Fix GCC -Wnarrowing warnings.

src/libespeak-ng/tr_languages.c

@@ -184,7 +184,8 @@ static const unsigned short chars_ignore_zwnj_hyphen[] = {
 	0, 0
 };
 
-const char string_ordinal[] = { 0xc2, 0xba, 0 }; // masculine ordinal character, UTF-8
+const unsigned char utf8_ordinal[] = { 0xc2, 0xba, 0 }; // masculine ordinal character, UTF-8
+const unsigned char utf8_null[] = { 0 }; // null string, UTF-8
 
 static Translator *NewTranslator(void)
 {
@@ -278,7 +279,7 @@ static Translator *NewTranslator(void)
 	tr->langopts.replace_chars = NULL;
 	tr->langopts.ascii_language[0] = 0; // Non-Latin alphabet languages, use this language to speak Latin words, default is English
 	tr->langopts.alt_alphabet_lang = L('e', 'n');
-	tr->langopts.roman_suffix = "";
+	tr->langopts.roman_suffix = utf8_null;
 
 	SetLengthMods(tr, 201);
 
@@ -720,7 +721,7 @@ Translator *SelectTranslator(const char *name)
 			tr->langopts.stress_flags = S_FINAL_SPANISH | S_FINAL_DIM_ONLY | S_FINAL_NO_2;
 			tr->langopts.numbers = NUM_SINGLE_STRESS | NUM_DECIMAL_COMMA | NUM_AND_UNITS | NUM_OMIT_1_HUNDRED | NUM_OMIT_1_THOUSAND | NUM_ROMAN | NUM_ROMAN_ORDINAL;
 			tr->langopts.numbers2 = NUM2_ORDINAL_NO_AND;
-			tr->langopts.roman_suffix = string_ordinal;
+			tr->langopts.roman_suffix = utf8_ordinal;
 		} else if (name2 == L_pap) {
 			// stress last syllable unless word ends with a vowel
 			tr->langopts.stress_rule = STRESSPOSN_1R;

src/libespeak-ng/translate.h

@@ -541,7 +541,7 @@ typedef struct {
 	int decimal_sep;
 	int max_digits;    // max number of digits which can be spoken as an integer number (rather than individual digits)
 	const char *ordinal_indicator;   // UTF-8 string
-	const char *roman_suffix;    // add this (ordinal) suffix to Roman numbers (LANG=an)
+	const unsigned char *roman_suffix;    // add this (ordinal) suffix to Roman numbers (LANG=an)
 
 	// bit 0, accent name before the letter name, bit 1 "capital" after letter name
 	int accents;