[352] | 1 | #include <iconv.h>
|
---|
| 2 | #include <errno.h>
|
---|
| 3 | #include <wchar.h>
|
---|
| 4 | #include <string.h>
|
---|
| 5 | #include <stdlib.h>
|
---|
| 6 | #include <limits.h>
|
---|
| 7 | #include <stdint.h>
|
---|
| 8 | #include "locale_impl.h"
|
---|
| 9 |
|
---|
| 10 | #define UTF_32BE 0300
|
---|
| 11 | #define UTF_16LE 0301
|
---|
| 12 | #define UTF_16BE 0302
|
---|
| 13 | #define UTF_32LE 0303
|
---|
| 14 | #define UCS2BE 0304
|
---|
| 15 | #define UCS2LE 0305
|
---|
| 16 | #define WCHAR_T 0306
|
---|
| 17 | #define US_ASCII 0307
|
---|
| 18 | #define UTF_8 0310
|
---|
| 19 | #define EUC_JP 0320
|
---|
| 20 | #define SHIFT_JIS 0321
|
---|
| 21 | #define GB18030 0330
|
---|
| 22 | #define GBK 0331
|
---|
| 23 | #define GB2312 0332
|
---|
| 24 | #define BIG5 0340
|
---|
| 25 | #define EUC_KR 0350
|
---|
| 26 |
|
---|
| 27 | /* Definitions of charmaps. Each charmap consists of:
|
---|
| 28 | * 1. Empty-string-terminated list of null-terminated aliases.
|
---|
| 29 | * 2. Special type code or number of elided entries.
|
---|
| 30 | * 3. Character table (size determined by field 2). */
|
---|
| 31 |
|
---|
| 32 | static const unsigned char charmaps[] =
|
---|
| 33 | "utf8\0char\0\0\310"
|
---|
| 34 | "wchart\0\0\306"
|
---|
| 35 | "ucs2\0ucs2be\0\0\304"
|
---|
| 36 | "ucs2le\0\0\305"
|
---|
| 37 | "utf16\0utf16be\0\0\302"
|
---|
| 38 | "utf16le\0\0\301"
|
---|
| 39 | "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
|
---|
| 40 | "ucs4le\0utf32le\0\0\303"
|
---|
| 41 | "ascii\0usascii\0iso646\0iso646us\0\0\307"
|
---|
| 42 | "eucjp\0\0\320"
|
---|
| 43 | "shiftjis\0sjis\0\0\321"
|
---|
| 44 | "gb18030\0\0\330"
|
---|
| 45 | "gbk\0\0\331"
|
---|
| 46 | "gb2312\0\0\332"
|
---|
| 47 | "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
|
---|
| 48 | "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
|
---|
| 49 | #include "codepages.h"
|
---|
| 50 | ;
|
---|
| 51 |
|
---|
| 52 | static const unsigned short legacy_chars[] = {
|
---|
| 53 | #include "legacychars.h"
|
---|
| 54 | };
|
---|
| 55 |
|
---|
| 56 | static const unsigned short jis0208[84][94] = {
|
---|
| 57 | #include "jis0208.h"
|
---|
| 58 | };
|
---|
| 59 |
|
---|
| 60 | static const unsigned short gb18030[126][190] = {
|
---|
| 61 | #include "gb18030.h"
|
---|
| 62 | };
|
---|
| 63 |
|
---|
| 64 | static const unsigned short big5[89][157] = {
|
---|
| 65 | #include "big5.h"
|
---|
| 66 | };
|
---|
| 67 |
|
---|
| 68 | static const unsigned short hkscs[] = {
|
---|
| 69 | #include "hkscs.h"
|
---|
| 70 | };
|
---|
| 71 |
|
---|
| 72 | static const unsigned short ksc[93][94] = {
|
---|
| 73 | #include "ksc.h"
|
---|
| 74 | };
|
---|
| 75 |
|
---|
| 76 | static int fuzzycmp(const unsigned char *a, const unsigned char *b)
|
---|
| 77 | {
|
---|
| 78 | for (; *a && *b; a++, b++) {
|
---|
| 79 | while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
|
---|
| 80 | if ((*a|32U) != *b) return 1;
|
---|
| 81 | }
|
---|
| 82 | return *a != *b;
|
---|
| 83 | }
|
---|
| 84 |
|
---|
| 85 | static size_t find_charmap(const void *name)
|
---|
| 86 | {
|
---|
| 87 | const unsigned char *s;
|
---|
| 88 | if (!*(char *)name) name=charmaps; /* "utf8" */
|
---|
| 89 | for (s=charmaps; *s; ) {
|
---|
| 90 | if (!fuzzycmp(name, s)) {
|
---|
| 91 | for (; *s; s+=strlen((void *)s)+1);
|
---|
| 92 | return s+1-charmaps;
|
---|
| 93 | }
|
---|
| 94 | s += strlen((void *)s)+1;
|
---|
| 95 | if (!*s) {
|
---|
| 96 | if (s[1] > 0200) s+=2;
|
---|
| 97 | else s+=2+(128U-s[1])/4*5;
|
---|
| 98 | }
|
---|
| 99 | }
|
---|
| 100 | return -1;
|
---|
| 101 | }
|
---|
| 102 |
|
---|
| 103 | iconv_t iconv_open(const char *to, const char *from)
|
---|
| 104 | {
|
---|
| 105 | size_t f, t;
|
---|
| 106 |
|
---|
| 107 | if ((t = find_charmap(to))==-1
|
---|
| 108 | || (f = find_charmap(from))==-1
|
---|
| 109 | || (charmaps[t] >= 0320)) {
|
---|
| 110 | errno = EINVAL;
|
---|
| 111 | return (iconv_t)-1;
|
---|
| 112 | }
|
---|
| 113 |
|
---|
| 114 | return (void *)(f<<16 | t);
|
---|
| 115 | }
|
---|
| 116 |
|
---|
| 117 | int iconv_close(iconv_t cd)
|
---|
| 118 | {
|
---|
| 119 | return 0;
|
---|
| 120 | }
|
---|
| 121 |
|
---|
| 122 | static unsigned get_16(const unsigned char *s, int e)
|
---|
| 123 | {
|
---|
| 124 | e &= 1;
|
---|
| 125 | return s[e]<<8 | s[1-e];
|
---|
| 126 | }
|
---|
| 127 |
|
---|
| 128 | static void put_16(unsigned char *s, unsigned c, int e)
|
---|
| 129 | {
|
---|
| 130 | e &= 1;
|
---|
| 131 | s[e] = c>>8;
|
---|
| 132 | s[1-e] = c;
|
---|
| 133 | }
|
---|
| 134 |
|
---|
| 135 | static unsigned get_32(const unsigned char *s, int e)
|
---|
| 136 | {
|
---|
| 137 | e &= 3;
|
---|
| 138 | return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
|
---|
| 139 | }
|
---|
| 140 |
|
---|
| 141 | static void put_32(unsigned char *s, unsigned c, int e)
|
---|
| 142 | {
|
---|
| 143 | e &= 3;
|
---|
| 144 | s[e^0] = c>>24;
|
---|
| 145 | s[e^1] = c>>16;
|
---|
| 146 | s[e^2] = c>>8;
|
---|
| 147 | s[e^3] = c;
|
---|
| 148 | }
|
---|
| 149 |
|
---|
| 150 | /* Adapt as needed */
|
---|
| 151 | #define mbrtowc_utf8 mbrtowc
|
---|
| 152 | #define wctomb_utf8 wctomb
|
---|
| 153 |
|
---|
| 154 | static unsigned legacy_map(const unsigned char *map, unsigned c)
|
---|
| 155 | {
|
---|
| 156 | unsigned x = c - 128 - map[-1];
|
---|
| 157 | x = legacy_chars[ map[x*5/4]>>2*x%8 |
|
---|
| 158 | map[x*5/4+1]<<8-2*x%8 & 1023 ];
|
---|
| 159 | return x ? x : c;
|
---|
| 160 | }
|
---|
| 161 |
|
---|
| 162 | size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
|
---|
| 163 | {
|
---|
| 164 | size_t x=0;
|
---|
| 165 | unsigned long cd = (unsigned long)cd0;
|
---|
| 166 | unsigned to = cd & 0xffff;
|
---|
| 167 | unsigned from = cd >> 16;
|
---|
| 168 | const unsigned char *map = charmaps+from+1;
|
---|
| 169 | const unsigned char *tomap = charmaps+to+1;
|
---|
| 170 | mbstate_t st = {0};
|
---|
| 171 | wchar_t wc;
|
---|
| 172 | unsigned c, d;
|
---|
| 173 | size_t k, l;
|
---|
| 174 | int err;
|
---|
| 175 | unsigned char type = map[-1];
|
---|
| 176 | unsigned char totype = tomap[-1];
|
---|
| 177 | locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
|
---|
| 178 |
|
---|
| 179 | if (!in || !*in || !*inb) return 0;
|
---|
| 180 |
|
---|
| 181 | *ploc = UTF8_LOCALE;
|
---|
| 182 |
|
---|
| 183 | for (; *inb; *in+=l, *inb-=l) {
|
---|
| 184 | c = *(unsigned char *)*in;
|
---|
| 185 | l = 1;
|
---|
| 186 |
|
---|
| 187 | if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
|
---|
| 188 | case UTF_8:
|
---|
| 189 | l = mbrtowc_utf8(&wc, *in, *inb, &st);
|
---|
| 190 | if (!l) l++;
|
---|
| 191 | else if (l == (size_t)-1) goto ilseq;
|
---|
| 192 | else if (l == (size_t)-2) goto starved;
|
---|
| 193 | c = wc;
|
---|
| 194 | break;
|
---|
| 195 | case US_ASCII:
|
---|
| 196 | goto ilseq;
|
---|
| 197 | case WCHAR_T:
|
---|
| 198 | l = sizeof(wchar_t);
|
---|
| 199 | if (*inb < l) goto starved;
|
---|
| 200 | c = *(wchar_t *)*in;
|
---|
| 201 | if (0) {
|
---|
| 202 | case UTF_32BE:
|
---|
| 203 | case UTF_32LE:
|
---|
| 204 | l = 4;
|
---|
| 205 | if (*inb < 4) goto starved;
|
---|
| 206 | c = get_32((void *)*in, type);
|
---|
| 207 | }
|
---|
| 208 | if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
|
---|
| 209 | break;
|
---|
| 210 | case UCS2BE:
|
---|
| 211 | case UCS2LE:
|
---|
| 212 | case UTF_16BE:
|
---|
| 213 | case UTF_16LE:
|
---|
| 214 | l = 2;
|
---|
| 215 | if (*inb < 2) goto starved;
|
---|
| 216 | c = get_16((void *)*in, type);
|
---|
| 217 | if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
|
---|
| 218 | if ((unsigned)(c-0xd800) < 0x400) {
|
---|
| 219 | if (type-UCS2BE < 2U) goto ilseq;
|
---|
| 220 | l = 4;
|
---|
| 221 | if (*inb < 4) goto starved;
|
---|
| 222 | d = get_16((void *)(*in + 2), type);
|
---|
| 223 | if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
|
---|
| 224 | c = ((c-0xd7c0)<<10) + (d-0xdc00);
|
---|
| 225 | }
|
---|
| 226 | break;
|
---|
| 227 | case SHIFT_JIS:
|
---|
| 228 | if (c-0xa1 <= 0xdf-0xa1) {
|
---|
| 229 | c += 0xff61-0xa1;
|
---|
| 230 | break;
|
---|
| 231 | }
|
---|
| 232 | l = 2;
|
---|
| 233 | if (*inb < 2) goto starved;
|
---|
| 234 | d = *((unsigned char *)*in + 1);
|
---|
| 235 | if (c-129 <= 159-129) c -= 129;
|
---|
| 236 | else if (c-224 <= 239-224) c -= 193;
|
---|
| 237 | else goto ilseq;
|
---|
| 238 | c *= 2;
|
---|
| 239 | if (d-64 <= 158-64) {
|
---|
| 240 | if (d==127) goto ilseq;
|
---|
| 241 | if (d>127) d--;
|
---|
| 242 | d -= 64;
|
---|
| 243 | } else if (d-159 <= 252-159) {
|
---|
| 244 | c++;
|
---|
| 245 | d -= 159;
|
---|
| 246 | }
|
---|
| 247 | c = jis0208[c][d];
|
---|
| 248 | if (!c) goto ilseq;
|
---|
| 249 | break;
|
---|
| 250 | case EUC_JP:
|
---|
| 251 | l = 2;
|
---|
| 252 | if (*inb < 2) goto starved;
|
---|
| 253 | d = *((unsigned char *)*in + 1);
|
---|
| 254 | if (c==0x8e) {
|
---|
| 255 | c = d;
|
---|
| 256 | if (c-0xa1 > 0xdf-0xa1) goto ilseq;
|
---|
| 257 | c += 0xff61 - 0xa1;
|
---|
| 258 | break;
|
---|
| 259 | }
|
---|
| 260 | c -= 0xa1;
|
---|
| 261 | d -= 0xa1;
|
---|
| 262 | if (c >= 84 || d >= 94) goto ilseq;
|
---|
| 263 | c = jis0208[c][d];
|
---|
| 264 | if (!c) goto ilseq;
|
---|
| 265 | break;
|
---|
| 266 | case GB2312:
|
---|
| 267 | if (c < 0xa1) goto ilseq;
|
---|
| 268 | case GBK:
|
---|
| 269 | case GB18030:
|
---|
| 270 | c -= 0x81;
|
---|
| 271 | if (c >= 126) goto ilseq;
|
---|
| 272 | l = 2;
|
---|
| 273 | if (*inb < 2) goto starved;
|
---|
| 274 | d = *((unsigned char *)*in + 1);
|
---|
| 275 | if (d < 0xa1 && type == GB2312) goto ilseq;
|
---|
| 276 | if (d-0x40>=191 || d==127) {
|
---|
| 277 | if (d-'0'>9 || type != GB18030)
|
---|
| 278 | goto ilseq;
|
---|
| 279 | l = 4;
|
---|
| 280 | if (*inb < 4) goto starved;
|
---|
| 281 | c = (10*c + d-'0') * 1260;
|
---|
| 282 | d = *((unsigned char *)*in + 2);
|
---|
| 283 | if (d-0x81>126) goto ilseq;
|
---|
| 284 | c += 10*(d-0x81);
|
---|
| 285 | d = *((unsigned char *)*in + 3);
|
---|
| 286 | if (d-'0'>9) goto ilseq;
|
---|
| 287 | c += d-'0';
|
---|
| 288 | c += 128;
|
---|
| 289 | for (d=0; d<=c; ) {
|
---|
| 290 | k = 0;
|
---|
| 291 | for (int i=0; i<126; i++)
|
---|
| 292 | for (int j=0; j<190; j++)
|
---|
| 293 | if (gb18030[i][j]-d <= c-d)
|
---|
| 294 | k++;
|
---|
| 295 | d = c+1;
|
---|
| 296 | c += k;
|
---|
| 297 | }
|
---|
| 298 | break;
|
---|
| 299 | }
|
---|
| 300 | d -= 0x40;
|
---|
| 301 | if (d>63) d--;
|
---|
| 302 | c = gb18030[c][d];
|
---|
| 303 | break;
|
---|
| 304 | case BIG5:
|
---|
| 305 | l = 2;
|
---|
| 306 | if (*inb < 2) goto starved;
|
---|
| 307 | d = *((unsigned char *)*in + 1);
|
---|
| 308 | if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
|
---|
| 309 | d -= 0x40;
|
---|
| 310 | if (d > 0x3e) d -= 0x22;
|
---|
| 311 | if (c-0xa1>=0xfa-0xa1) {
|
---|
| 312 | if (c-0x87>=0xff-0x87) goto ilseq;
|
---|
| 313 | if (c < 0xa1) c -= 0x87;
|
---|
| 314 | else c -= 0x87 + (0xfa-0xa1);
|
---|
| 315 | c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
|
---|
| 316 | | hkscs[c*157+d];
|
---|
| 317 | /* A few HKSCS characters map to pairs of UCS
|
---|
| 318 | * characters. These are mapped to surrogate
|
---|
| 319 | * range in the hkscs table then hard-coded
|
---|
| 320 | * here. Ugly, yes. */
|
---|
| 321 | if (c/256 == 0xdc) {
|
---|
| 322 | if (totype-0300U > 8) k = 2;
|
---|
| 323 | else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
|
---|
| 324 | if (k > *outb) goto toobig;
|
---|
| 325 | x += iconv((iconv_t)(uintptr_t)to,
|
---|
| 326 | &(char *){"\303\212\314\204"
|
---|
| 327 | "\303\212\314\214"
|
---|
| 328 | "\303\252\314\204"
|
---|
| 329 | "\303\252\314\214"
|
---|
| 330 | +c%256}, &(size_t){4},
|
---|
| 331 | out, outb);
|
---|
| 332 | continue;
|
---|
| 333 | }
|
---|
| 334 | if (!c) goto ilseq;
|
---|
| 335 | break;
|
---|
| 336 | }
|
---|
| 337 | c -= 0xa1;
|
---|
| 338 | c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
|
---|
| 339 | if (!c) goto ilseq;
|
---|
| 340 | break;
|
---|
| 341 | case EUC_KR:
|
---|
| 342 | l = 2;
|
---|
| 343 | if (*inb < 2) goto starved;
|
---|
| 344 | d = *((unsigned char *)*in + 1);
|
---|
| 345 | c -= 0xa1;
|
---|
| 346 | d -= 0xa1;
|
---|
| 347 | if (c >= 93 || d >= 94) {
|
---|
| 348 | c += (0xa1-0x81);
|
---|
| 349 | d += 0xa1;
|
---|
| 350 | if (c >= 93 || c>=0xc6-0x81 && d>0x52)
|
---|
| 351 | goto ilseq;
|
---|
| 352 | if (d-'A'<26) d = d-'A';
|
---|
| 353 | else if (d-'a'<26) d = d-'a'+26;
|
---|
| 354 | else if (d-0x81<0xff-0x81) d = d-0x81+52;
|
---|
| 355 | else goto ilseq;
|
---|
| 356 | if (c < 0x20) c = 178*c + d;
|
---|
| 357 | else c = 178*0x20 + 84*(c-0x20) + d;
|
---|
| 358 | c += 0xac00;
|
---|
| 359 | for (d=0xac00; d<=c; ) {
|
---|
| 360 | k = 0;
|
---|
| 361 | for (int i=0; i<93; i++)
|
---|
| 362 | for (int j=0; j<94; j++)
|
---|
| 363 | if (ksc[i][j]-d <= c-d)
|
---|
| 364 | k++;
|
---|
| 365 | d = c+1;
|
---|
| 366 | c += k;
|
---|
| 367 | }
|
---|
| 368 | break;
|
---|
| 369 | }
|
---|
| 370 | c = ksc[c][d];
|
---|
| 371 | if (!c) goto ilseq;
|
---|
| 372 | break;
|
---|
| 373 | default:
|
---|
| 374 | if (c < 128+type) break;
|
---|
| 375 | c = legacy_map(map, c);
|
---|
| 376 | if (c==1) goto ilseq;
|
---|
| 377 | }
|
---|
| 378 |
|
---|
| 379 | switch (totype) {
|
---|
| 380 | case WCHAR_T:
|
---|
| 381 | if (*outb < sizeof(wchar_t)) goto toobig;
|
---|
| 382 | *(wchar_t *)*out = c;
|
---|
| 383 | *out += sizeof(wchar_t);
|
---|
| 384 | *outb -= sizeof(wchar_t);
|
---|
| 385 | break;
|
---|
| 386 | case UTF_8:
|
---|
| 387 | if (*outb < 4) {
|
---|
| 388 | char tmp[4];
|
---|
| 389 | k = wctomb_utf8(tmp, c);
|
---|
| 390 | if (*outb < k) goto toobig;
|
---|
| 391 | memcpy(*out, tmp, k);
|
---|
| 392 | } else k = wctomb_utf8(*out, c);
|
---|
| 393 | *out += k;
|
---|
| 394 | *outb -= k;
|
---|
| 395 | break;
|
---|
| 396 | case US_ASCII:
|
---|
| 397 | if (c > 0x7f) subst: x++, c='*';
|
---|
| 398 | default:
|
---|
| 399 | if (*outb < 1) goto toobig;
|
---|
| 400 | if (c < 128+totype || (c<256 && c==legacy_map(tomap, c))) {
|
---|
| 401 | revout:
|
---|
| 402 | *(*out)++ = c;
|
---|
| 403 | *outb -= 1;
|
---|
| 404 | break;
|
---|
| 405 | }
|
---|
| 406 | d = c;
|
---|
| 407 | for (c=128+totype; c<256; c++) {
|
---|
| 408 | if (d == legacy_map(tomap, c)) {
|
---|
| 409 | goto revout;
|
---|
| 410 | }
|
---|
| 411 | }
|
---|
| 412 | goto subst;
|
---|
| 413 | case UCS2BE:
|
---|
| 414 | case UCS2LE:
|
---|
| 415 | case UTF_16BE:
|
---|
| 416 | case UTF_16LE:
|
---|
| 417 | if (c < 0x10000 || type-UCS2BE < 2U) {
|
---|
| 418 | if (c >= 0x10000) c = 0xFFFD;
|
---|
| 419 | if (*outb < 2) goto toobig;
|
---|
| 420 | put_16((void *)*out, c, totype);
|
---|
| 421 | *out += 2;
|
---|
| 422 | *outb -= 2;
|
---|
| 423 | break;
|
---|
| 424 | }
|
---|
| 425 | if (*outb < 4) goto toobig;
|
---|
| 426 | c -= 0x10000;
|
---|
| 427 | put_16((void *)*out, (c>>10)|0xd800, totype);
|
---|
| 428 | put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
|
---|
| 429 | *out += 4;
|
---|
| 430 | *outb -= 4;
|
---|
| 431 | break;
|
---|
| 432 | case UTF_32BE:
|
---|
| 433 | case UTF_32LE:
|
---|
| 434 | if (*outb < 4) goto toobig;
|
---|
| 435 | put_32((void *)*out, c, totype);
|
---|
| 436 | *out += 4;
|
---|
| 437 | *outb -= 4;
|
---|
| 438 | break;
|
---|
| 439 | }
|
---|
| 440 | }
|
---|
| 441 | *ploc = loc;
|
---|
| 442 | return x;
|
---|
| 443 | ilseq:
|
---|
| 444 | err = EILSEQ;
|
---|
| 445 | x = -1;
|
---|
| 446 | goto end;
|
---|
| 447 | toobig:
|
---|
| 448 | err = E2BIG;
|
---|
| 449 | x = -1;
|
---|
| 450 | goto end;
|
---|
| 451 | starved:
|
---|
| 452 | err = EINVAL;
|
---|
| 453 | x = -1;
|
---|
| 454 | end:
|
---|
| 455 | errno = err;
|
---|
| 456 | *ploc = loc;
|
---|
| 457 | return x;
|
---|
| 458 | }
|
---|