt_E)
    byte_count= out_bps * (S64)(t_per0(now, fast_tim1) * 0.001 * out_rate /
				(fast ? fast_mult : 1));

  // Do byte-swapping if bigendian and outputting to a file or stream
  if ((opt_O || opt_o) &&
      out_mode == 1 && bigendian)
     out_mode= 2;

  if (opt_W)
    writeWAV();

  if (!opt_Q) fprintf(stderr, "\n");
  corrVal(0);		// Get into correct period
  dispCurrPer(stderr);	// Display
  status(0);

  while (1) {
    for (c= 0; c < cnt; c++) {
      corrVal(1);
      outChunk();
      ms_inc= out_buf_ms + err;
      now_lo += out_buf_lo + err_lo;
      if (now_lo >= 0x10000) { ms_inc += now_lo >> 16; now_lo &= 0xFFFF; }
      now += ms_inc;
      if (now > H24) now -= H24;
      if (vfast && (c&1)) status(0);
    }

    if (fast) {
      if (!vfast) status(0);
    }
    else {
      // Synchronize with real clock, gently over the next second or so
      char buf[32];
      int diff= calcNow() - now;
      if (abs(diff) > H12) diff= 0;
      sprintf(buf, "(%d)", diff);

      err_lo= diff * 0x10000 / cnt;
      err= err_lo >> 16;
      err_lo &= 0xFFFF;

      if (DEBUG_CHK_UTIME) {
	int prev= utime;
	utime= userTime();
	sprintf(buf, "%d ticks", utime-prev);		// Replaces standard message
      }
      status(buf);
    }
  }
}


//
//	Output a chunk of sound (a buffer-ful), then return
//
//	Note: Optimised for 16-bit output.  Eight-bit output is
//	slower, but then it probably won't have to run at as high a
//	sample rate.
//

int rand0, rand1;

void
outChunk() {
   int off= 0;

   if (mix_in) {
      int rv= inbuf_read(tmp_buf, out_blen);
      if (rv == 0) {
	 if (!opt_Q) warn("\nEnd of mix input audio stream");
	 exit(0);
      }
      while (rv < out_blen) tmp_buf[rv++]= 0;
   }

   while (off < out_blen) {
      int ns= noise2();		// Use same pink noise source for everything
      int tot1, tot2;		// Left and right channels
      int mix1, mix2;		// Incoming mix signals
      int val, a;
      Channel *ch;
      int *tab;

      mix1= tmp_buf[off];
      mix2= tmp_buf[off+1];

      // Do default mixing at 100% if no mix/* stuff is present
      if (!mix_flag) {
	 tot1= mix1 << 12;
	 tot2= mix2 << 12;
      } else {
	 tot1= tot2= 0;
      }

      ch= &chan[0];
      for (a= 0; a<N_CH; a++, ch++) switch (ch->typ) {
       case 0:
	  break;
       case 1:	// Binaural tones
	  ch->off1 += ch->inc1;
	  ch->off1 &= (ST_SIZ << 16) - 1;
	  tot1 += ch->amp * sin_table[ch->off1 >> 16];
	  ch->off2 += ch->inc2;
	  ch->off2 &= (ST_SIZ << 16) - 1;
	  tot2 += ch->amp2 * sin_table[ch->off2 >> 16];
	  break;
       case 2:	// Pink noise
	  val= ns * ch->amp;
	  tot1 += val;
	  tot2 += val;
	  break;
       case 3:	// Bell
	  if (ch->off2) {
	     ch->off1 += ch->inc1;
	     ch->off1 &= (ST_SIZ << 16) - 1;
	     val= ch->off2 * sin_table[ch->off1 >> 16];
	     tot1 += val; tot2 += val;
	     if (--ch->inc2 < 0) {
		ch->inc2= out_rate/20;
		ch->off2 -= 1 + ch->off2 / 12;	// Knock off 10% each 50 ms
	     }
	  }
	  break;
       case 4:	// Spinning pink noise
	  ch->off1 += ch->inc1;
	  ch->off1 &= (ST_SIZ << 16) - 1;
	  val= (ch->inc2 * sin_table[ch->off1 >> 16]) >> 24;
	  tot1 += ch->amp * noise_buf[(uchar)(noise_off+128+val)];
	  tot2 += ch->amp * noise_buf[(uchar)(noise_off+128-val)];
	  break;
       case 5:	// Mix level
	  tot1 += mix1 * ch->amp;
	  tot2 += mix2 * ch->amp;
	  break;
       default:	// Waveform-based binaural tones
	  tab= waves[-1 - ch->typ];
	  ch->off1 += ch->inc1;
	  ch->off1 &= (ST_SIZ << 16) - 1;
	  tot1 += ch->amp * tab[ch->off1 >> 16];
	  ch->off2 += ch->inc2;
	  ch->off2 &= (ST_SIZ << 16) - 1;
	  tot2 += ch->amp * tab[ch->off2 >> 16];
	  break;
      }


      // // Add pink noise as dithering
      // tot1 += (ns >> NS_DITHER) + 0x8000;
      // tot2 += (ns >> NS_DITHER) + 0x8000;

      // // Add white noise as dithering
      // tot1 += (seed >> 1) + 0x8000;
      // tot2 += (seed >> 1) + 0x8000;

      // White noise dither; you could also try (rand0-rand1) for a
      // dither with more high frequencies
      rand0= rand1;
      rand1= (rand0 * 0x660D + 0xF35F) & 0xFFFF;
      if (tot1 <= 0x7FFF0000) tot1 += rand0;
      if (tot2 <= 0x7FFF0000) tot2 += rand0;

      out_buf[off++]= tot1 >> 16;
      out_buf[off++]= tot2 >> 16;
  }

  // Generate debugging amplitude output
  if (DEBUG_DUMP_AMP) {
    short *sp= out_buf;
    short *end= out_buf + out_blen;
    int max= 0;
    while (sp < end) {
       int val= (int)sp[0] + (int)sp[1]; sp += 2;
       if (val < 0) val= -val;
       if (val > max) max= val;
    }
    max /= 328;
    while (max-- > 0) putc('#', stdout);
    printf("\n"); fflush(stdout);
  }

  // Rewrite buffer for 8-bit mode
  if (out_mode == 0) {
    short *sp= out_buf;
    short *end= out_buf + out_blen;
    char *cp= (char*)out_buf;
    while (sp < end) *cp++= (*sp++ >> 8) + 128;
  }

  // Rewrite buffer for 16-bit byte-swapping
  if (out_mode == 2) {
    char *cp= (char*)out_buf;
    char *end= (char*)(out_buf + out_blen);
    while (cp < end) { char tmp= *cp++; cp[-1]= cp[0]; *cp++= tmp; }
  }

  // Check and update the byte count if necessary
  if (byte_count > 0) {
    if (byte_count <= out_bsiz) {
      writeOut((char*)out_buf, byte_count);
      exit(0);		// All done
    }
    else {
      writeOut((char*)out_buf, out_bsiz);
      byte_count -= out_bsiz;
    }
  }
  else
    writeOut((char*)out_buf, out_bsiz);
}

void
writeOut(char *buf, int siz) {
  int rv;

#ifdef WIN_AUDIO
  if (out_fd == -9999) {
     // Win32 output: write it to a header and send it off
     MMRESULT rv;

     //debug_win32_buffer_status();

     //while (aud_cnt == BUFFER_COUNT) {
     //while (aud_head[aud_current]->dwFlags & WHDR_INQUEUE) {
     while (!(aud_head[aud_current]->dwFlags & WHDR_DONE)) {
	//debug("SLEEP %d", out_buf_ms / 2 + 1);
	Sleep(out_buf_ms / 2 + 1);
	//debug_win32_buffer_status();
     }

     memcpy(aud_head[aud_current]->lpData, buf, siz);
     aud_head[aud_current]->dwBufferLength= (DWORD)siz;

     //debug("Output buffer %d", aud_current);
     rv= waveOutWrite(aud_handle, aud_head[aud_current], sizeof(WAVEHDR));

     if (rv != MMSYSERR_NOERROR) {
        char buf[255];
        waveOutGetErrorText(rv, buf, sizeof(buf)-1);
        error("Error writing a fragment to the audio device:\n  %s", buf);
     }

     aud_cnt++;
     aud_current++;
     aud_current %= BUFFER_COUNT;

     return;
  }
#endif

#ifdef MAC_AUDIO
  if (out_fd == -9999) {
    int new_wr= (aud_wr + 1) % BUFFER_COUNT;

    // Wait until there is space
    while (new_wr == aud_rd) delay(20);

    memcpy(aud_buf[aud_wr], buf, siz);
    aud_wr= new_wr;

    return;
  }
#endif

  while (-1 != (rv= write(out_fd, buf, siz))) {
    if (0 == (siz -= rv)) return;
    buf += rv;
  }
  error("Output error");
}

//
//	Calculate amplitude adjustment factor for frequency 'freq'
//

double
ampAdjust(double freq) {
   int a;
   struct AmpAdj *p0, *p1;

   if (!opt_c) return 1.0;
   if (freq <= ampadj[0].freq) return ampadj[0].adj;
   if (freq >= ampadj[opt_c-1].freq) return ampadj[opt_c-1].adj;

   for (a= 1; a<opt_c; a++)
      if (freq < ampadj[a].freq)
	 break;

   p0= &ampadj[a-1];
   p1= &ampadj[a];

   return p0->adj + (p1->adj - p0->adj) * (freq - p0->freq) / (p1->freq - p0->freq);
}


//
//	Correct channel values and types according to current period,
//	and current time
//

void
corrVal(int running) {
   int a;
   int t0= per->tim;
   int t1= per->nxt->tim;
   Channel *ch;
   Voice *v0, *v1, *vv;
   double rat0, rat1;
   int trigger= 0;

   // Move to the correct period
   while ((now >= t0) ^ (now >= t1) ^ (t1 > t0)) {
      per= per->nxt;
      t0= per->tim;
      t1= per->nxt->tim;
      if (running) {
	 if (tty_erase) {
#ifdef ANSI_TTY
	    fprintf(stderr, "\033[K");
#else
	    fprintf(stderr, "%*s\r", tty_erase, "");
	    tty_erase= 0;
#endif
	 }
	 dispCurrPer(stderr); status(0);
      }
      trigger= 1;		// Trigger bells or whatever
   }

   // Run through to calculate voice settings for current time
   rat1= t_per0(t0, now) / (double)t_per24(t0, t1);
   rat0= 1 - rat1;
   for (a= 0; a<N_CH; a++) {
      ch= &chan[a];
      v0= &per->v0[a];
      v1= &per->v1[a];
      vv= &ch->v;

      if (vv->typ != v0->typ) {
	 switch (vv->typ= ch->typ= v0->typ) {
	  case 1:
	     ch->off1= ch->off2= 0; break;
	  case 2:
	     break;
	  case 3:
	     ch->off1= ch->off2= 0; break;
	  case 4:
	     ch->off1= ch->off2= 0; break;
	  case 5:
	     break;
	  default:
	     ch->off1= ch->off2= 0; break;
	 }
      }

      // Setup vv->*
      switch (vv->typ) {
       case 1:
	  vv->amp= rat0 * v0->amp + rat1 * v1->amp;
	  vv->carr= rat0 * v0->carr + rat1 * v1->carr;
	  vv->res= rat0 * v0->res + rat1 * v1->res;
	  break;
       case 2:
	  vv->amp= rat0 * v0->amp + rat1 * v1->amp;
	  break;
       case 3:
	  vv->amp= v0->amp;		// No need to slide, as bell only rings briefly
	  vv->carr= v0->carr;
	  break;
       case 4:
	  vv->amp= rat0 * v0->amp + rat1 * v1->amp;
	  vv->carr= rat0 * v0->carr + rat1 * v1->carr;
	  vv->res= rat0 * v0->res + rat1 * v1->res;
	  if (vv->carr > spin_carr_max) vv->carr= spin_carr_max; // Clipping sweep width
	  if (vv->carr < -spin_carr_max) vv->carr= -spin_carr_max;
	  break;
       case 5:
	  vv->amp= rat0 * v0->amp + rat1 * v1->amp;
	  break;
       default:		// Waveform based binaural
	  vv->amp= rat0 * v0->amp + rat1 * v1->amp;
	  vv->carr= rat0 * v0->carr + rat1 * v1->carr;
	  vv->res= rat0 * v0->res + rat1 * v1->res;
	  break;
      }
   }

   // Check and limit amplitudes if -c option in use
   if (opt_c) {
      double tot_beat= 0, tot_other= 0;
      for (a= 0; a<N_CH; a++) {
	 vv= &chan[a].v;
	 if (vv->typ == 1) {
	    double adj1= ampAdjust(vv->carr + vv->res/2);
	    double adj2= ampAdjust(vv->carr - vv->res/2);
	    if (adj2 > adj1) adj1= adj2;
	    tot_beat += vv->amp * adj1;
	 } else if (vv->typ) {
	    tot_other += vv->amp;
	 }
      }
      if (tot_beat + tot_other > 4096) {
	 double adj_beat= (tot_beat > 4096) ? 4096 / tot_beat : 1.0;
	 double adj_other= (4096 - tot_beat * adj_beat) / tot_other;
	 for (a= 0; a<N_CH; a++) {
	    vv= &chan[a].v;
	    if (vv->typ == 1)
	       vv->amp *= adj_beat;
	    else if (vv->typ)
	       vv->amp *= adj_other;
	 }
      }
   }

   // Setup Channel data from Voice data
   for (a= 0; a<N_CH; a++) {
      ch= &chan[a];
      vv= &ch->v;

      // Setup ch->* from vv->*
      switch (vv->typ) {
	 double freq1, freq2;
       case 1:
	  freq1= vv->carr + vv->res/2;
	  freq2= vv->carr - vv->res/2;
	  if (opt_c) {
	     ch->amp= vv->amp * ampAdjust(freq1);
	     ch->amp2= vv->amp * ampAdjust(freq2);
	  } else
	     ch->amp= ch->amp2= (int)vv->amp;
	  ch->inc1= (int)(freq1 / out_rate * ST_SIZ * 65536);
	  ch->inc2= (int)(freq2 / out_rate * ST_SIZ * 65536);
	  break;
       case 2:
	  ch->amp= (int)vv->amp;
	  break;
       case 3:
	  ch->amp= (int)vv->amp;
	  ch->inc1= (int)(vv->carr / out_rate * ST_SIZ * 65536);
	  if (trigger) {		// Trigger the bell only on entering the period
	     ch->off2= ch->amp;
	     ch->inc2= out_rate/20;
	  }
	  break;
       case 4:
	  ch->amp= (int)vv->amp;
	  ch->inc1= (int)(vv->res / out_rate * ST_SIZ * 65536);
	  ch->inc2= (int)(vv->carr * 1E-6 * out_rate * (1<<24) / ST_AMP);
	  break;
       case 5:
	  ch->amp= (int)vv->amp;
	  break;
       default:		// Waveform based binaural
	  ch->amp= (int)vv->amp;
	  ch->inc1= (int)((vv->carr + vv->res/2) / out_rate * ST_SIZ * 65536);
	  ch->inc2= (int)((vv->carr - vv->res/2) / out_rate * ST_SIZ * 65536);
	  if (ch->inc1 > ch->inc2)
	     ch->inc2= -ch->inc2;
	  else
	     ch->inc1= -ch->inc1;
	  break;
      }
   }
}

//
//	Setup audio device
//

void
setup_device(void) {

  // Handle output to files and pipes
  if (opt_O || opt_o) {
    if (opt_O)
      out_fd= 1;		// stdout
    else {
      FILE *out;		// Need to create a stream to set binary mode for DOS
      if (!(out= fopen(opt_o, "wb")))
	error("Can't open \"%s\", errno %d", opt_o, errno);
      out_fd= fileno(out);
    }
    out_blen= out_rate * 2 / out_prate;		// 10 fragments a second by default
    while (out_blen & (out_blen-1)) out_blen &= out_blen-1;		// Make power of two
    out_bsiz= out_blen * (out_mode ? 2 : 1);
    out_bps= out_mode ? 4 : 2;
    out_buf= (short*)Alloc(out_blen * sizeof(short));
    out_buf_lo= (int)(0x10000 * 1000.0 * 0.5 * out_blen / out_rate);
    out_buf_ms= out_buf_lo >> 16;
    out_buf_lo &= 0xFFFF;
    tmp_buf= (int*)Alloc(out_blen * sizeof(int));

    if (!opt_Q && !opt_W)		// Informational message for opt_W is written later
       warn("Outputting %d-bit raw audio data at %d Hz with %d-sample blocks, %d ms per block",
	    out_mode ? 16 : 8, out_rate, out_blen/2, out_buf_ms);
    return;
  }

#ifdef OSS_AUDIO
  // Normal /dev/dsp output
  {
    int stereo, rate, fragsize, numfrags, enc;
    int targ_ms= 400;	// How much buffering we want, ideally
    int afmt_req, afmt;
    int test= 1;
    audio_buf_info info;
    int retry= 0;

    fragsize= 14;  // Ask for fragments of 2^14 == 16384 bytes == 4096 samples

    while (1) {
       if (0 > (out_fd= open(opt_d, O_WRONLY)))
	  error("Can't open %s, errno %d", opt_d, errno);

       afmt= afmt_req= ((out_mode == 0) ? AFMT_U8 :
			((char*)&test)[0] ? AFMT_S16_LE : AFMT_S16_BE);
       stereo= 1;
       rate= out_rate;
       numfrags= (out_rate * 4 * targ_ms / 1000) >> fragsize;
       if (numfrags < 1) numfrags= 1;
       enc= (numfrags<<16) | fragsize;

       if (0 > ioctl(out_fd, SNDCTL_DSP_SETFRAGMENT, &enc) ||
	   0 > ioctl(out_fd, SNDCTL_DSP_SAMPLESIZE, &afmt) ||
	   0 > ioctl(out_fd, SNDCTL_DSP_STEREO, &stereo) ||
	   0 > ioctl(out_fd, SNDCTL_DSP_SPEED, &rate))
	  error("Can't configure %s, errno %d", opt_d, errno);

       if (afmt != afmt_req)
	  error("Can't open device in %d-bit mode", out_mode ? 16 : 8);
       if (!stereo)
	  error("Can't open device in stereo");

       out_rate= rate;

       if (-1 == ioctl(out_fd, SNDCTL_DSP_GETOSPACE, &info))
	  error("Can't get audio buffer info, errno %d", errno);

       if (!retry && info.fragsize != (1<<fragsize)) {
	  // We've received a different fragment size to what we asked
	  // for.  This means that the numfrags calculation is wrong.
	  // Re-open based on this fragsize.
	  close(out_fd);
	  for (fragsize= 1; (1<<fragsize) < info.fragsize; fragsize++) ;
	  retry= 1;
	  //warn("Retrying /dev/dsp open for fragsize %d", fragsize);
	  continue;
       }
       break;
    }

    out_bsiz= info.fragsize;
    out_blen= out_mode ? out_bsiz/2 : out_bsiz;
    out_bps= out_mode ? 4 : 2;
    out_buf= (short*)Alloc(out_blen * sizeof(short));
    out_buf_lo= (int)(0x10000 * 1000.0 * 0.5 * out_blen / out_rate);
    out_buf_ms= out_buf_lo >> 16;
    out_buf_lo &= 0xFFFF;
    tmp_buf= (int*)Alloc(out_blen * sizeof(int));

    if (!opt_Q)
       warn("Outputting %d-bit audio at %d Hz with %d %d-sample fragments, %d ms per fragment",
	    out_mode ? 16 : 8, out_rate, info.fragstotal, out_blen/2, out_buf_ms);
  }
#endif
#ifdef WIN_AUDIO
  // Output using Win32 calls
  {
     MMRESULT rv;
     WAVEFORMATEX fmt;
     int a;

     fmt.wFormatTag= WAVE_FORMAT_PCM;
     fmt.nChannels= 2;
     fmt.nSamplesPerSec= out_rate;
     fmt.wBitsPerSample= out_mode ? 16 : 8;
     fmt.nBlockAlign= fmt.nChannels * (fmt.wBitsPerSample/8);
     fmt.nAvgBytesPerSec= fmt.nSamplesPerSec * fmt.nBlockAlign;
     fmt.cbSize= 0;
     aud_handle= NULL;

     // if (MMSYSERR_NOERROR !=
     //    waveOutOpen(&aud_handle, WAVE_MAPPER, &fmt, 0,
     //                0L, WAVE_FORMAT_QUERY))
     //    error("Windows is rejecting our audio request (%d-bit stereo, %dHz)",
     //          out_mode ? 16 : 8, out_rate);

     if (MMSYSERR_NOERROR !=
	 (rv= waveOutOpen(&aud_handle, WAVE_MAPPER,
			  (WAVEFORMATEX*)&fmt, (DWORD)win32_audio_callback,
			  (DWORD)0, CALLBACK_FUNCTION))) {
	char buf[255];
	waveOutGetErrorText(rv, buf, sizeof(buf)-1);
	error("Can't open audio device (%d-bit stereo, %dHz):\n  %s",
               out_mode ? 16 : 8, out_rate, buf);
     }

     if (fmt.nChannels != 2)
	error("Can't open audio device in stereo");
     if (fmt.wBitsPerSample != (out_mode ? 16 : 8))
	error("Can't open audio device in %d-bit mode", out_mode ? 16 : 8);

     aud_current= 0;
     aud_cnt= 0;

     for (a= 0; a<BUFFER_COUNT; a++) {
	char *p= (char *)Alloc(sizeof(WAVEHDR) + BUFFER_SIZE);
	WAVEHDR *w= aud_head[a]= (WAVEHDR*)p;

	w->lpData= (LPSTR)p + sizeof(WAVEHDR);
	w->dwBufferLength= (DWORD)BUFFER_SIZE;
	w->dwBytesRecorded= 0L;
	w->dwUser= 0;
	w->dwFlags= 0;
	w->dwLoops= 0;
	w->lpNext= 0;
	w->reserved= 0;

	rv= waveOutPrepareHeader(aud_handle, w, sizeof(WAVEHDR));
	if (rv != MMSYSERR_NOERROR) {
	   char buf[255];
	   waveOutGetErrorText(rv, buf, sizeof(buf)-1);
	   error("Can't setup a wave header %d:\n  %s", a, buf);
	}
	w->dwFlags |= WHDR_DONE;
     }

     out_rate= fmt.nSamplesPerSec;
     out_bsiz= BUFFER_SIZE;
     out_blen= out_mode ? out_bsiz/2 : out_bsiz;
     out_bps= out_mode ? 4 : 2;
     out_buf= (short*)Alloc(out_blen * sizeof(short));
     out_buf_lo= (int)(0x10000 * 1000.0 * 0.5 * out_blen / out_rate);
     out_buf_ms= out_buf_lo >> 16;
     out_buf_lo &= 0xFFFF;
     out_fd= -9999;
     tmp_buf= (int*)Alloc(out_blen * sizeof(int));

     if (!opt_Q)
	warn("Outputting %d-bit audio at %d Hz with %d %d-sample fragments, "
	     "%d ms per fragment", out_mode ? 16 : 8,
	     out_rate, BUFFER_COUNT, out_blen/2, out_buf_ms);
  }
#endif
#ifdef MAC_AUDIO
  // Mac CoreAudio for OS X
  {
    char deviceName[256];
    OSStatus err;
    UInt32 propertySize, bufferByteCount;
    struct AudioStreamBasicDescription streamDesc;
    int old_out_rate= out_rate;

    out_bsiz= BUFFER_SIZE;
    out_blen= out_mode ? out_bsiz/2 : out_bsiz;
    out_bps= out_mode ? 4 : 2;
    out_buf= (short*)Alloc(out_blen * sizeof(short));
    out_buf_lo= (int)(0x10000 * 1000.0 * 0.5 * out_blen / out_rate);
    out_buf_ms= out_buf_lo >> 16;
    out_buf_lo &= 0xFFFF;
    tmp_buf= (int*)Alloc(out_blen * sizeof(int));

    // N.B.  Both -r and -b flags are totally ignored for CoreAudio --
    // we just use whatever the default device is set to, and feed it
    // floats.
    out_mode= 1;
    out_fd= -9999;

    // Find default device
    propertySize= sizeof(aud_dev);
    if ((err= AudioHardwareGetProperty(kAudioHardwarePropertyDefaultOutputDevice,
				       &propertySize, &aud_dev)))
      error("Get default output device failed, status = %d", (int)err);

    if (aud_dev == kAudioDeviceUnknown)
      error("No default audio device found");

    // Get device name
    propertySize= sizeof(deviceName);
    if ((err= AudioDeviceGetProperty(aud_dev, 1, 0,
				     kAudioDevicePropertyDeviceName,
				     &propertySize, deviceName)))
      error("Get audio device name failed, status = %d", (int)err);

    // Get device properties
    propertySize= sizeof(streamDesc);
    if ((err= AudioDeviceGetProperty(aud_dev, 1, 0,
				     kAudioDevicePropertyStreamFormat,
				     &propertySize, &streamDesc)))
      error("Get audio device properties failed, status = %d", (int)err);

    out_rate= (int)streamDesc.mSampleRate;

    if (streamDesc.mChannelsPerFrame != 2)
      error("SBaGen requires a stereo output device -- \n"
	    "default output has %d channels",
	    streamDesc.mChannelsPerFrame);

    if (streamDesc.mFormatID != kAudioFormatLinearPCM ||
	!(streamDesc.mFormatFlags & kLinearPCMFormatFlagIsFloat))
      error("Expecting a 32-bit float linear PCM output stream -- \n"
	    "default output uses another format");

    // Set buffer size
    bufferByteCount= BUFFER_SIZE / 2 * sizeof(float);
    propertySize= sizeof(bufferByteCount);
    if ((err= AudioDeviceSetProperty(aud_dev, 0, 0, 0,
				     kAudioDevicePropertyBufferSize,
				     propertySize, &bufferByteCount)))
      error("Set audio output buffer size failed, status = %d", (int)err);

    // Setup callback and start it
    err= AudioDeviceAddIOProc(aud_dev, mac_callback, (void *)1);
    err= AudioDeviceStart(aud_dev, mac_callback);

    // Report settings
    if (!opt_Q) {
       if (old_out_rate != out_rate && !out_rate_def)
	  warn("*** WARNING: Non-default sampling rates not yet supported on OS X ***");
       warn("Outputting %d-bit audio at %d Hz to \"%s\",\n"
	    "  using %d %d-sample fragments, %d ms per fragment",
	    (int)streamDesc.mBitsPerChannel, out_rate, deviceName,
	    BUFFER_COUNT, out_blen/2, out_buf_ms);
    }
  }
#endif
#ifdef NO_AUDIO
  error("Direct output to soundcard not supported on this platform.\n"
	"Use -o or -O to write raw data, or -Wo or -WO to write a WAV file.");
#endif
}

//
//	Audio callback for Win32
//

#ifdef WIN_AUDIO
void CALLBACK
win32_audio_callback(HWAVEOUT hand, UINT uMsg,
		     DWORD dwInstance, DWORD dwParam1, DWORD dwParam2) {
   switch (uMsg) {
    case WOM_CLOSE:
       break;
    case WOM_OPEN:
       break;
    case WOM_DONE:
       aud_cnt--;
       //debug("Buffer done (cnt==%d)", aud_cnt);
       //debug_win32_buffer_status();
       break;
   }
}

void debug_win32_buffer_status() {
   char tmp[80];
   char *p= tmp;
   int a;
   for (a= 0; a<BUFFER_COUNT; a++) {
      *p++= (aud_head[a]->dwFlags & WHDR_INQUEUE) ? 'I' : '-';
      *p++= (aud_head[a]->dwFlags & WHDR_DONE) ? 'D' : '-';
      *p++= ' ';
   }
   p[-1]= 0;
   debug(tmp);
}
#endif

//
//	Audio callback for Mac OS X
//

#ifdef MAC_AUDIO
OSStatus mac_callback(AudioDeviceID inDevice,
		      const AudioTimeStamp *inNow,
		      const AudioBufferList *inInputData,
		      const AudioTimeStamp *inInputTime,
		      AudioBufferList *outOutputData,
		      const AudioTimeStamp *inOutputTime,
		      void *inClientData)
{
  float *fp= outOutputData->mBuffers[0].mData;
  int cnt= BUFFER_SIZE / 2;
  short *sp;

  if (aud_rd == aud_wr) {
    // Nothing in buffer list, so fill with silence
    while (cnt-- > 0) *fp++= 0.0;
  } else {
    // Consume a buffer
    sp= (short*)aud_buf[aud_rd];
    while (cnt-- > 0) *fp++= *sp++ * (1/32768.0);
    aud_rd= (aud_rd + 1) % BUFFER_COUNT;
  }

  return kAudioHardwareNoError;
}
#endif

//
//	Write a WAV header, and setup out_mode if byte-swapping is
//	required.  'byte_count' should have been set up by this point.
//

#define addU4(xx) { int a= xx; *p++= a; *p++= (a >>= 8); *p++= (a >>= 8); *p++= (a >>= 8); }
#define addStr(xx) { char *q= xx; *p++= *q++; *p++= *q++; *p++= *q++; *p++= *q++; }

void
writeWAV() {
  char buf[44], *p= buf;

  if (byte_count + 36 != (int)(byte_count + 36)) {
     int tmp;
     byte_count= 0xFFFFFFF8-36;
     tmp= byte_count/out_bps/out_rate;
     warn("WARNING: Selected length is too long for the WAV format; truncating to %dh%02dm%02ds",
	  tmp/3600, tmp/60%60, tmp%60);
  }

  addStr("RIFF");
  addU4(byte_count + 36);
  addStr("WAVE");
  addStr("fmt ");
  addU4(16);
  addU4(0x00020001);
  addU4(out_rate);
  addU4(out_rate * out_bps);
  addU4(0x0004 + 0x10000*(out_bps*4));	// 2,4 -> 8,16 - always assume stereo
  addStr("data");
  addU4(byte_count);
  writeOut(buf, 44);

  if (!opt_Q)
     warn("Outputting %d-bit WAV data at %d Hz, file size %d bytes",
	  out_mode ? 16 : 8, out_rate, byte_count + 44);
}

//
//	Read a line, discarding blank lines and comments.  Rets:
//	Another line?  Comments starting with '##' are displayed on
//	stderr.
//

int
readLine() {
   char *p;
   lin= buf;

   while (1) {
      if (!fgets(lin, sizeof(buf), in)) {
	 if (feof(in)) return 0;
	 error("Read error on sequence file");
      }

      in_lin++;

      while (isspace(*lin)) lin++;
      p= strchr(lin, '#');
      if (p && p[1] == '#') fprintf(stderr, "%s", p);
      p= p ? p : strchr(lin, 0);
      while (p > lin && isspace(p[-1])) p--;
      if (p != lin) break;
   }
   *p= 0;
   lin_copy= buf_copy;
   strcpy(lin_copy, lin);
   return 1;
}

//
//	Get next word at '*lin', moving lin onwards, or return 0
//

char *
getWord() {
  char *rv, *end;
  while (isspace(*lin)) lin++;
  if (!*lin) return 0;

  rv= lin;
  while (*lin && !isspace(*lin)) lin++;
  end= lin;
  if (*lin) lin++;
  *end= 0;

  return rv;
}

//
//	Bad sequence file
//

void
badSeq() {
  error("Bad sequence file content at line: %d\n  %s", in_lin, lin_copy);
}

// Convenience for situations where buffer is being filled by
// something other than readLine()
void
readNameDef2() {
   lin= buf; lin_copy= buf_copy;
   strcpy(lin_copy, lin);
   readNameDef();
}
void
readTimeLine2() {
   lin= buf; lin_copy= buf_copy;
   strcpy(lin_copy, lin);
   readTimeLine();
}

// Convenience for creating sequences on the fly
void
formatNameDef(char *fmt, ...) {
   va_list ap;
   va_start(ap, fmt);
   vsnprintf(buf, sizeof(buf), fmt, ap);
   readNameDef2();
}
void
formatTimeLine(int tim, char *fmt, ...) {
   va_list ap;
   char *p= buf + sprintf(buf, "%02d:%02d:%02d ", tim/3600, tim/60%60, tim%60);
   va_start(ap, fmt);
   vsnprintf(p, buf + sizeof(buf) - p, fmt, ap);
   readTimeLine2();
}

//
//	Generate a list of Period structures, based on the tone-specs
//	passed in (ac,av)
//

void
readSeqImm(int ac, char **av) {
   char *p= buf;

   in_lin= 0;
   p += sprintf(p, "immediate:");
   while (ac-- > 0) p += sprintf(p, " %s", *av++);
   readNameDef2();

   strcpy(buf, "00:00 immediate");
   readTimeLine2();

   correctPeriods();
}

//
//	Read a list of sequence files, and generate a list of Period
//	structures
//

void
readSeq(int ac, char **av) {
   // Setup a 'now' value to use for NOW in the sequence file
   now= calcNow();

   while (ac-- > 0) {
      char *fnam= *av++;
      int start= 1;

      in= (0 == strcmp("-", fnam)) ? stdin : fopen(fnam, "r");
      if (!in) error("Can't open sequence file: %s", fnam);

      in_lin= 0;

      while (readLine()) {
	 char *p= lin;

	 // Blank lines
	 if (!*p) continue;

	 // Look for options
	 if (*p == '-') {
	    if (!start)
	       error("Options are only permitted at start of sequence file:\n  %s", p);
	    handleOptions(p);
	    continue;
	 }

	 // Check to see if it fits the form of <name>:<white-space>
	 start= 0;
	 if (!isalpha(*p))
	    p= 0;
	 else {
	    while (isalnum(*p) || *p == '_' || *p == '-') p++;
	    if (*p++ != ':' || !isspace(*p))
	       p= 0;
	 }

	 if (p)
	    readNameDef();
	 else
	    readTimeLine();
      }

      if (in != stdin) fclose(in);
   }

   correctPeriods();
}


//
//	Fill in all the correct information for the Periods, assuming
//	they have just been loaded using readTimeLine()
//


void
correctPeriods() {
  // Get times all correct
  {
    Period *pp= per;
    do {
      if (pp->fi == -2) {
	pp->tim= pp->nxt->tim;
	pp->fi= -1;
      }

      pp= pp->nxt;
    } while (pp != per);
  }

  // Make sure that the transitional periods each have enough time
  {
    Period *pp= per;
    do {
      if (pp->fi == -1) {
	int per= t_per0(pp->tim, pp->nxt->tim);
	if (per < fade_int) {
	  int adj= (fade_int - per) / 2, adj0, adj1;
	  adj0= t_per0(pp->prv->tim, pp->tim);
	  adj0= (adj < adj0) ? adj : adj0;
	  adj1= t_per0(pp->nxt->tim, pp->nxt->nxt->tim);
	  adj1= (adj < adj1) ? adj : adj1;
	  pp->tim= (pp->tim - adj0 + H24) % H24;
	  pp->nxt->tim= (pp->nxt->tim + adj1) % H24;
	}
      }

      pp= pp->nxt;
    } while (pp != per);
  }

  // Fill in all the voice arrays, and sort out details of
  // transitional periods
  {
    Period *pp= per;
    do {
      if (pp->fi < 0) {
	int fo, fi;
	int a;
	int midpt= 0;

	Period *qq= (Period*)Alloc(sizeof(*qq));
	qq->prv= pp; qq->nxt= pp->nxt;
	qq->prv->nxt= qq->nxt->prv= qq;

	qq->tim= t_mid(pp->tim, qq->nxt->tim);

	memcpy(pp->v0, pp->prv->v1, sizeof(pp->v0));
	memcpy(qq->v1, qq->nxt->v0, sizeof(qq->v1));

	// Special handling for bells
	for (a= 0; a<N_CH; a++) {
	  if (pp->v0[a].typ == 3 && pp->fi != -3)
	    pp->v0[a].typ= 0;

	  if (qq->v1[a].typ == 3 && pp->fi == -3)
	    qq->v1[a].typ= 0;
	}

	fo= pp->prv->fo;
	fi= qq->nxt->fi;

	// Special handling for -> slides:
	//   always slide, and stretch slide if possible
	if (pp->fi == -3) {
	  fo= fi= 2;		// Force slides for ->
	  for (a= 0; a<N_CH; a++) {
	    Voice *vp= &pp->v0[a];
	    Voice *vq= &qq->v1[a];
	    if (vp->typ == 0 && vq->typ != 0 && vq->typ != 3) {
	      memcpy(vp, vq, sizeof(*vp)); vp->amp= 0;
	    }
	    else if (vp->typ != 0 && vq->typ == 0) {
	      memcpy(vq, vp, sizeof(*vq)); vq->amp= 0;
	    }
	  }
	}

	memcpy(pp->v1, pp->v0, sizeof(pp->v1));
	memcpy(qq->v0, qq->v1, sizeof(qq->v0));

	for (a= 0; a<N_CH; a++) {
	  Voice *vp= &pp->v1[a];
	  Voice *vq= &qq->v0[a];
	  if ((fo == 0 || fi == 0) ||		// Fade in/out to silence
	      (vp->typ != vq->typ) ||		// Different types
	      ((fo == 1 || fi == 1) &&		// Fade thru, but different pitches
	       (vp->typ == 1 || vp->typ < 0) &&
	       (vp->carr != vq->carr || vp->res != vq->res))
	      ) {
	    vp->amp= vq->amp= 0;		// To silence
	    midpt= 1;				// Definitely need the mid-point

	    if (vq->typ == 3) {	 		// Special handling for bells
	      vq->amp= qq->v1[a].amp;
	      qq->nxt->v0[a].typ= qq->nxt->v1[a].typ= 0;
	    }
	  }
	  else if (vp->typ == 3) {		// Else smooth transition - for bells not so smooth
	    qq->v0[a].typ= qq->v1[a].typ= 0;
	  }
	  else {				// Else smooth transition
	    vp->amp= vq->amp= (vp->amp + vq->amp) / 2;
	    if (vp->typ == 1 || vp->typ == 4 || vp->typ < 0) {
	      vp->carr= vq->carr= (vp->carr + vq->carr) / 2;
	      vp->res= vq->res= (vp->res + vq->res) / 2;
	    }
	  }
	}

	// If we don't really need the mid-point, then get rid of it
	if (!midpt) {
	  memcpy(pp->v1, qq->v1, sizeof(pp->v1));
	  qq->prv->nxt= qq->nxt;
	  qq->nxt->prv= qq->prv;
	  free(qq);
	}
	else pp= qq;
      }

      pp= pp->nxt;
    } while (pp != per);
  }

  // Clear out zero length sections, and duplicate sections
  {
    Period *pp;
    while (per != per->nxt) {
      pp= per;
      do {
	if (voicesEq(pp->v0, pp->v1) &&
	    voicesEq(pp->v0, pp->nxt->v0) &&
	    voicesEq(pp->v0, pp->nxt->v1))
	  pp->nxt->tim= pp->tim;

	if (pp->tim == pp->nxt->tim) {
	  if (per == pp) per= per->prv;
	  pp->prv->nxt= pp->nxt;
	  pp->nxt->prv= pp->prv;
	  free(pp);
	  pp= 0;
	  break;
	}
	pp= pp->nxt;
      } while (pp != per);
      if (pp) break;
    }
  }

  // Make sure that the total is 24 hours only (not more !)
  if (per->nxt != per) {
    int tot= 0;
    Period *pp= per;

    do {
      tot += t_per0(pp->tim, pp->nxt->tim);
      pp= pp->nxt;
    } while (pp != per);

    if (tot > H24) {
      warn("Total time is greater than 24 hours.  Probably two times are\n"
	   "out of order.  Suspicious intervals are:\n");
      pp= per;
      do {
	if (t_per0(pp->tim, pp->nxt->tim) >= H12)
	   warn("  %02d:%02d:%02d -> %02d:%02d:%02d",
		pp->tim % 86400000 / 3600000,
		pp->tim % 3600000 / 60000,
		pp->tim % 60000 / 1000,
		pp->nxt->tim % 86400000 / 3600000,
		pp->nxt->tim % 3600000 / 60000,
		pp->nxt->tim % 60000 / 1000);
	pp= pp->nxt;
      } while (pp != per);
      error("\nCheck the sequence around these times and try again");
    }
  }

  // Print the whole lot out
  if (opt_D) {
    Period *pp;
    if (per->nxt != per)
      while (per->prv->tim < per->tim) per= per->nxt;

    pp= per;
    do {
      dispCurrPer(stdout);
      per= per->nxt;
    } while (per != pp);
    printf("\n");

    exit(0);		// All done
  }
}

int
voicesEq(Voice *v0, Voice *v1) {
  int a= N_CH;

  while (a-- > 0) {
    if (v0->typ != v1->typ) return 0;
    switch (v0->typ) {
     case 1:
     case 4:
     default:
       if (v0->amp != v1->amp ||
	   v0->carr != v1->carr ||
	   v0->res != v1->res)
	 return 0;
       break;
     case 2:
     case 5:
       if (v0->amp != v1->amp)
	 return 0;
       break;
     case 3:
       if (v0->amp != v1->amp ||
	   v0->carr != v1->carr)
	 return 0;
       break;
    }
    v0++; v1++;
  }
  return 1;
}

//
//	Read a name definition
//

void
readNameDef() {
  char *p, *q;
  NameDef *nd;
  int ch;

  if (!(p= getWord())) badSeq();

  q= strchr(p, 0) - 1;
  if (*q != ':') badSeq();
  *q= 0;
  for (q= p; *q; q++) if (!isalnum(*q) && *q != '-' && *q != '_')
    error("Bad name \"%s\" in definition, line %d:\n  %s", p, in_lin, lin_copy);

  // Waveform definition ?
  if (0 == memcmp(p, "wave", 4) &&
      isdigit(p[4]) &&
      isdigit(p[5]) &&
      !p[6]) {
     int ii= (p[4] - '0') * 10 + (p[5] - '0');
     int siz= ST_SIZ * sizeof(int);
     int *arr= (int*)Alloc(siz);
     double *dp0= (double*)arr;
     double *dp1= (double*)(siz + (char*)arr);
     double *dp= dp0;
     double dmax= 0, dmin= 1;
     int np;

     if (waves[ii])
	error("Waveform %02d already defined, line %d:\n  %s",
	      ii, in_lin, lin_copy);
     waves[ii]= arr;

     while ((p= getWord())) {
	double dd;
	char dmy;
	if (1 != sscanf(p, "%lf %c", &dd, &dmy))
	   error("Expecting floating-point numbers on this waveform "
		 "definition line, line %d:\n  %s",
		 in_lin, lin_copy);
	if (dp >= dp1)
	   error("Too many samples on line (maximum %d), line %d:\n  %s",
		 dp1-dp0, in_lin, lin_copy);
	*dp++= dd;
	if (dmax < dmin) dmin= dmax= dd;
	else {
	   if (dd > dmax) dmax= dd;
	   if (dd < dmin) dmin= dd;
	}
     }
     dp1= dp;
     np= dp1 - dp0;
     if (np < 2)
	error("Expecting at least two samples in the waveform, line %d:\n  %s",
	      in_lin, lin_copy);

     // Adjust to range 0-1
     for (dp= dp0; dp < dp1; dp++)
	*dp= (*dp - dmin) / (dmax - dmin);

     sinc_interpolate(dp0, np, arr);

     if (DEBUG_DUMP_WAVES) {
	int a;
	printf("Dumping wave%02d:\n", ii);
	for (a= 0; a<ST_SIZ; a++)
	   printf("%d %g\n", a, arr[a] * 1.0 / ST_AMP);
     }
     return;
  }

  // Must be block or tone-set, then, so put into a NameDef
  nd= (NameDef*)Alloc(sizeof(NameDef));
  nd->name= StrDup(p);

  // Block definition ?
  if (*lin == '{') {
    BlockDef *bd, **prvp;
    if (!(p= getWord()) ||
	0 != strcmp(p, "{") ||
	0 != (p= getWord()))
      badSeq();

    prvp= &nd->blk;

    while (readLine()) {
      if (*lin == '}') {
	if (!(p= getWord()) ||
	    0 != strcmp(p, "}") ||
	    0 != (p= getWord()))
	  badSeq();
	if (!nd->blk) error("Empty blocks not permitted, line %d:\n  %s", in_lin, lin_copy);
	nd->nxt= nlist; nlist= nd;
	return;
      }

      if (*lin != '+')
	error("All lines in the block must have relative time, line %d:\n  %s",
	      in_lin, lin_copy);

      bd= (BlockDef*) Alloc(sizeof(*bd));
      *prvp= bd; prvp= &bd->nxt;
      bd->lin= StrDup(lin);
    }

    // Hit EOF before }
    error("End-of-file within block definition (missing '}')");
  }

  // Normal line-definition
  for (ch= 0; ch < N_CH && (p= getWord()); ch++) {
    char dmy;
    double amp, carr, res;
    int wave;

    // Interpret word into Voice nd->vv[ch]
    if (0 == strcmp(p, "-")) continue;
    if (1 == sscanf(p, "pink/%lf %c", &amp, &dmy)) {
       nd->vv[ch].typ= 2;
       nd->vv[ch].amp= AMP_DA(amp);
       continue;
    }
    if (2 == sscanf(p, "bell%lf/%lf %c", &carr, &amp, &dmy)) {
       nd->vv[ch].typ= 3;
       nd->vv[ch].carr= carr;
       nd->vv[ch].amp= AMP_DA(amp);
       continue;
    }
    if (1 == sscanf(p, "mix/%lf %c", &amp, &dmy)) {
       nd->vv[ch].typ= 5;
       nd->vv[ch].amp= AMP_DA(amp);
       mix_flag= 1;
       continue;
    }
    if (4 == sscanf(p, "wave%d:%lf%lf/%lf %c", &wave, &carr, &res, &amp, &dmy)) {
       if (wave < 0 || wave >= 100)
	  error("Only wave00 to wave99 is permitted at line: %d\n  %s", in_lin, lin_copy);
       if (!waves[wave])
	  error("Waveform %02d has not been defined, line: %d\n  %s", wave, in_lin, lin_copy);
       nd->vv[ch].typ= -1-wave;
       nd->vv[ch].carr= carr;
       nd->vv[ch].res= res;
       nd->vv[ch].amp= AMP_DA(amp);
       continue;
    }
    if (3 == sscanf(p, "%lf%lf/%lf %c", &carr, &res, &amp, &dmy)) {
      nd->vv[ch].typ= 1;
      nd->vv[ch].carr= carr;
      nd->vv[ch].res= res;
      nd->vv[ch].amp= AMP_DA(amp);
      continue;
    }
    if (2 == sscanf(p, "%lf/%lf %c", &carr, &amp, &dmy)) {
      nd->vv[ch].typ= 1;
      nd->vv[ch].carr= carr;
      nd->vv[ch].res= 0;
      nd->vv[ch].amp= AMP_DA(amp);
      continue;
    }
    if (3 == sscanf(p, "spin:%lf%lf/%lf %c", &carr, &res, &amp, &dmy)) {
      nd->vv[ch].typ= 4;
      nd->vv[ch].carr= carr;
      nd->vv[ch].res= res;
      nd->vv[ch].amp= AMP_DA(amp);
      continue;
    }
    badSeq();
  }
  nd->nxt= nlist; nlist= nd;
}

//
//	Bad time
//

void
badTime(char *tim) {
  error("Badly constructed time \"%s\", line %d:\n  %s", tim, in_lin, lin_copy);
}

//
//	Read a time-line of either type
//

void
readTimeLine() {
  char *p, *tim_p;
  int nn;
  int fo, fi;
  Period *pp;
  NameDef *nd;
  static int last_abs_time= -1;
  int tim, rtim = 0;

  if (!(p= getWord())) badSeq();
  tim_p= p;

  // Read the time represented
  tim= -1;
  if (0 == memcmp(p, "NOW", 3)) {
    last_abs_time= tim= now;
    p += 3;
  }

  while (*p) {
    if (*p == '+') {
      if (tim < 0) {
	if (last_abs_time < 0)
	  error("Relative time without previous absolute time, line %d:\n  %s", in_lin, lin_copy);
	tim= last_abs_time;
      }
      p++;
    }
    else if (tim != -1) badTime(tim_p);

    if (0 == (nn= readTime(p, &rtim))) badTime(tim_p);
    p += nn;

    if (tim == -1)
      last_abs_time= tim= rtim;
    else
      tim= (tim + rtim) % H24;
  }

  if (fast_tim0 < 0) fast_tim0= tim;		// First time
  fast_tim1= tim;				// Last time

  if (!(p= getWord())) badSeq();

  fi= fo= 1;
  if (!isalpha(*p)) {
    switch (p[0]) {
     case '<': fi= 0; break;
     case '-': fi= 1; break;
     case '=': fi= 2; break;
     default: badSeq();
    }
    switch (p[1]) {
     case '>': fo= 0; break;
     case '-': fo= 1; break;
     case '=': fo= 2; break;
     default: badSeq();
    }
    if (p[2]) badSeq();

    if (!(p= getWord())) badSeq();
  }

  for (nd= nlist; nd && 0 != strcmp(p, nd->name); nd= nd->nxt) ;
  if (!nd) error("Name \"%s\" not defined, line %d:\n  %s", p, in_lin, lin_copy);

  // Check for block name-def
  if (nd->blk) {
    char *prep= StrDup(tim_p);		// Put this at the start of each line
    BlockDef *bd= nd->blk;

    while (bd) {
      lin= buf; lin_copy= buf_copy;
      sprintf(lin, "%s%s", prep, bd->lin);
      strcpy(lin_copy, lin);
      readTimeLine();		// This may recurse, and that's why we're StrDuping the string
      bd= bd->nxt;
    }
    free(prep);
    return;
  }

  // Normal name-def
  pp= (Period*)Alloc(sizeof(*pp));
  pp->tim= tim;
  pp->fi= fi;
  pp->fo= fo;

  memcpy(pp->v0, nd->vv, N_CH * sizeof(Voice));
  memcpy(pp->v1, nd->vv, N_CH * sizeof(Voice));

  if (!per)
    per= pp->nxt= pp->prv= pp;
  else {
    pp->nxt= per; pp->prv= per->prv;
    pp->prv->nxt= pp->nxt->prv= pp;
  }

  // Automatically add a transitional period
  pp= (Period*)Alloc(sizeof(*pp));
  pp->fi= -2;		// Unspecified transition
  pp->nxt= per; pp->prv= per->prv;
  pp->prv->nxt= pp->nxt->prv= pp;

  if (0 != (p= getWord())) {
    if (0 != strcmp(p, "->")) badSeq();
    pp->fi= -3;		// Special '->' transition
    pp->tim= tim;
  }
}

int
readTime(char *p, int *timp) {		// Rets chars consumed, or 0 error
  int nn, hh, mm, ss;

  if (3 > sscanf(p, "%2d:%2d:%2d%n", &hh, &mm, &ss, &nn)) {
    ss= 0;
    if (2 > sscanf(p, "%2d:%2d%n", &hh, &mm, &nn)) return 0;
  }

  if (hh < 0 || hh >= 24 ||
      mm < 0 || mm >= 60 ||
      ss < 0 || ss >= 60) return 0;

  *timp= ((hh * 60 + mm) * 60 + ss) * 1000;
  return nn;
}

//
//	Takes a set of points and repeats them twice, inverting the
//	second set, and then interpolates them using a periodic sinc
//	function (see http://www-ccrma.stanford.edu/~jos/resample/)
//	and writes them to arr[] in the same format as the sin_table[].
//

void sinc_interpolate(double *dp, int np, int *arr) {
   double *sinc;	// Temporary sinc-table
   double *out;		// Temporary output table
   int a, b;
   double dmax, dmin;
   double adj, off;

   // Generate a modified periodic sin(x)/x function to be used for
   // each of the points.  Really this should be sin(x)/x modified
   // by the sum of an endless series.  However, this doesn't
   // converge very quickly, so to save time I'm approximating this
   // series by 1-4*t*t where t ranges from 0 to 0.5 over the first
   // half of the periodic cycle.  If you do the maths, this is at
   // most 5% out.  This will have to do - it's smooth, and I don't
   // know enough maths to make this series converge quicker.
   sinc= (double *)Alloc(ST_SIZ * sizeof(double));
   sinc[0]= 1.0;
   for (a= ST_SIZ/2; a>0; a--) {
      double tt= a * 1.0 / ST_SIZ;
      double t2= tt*tt;
      double adj= 1 - 4 * t2;
      double xx= 2 * np * 3.14159265358979323846 * tt;
      double vv= adj * sin(xx) / xx;
      sinc[a]= vv;
      sinc[ST_SIZ-a]= vv;
   }

   // Build waveform into buffer
   out= (double *)Alloc(ST_SIZ * sizeof(double));
   for (b= 0; b<np; b++) {
      int off= b * ST_SIZ / np / 2;
      double val= dp[b];
      for (a= 0; a<ST_SIZ; a++) {
	 out[(a + off)&(ST_SIZ-1)] += sinc[a] * val;
	 out[(a + off + ST_SIZ/2)&(ST_SIZ-1)] -= sinc[a] * val;
      }
   }

   // Look for maximum for normalization
   dmax= dmin= 0;
   for (a= 0; a<ST_SIZ; a++) {
      if (out[a] > dmax) dmax= out[a];
      if (out[a] < dmin) dmin= out[a];
   }

   // Write out to output buffer
   off= -0.5 * (dmax + dmin);
   adj= ST_AMP / ((dmax - dmin) / 2);
   for (a= 0; a<ST_SIZ; a++)
      arr[a]= (int)((out[a] + off) * adj);

   free(sinc);
   free(out);
}

//
//	Handling pre-programmed sequences
//

void
readPreProg(int ac, char **av) {
   if (ac < 1)
      error("Expecting a pre-programmed sequence description.  Examples:"
	    NL "  drop 25ds+ pink/30"
	    NL "  drop 25gs+/2 mix/60"
	    );

   // Handle 'drop'
   if (0 == strcmp(av[0], "drop")) {
      ac--; av++;
      create_drop(ac, av);
      return;
   }

   // Handle 'slide'
   if (0 == strcmp(av[0], "slide")) {
      ac--; av++;
      create_slide(ac, av);
      return;
   }

   error("Unknown pre-programmed sequence type: %s", av[0]);
}

//
//	Error for bad p-drop args
//

void
bad_drop() {
   error("Bad arguments: expecting -p drop [<time-spec>] <drop-spec> [<tone-specs...>]"
	 NL "<drop-spec> is <digit><digit>[.<digit>...]<a-l>[s|k][+][^][/<amp>]"
	 NL "The optional <time-spec> is t<drop-time>,<hold-time>,<wake-time>, all times"
	 NL "  in minutes (the default is equivalent to 't30,30,3')."
	 NL "The optional <tone-specs...> let you mix other stuff with the drop"
	 NL "  sequence like pink noise or a mix soundtrack, e.g 'pink/20' or 'mix/60'");
}

//
//	Generate a p-drop sequence
//
//	Credits: Jonathan Bisson created the first version of this C
//	code.  This is a rewrite to make it fit with the rest of the
//	code better.
//

void
create_drop(int ac, char **av) {
   char *fmt;
   char *p, *q;
   int a;
   int slide, n_step, islong, wakeup;
   double carr, amp, c0, c2;
   double beat_target;
   double beat[40];
   static double beat_targets[]= {
      4.4, 3.7, 3.1, 2.5, 2.0, 1.5, 1.2, 0.9, 0.7, 0.5, 0.4, 0.3
   };
   char extra[256];
   int len, len0= 1800, len1= 1800, len2= 180;
   int steplen, end;

#define BAD bad_drop()

   // Pick up optional time-spec
   if (ac < 1) BAD;
   if (av[0][0] == 't') {
      double v0, v1, v2;
      char dmy;
      if (3 != sscanf(av[0]+1, "%lf,%lf,%lf %c", &v0, &v1, &v2, &dmy)) BAD;
      len0= 60 * (int)v0;	// Whole minutes only
      len1= 60 * (int)v1;
      len2= 60 * (int)v2;
      ac--; av++;
   }

   // Handle argument list
   if (ac < 1) BAD;
   fmt= *av++; ac--;
   p= extra; *p= 0;
   while (ac > 0) {
      if (p + strlen(av[0]) + 2 > extra + sizeof(extra))
	 error("Too many extra tone-specs after -p drop");
      p += sprintf(p, " %s", av[0]);
      ac--; av++;
   }

   // Scan the format
   carr= 200 - 2 * strtod(fmt, &p);
   if (p == fmt || carr < 0) BAD;

   a= tolower(*p) - 'a'; p++;
   if (a < 0 || a >= sizeof(beat_targets) / sizeof(beat_targets[0])) BAD;
   beat_target= beat_targets[a];

   slide= 0;
   steplen= 180;
   if (*p == 's') { p++; slide= 1; steplen= 60; }
   else if (*p == 'k') { p++; steplen= 60; }
   n_step= 1 + (len0-1) / steplen;	// Round up
   len0= n_step * steplen;
   if (!slide) len1= (1 + (len1-1) / steplen) * steplen;

   islong= 0;
   if (*p == '+') { islong= 1; p++; }

   wakeup= 0;
   if (*p == '^') { wakeup= 1; p++; }

   amp= 1.0;
   if (*p == '/') {
      p++; q= p;
      amp= strtod(p, &p);
      if (p == q) BAD;
   }

   while (isspace(*p)) p++;
   if (*p) error("Trailing rubbish after -p drop spec: \"%s\"", p);

#undef BAD

   // Sort out carriers
   len= islong ? len0 + len1 : len0;
   c0= carr + 5.0;
   c2= carr;

   // Calculate beats
   for (a= 0; a<n_step; a++)
      beat[a]= 10 * exp(log(beat_target/10) * a / (n_step-1));

   // Display summary
   warn("DROP summary:");
   if (slide) {
      warn(" Carrier slides from %gHz to %gHz over %d minutes",
	   c0, c2, len/60);
      warn(" Beat frequency slides from %gHz to %gHz over %d minutes",
	   beat[0], beat[n_step-1], len0/60);
   } else {
      warn(" Carrier steps from %gHz to %gHz over %d minutes",
	   c0, c2, len/60);
      warn(" Beat frequency steps from %gHz to %gHz over %d minutes:",
	   beat[0], beat[n_step-1], len0/60);
      fprintf(stderr, "   ");
      for (a= 0; a<n_step; a++) fprintf(stderr, " %.2f", beat[a]);
      fprintf(stderr, "\n");
   }
   if (wakeup) {
      warn(" Final wake-up of %d minutes, to return to initial frequencies", len2/60);
   }

   // Start generating sequence
   handleOptions("-SE");
   in_lin= 0;

   formatNameDef("off: -");
   formatTimeLine(86395, "== off ->");		// 23:59:55

   if (slide) {
      // Slide version
      for (a= 0; a<n_step; a++) {
	 int tim= a * len0 / (n_step-1);
	 formatNameDef("ts%02d: %g+%g/%g %s", a,
		       c0 + (c2-c0) * tim * 1.0 / len,
		       beat[a], amp, extra);
	 formatTimeLine(tim, "== ts%02d ->", a);
      }

      if (islong) {
	 formatNameDef("tsend: %g+%g/%g %s",
		       c2, beat[n_step-1], amp, extra);
	 formatTimeLine(len, "== tsend ->");
      }
      end= len;
   } else {
      // Step version
      int lim= len / steplen;
      int stepslide= steplen < 90 ? 5 : 10;    // Seconds slide between steps
      for (a= 0; a<lim; a++) {
	 int tim0= a * steplen;
	 int tim1= (a+1) * steplen;
	 formatNameDef("ts%02d: %g+%g/%g %s", a,
		       c0 + (c2-c0) * tim1/len,
		       beat[(a>=n_step) ? n_step-1 : a],
		       amp, extra);
	 formatTimeLine(tim0, "== ts%02d ->", a);
	 formatTimeLine(tim1-stepslide, "== ts%02d ->", a);
      }
      end= len-stepslide;
   }

   // Wake-up and ending
   if (wakeup) {
      formatNameDef("tswake: %g+%g/%g %s",
		    c0, beat[0], amp, extra);
      formatTimeLine(end+len2, "== tswake ->");
      end += len2;
   }
   formatTimeLine(end+10, "== off");

   correctPeriods();
}

//
//	Generate a -p slide sequence
//
//	The idea of this is to hold the beat frequency constant, but
//	to slide down through the carrier frequencies from about 200Hz.
//
//	-p slide [t<duration-minutes>] <carr>+<beat>/<amp> [extra tone-sets]

void
bad_slide() {
   error("Bad arguments: expecting -p slide [<time-spec>] <slide-spec> [<tone-specs...>]"
	 NL "<slide-spec> is just like a tone-spec: <carrier><sign><beat>/<amp>"
	 NL "The optional <time-spec> is t<slide-time>, giving length of session in"
	 NL "  minutes (the default is equivalent to 't30')."
	 NL "The optional <tone-specs...> let you mix other stuff with the drop"
	 NL "  sequence like pink noise or a mix soundtrack, e.g 'pink/20' or 'mix/60'");
}

void
create_slide(int ac, char **av) {
   int len= 1800;
   char *p, dmy;
   double val, c0, c1, beat, amp;
   char extra[256];

#define BAD bad_slide()

   // Handle arguments
   if (ac < 1) BAD;
   if (av[0][0] == 't') {
      val= strtod(av[0]+1, &p);
      if (p == av[0] + 1 || *p) BAD;
      len= 60.0 * val;
      ac--; av++;
   }

   if (ac < 1) BAD;
   if (3 != sscanf(av[0], "%lf%lf/%lf %c", &c0, &beat, &amp, &dmy)) BAD;
   c1= beat/2;
   ac--; av++;

#undef BAD

   // Gather 'extra'
   p= extra; *p= 0;
   while (ac > 0) {
      if (p + strlen(av[0]) + 2 > extra + sizeof(extra))
	 error("Too many extra tone-specs after -p slide");
      p += sprintf(p, " %s", av[0]);
      ac--; av++;
   }

   // Summary
   warn("SLIDE summary:");
   warn(" Sliding carrier from %gHz to %gHz over %g minutes",
	c0, c1, len/60.0);
   warn(" Holding beat constant at %gHz", beat);

   // Generate sequence
   handleOptions("-SE");
   in_lin= 0;

   formatNameDef("off: -");
   formatTimeLine(86395, "== off ->");		// 23:59:55
   formatNameDef("ts0: %g%+g/%g %s", c0, beat, amp, extra);
   formatTimeLine(0, "== ts0 ->");
   formatNameDef("ts1: %g%+g/%g %s", c1, beat, amp, extra);
   formatTimeLine(len, "== ts1 ->");
   formatTimeLine(len+10, "== off");

   correctPeriods();
}


// END //