Merge branch 'master' into build

src/compiledict.cpp

@@ -524,7 +524,7 @@ step=1;  // TEST
 				{
 					multiple_numeric_hyphen = 1;
 				}
-				else
+//				else  // ???
 				{
 					flag_codes[n_flag_codes++] = BITNUM_FLAG_HYPHENATED;
 				}
@@ -739,7 +739,7 @@ step=1;  // TEST
 			ix = multiple_string_end - multiple_string;
 			if(multiple_numeric_hyphen)
 			{
-				dict_line[length++] = ' ';
+				dict_line[length++] = ' ';   // ???
 			}
 			memcpy(&dict_line[length],multiple_string,ix);
 			length += ix;

src/dictionary.cpp

@@ -1185,8 +1185,8 @@ void SetWordStress(Translator *tr, char *output, unsigned int *dictionary_flags,
 	switch(tr->langopts.stress_rule)
 	{
 	case 8:
-		// stress on first syllable, unless it is a light syllable
-		if(syllable_weight[1] > 0)
+		// stress on first syllable, unless it is a light syllable followed by a heavy syllable
+		if((syllable_weight[1] > 0) || (syllable_weight[2] == 0))
 			break;
 		// else drop through to case 1
 	case 1:
@@ -3250,7 +3250,7 @@ int LookupDictList(Translator *tr, char **wordptr, char *ph_out, unsigned int *f
 		}
 	}
 
-	for(length=0; length<N_WORD_BYTES; length++)
+	for(length=0; length<(N_WORD_BYTES-1); length++)
 	{
 		if(((c = *word1++)==0) || (c == ' '))
 			break;

src/phonemelist.cpp

@@ -171,6 +171,7 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
 		// set consonant clusters to all voiced or all unvoiced
 		// Regressive
 		int type;
+		int word_end_devoice = 0;
 		voicing = 0;
 
 		for(j=n_ph_list2-1; j>=0; j--)
@@ -190,7 +191,12 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
 			{
 				// LANG=Russian, [v] amd [v;] don't cause regression, or [R^]
 				if((ph->mnemonic == 'v') || (ph->mnemonic == ((';'<<8)+'v')) || ((ph->mnemonic & 0xff)== 'R'))
-					type = phLIQUID;
+				{
+					if(word_end_devoice == 1)
+						voicing = 0;
+					else
+						type = phLIQUID;
+				}
 			}
 
 			if((type==phSTOP) || type==(phFRICATIVE))
@@ -231,6 +237,8 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
 					voicing = 0;
 				}
 			}
+
+			word_end_devoice = 0;
 			if(plist2[j].sourceix)
 			{
 				if(regression & 0x04)
@@ -242,7 +250,10 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
 				{
 					// devoice word-final consonants, unless propagating voiced
 					if(voicing == 0)
+					{
 						voicing = 1;
+						word_end_devoice = 1;
+					}
 				}
 			}
 		}

src/readclause.cpp

@@ -364,10 +364,10 @@ int towlower2(unsigned int c)
 		{
 			c = 0x131;   // I -> ı
 		}
-	}
-#ifdef __WIN32__
-	if(c == 0x130)   // uppercase i-dot
-		return('i');
+	}
+#ifdef __WIN32__
+	if(c == 0x130)   // uppercase i-dot
+		return('i');
 #endif
 	return(towlower(c));
 }
@@ -901,14 +901,21 @@ static int AnnouncePunctuation(Translator *tr, int c1, int *c2_ptr, char *output
 			else
 			if(punct_count < 4)
 			{
-				sprintf(buf,"\001+10S");
+				buf[0] = 0;
+				if(embedded_value[EMBED_S] < 300)
+					sprintf(buf,"\001+10S");  // Speak punctuation name faster, unless we are already speaking fast.  It would upset Sonic SpeedUp
+
 				while(punct_count-- > 0)
 				{
 					sprintf(buf2," %s",punctname);
 					strcat(buf, buf2);
 				}
-				sprintf(buf2," \001-10S");
-				strcat(buf, buf2);
+
+				if(embedded_value[EMBED_S] < 300)
+				{
+					sprintf(buf2," \001-10S");
+					strcat(buf, buf2);
+				}
 			}
 			else
 			{
@@ -2000,6 +2007,15 @@ terminator=0;  // ??  Sentence intonation, but no pause ??
 }  // end of ProcessSsmlTag
 
 
+static void RemoveChar(char *p)
+{//=======================
+// Replace a UTF-8 character by spaces
+	int c;
+
+	memset(p, ' ', utf8_in(&c, p));
+}  // end of RemoveChar
+
+
 static MNEM_TAB xml_char_mnemonics[] = {
 	{"gt",'>'},
 	{"lt", 0xe000 + '<'},   // private usage area, to avoid confusion with XML tag
@@ -2354,6 +2370,15 @@ f_input = f_in;  // for GetC etc
 					}
 				}
 			}
+
+			if(c1 == 0xd4d)
+			{
+				// Malayalam virama, check if next character is Zero-width-joiner
+				if(c2 == 0x200d)
+				{
+					c1 = 0xd4e;   // use this unofficial code for chillu-virama
+				}
+			}
 		}
 
 		if(iswupper(c1))
@@ -2408,8 +2433,9 @@ f_input = f_in;  // for GetC etc
 				UngetC(c2);
 
 				if(end_clause_after_tag)
-					ix = end_clause_index;  // delete clause-end punctiation
-
+				{
+					RemoveChar(&buf[end_clause_index]);  // delete clause-end punctiation
+				}
 				buf[ix] = ' ';
 				buf[ix+1] = 0;
 				if(parag > 3)
@@ -2485,7 +2511,7 @@ if(option_ssml) parag=1;
 					continue;
 				}
 
-				if((iswspace(c2) || (punct_data & 0x8000) || IsBracket(c2) || (c2=='?') || Eof()))    // don't check for '-' because if prevent recognizing ':-)'
+				if((iswspace(c2) || (punct_data & 0x8000) || IsBracket(c2) || (c2=='?') || Eof() || (c2 == ctrl_embedded)))    // don't check for '-' because it prevents recognizing ':-)'
 //				if((iswspace(c2) || (punct_data & 0x8000) || IsBracket(c2) || (c2=='?') || (c2=='-') || Eof()))
 				{
 					// note: (c2='?') is for when a smart-quote has been replaced by '?'
@@ -2682,7 +2708,7 @@ if(option_ssml) parag=1;
 	}
 	if(end_clause_after_tag)
 	{
-		ix = end_clause_index;  // delete clause-end punctuation
+		RemoveChar(&buf[end_clause_index]);  // delete clause-end punctiation
 	}
 	buf[ix] = ' ';
 	buf[ix+1] = 0;
@@ -2707,6 +2733,7 @@ void InitText2(void)
 	int param;
 
 	ungot_char = 0;
+	ungot_char2 = 0;
 
 	n_ssml_stack =1;
 	n_param_stack = 1;

src/setlengths.cpp

@@ -141,6 +141,7 @@ void SetSpeed(int control)
 	int s1;
 	int wpm;
 	int wpm2;
+	int wpm_value;
 	double sonic;
 
 	speed.loud_consonants = 0;
@@ -153,6 +154,8 @@ void SetSpeed(int control)
 	if(control == 2)
 		wpm = embedded_value[EMBED_S2];
 
+	wpm_value = wpm;
+
 	if(voice->speed_percent > 0)
 	{
 		wpm = (wpm * voice->speed_percent)/100;
@@ -162,7 +165,7 @@ void SetSpeed(int control)
 	{
 		DoSonicSpeed(1 * 1024);
 	}
-	if((wpm > 450) || ((wpm > speed.fast_settings[0]) && (wpm > 350)))
+	if((wpm_value > 450) || ((wpm_value > speed.fast_settings[0]) && (wpm > 350)))
 	{
 		wpm2 = wpm;
 		wpm = 175;

src/synth_mbrola.cpp

@@ -151,7 +151,10 @@ espeak_ERROR LoadMbrolaTable(const char *mbrola_voice, const char *phtrans, int
 #endif
 #ifdef PLATFORM_WINDOWS
 	if(load_MBR() == FALSE)     // load mbrola.dll
-		return(EE_INTERNAL_ERROR); 
+	{
+		fprintf(stderr, "Can't load mbrola.dll\n");
+		return(EE_INTERNAL_ERROR);
+	} 
 #endif
 
 	if(init_MBR(path) != 0)      // initialise the required mbrola voice

src/synthdata.cpp

@@ -35,7 +35,7 @@
 #include "translate.h"
 #include "wave.h"
 
-const char *version_string = "1.45.12  01.Apr.11";
+const char *version_string = "1.45.04  25.Apr.11";
 const int version_phdata  = 0x014500;
 
 int option_device_number = -1;

src/tr_languages.cpp

@@ -385,6 +385,7 @@ Translator *SelectTranslator(const char *name)
 
 	case L('a','r'):   // Arabic
 			tr->letter_bits_offset = OFFSET_ARABIC;
+			tr->langopts.numbers = NUM_SWAP_TENS | NUM_AND_UNITS | NUM_HUNDRED_AND | NUM_OMIT_1_HUNDRED | NUM_AND_HUNDRED | NUM_THOUSAND_AND | NUM_OMIT_1_THOUSAND;
 			tr->langopts.param[LOPT_UNPRONOUNCABLE] = 1;   // disable check for unpronouncable words
 		break;
 
@@ -393,7 +394,7 @@ Translator *SelectTranslator(const char *name)
 			SetCyrillicLetters(tr);
 			SetLetterVowel(tr,0x2a);
 			tr->langopts.param[LOPT_UNPRONOUNCABLE] = 0x432;    // [v]  don't count this character at start of word
-			tr->langopts.param[LOPT_REGRESSIVE_VOICING] = 0x10;  // devoice at end of word
+			tr->langopts.param[LOPT_REGRESSIVE_VOICING] = 0x17;  // devoice at end of word, and change voicing to match a following consonant (except v)
 			tr->langopts.param[LOPT_REDUCE] = 2;
 			tr->langopts.stress_rule = STRESSPOSN_2R;
 			tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_ALLOW_SPACE | NUM_OMIT_1_HUNDRED | NUM_HUNDRED_AND | NUM_AND_UNITS | NUM_SINGLE_AND | NUM_ROMAN | NUM_ROMAN_ORDINAL | NUM_ROMAN_CAPITALS ;
@@ -470,7 +471,8 @@ Translator *SelectTranslator(const char *name)
 			tr->langopts.param[LOPT_REGRESSIVE_VOICING] = 0x10;  // devoice at end of word
 			tr->langopts.param[LOPT_LONG_VOWEL_THRESHOLD] = 175/2;
 		
-			tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_OMIT_1_HUNDRED | NUM_OMIT_1_THOUSAND | NUM_ALLOW_SPACE | NUM_ORDINAL_DOT | NUM_ROMAN;
+			tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_ALLOW_SPACE | NUM_ORDINAL_DOT | NUM_ROMAN;
+//			tr->langopts.numbers = NUM_DECIMAL_COMMA | NUM_SWAP_TENS | NUM_OMIT_1_HUNDRED | NUM_OMIT_1_THOUSAND | NUM_ALLOW_SPACE | NUM_ORDINAL_DOT | NUM_ROMAN;
 			SetLetterVowel(tr,'y');
 			tr->langopts.param[LOPT_UNPRONOUNCABLE] = 2;   // use de_rules for unpronouncable rules
 		}
@@ -1218,6 +1220,7 @@ SetLengthMods(tr,3);  // all equal
 			}
 			tr->langopts.param[LOPT_WORD_MERGE] = 1;   // don't break vowels betwen words
 			SetIndicLetters(tr);   // call this after setting OFFSET_
+				SetLetterBitsRange(tr,LETTERGP_B,0x4e,0x4e);   // chillu-virama (unofficial)
 		}
 		break;
 

src/translate.cpp

@@ -122,7 +122,7 @@ static const unsigned short brackets[] = {
 static const unsigned short breaks[] = {'_', 0};
 
 // treat these characters as spaces, in addition to iswspace()
-static const wchar_t chars_space[] = {0x2500,0};  // box drawing horiz
+// static const wchar_t chars_space[] = {0x2500,0x2501,0};  // box drawing horiz
 
 
 // Translate character codes 0xA0 to 0xFF into their unicode values
@@ -432,8 +432,10 @@ int IsSpace(unsigned int c)
 {//========================
 	if(c == 0)
 		return(0);
-	if(wcschr(chars_space,c))
-		return(1);
+	if((c >= 0x2500) && (c < 0x25a0))
+		return(1);  // box drawing characters
+//	if(wcschr(chars_space,c))
+//		return(1);
 	return(iswspace(c));
 }
 
@@ -867,6 +869,8 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
 	}
 
 	word_copy_length = wordx - word_start;
+	if(word_copy_length >= N_WORD_BYTES)
+		word_copy_length = N_WORD_BYTES-1;
 	memcpy(word_copy2, word_start, word_copy_length);
 
 	spell_word = 0;
@@ -3017,7 +3021,8 @@ if((c == '/') && (tr->langopts.testing & 2) && IsDigit09(next_in) && IsAlpha(pre
 		}
 		else
 		{
-			ix += utf8_out(c,&sbuf[ix]);   // sbuf[ix++] = c;
+			if((ix < (N_TR_SOURCE - 4)))
+				ix += utf8_out(c,&sbuf[ix]);   // sbuf[ix++] = c;
 		}
 		if(pre_pause_add > pre_pause)
 			pre_pause = pre_pause_add;

src/wavegen.cpp

@@ -1540,14 +1540,32 @@ static int SetWithRange0(int value, int max)
 }
 
 
+static void SetPitchFormants()
+{//===========================
+	int ix;
+	int factor;
+	int pitch_value;
+
+	// adjust formants to give better results for a different voice pitch
+	if((pitch_value = embedded_value[EMBED_P]) > MAX_PITCH_VALUE)
+		pitch_value = MAX_PITCH_VALUE;
+
+	factor = 256 + (25 * (pitch_value - 50))/50;
+	for(ix=0; ix<=5; ix++)
+	{
+		wvoice->freq[ix] = (wvoice->freq2[ix] * factor)/256;
+	}
+	factor = embedded_value[EMBED_T]*3;
+	wvoice->height[0] = (wvoice->height2[0] * (256 - factor*2))/256;
+	wvoice->height[1] = (wvoice->height2[1] * (256 - factor))/256;
+}
+
+
 void SetEmbedded(int control, int value)
 {//=====================================
 	// there was an embedded command in the text at this point
 	int sign=0;
 	int command;
-	int ix;
-	int factor;
-	int pitch_value;
 
 	command = control & 0x1f;
 	if((control & 0x60) == 0x60)
@@ -1570,18 +1588,7 @@ void SetEmbedded(int control, int value)
 	case EMBED_T:
 		WavegenSetEcho();   // and drop through to case P
 	case EMBED_P:
-		// adjust formants to give better results for a different voice pitch
-		if((pitch_value = embedded_value[EMBED_P]) > MAX_PITCH_VALUE)
-			pitch_value = MAX_PITCH_VALUE;
-
-		factor = 256 + (25 * (pitch_value - 50))/50;
-		for(ix=0; ix<=5; ix++)
-		{
-			wvoice->freq[ix] = (wvoice->freq2[ix] * factor)/256;
-		}
-		factor = embedded_value[EMBED_T]*3;
-		wvoice->height[0] = (wvoice->height2[0] * (256 - factor*2))/256;
-		wvoice->height[1] = (wvoice->height2[1] * (256 - factor))/256;
+		SetPitchFormants();
 		break;
 
 	case EMBED_A:  // amplitude
@@ -1618,6 +1625,7 @@ void WavegenSetVoice(voice_t *v)
 		option_harmonic1 = 6;
 	}
 	WavegenSetEcho();
+	SetPitchFormants();
 	MarkerEvent(espeakEVENT_SAMPLERATE,0,wvoice->samplerate,out_ptr);
 //	WVoiceChanged(wvoice);
 }