Merge branch 'upstream' into development

Conflicts:
	dictsource/dict_phonemes
	phsource/compile_report

dictsource/an_rules

 .L03 f h j s y z r      // allow [B] [D] [Q] before these consonants
 .L04 b v f p      // labials, nasals are [m]
 .L05 g k          // velars, nasals are [N]
-
+.L06 a e o
 
 .group a
         a          a
      i) e          E
      u) e          E
      y) e          E
+     
+  L06_) e          e#   // unstressed 'e' silent at start of word after a,e,o
 
      _) ent' (AP4  ent
      _) enta'      ,ent%a
      &) s (_S1     s
 
      _) s' (AP2    s
+     _) s (C@      %es
+     _) s (hA      s
+     _) s (y       s
+    A_) s (C       s
 
 
 .group t

dictsource/de_list

 teaser		_^_EN
 tech		_^_EN
 terminal	_^_EN
+time		_^_EN
 trip		_^_EN
 tuner		_^_EN
 unix		_^_EN
 ersten          e:rst@n
 erstmal         e:rstma:l
 erzgebirg       e:rtsg@bIrg
+expedient       Ekspe:dI'Ent
 extravagant      $1
 extrem          Ekstr'e:m
 

dictsource/de_rules

     @t) ien (_     'i:@n
      &) iens (_    =I;@ns
         ient (_    I'Ent
+     d) ient (_     i:nt
      &) ienz (_    I'Ents
    sch) ie (n      i:
      @) ie (reX    'i:

dictsource/kn_rules

 ೮	8
 ೯	9
 
+.L01 ಕ ಖ ಗ ಘ ಙ   // velar
+
+.L02 ಚ ಛ ಜ ಝ ಞ  // palatal
+
+.L03 ಟ ಠ ಡ ಢ ಣ  // retroflex
+
+.L04 ತ ಥ ದ ಧ ನ   // dental
+
+.L05 ಪ ಫ ಬ ಭ ಮ  // labial
+
 
 .group ಕ
 	ಕ	ka
 
 .group 0xe0b2   // characters which start with UTF-8 bytes: [e0 b2]
 
-	ಂ	m	// anusvara (this should nasalize the vowel)
+	ಂ(L01	N	// anusvara (this should nasalize the vowel)
+	ಂ(L02	n^
+	ಂ(L03	n.
+	ಂ(L04	n
+	ಂ	m
 
 	ಃ	h	// visarga
 

dictsource/nl_list

 boezemen	$alt
 ademen		$alt
 bliksemen	$alt
-wasemen		$alt
+wasemen		v#a:s@m@n $alt
 
 // words where -er is not a stressed verb ending
 
 wakker		$alt
 wapper		$alt
 water		$alt
-wedijver	v#Ed'EIv@r $alt
+wedijver	v#'EtEIv@r $alt
 weelder		$alt 
 wegwijzer	$alt
 weiger		$alt
 vergezicht	vErQ@zIxt
 verre		vEr@
 verrekijker	vEr@kEIk@r
-versa		vErs'a:
+versa		v'Erza:
 verste		vErst@
 verte		vErt@
 volledig	$2

dictsource/nl_rules

           _) ge (tenL02   Qe:
           _) geu          QY:
           _) ge (ven      Qe:
-          _) ge (ver      Qe:
+             ge (vend_    Qe:
+             ge (ver      Qe:
              ge (val      Q@
-          _) ge (ving     Qe:
+             ge (ving     Qe:
 
              ge (sprek    Q@  // in compounds
 
           &) iko (_       =iko:
         fam) i (lie       'i
           _) in (C@P2     'In
+          _) in (a@P2     'In
           _) ing (eC_      IN   // bewerkingen
           _) in (do        In
           _) in (forma     In

dictsource/sk_list

 £	libRa
 ¤	mena
 ©	kopiRajt
-®	Registrovane:
+®	RegistRovane:
+¶	'oznatSenie||'odseku
 ¥	jen
 −	mi:nus
 ±	plus||m'i:nus

dictsource/ta_list

 
 _dpt	_pul.l.i
 
-// drop final [U] when followed by an ordinal suffix
+// drop final [U] when followed by an ordinal suffix or thousands
 _1o	onR
 _2o	irVn.d.
 _3o	mu:nR
 _7Co	;ez.Unu:R
 _8Co	;en.n.u:R
 
+
+
 _#ம்	a:m  // ordinal suffix
 _#ல்	il
 _#வது	a:vVdU
 _#வதும்	a:vVdUm
 _#வதுமாக a:vVdUma:gV
 
+_#ஆம்	a:m  // including ஆ
+_#ஆவது	a:vVdU
+_#ஆவதும்	a:vVdUm
+_#ஆவதுமாக a:vVdUma:gV
+
 _#-ம்	a:m  // ordinal suffix with hyphen
 _#-ல்	il
 _#-வது	a:vVdU
 _#-வதும்	a:vVdUm
 _#-வதுமாக a:vVdUma:gV
 
+_#-ஆம்	a:m  // including ஆ
+_#-ஆவது	a:vVdU
+_#-ஆவதும்	a:vVdUm
+_#-ஆவதுமாக a:vVdUma:gV
+
 // English words
 tamil	தமிழ்	$text
 
 பாதிக்காம	$alt
 பனியா	$alt
 பாம்	$alt
+போத 	$alt
 
 
 	// initial த as [d]
 போன்	fo:n
 போனில்	fo:nil
 //பைலை 	faIlaI
-பாதர்	fa:dVr
 பாதரிடம்	fa:dVrid.Vm
 பாதருடன்	fa:dVrUd.Vn
 பாதரில்லாமல்	fa:dVrilla:mVl

dictsource/ta_rules

 // endings after போர்ட  (board)
 .L31  ்_  ிங்  ின  ை  ுக்க  ில  ும 
 
+// words after அர்த்த
+.L32  ராத்திரி  ஜாம  நிசி  வேளை 
+
 
 // consonants
 
      _) க (ால்ஃப்		g
      _) கால்ப்		ga:lf
      _) க (ிஃப்ட		g
-     _) கிட (ார		git.
      _) க (ிடுகிடு	g
      _) க (ிண்ணென்ற	g
      _) க (ின்னஸ	g
      _) க் (ளாஸ்_ 	g
      _) க்வ (A		kw
      _) க் (வாலியர	g
+     _) க (ுட்கா		g
+     _) கோவர்த்த		go:vVrddV#
+     _) கோர்க (ாலாந்த	go:rk
+     _) குருமுக (ி	gurumuk
+     _) க (ோஸ்வாமி	g
+     _) க (ோசுவாமி	g
+     _) க (ிடார்		g
+     _) க (ிடாரை	g
+     _) க (ிடாரில	g
+     _) க (ிடாரின	g
+     _) க (ிடாரிட	g
 //endsort
 
 //sort
+  _முட்) க (டிகார	gV
+ _தமிழ்) க (ூற		k
+   _து) க (ாராம		k
+ _நாகர்) க (ோவில	k
  _கோபி) க		kV#
  _ஸ்டா) க		kV#
 _ஜமாய்) க		kV#
      _) ச (ௌஹான	tS
     _ந) ச் (னு		tS
    _கா) ச்மூச		tSmu:tSV#
+    அ) சு (வமேத	s
+    அ) சோக (்		s.o:k
+_பாஸ்கரா) ச (ாரி		tS
+ _பரமா) ச (ாரி		tS
+   _மி) ச (ோராம	z
+     _) சு (வாதி	s
 //endsort
 
 
      அ) ட்லாண்ட		t.la:n.t.V#  // atlant
      _) ட்விங்க (ிள	t.wiNk
      _) ட்வ (ிட்டர	t.w
+   _ஸ்) டி (ரைக்		t.
+     _) ட (ூல		t.
 //endsort
 
 
      _) த (ோஹா		d
      _) தௌலதாப (ாத	daUlVd.a:b
      _) த (்ராட்சை	d
+     _) த (ாவீத		d
+     _) த (ுரியானந்த	d
+     _) த (ாவூத		d
 //endsort
 
 //sort
+   _அர்) த்த (_L32	ddV#
   _நிசப்) த		dV#
   அனந்) த		tV#
 _வசந்) த		tV#
      _) ப் (ளூஸ		b
      _) ப் (ளோரிடா	f
      _) ப் (ளௌஸ	b
+     _) புத்த (கய	buddV
+     _) பாப (ுஜி	ba:p
+     _) பிக (ார		biH
+     _) ப (ிஹார		b
+     _) ப (வ்விய	bV
+     _) பி (ளட்		b
+   L03) ப்ப (ூதகண	bb
+   L03) ப்ப (ூதங்கள	bb
+     _) புளூடூத்	bl.u:t.u:T
+     _) ப்ளூடூத்		bl.u:t.u:T
+     _) பி (ரவுன	b
+     _) ப் (ரவுன		b
+     _) ப (ட்டாச்சாரி	bV
+     _) ப (ிசினச	b
+     _) ப (ிசினெச	b
+     _) ப (ூசுக்த	b
+     _) ப (ேலூர		b
+     _) பி (ரஷர		p
+     _) பி (ரத்தியட்ச	p
+     _) பி (ரத்தியக்ஷ	p
+     _) ப (ோதத்		b
+     _) ப (ோதங்கள	b
+_ஜுகல்_) ப (ந்தி		bV
+     _) புத்த (ியக்க	putt
+     _) ப (ாணம		b
+     _) ப (ாணத்த	b
 //endsort
 
 //sort
+   _சுக) ப (ோக_		b
+    _க) ப (ளீகர		bV
+   _ஜா) ப (ர		fV
+_முரண்) ப (ாட		p
+  _வழி) ப (ார்		p
+  _வழி) ப (ோட		p
+  _பழி) ப (ோட		p
+   _துர்) ப்ப (ுத்தி	bb
+  அவிர்) ப்ப (ாக		bb
+   L03) ப்பி (ரேத	pp
+   L03) ப்பி (ரதே	pp
+_துண்டு) ப்பி (ரசுர	pp
+  _கை) ப்பி (ரசுர	pp
    _நீர்) ப		pV#
    _ப்ரூ) ப		fV#
    இடர்) ப		pV#
      _) வி (யா		v
 	வி (யூக	v
 இண்டர்) வி (யூ		v
-
+ L03வ்) வி (யா		v
 
 .group ஶ
 	ஶ	SV
    _ம) ஸ்த (ூர	zd
    அரி) ஸ (ோன		z
   _சூய) ஸ	zV#
+  _ஹி) ஸ்ப (ுல்லா	zb
 
 .group ஹ
 	ஹ	HV
 	இலேக்க (ிய	le:gg
 	இ (லேச
 	இ (லையோலா
+	இ (ராத்_
+	இ (ராப்பகல
 //endsort
 
 .group உ

phsource/ph_aragon

   FMT(vowel/ee_1)
 endphoneme
 
+phoneme e#
+  vowel starttype #e endtype #e
+  length 200
+  ChangeIfStressed(e)
+  ChangePhoneme(NULL)
+endphoneme
 
 phoneme O
   vowel starttype #o endtype #o

phsource/ph_dutch

   CALL base/z
 endphoneme
 
+
+phoneme l
+  import_phoneme en/l
+endphoneme
+
+phoneme x2
+  import_phoneme af/x2
+endphoneme

phsource/ph_english_us

   Vowelin f1=0  f2=1700 -300 300  f3=-100 80
   Vowelout f1=0 f2=1700 -300 250  f3=-100 80  rms=20
 
-  IF nextPhW(isVowel) AND nextPh(isUnstressed) AND NOT next2PhW(n) THEN
-    IF prevPhW(isVowel) THEN
-      ChangePhoneme(t#)
-    ELIF prevPhW(n) THEN
-      WAV(ustop/t, 60)
+  IF nextPhW(isVowel) AND nextPh(isUnstressed) THEN
+    IF NOT next2PhW(n) OR nextPhW(3:) THEN
+      IF prevPhW(isVowel) AND NOT prevPhW(aI@) THEN
+        ChangePhoneme(t#)
+      ELIF prevPhW(n) THEN
+        WAV(ustop/t, 60)
+      ENDIF
     ENDIF
   ENDIF
 

phsource/ph_malayalam

 endphoneme
 
 
+
+// suggested by Prasenjit Sen, but perhaps not needed?
+// ml_rules changes [p], [t], [k] to [b], [d], [g] in non-initial
+// positions
+
+phoneme p
+  import_phoneme consonants/p-
+  voicingswitch b
+
+  IF thisPh(isWordStart) THEN
+      CALL base/p
+  ELIF thisPh(isWordEnd) AND NOT thisPh(isFirstVowel) THEN
+      ChangePhoneme(b)
+  ENDIF
+endphoneme
+
+
+phoneme t 
+  import_phoneme base/t[
+  voicingswitch d
+
+  IF thisPh(isWordStart) THEN
+      CALL base/t
+  ELIF thisPh(isWordEnd) AND NOT thisPh(isFirstVowel) THEN
+      ChangePhoneme(d)
+  ENDIF
+endphoneme
+
+
+phoneme k
+  import_phoneme consonants/k-
+  voicingswitch g
+
+  IF thisPh(isWordStart) THEN
+      CALL base/k
+  ELIF thisPh(isWordEnd) AND NOT thisPh(isFirstVowel) THEN
+      ChangePhoneme(g)
+  ENDIF
+endphoneme
+
+

phsource/phonemes

 phonemetable as bn
 include ph_assamese
 
+phonemetable or bn
+include ph_oriya
+
+phonemetable mni bn
+include ph_manipuri
 
 
 

src/compiledata.cpp

 
 	sprintf(fname_log,"%s%s",path_dsource,"dict_log");
 	log = fopen(fname_log,"w");
+	if(log != 0)
+	{
+		fprintf(log, "%s", utf8_bom);
+	}
 	sprintf(fname_log,"%s%s",path_dsource,"dict_phonemes");
 	f_phused = fopen(fname_log,"w");
 
 		else
 		{
 			error("Unexpected keyword '%s'",item_string);
+
+			if((strcmp(item_string, "phoneme") == 0) || (strcmp(item_string, "endphoneme") == 0))
+				return(-1);
 		}
 
 		// output the word

src/compiledict.cpp

 	char *mnemptr;
 	unsigned char flag_codes[100];
 	char encoded_ph[200];
-	unsigned char bad_phoneme[4];
+	char bad_phoneme_str[4];
+	int bad_phoneme;
 	static char nullstring[] = {0};
 
 	text_not_phonemes = 0;
 	}
 	else
 	{
-		EncodePhonemes(phonetic,encoded_ph,bad_phoneme);
+		EncodePhonemes(phonetic,encoded_ph,&bad_phoneme);
 		if(strchr(encoded_ph,phonSWITCH) != 0)
 		{
 			flag_codes[n_flag_codes++] = BITNUM_FLAG_ONLY_S;  // don't match on suffixes (except 's') when switching languages
 		}
 
 		// check for errors in the phonemes codes
-		if(bad_phoneme[0] != 0)
+		if(bad_phoneme != 0)
 		{
 			// unrecognised phoneme, report error
-			fprintf(f_log,"%5d: Bad phoneme [%c] (0x%x) in: %s  %s\n",linenum,bad_phoneme[0],bad_phoneme[0],word,phonetic);
+			bad_phoneme_str[utf8_out(bad_phoneme, bad_phoneme_str)] = 0;
+			fprintf(f_log,"%5d: Bad phoneme [%s] (U+%x) in: %s  %s\n",linenum,bad_phoneme_str,bad_phoneme,word,phonetic);
 			error_count++;
 		}
 	}
 	int finish=0;
 	char buf[80];
 	char output[150];
-	unsigned char bad_phoneme[4];
+	int bad_phoneme;
+	char bad_phoneme_str[4];
 
 	buf[0]=0;
 	rule_cond[0]=0;
 		return(NULL);
 	}
 
-	EncodePhonemes(rule_phonemes,buf,bad_phoneme);
-	if(bad_phoneme[0] != 0)
+	EncodePhonemes(rule_phonemes,buf,&bad_phoneme);
+	if(bad_phoneme != 0)
 	{
-		fprintf(f_log,"%5d: Bad phoneme [%c] in %s\n",linenum,bad_phoneme[0],input);
+		bad_phoneme_str[utf8_out(bad_phoneme, bad_phoneme_str)] = 0;
+		fprintf(f_log,"%5d: Bad phoneme [%s] (U+%x) in: %s\n",linenum,bad_phoneme_str,bad_phoneme,input);
 		error_count++;
 	}
 	strcpy(output,buf);

src/dictionary.cpp

 
 
 
-const char *EncodePhonemes(const char *p, char *outptr, unsigned char *bad_phoneme)
-/***************************************************************************/
+const char *EncodePhonemes(const char *p, char *outptr, int *bad_phoneme)
+/******************************************************************/
 /* Translate a phoneme string from ascii mnemonics to internal phoneme numbers,
    from 'p' up to next blank .
    Returns advanced 'p'
 	unsigned int  mnemonic_word;
 
 	if(bad_phoneme != NULL)
-		bad_phoneme[0] = 0;
+		*bad_phoneme = 0;
 
 	// skip initial blanks
 	while(isspace(*p))
 				// not recognised, report and ignore
 				if(bad_phoneme != NULL)
 				{
-					bad_phoneme[0] = *p;
-					bad_phoneme[1] = 0;
+					utf8_in(bad_phoneme, p);
 				}
 				*outptr++ = 0;
 				return(p+1);

src/espeakedit.cpp

 	int debug_flag=0;
 	char fname_log[sizeof(path_dsource)+12];
 	char err_fname[sizeof(path_home)+15];
+	static const char utf8_bom[] = {0xef,0xbb,0xbf,0};
 
 	switch(event.GetId())
 	{
 	case MENU_COMPILE_DICT:
 		sprintf(fname_log,"%s%s",path_dsource,"dict_log");
 		log = fopen(fname_log,"w");
+		if(log != NULL)
+		{
+			fprintf(log, "%s", utf8_bom);
+		}
 
 		LoadDictionary(translator, translator->dictionary_name, 0);
 		if((err = CompileDictionary(path_dsource,translator->dictionary_name,log,err_fname,debug_flag)) < 0)

src/event.cpp

 
 
 //<includes
+#ifndef PLATFORM_WINDOWS
 #include <unistd.h>
+#endif
 #include <assert.h>
 #include <string.h>
 #include <stdlib.h>
 static sem_t my_sem_stop_is_acknowledged;
 // my_thread: polls the audio duration and compares it to the duration of the first event.
 static pthread_t my_thread;
+static bool thread_inited;
 
 static t_espeak_callback* my_callback = NULL;
 static int my_event_is_running=0;
   assert(-1 != sem_init(&my_sem_stop_is_acknowledged, 0, 0));
 
   pthread_attr_t a_attrib;
-  if (pthread_attr_init (& a_attrib)
-      || pthread_attr_setdetachstate(&a_attrib, PTHREAD_CREATE_JOINABLE)
-      || pthread_create( &my_thread,
-			 & a_attrib,
-			 polling_thread,
-			 (void*)NULL))
-    {
-      assert(0);
-    }
 
+  if (pthread_attr_init (&a_attrib) == 0
+      && pthread_attr_setdetachstate(&a_attrib, PTHREAD_CREATE_JOINABLE) == 0)
+  {
+    thread_inited = (0 == pthread_create(&my_thread,
+      &a_attrib,
+      polling_thread,
+      (void*)NULL));
+  }
+  assert(thread_inited);
   pthread_attr_destroy(&a_attrib);
 }
 //>
 		"MARK",
 		"PLAY",
 		"END",
-		"MSG_TERMINATED"
+		"MSG_TERMINATED",
+		"PHONEME",
+		"SAMPLERATE",
+		"??"
 		};
 
 		SHOW("event_display > event=0x%x\n",event);
 {
 ENTER("event_terminate");
 
-	if (my_thread)
+	if (thread_inited)
 	{
 		pthread_cancel(my_thread);
 		pthread_join(my_thread,NULL);
 		sem_destroy(&my_sem_stop_is_required);
 		sem_destroy(&my_sem_stop_is_acknowledged);
 		init(); // purge event
+		thread_inited = 0;
 	}
 }
 

src/fifo.cpp

 
 //<includes
 
+#ifndef PLATFORM_WINDOWS
 #include <unistd.h>
+#endif
 #include <assert.h>
 #include <string.h>
 #include <stdlib.h>

src/numbers.cpp

 				else
 				{
 					// followed by hundreds or thousands etc
-					sprintf(string,"_%da",value);
+					if(tr->langopts.numbers2 & NUM2_ORDINAL_AND_THOUSANDS)
+						sprintf(string, "_%do", value);
+					else
+						sprintf(string, "_%da", value);
 					found = Lookup(tr, string, ph_digits);
 				}
 
 							else if(((control & 2) == 0) || ((tr->langopts.numbers & NUM_SWAP_TENS) != 0))
 							{
 								// followed by hundreds or thousands (or tens)
-								sprintf(string,"_%da",units);
+								if(tr->langopts.numbers2 & NUM2_ORDINAL_AND_THOUSANDS)
+									sprintf(string, "_%do", units);
+								else
+									sprintf(string, "_%da", units);
 								found = Lookup(tr, string, ph_digits);
 							}
 						}

src/phoneme.h

 #define phSINGLE_INSTN  0x1000    // this phoneme has a single instruction program, with an implicit Return
 #define phBRKAFTER 0x4000  // [*] add a post-pause
 
-#define phNONSYLLABIC  0x100000   // don't count this vowel as a syllable when finding the stress position 
+#define phNONSYLLABIC  0x100000   // don't count this vowel as a syllable when finding the stress position
 #define phLONG         0x200000
 #define phLENGTHENSTOP 0x400000  // make the pre-pause slightly longer
 #define phRHOTIC       0x800000  // bit 23
 	unsigned char  end_type;
 	unsigned char  std_length;   // for vowels, in mS/2;  for phSTRESS phonemes, this is the stress/tone type
 	unsigned char  length_mod;   // a length_mod group number, used to access length_mod_tab
-	
+
 } PHONEME_TAB;
 
 
 
 
 
-#define PH(c1,c2)  (c2<<8)+c1          // combine two characters into an integer for phoneme name 
+#define PH(c1,c2)  (c2<<8)+c1          // combine two characters into an integer for phoneme name
 #define PH3(c1,c2,c3) (c3<<16)+(c2<<8)+c1
 #define PhonemeCode2(c1,c2)  PhonemeCode((c2<<8)+c1)
 int LookupPhonemeString(const char *string);
 int PhonemeCode(unsigned int mnem);
 
-const char *EncodePhonemes(const char *p, char *outptr, unsigned char *bad_phoneme);
+const char *EncodePhonemes(const char *p, char *outptr, int *bad_phoneme);
 void DecodePhonemes(const char *inptr, char *outptr);
 
 extern const char *WordToString(unsigned int word);

src/speak_lib.cpp

 #include "speech.h"
 
 #include <sys/stat.h>
-#ifndef PLATFORM_WINDOWS
+#ifdef PLATFORM_WINDOWS
+#include <fcntl.h>
+#include <io.h>
+#include <windows.h>
+#include <winreg.h>
+#else  /* PLATFORM_POSIX */
 #include <unistd.h>
 #endif
 
   "SENTENCE",
   "MARK",
   "PLAY",
-  "END"};
+  "END",
+  "MSG_TERMINATED",
+  "PHONEME",
+  "SAMPLERATE",
+  "??" };
 #endif
 
 

src/synthdata.cpp

 #include "translate.h"
 #include "wave.h"
 
-const char *version_string = "1.47.12  23.Sep.13";
+const char *version_string = "1.47.13  22.Oct.13";
 const int version_phdata  = 0x014709;
 
 int option_device_number = -1;

src/tr_languages.cpp

 				SetupTranslator(tr,stress_lengths_ta, NULL);
 				tr->letter_bits_offset = OFFSET_TAMIL;
 				tr->langopts.numbers =  NUM_OMIT_1_THOUSAND ;
+				tr->langopts.numbers2 = NUM2_ORDINAL_AND_THOUSANDS;
 				tr->langopts.param[LOPT_WORD_MERGE] = 1;   // don't break vowels betwen words
 			}
 			else
 		break;
 
 	case L('u','r'):   // Urdu
+	case L('s','d'):   // Sindhi
 		{
 			tr->letter_bits_offset = OFFSET_ARABIC;
 			tr->langopts.param[LOPT_UNPRONOUNCABLE] = 1;   // disable check for unpronouncable words

src/translate.cpp

 					prefix_flags = 1;
 				if(found == 0)
 				{
-					end_type = TranslateRules(tr, wordx, phonemes, N_WORD_PHONEMES, end_phonemes, 0, dictionary_flags);
+					end_type = TranslateRules(tr, wordx, phonemes, N_WORD_PHONEMES, end_phonemes, wflags & FLAG_HYPHEN_AFTER, dictionary_flags);
 
 					if(phonemes[0] == phonSWITCH)
 					{
 	int ix;
 	int sylimit;        // max. number of syllables in a word to be combined with a preceding preposition
 	const char *new_language;
-	unsigned char bad_phoneme[4];
+	int bad_phoneme;
 	int word_flags;
 	int word_copy_len;
 	char word_copy[N_WORD_BYTES+1];
 		}
 		else
 		{
-			EncodePhonemes(word,word_phonemes,bad_phoneme);
+			EncodePhonemes(word,word_phonemes,&bad_phoneme);
 		}
 		flags = FLAG_FOUND;
 	}

src/translate.h

 #define NUM2_ENGLISH_NUMERALS   0x8000
 #define NUM2_PERCENT_BEFORE     0x10000
 #define NUM2_OMIT_1_HUNDRED_ONLY 0x20000
+#define NUM2_ORDINAL_AND_THOUSANDS 0x40000
 	// bits 1-4  use variant form of numbers before thousands,millions,etc.
 	// bits 6-8  use different forms of thousand, million, etc (M MA MB)
 	// bit9=(LANG=rw) say "thousand" and "million" before its number, not after
 	// bit15=(LANG=ne)  speak (non-replaced) English numerals in English
 	// bit16=(LANG=si)  say "%" before the number
 	// bit17=(LANG=ml)  omit "one" before hundred only if there are no previous digits
+	// bit18=(LANG=ta)  same variant for ordinals and thousands (#o = #a)
 	int numbers2;
 
 #define BREAK_THOUSANDS   0x49249248

src/wave.cpp

 #include <time.h>
 
 #include "portaudio.h"
-#ifndef PLATFORM_WINDOWS
+#ifdef PLATFORM_WINDOWS
+#include <windows.h>
+#else
 #include <unistd.h>
 #endif
 #include "wave.h"
 
 //<Definitions
 
+#ifdef NEED_STRUCT_TIMESPEC
+#define HAVE_STRUCT_TIMESPEC 1
+struct timespec {
+        long tv_sec;
+        long tv_nsec;
+};
+#endif /* HAVE_STRUCT_TIMESPEC */
+
+
 enum {ONE_BILLION=1000000000};
 
 #ifdef USE_PORTAUDIO
 #endif
 
 
- 
- 
+
+
 #ifdef USE_PULSEAUDIO
 // create some wrappers for runtime detection
 
 #define BUFFER_LENGTH (MAX_SAMPLE_RATE*2*sizeof(uint16_t))
 //#define THRESHOLD (BUFFER_LENGTH/5)
 static char myBuffer[BUFFER_LENGTH];
-static char* myRead=NULL; 
-static char* myWrite=NULL; 
+static char* myRead=NULL;
+static char* myWrite=NULL;
 static int out_channels=1;
 static int my_stream_could_start=0;
 static int wave_samplerate;
 
 // time measurement
 // The read and write position audio stream in the audio stream are measured in ms.
-// 
+//
 // * When the stream is opened, myReadPosition and myWritePosition are cleared.
 // * myWritePosition is updated in wave_write.
 // * myReadPosition is updated in pa_callback (+ sample delay).
   PaError err;
   SHOW_TIME("start_stream");
 
-  my_stream_could_start=0;  
+  my_stream_could_start=0;
   mInCallbackFinishedState = false;
 
   err = Pa_StartStream(pa_stream);
   SHOW("start_stream > Pa_StartStream=%d (%s)\n", err, Pa_GetErrorText(err));
-  
+
 #if USE_PORTAUDIO == 19
   if(err == paStreamIsNotStopped)
     {
 				memcpy(outputBuffer, myRead, aUsedMem);
 			}
 			char* p = (char*)outputBuffer + aUsedMem;
-			memset(p, 0, n - aUsedMem); 
+			memset(p, 0, n - aUsedMem);
 			//	  myReadPosition += aUsedMem/(out_channels*sizeof(uint16_t));
 			myRead = aWrite;
 		}
 			myRead = myBuffer + aRest;
 		}
 		else
-		{ 
+		{
 			SHOW_TIME("pa_callback > underflow");
 			aResult=1; // paComplete;
 
 
 			size_t aUsedMem = aTopMem + aRest;
 			char* p = (char*)outputBuffer + aUsedMem;
-			memset(p, 0, n - aUsedMem); 
+			memset(p, 0, n - aUsedMem);
 			//	  myReadPosition += aUsedMem/(out_channels*sizeof(uint16_t));
 			myRead = aWrite;
 		}
 				//paClipOff | paDitherOff,
 				paNoFlag,
 				pa_callback, (void *)userdata);
-   
-      SHOW("wave_open_sound > Pa_OpenDefaultStream(1): err=%d (%s)\n",err, Pa_GetErrorText(err));      
+
+      SHOW("wave_open_sound > Pa_OpenDefaultStream(1): err=%d (%s)\n",err, Pa_GetErrorText(err));
 
       if(err == paInvalidChannelCount)
 	{
 			  //			  paClipOff | paDitherOff,
 			  pa_callback,
 			  (void *)userdata);
-      if ((err!=paNoError) 
+      if ((err!=paNoError)
 	  && (err!=paInvalidChannelCount)) //err==paUnanticipatedHostError
 	{
 	  fprintf(stderr, "wave_open_sound > Pa_OpenStream : err=%d (%s)\n",err,Pa_GetErrorText(err));
   myOutputParameters.channelCount = 1;
   myOutputParameters.sampleFormat = paInt16;
 
-  // Latency greater than 100ms for avoiding glitches 
+  // Latency greater than 100ms for avoiding glitches
   // (e.g. when moving a window in a graphical desktop)
   //  deviceInfo = Pa_GetDeviceInfo(selectedDevice);
   if (deviceInfo)
 		for( i=0; i<numDevices; i++ )
 		{
 		deviceInfo = Pa_GetDeviceInfo( i );
-	
+
 			if (deviceInfo == NULL)
 			{
 				break;
 			}
 			const PaHostApiInfo *hostInfo = Pa_GetHostApiInfo( deviceInfo->hostApi );
-	
+
 			if (hostInfo && hostInfo->type == paALSA)
-			{ 
+			{
 				// Check (once) the default output device
 				if (defaultAlsaIndex == numDevices)
 				{
 					update_output_parameters(defaultAlsaIndex, deviceInfo);
 					if (Pa_IsFormatSupported(NULL, &myOutputParameters, wave_samplerate) == 0)
 					{
-						SHOW( "select_device > ALSA (default), name=%s (#%d)\n", deviceInfo->name, defaultAlsaIndex);		  
+						SHOW( "select_device > ALSA (default), name=%s (#%d)\n", deviceInfo->name, defaultAlsaIndex);
 						selectedIndex = defaultAlsaIndex;
 						selectedDeviceInfo = deviceInfo;
 						break;
 					}
 				}
-	
-				// if the default output device does not match, 
-				// look for the device with the highest number of output channels 
+
+				// if the default output device does not match,
+				// look for the device with the highest number of output channels
 				SHOW( "select_device > ALSA, i=%d (numDevices=%d)\n", i, numDevices);
-	
+
 				update_output_parameters(i, deviceInfo);
-	
+
 				if (Pa_IsFormatSupported(NULL, &myOutputParameters, wave_samplerate) == 0)
 				{
 					SHOW( "select_device > ALSA, name=%s (#%d)\n", deviceInfo->name, i);
-	
+
 					if (!selectedDeviceInfo
 						|| (selectedDeviceInfo->maxOutputChannels < deviceInfo->maxOutputChannels))
 					{
 	{
 		update_output_parameters(selectedIndex, selectedDeviceInfo);
 	}
-	else 
+	else
 	{
 		i = Pa_GetDefaultOutputDevice();
 		deviceInfo = Pa_GetDeviceInfo( i );
 		update_output_parameters(i, deviceInfo);
 	}
 
-#endif    
+#endif
 }
 
 //>
 //<copyBuffer
 
 
-static size_t copyBuffer(char* dest, char* src, const size_t theSizeInBytes) 
-{ 
+static size_t copyBuffer(char* dest, char* src, const size_t theSizeInBytes)
+{
 	size_t bytes_written = 0;
 	unsigned int i = 0;
 	uint16_t* a_dest = NULL;
 	uint16_t* a_src = NULL;
- 
+
 	if ((src != NULL) && dest != NULL)
-	{ 
+	{
 		// copy for one channel (mono)?
 		if(out_channels==1)
-		{ 
+		{
 			SHOW("copyBuffer > 1 channel > memcpy %x (%d bytes)\n", (int)myWrite, theSizeInBytes);
 			memcpy(dest, src, theSizeInBytes);
 			bytes_written = theSizeInBytes;
 			i = 0;
 			a_dest = (uint16_t* )dest;
 			a_src = (uint16_t* )src;
- 
+
 			for(i=0; i<theSizeInBytes/2; i++)
 			{
 				a_dest[2*i] = a_src[i];
 			bytes_written = 2*theSizeInBytes;
 		} // end if(out_channels==1)
 	} // end if ((src != NULL) && dest != NULL)
- 
-	return bytes_written; 
+
+	return bytes_written;
 }
 
 //>
 	// space in ringbuffer for the sample needed: 1x mono channel but 2x for 1 stereo channel
 	size_t bytes_to_write = (out_channels==1) ? theSize : theSize*2;
 	my_stream_could_start = 0;
- 
+
 	if(pa_stream == NULL)
 	{
 		SHOW_TIME("wave_write > wave_open_sound\n");
 		my_stream_could_start = 1;
 	}
 	assert(BUFFER_LENGTH >= bytes_to_write);
- 
+
 	if (myWrite >= myBuffer + BUFFER_LENGTH)
 	{
 		myWrite = myBuffer;
 	} // end if (myWrite >= myBuffer + BUFFER_LENGTH)
- 
+
 	size_t aTotalFreeMem=0;
 	char* aRead = myRead;
 	SHOW("wave_write > aRead=%x, myWrite=%x\n", (int)aRead, (int)myWrite);
- 
+
 	while (1)
 	{
 		if (my_callback_is_output_enabled && (0==my_callback_is_output_enabled()))
 			SHOW_TIME("wave_write > my_callback_is_output_enabled: no!");
 			return 0;
 		}
- 
+
 		aRead = myRead;
- 
+
 		// write pointer is before read pointer?
 		if (myWrite >= aRead)
 		{
 		{
 			aTotalFreeMem = aRead - myWrite;
 		} // end if (myWrite >= aRead)
- 
+
 		if (aTotalFreeMem>1)
 		{
 			// -1 because myWrite must be different of aRead
 			// otherwise buffer would be considered as empty
 			aTotalFreeMem -= 1;
 		} // end if (aTotalFreeMem>1)
- 
+
 		if (aTotalFreeMem >= bytes_to_write)
 		{
 			break;
 		} // end if (aTotalFreeMem >= bytes_to_write)
-		
+
 		//SHOW_TIME("wave_write > wait");
 		SHOW("wave_write > wait: aTotalFreeMem=%d\n", aTotalFreeMem);
 		SHOW("wave_write > aRead=%x, myWrite=%x\n", (int)aRead, (int)myWrite);
 		usleep(10000);
 	} // end while (1)
- 
+
 	aRead = myRead;
- 
+
 	// write pointer is ahead the read pointer?
 	if (myWrite >= aRead)
 	{
 		SHOW_TIME("wave_write > myWrite <= aRead");
 		myWrite += copyBuffer(myWrite, theMono16BitsWaveBuffer, theSize);
 	} // end if (myWrite >= aRead)
- 
+
 	bytes_written = bytes_to_write;
 	myWritePosition += theSize/sizeof(uint16_t); // add number of samples
- 
+
 	if (my_stream_could_start && (get_used_mem() >= out_channels * sizeof(uint16_t) * FRAMES_PER_BUFFER))
 	{
 		start_stream();
 	} // end if (my_stream_could_start && (get_used_mem() >= out_channels * sizeof(uint16_t) * FRAMES_PER_BUFFER))
- 
+
 	SHOW_TIME("wave_write > LEAVE");
- 
+
 	return bytes_written;
 }
 
       SHOW_TIME("wave_close > LEAVE (NULL stream)");
       return 0;
     }
-      
+
   if( Pa_IsStreamStopped( pa_stream ) )
     {
       SHOW_TIME("wave_close > LEAVE (stopped)");
       SHOW_TIME("wave_close > LEAVE (NULL stream)");
       return 0;
     }
-  
+
   if( Pa_StreamActive( pa_stream ) == false && mInCallbackFinishedState == false )
     {
       SHOW_TIME("wave_close > LEAVE (not active)");
       SHOW_TIME("wave_close > LEAVE (stopStreamCount)");
       return 0;
     }
-  
-  // Comment from Audacity-1.2.4b adapted to the eSpeak context. 
+
+  // Comment from Audacity-1.2.4b adapted to the eSpeak context.
   //
   // We got here in one of two ways:
   //
-  // 1. The calling program calls the espeak_Cancel function and we 
-  //    therefore want to stop as quickly as possible.  
+  // 1. The calling program calls the espeak_Cancel function and we
+  //    therefore want to stop as quickly as possible.
   //    So we use AbortStream().  If this is
   //    the case the portaudio stream is still in the Running state
   //    (see PortAudio state machine docs).
   //
   // 2. The callback told PortAudio to stop the stream since it had
-  //    reached the end of the selection.  
-  //    The event polling thread discovered this by noticing that 
+  //    reached the end of the selection.
+  //    The event polling thread discovered this by noticing that
   //    wave_is_busy() returned false.
   //    wave_is_busy() (which calls Pa_GetStreamActive()) will not return
   //    false until all buffers have finished playing, so we can call
   // call StopStream if the callback brought us here, and AbortStream
   // if the user brought us here.
   //
-  
+
 #if (USE_PORTAUDIO == 19)
   if (pa_stream)
     {
   if (pa_stream)
     {
 #if USE_PORTAUDIO == 18
-      active = Pa_StreamActive(pa_stream) 
+      active = Pa_StreamActive(pa_stream)
 	&& (mInCallbackFinishedState == false);
 #else
       active = Pa_IsStreamActive(pa_stream)
 	&& (mInCallbackFinishedState == false);
 #endif
     }
-  
+
   SHOW("wave_is_busy: %d\n",active);
 
 
 
   if (!time || !pa_stream)
     {
-      SHOW("event get_remaining_time> %s\n","audio device not available");	  
+      SHOW("event get_remaining_time> %s\n","audio device not available");
       return -1;
     }
 
     {
       SHOW("event > add_time_in_ms ns: %d sec %Lu nsec \n", ts->tv_sec, t_ns);
       ts->tv_sec += 1;
-      t_ns -= ONE_BILLION;	  
+      t_ns -= ONE_BILLION;
     }
   ts->tv_nsec = (long int)t_ns;
 }