From: tomo Date: Fri, 1 Oct 1999 05:27:38 +0000 (+0000) Subject: (Vcharset_hiragana_jisx0208): New variable. X-Git-Tag: r21-2-19-utf-2000-0_8-2~25 X-Git-Url: http://git.chise.org/gitweb/?a=commitdiff_plain;h=b150fee929ac6d9737f38b53768d0708d667d037;p=chise%2Fxemacs-chise.git.1 (Vcharset_hiragana_jisx0208): New variable. (Vcharset_katakana_jisx0208): New variable. (Qhiragana_jisx0208): New variable. (Qkatakana_jisx0208): New variable. (make_charset): Add new argument `byte_offset'. (charset_get_byte1): Modify for new coded-charset definition; use `XCHARSET_UCS_MIN', `XCHARSET_UCS_MAX', `XCHARSET_CODE_OFFSET' and `XCHARSET_BYTE_OFFSET'. (Fmake_charset): Modify for `make_charset'. (Fmake_reverse_direction_charset): Likewise. (syms_of_mule_charset): Add new symbols `hiragana-jisx0208' and `katakana-jisx0208'. (complex_vars_of_mule_charset): Modify for `make_charset'; quote `.' in font registry of charset `katakana-jisx0201', `latin-jisx0201', `vietnamese-viscii-lower' and `vietnamese-viscii-upper'; modify DOC-string of charset `japanese-jisx0208-1978' and `japanese-jisx0208'; modify font registry of charset `japanese-jisx0208' not to use font for JIS X0208:1990; add new charset `hiragana-jisx0208' and `katakana-jisx0208'. --- diff --git a/src/mule-charset.c b/src/mule-charset.c index aca5fcb..6177211 100644 --- a/src/mule-charset.c +++ b/src/mule-charset.c @@ -61,6 +61,8 @@ Lisp_Object Vcharset_chinese_cns11643_2; Lisp_Object Vcharset_ucs_bmp; Lisp_Object Vcharset_latin_viscii_lower; Lisp_Object Vcharset_latin_viscii_upper; +Lisp_Object Vcharset_hiragana_jisx0208; +Lisp_Object Vcharset_katakana_jisx0208; #endif Lisp_Object Vcharset_chinese_big5_1; Lisp_Object Vcharset_chinese_big5_2; @@ -342,6 +344,8 @@ Lisp_Object Qascii, Qucs_bmp, Qlatin_viscii_lower, Qlatin_viscii_upper, + Qhiragana_jisx0208, + Qkatakana_jisx0208, #endif Qchinese_big5_1, Qchinese_big5_2, @@ -769,7 +773,8 @@ make_charset (Charset_ID id, Lisp_Object name, Lisp_Object long_name, Lisp_Object doc, Lisp_Object reg, Lisp_Object decoding_table, - Emchar ucs_min, Emchar ucs_max, Emchar code_offset) + Emchar ucs_min, Emchar ucs_max, + Emchar code_offset, unsigned char byte_offset) { Lisp_Object obj; struct Lisp_Charset *cs = @@ -794,6 +799,7 @@ make_charset (Charset_ID id, Lisp_Object name, CHARSET_UCS_MIN(cs) = ucs_min; CHARSET_UCS_MAX(cs) = ucs_max; CHARSET_CODE_OFFSET(cs) = code_offset; + CHARSET_BYTE_OFFSET(cs) = byte_offset; #endif switch ( CHARSET_TYPE (cs) ) @@ -893,8 +899,11 @@ make_charset (Charset_ID id, Lisp_Object name, ASCII, Control-1, Composite, and the two faux private charsets. */ #if UTF2000 - assert (NILP (charset_by_attributes[type][final])); - charset_by_attributes[type][final] = obj; + if (code_offset == 0) + { + assert (NILP (charset_by_attributes[type][final])); + charset_by_attributes[type][final] = obj; + } #else assert (NILP (charset_by_attributes[type][final][direction])); charset_by_attributes[type][final][direction] = obj; @@ -953,44 +962,61 @@ charset_get_byte1 (Lisp_Object charset, Emchar ch) if ((table = XCHARSET_TO_BYTE1_TABLE (charset)) != NULL) return get_byte_from_character_table (ch, table); - else if ((CHARSET_UCS_MIN (XCHARSET (charset)) <= ch) - && (ch <= CHARSET_UCS_MAX (XCHARSET (charset)))) - return ch - CHARSET_UCS_MIN (XCHARSET (charset)) - + CHARSET_CODE_OFFSET (XCHARSET (charset)); - else if (XCHARSET_DIMENSION (charset) == 1) - { - if (XCHARSET_CHARS (charset) == 94) - { - if (((d = ch - (MIN_CHAR_94 - + (XCHARSET_FINAL (charset) - '0') * 94)) >= 0) - && (d < 94)) - return d + 33; - } - else if (XCHARSET_CHARS (charset) == 96) - { - if (((d = ch - (MIN_CHAR_96 - + (XCHARSET_FINAL (charset) - '0') * 96)) >= 0) - && (d < 96)) - return d + 32; - } - else - return 0; - } - else if (XCHARSET_DIMENSION (charset) == 2) + else if ((XCHARSET_UCS_MIN (charset) <= ch) + && (ch <= XCHARSET_UCS_MAX (charset))) + return (ch - XCHARSET_UCS_MIN (charset) + + XCHARSET_CODE_OFFSET (charset)) + / (XCHARSET_DIMENSION (charset) == 1 ? + 1 + : + XCHARSET_DIMENSION (charset) == 2 ? + XCHARSET_CHARS (charset) + : + XCHARSET_DIMENSION (charset) == 3 ? + XCHARSET_CHARS (charset) * XCHARSET_CHARS (charset) + : + XCHARSET_CHARS (charset) + * XCHARSET_CHARS (charset) * XCHARSET_CHARS (charset)) + + XCHARSET_BYTE_OFFSET (charset); + else if (XCHARSET_CODE_OFFSET (charset) == 0) { - if (XCHARSET_CHARS (charset) == 94) + if (XCHARSET_DIMENSION (charset) == 1) { - if (((d = ch - (MIN_CHAR_94x94 - + (XCHARSET_FINAL (charset) - '0') * 94 * 94)) >= 0) - && (d < 94 * 94)) - return (d / 94) + 33; + if (XCHARSET_CHARS (charset) == 94) + { + if (((d = ch - (MIN_CHAR_94 + + (XCHARSET_FINAL (charset) - '0') * 94)) >= 0) + && (d < 94)) + return d + 33; + } + else if (XCHARSET_CHARS (charset) == 96) + { + if (((d = ch - (MIN_CHAR_96 + + (XCHARSET_FINAL (charset) - '0') * 96)) >= 0) + && (d < 96)) + return d + 32; + } + else + return 0; } - else if (XCHARSET_CHARS (charset) == 96) + else if (XCHARSET_DIMENSION (charset) == 2) { - if (((d = ch - (MIN_CHAR_96x96 - + (XCHARSET_FINAL (charset) - '0') * 96 * 96)) >= 0) - && (d < 96 * 96)) - return (d / 96) + 32; + if (XCHARSET_CHARS (charset) == 94) + { + if (((d = ch - (MIN_CHAR_94x94 + + (XCHARSET_FINAL (charset) - '0') * 94 * 94)) + >= 0) + && (d < 94 * 94)) + return (d / 94) + 33; + } + else if (XCHARSET_CHARS (charset) == 96) + { + if (((d = ch - (MIN_CHAR_96x96 + + (XCHARSET_FINAL (charset) - '0') * 96 * 96)) + >= 0) + && (d < 96 * 96)) + return (d / 96) + 32; + } } } return 0; @@ -1007,8 +1033,19 @@ charset_get_byte2 (Lisp_Object charset, Emchar ch) if ((table = XCHARSET_TO_BYTE2_TABLE (charset)) != NULL) return get_byte_from_character_table (ch, table); - else if (EQ (charset, Vcharset_ucs_bmp)) - return (ch >> 8) & 0xff; + else if ((XCHARSET_UCS_MIN (charset) <= ch) + && (ch <= XCHARSET_UCS_MAX (charset))) + return ((ch - XCHARSET_UCS_MIN (charset) + + XCHARSET_CODE_OFFSET (charset)) + / (XCHARSET_DIMENSION (charset) == 2 ? + 1 + : + XCHARSET_DIMENSION (charset) == 3 ? + XCHARSET_CHARS (charset) + : + XCHARSET_CHARS (charset) * XCHARSET_CHARS (charset))) + % XCHARSET_CHARS (charset) + + XCHARSET_BYTE_OFFSET (charset); else if (XCHARSET_CHARS (charset) == 94) return (MIN_CHAR_94x94 + (XCHARSET_FINAL (charset) - '0') * 94 * 94 <= ch) @@ -1336,7 +1373,7 @@ character set. Recognized properties are: charset = make_charset (id, name, type, columns, graphic, final, direction, short_name, long_name, doc_string, registry, - Qnil, 0, 0, 0); + Qnil, 0, 0, 0, 0); if (!NILP (ccl_program)) XCHARSET_CCL_PROGRAM (charset) = ccl_program; return charset; @@ -1388,7 +1425,8 @@ NEW-NAME is the name of the new charset. Return the new charset. CHARSET_DECODING_TABLE(cs), CHARSET_UCS_MIN(cs), CHARSET_UCS_MAX(cs), - CHARSET_CODE_OFFSET(cs) + CHARSET_CODE_OFFSET(cs), + CHARSET_BYTE_OFFSET(cs) #else Qnil, 0, 0, 0 #endif @@ -1856,6 +1894,8 @@ syms_of_mule_charset (void) defsymbol (&Qucs_bmp, "ucs-bmp"); defsymbol (&Qlatin_viscii_lower, "vietnamese-viscii-lower"); defsymbol (&Qlatin_viscii_upper, "vietnamese-viscii-upper"); + defsymbol (&Qhiragana_jisx0208, "hiragana-jisx0208"); + defsymbol (&Qkatakana_jisx0208, "katakana-jisx0208"); #endif defsymbol (&Qchinese_big5_1, "chinese-big5-1"); defsymbol (&Qchinese_big5_2, "chinese-big5-2"); @@ -3163,7 +3203,7 @@ complex_vars_of_mule_charset (void) build_string ("BMP"), build_string ("BMP"), build_string (""), - Qnil, 0, 0xFFFF, 0); + Qnil, 0, 0xFFFF, 0, 0); #else # define latin_iso8859_2_to_ucs NULL # define latin_iso8859_3_to_ucs NULL @@ -3189,7 +3229,7 @@ complex_vars_of_mule_charset (void) build_string ("ASCII)"), build_string ("ASCII (ISO646 IRV)"), build_string ("\\(iso8859-[0-9]*\\|-ascii\\)"), - Qnil, 0, 0x7F, 0); + Qnil, 0, 0x7F, 0, 0); Vcharset_control_1 = make_charset (LEADING_BYTE_CONTROL_1, Qcontrol_1, CHARSET_TYPE_94, 1, 1, 0, @@ -3198,7 +3238,7 @@ complex_vars_of_mule_charset (void) build_string ("Control characters"), build_string ("Control characters 128-191"), build_string (""), - Qnil, 0x80, 0x9F, 0); + Qnil, 0x80, 0x9F, 0, 0); Vcharset_latin_iso8859_1 = make_charset (LEADING_BYTE_LATIN_ISO8859_1, Qlatin_iso8859_1, CHARSET_TYPE_96, 1, 1, 'A', @@ -3207,7 +3247,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-1 (Latin-1)"), build_string ("ISO8859-1 (Latin-1)"), build_string ("iso8859-1"), - Qnil, 0xA0, 0xFF, 32); + Qnil, 0xA0, 0xFF, 0, 32); Vcharset_latin_iso8859_2 = make_charset (LEADING_BYTE_LATIN_ISO8859_2, Qlatin_iso8859_2, CHARSET_TYPE_96, 1, 1, 'B', @@ -3216,7 +3256,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-2 (Latin-2)"), build_string ("ISO8859-2 (Latin-2)"), build_string ("iso8859-2"), - latin_iso8859_2_to_ucs, 0, 0, 32); + latin_iso8859_2_to_ucs, 0, 0, 0, 32); Vcharset_latin_iso8859_3 = make_charset (LEADING_BYTE_LATIN_ISO8859_3, Qlatin_iso8859_3, CHARSET_TYPE_96, 1, 1, 'C', @@ -3225,7 +3265,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-3 (Latin-3)"), build_string ("ISO8859-3 (Latin-3)"), build_string ("iso8859-3"), - latin_iso8859_3_to_ucs, 0, 0, 32); + latin_iso8859_3_to_ucs, 0, 0, 0, 32); Vcharset_latin_iso8859_4 = make_charset (LEADING_BYTE_LATIN_ISO8859_4, Qlatin_iso8859_4, CHARSET_TYPE_96, 1, 1, 'D', @@ -3234,7 +3274,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-4 (Latin-4)"), build_string ("ISO8859-4 (Latin-4)"), build_string ("iso8859-4"), - latin_iso8859_4_to_ucs, 0, 0, 32); + latin_iso8859_4_to_ucs, 0, 0, 0, 32); Vcharset_thai_tis620 = make_charset (LEADING_BYTE_THAI_TIS620, Qthai_tis620, CHARSET_TYPE_96, 1, 1, 'T', @@ -3243,7 +3283,7 @@ complex_vars_of_mule_charset (void) build_string ("TIS620 (Thai)"), build_string ("TIS620.2529 (Thai)"), build_string ("tis620"), - Qnil, MIN_CHAR_THAI, MAX_CHAR_THAI, 32); + Qnil, MIN_CHAR_THAI, MAX_CHAR_THAI, 0, 32); Vcharset_greek_iso8859_7 = make_charset (LEADING_BYTE_GREEK_ISO8859_7, Qgreek_iso8859_7, CHARSET_TYPE_96, 1, 1, 'F', @@ -3252,7 +3292,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-7 (Greek)"), build_string ("ISO8859-7 (Greek)"), build_string ("iso8859-7"), - Qnil, MIN_CHAR_GREEK, MAX_CHAR_GREEK, 32); + Qnil, MIN_CHAR_GREEK, MAX_CHAR_GREEK, 0, 32); Vcharset_arabic_iso8859_6 = make_charset (LEADING_BYTE_ARABIC_ISO8859_6, Qarabic_iso8859_6, CHARSET_TYPE_96, 1, 1, 'G', @@ -3261,7 +3301,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-6 (Arabic)"), build_string ("ISO8859-6 (Arabic)"), build_string ("iso8859-6"), - Qnil, 0, 0, 32); + Qnil, 0, 0, 0, 32); Vcharset_hebrew_iso8859_8 = make_charset (LEADING_BYTE_HEBREW_ISO8859_8, Qhebrew_iso8859_8, CHARSET_TYPE_96, 1, 1, 'H', @@ -3270,7 +3310,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-8 (Hebrew)"), build_string ("ISO8859-8 (Hebrew)"), build_string ("iso8859-8"), - Qnil, MIN_CHAR_HEBREW, MAX_CHAR_HEBREW, 32); + Qnil, MIN_CHAR_HEBREW, MAX_CHAR_HEBREW, 0, 32); Vcharset_katakana_jisx0201 = make_charset (LEADING_BYTE_KATAKANA_JISX0201, Qkatakana_jisx0201, CHARSET_TYPE_94, 1, 1, 'I', @@ -3278,10 +3318,10 @@ complex_vars_of_mule_charset (void) build_string ("JISX0201 Kana"), build_string ("JISX0201.1976 (Japanese Kana)"), build_string ("JISX0201.1976 Japanese Kana"), - build_string ("jisx0201.1976"), + build_string ("jisx0201\\.1976"), Qnil, MIN_CHAR_HALFWIDTH_KATAKANA, - MAX_CHAR_HALFWIDTH_KATAKANA, 33); + MAX_CHAR_HALFWIDTH_KATAKANA, 0, 33); Vcharset_latin_jisx0201 = make_charset (LEADING_BYTE_LATIN_JISX0201, Qlatin_jisx0201, CHARSET_TYPE_94, 1, 0, 'J', @@ -3289,8 +3329,8 @@ complex_vars_of_mule_charset (void) build_string ("JISX0201 Roman"), build_string ("JISX0201.1976 (Japanese Roman)"), build_string ("JISX0201.1976 Japanese Roman"), - build_string ("jisx0201.1976"), - latin_jisx0201_to_ucs, 0, 0, 33); + build_string ("jisx0201\\.1976"), + latin_jisx0201_to_ucs, 0, 0, 0, 33); Vcharset_cyrillic_iso8859_5 = make_charset (LEADING_BYTE_CYRILLIC_ISO8859_5, Qcyrillic_iso8859_5, CHARSET_TYPE_96, 1, 1, 'L', @@ -3299,7 +3339,7 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-5 (Cyrillic)"), build_string ("ISO8859-5 (Cyrillic)"), build_string ("iso8859-5"), - Qnil, MIN_CHAR_CYRILLIC, MAX_CHAR_CYRILLIC, 32); + Qnil, MIN_CHAR_CYRILLIC, MAX_CHAR_CYRILLIC, 0, 32); Vcharset_latin_iso8859_9 = make_charset (LEADING_BYTE_LATIN_ISO8859_9, Qlatin_iso8859_9, CHARSET_TYPE_96, 1, 1, 'M', @@ -3308,17 +3348,17 @@ complex_vars_of_mule_charset (void) build_string ("ISO8859-9 (Latin-5)"), build_string ("ISO8859-9 (Latin-5)"), build_string ("iso8859-9"), - latin_iso8859_9_to_ucs, 0, 0, 32); + latin_iso8859_9_to_ucs, 0, 0, 0, 32); Vcharset_japanese_jisx0208_1978 = make_charset (LEADING_BYTE_JAPANESE_JISX0208_1978, Qjapanese_jisx0208_1978, CHARSET_TYPE_94X94, 2, 0, '@', CHARSET_LEFT_TO_RIGHT, - build_string ("JISX0208.1978"), - build_string ("JISX0208.1978 (Japanese)"), + build_string ("JIS X0208:1978"), + build_string ("JIS X0208:1978 (Japanese)"), build_string - ("JISX0208.1978 Japanese Kanji (so called \"old JIS\")"), + ("JIS X0208:1978 Japanese Kanji (so called \"old JIS\")"), build_string ("\\(jisx0208\\|jisc6226\\)\\.1978"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); Vcharset_chinese_gb2312 = make_charset (LEADING_BYTE_CHINESE_GB2312, Qchinese_gb2312, CHARSET_TYPE_94X94, 2, 0, 'A', @@ -3327,16 +3367,16 @@ complex_vars_of_mule_charset (void) build_string ("GB2312)"), build_string ("GB2312 Chinese simplified"), build_string ("gb2312"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); Vcharset_japanese_jisx0208 = make_charset (LEADING_BYTE_JAPANESE_JISX0208, Qjapanese_jisx0208, CHARSET_TYPE_94X94, 2, 0, 'B', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0208"), - build_string ("JISX0208.1983/1990 (Japanese)"), - build_string ("JISX0208.1983/1990 Japanese Kanji"), - build_string ("jisx0208.19\\(83\\|90\\)"), - Qnil, 0, 0, 33); + build_string ("JIS X0208:1983 (Japanese)"), + build_string ("JIS X0208:1983 Japanese Kanji"), + build_string ("jisx0208\\.1983"), + Qnil, 0, 0, 0, 33); Vcharset_korean_ksc5601 = make_charset (LEADING_BYTE_KOREAN_KSC5601, Qkorean_ksc5601, CHARSET_TYPE_94X94, 2, 0, 'C', @@ -3345,7 +3385,7 @@ complex_vars_of_mule_charset (void) build_string ("KSC5601 (Korean"), build_string ("KSC5601 Korean Hangul and Hanja"), build_string ("ksc5601"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); Vcharset_japanese_jisx0212 = make_charset (LEADING_BYTE_JAPANESE_JISX0212, Qjapanese_jisx0212, CHARSET_TYPE_94X94, 2, 0, 'D', @@ -3354,7 +3394,7 @@ complex_vars_of_mule_charset (void) build_string ("JISX0212 (Japanese)"), build_string ("JISX0212 Japanese Supplement"), build_string ("jisx0212"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); #define CHINESE_CNS_PLANE_RE(n) "cns11643[.-]\\(.*[.-]\\)?" n "$" Vcharset_chinese_cns11643_1 = @@ -3366,7 +3406,7 @@ complex_vars_of_mule_charset (void) build_string ("CNS 11643 Plane 1 Chinese traditional"), build_string (CHINESE_CNS_PLANE_RE("1")), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); Vcharset_chinese_cns11643_2 = make_charset (LEADING_BYTE_CHINESE_CNS11643_2, Qchinese_cns11643_2, CHARSET_TYPE_94X94, 2, 0, 'H', @@ -3376,7 +3416,7 @@ complex_vars_of_mule_charset (void) build_string ("CNS 11643 Plane 2 Chinese traditional"), build_string (CHINESE_CNS_PLANE_RE("2")), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); #ifdef UTF2000 Vcharset_latin_viscii_lower = make_charset (LEADING_BYTE_LATIN_VISCII_LOWER, Qlatin_viscii_lower, @@ -3385,8 +3425,8 @@ complex_vars_of_mule_charset (void) build_string ("VISCII lower"), build_string ("VISCII lower (Vietnamese)"), build_string ("VISCII lower (Vietnamese)"), - build_string ("VISCII1.1"), - latin_viscii_lower_to_ucs, 0, 0, 32); + build_string ("VISCII1\\.1"), + latin_viscii_lower_to_ucs, 0, 0, 0, 32); Vcharset_latin_viscii_upper = make_charset (LEADING_BYTE_LATIN_VISCII_UPPER, Qlatin_viscii_upper, CHARSET_TYPE_96, 1, 1, '2', @@ -3394,8 +3434,28 @@ complex_vars_of_mule_charset (void) build_string ("VISCII upper"), build_string ("VISCII upper (Vietnamese)"), build_string ("VISCII upper (Vietnamese)"), - build_string ("VISCII1.1"), - latin_viscii_upper_to_ucs, 0, 0, 32); + build_string ("VISCII1\\.1"), + latin_viscii_upper_to_ucs, 0, 0, 0, 32); + Vcharset_hiragana_jisx0208 = + make_charset (LEADING_BYTE_HIRAGANA_JISX0208, Qhiragana_jisx0208, + CHARSET_TYPE_94X94, 2, 0, 'B', + CHARSET_LEFT_TO_RIGHT, + build_string ("Hiragana"), + build_string ("Hiragana of JIS X0208"), + build_string ("Japanese Hiragana of JIS X0208"), + build_string ("jisx0208\\.19\\(78\\|83\\|90\\)"), + Qnil, MIN_CHAR_HIRAGANA, MAX_CHAR_HIRAGANA, + (0x24 - 33) * 94 + (0x21 - 33), 33); + Vcharset_katakana_jisx0208 = + make_charset (LEADING_BYTE_KATAKANA_JISX0208, Qkatakana_jisx0208, + CHARSET_TYPE_94X94, 2, 0, 'B', + CHARSET_LEFT_TO_RIGHT, + build_string ("Katakana"), + build_string ("Katakana of JIS X0208"), + build_string ("Japanese Katakana of JIS X0208"), + build_string ("jisx0208\\.19\\(78\\|83\\|90\\)"), + Qnil, MIN_CHAR_KATAKANA, MAX_CHAR_KATAKANA, + (0x25 - 33) * 94 + (0x21 - 33), 33); #endif Vcharset_chinese_big5_1 = make_charset (LEADING_BYTE_CHINESE_BIG5_1, Qchinese_big5_1, @@ -3406,7 +3466,7 @@ complex_vars_of_mule_charset (void) build_string ("Big5 Level-1 Chinese traditional"), build_string ("big5"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); Vcharset_chinese_big5_2 = make_charset (LEADING_BYTE_CHINESE_BIG5_2, Qchinese_big5_2, CHARSET_TYPE_94X94, 2, 0, '1', @@ -3416,7 +3476,7 @@ complex_vars_of_mule_charset (void) build_string ("Big5 Level-2 Chinese traditional"), build_string ("big5"), - Qnil, 0, 0, 33); + Qnil, 0, 0, 0, 33); #ifdef ENABLE_COMPOSITE_CHARS /* #### For simplicity, we put composite chars into a 96x96 charset.