nsz Git - musl/blob - src/locale/iconv.c

   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 struct stateful_cd {
 104         iconv_t base_cd;
 105         unsigned state;
 106 };
 107
 108 static iconv_t combine_to_from(size_t t, size_t f)
 109 {
 110         return (void *)(f<<16 | t<<1 | 1);
 111 }
 112
 113 static size_t extract_from(iconv_t cd)
 114 {
 115         return (size_t)cd >> 16;
 116 }
 117
 118 static size_t extract_to(iconv_t cd)
 119 {
 120         return (size_t)cd >> 1 & 0x7fff;
 121 }
 122
 123 iconv_t iconv_open(const char *to, const char *from)
 124 {
 125         size_t f, t;
 126
 127         if ((t = find_charmap(to))==-1
 128          || (f = find_charmap(from))==-1
 129          || (charmaps[t] >= 0320)) {
 130                 errno = EINVAL;
 131                 return (iconv_t)-1;
 132         }
 133         iconv_t cd = combine_to_from(t, f);
 134
 135         if (0) {
 136                 struct stateful_cd *scd = malloc(sizeof *scd);
 137                 if (!scd) return (iconv_t)-1;
 138                 scd->base_cd = cd;
 139                 scd->state = 0;
 140                 cd = (iconv_t)scd;
 141         }
 142
 143         return cd;
 144 }
 145
 146 static unsigned get_16(const unsigned char *s, int e)
 147 {
 148         e &= 1;
 149         return s[e]<<8 | s[1-e];
 150 }
 151
 152 static void put_16(unsigned char *s, unsigned c, int e)
 153 {
 154         e &= 1;
 155         s[e] = c>>8;
 156         s[1-e] = c;
 157 }
 158
 159 static unsigned get_32(const unsigned char *s, int e)
 160 {
 161         e &= 3;
 162         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 163 }
 164
 165 static void put_32(unsigned char *s, unsigned c, int e)
 166 {
 167         e &= 3;
 168         s[e^0] = c>>24;
 169         s[e^1] = c>>16;
 170         s[e^2] = c>>8;
 171         s[e^3] = c;
 172 }
 173
 174 /* Adapt as needed */
 175 #define mbrtowc_utf8 mbrtowc
 176 #define wctomb_utf8 wctomb
 177
 178 static unsigned legacy_map(const unsigned char *map, unsigned c)
 179 {
 180         unsigned x = c - 128 - map[-1];
 181         x = legacy_chars[ map[x*5/4]>>2*x%8 |
 182                 map[x*5/4+1]<<8-2*x%8 & 1023 ];
 183         return x ? x : c;
 184 }
 185
 186 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 187 {
 188         size_t x=0;
 189         struct stateful_cd *scd=0;
 190         if (!((size_t)cd & 1)) {
 191                 scd = (void *)cd;
 192                 cd = scd->base_cd;
 193         }
 194         unsigned to = extract_to(cd);
 195         unsigned from = extract_from(cd);
 196         const unsigned char *map = charmaps+from+1;
 197         const unsigned char *tomap = charmaps+to+1;
 198         mbstate_t st = {0};
 199         wchar_t wc;
 200         unsigned c, d;
 201         size_t k, l;
 202         int err;
 203         unsigned char type = map[-1];
 204         unsigned char totype = tomap[-1];
 205         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 206
 207         if (!in || !*in || !*inb) return 0;
 208
 209         *ploc = UTF8_LOCALE;
 210
 211         for (; *inb; *in+=l, *inb-=l) {
 212                 c = *(unsigned char *)*in;
 213                 l = 1;
 214
 215                 switch (type) {
 216                 case UTF_8:
 217                         if (c < 128) break;
 218                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 219                         if (l == (size_t)-1) goto ilseq;
 220                         if (l == (size_t)-2) goto starved;
 221                         c = wc;
 222                         break;
 223                 case US_ASCII:
 224                         if (c >= 128) goto ilseq;
 225                         break;
 226                 case WCHAR_T:
 227                         l = sizeof(wchar_t);
 228                         if (*inb < l) goto starved;
 229                         c = *(wchar_t *)*in;
 230                         if (0) {
 231                 case UTF_32BE:
 232                 case UTF_32LE:
 233                         l = 4;
 234                         if (*inb < 4) goto starved;
 235                         c = get_32((void *)*in, type);
 236                         }
 237                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 238                         break;
 239                 case UCS2BE:
 240                 case UCS2LE:
 241                 case UTF_16BE:
 242                 case UTF_16LE:
 243                         l = 2;
 244                         if (*inb < 2) goto starved;
 245                         c = get_16((void *)*in, type);
 246                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 247                         if ((unsigned)(c-0xd800) < 0x400) {
 248                                 if (type-UCS2BE < 2U) goto ilseq;
 249                                 l = 4;
 250                                 if (*inb < 4) goto starved;
 251                                 d = get_16((void *)(*in + 2), type);
 252                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 253                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 254                         }
 255                         break;
 256                 case SHIFT_JIS:
 257                         if (c < 128) break;
 258                         if (c-0xa1 <= 0xdf-0xa1) {
 259                                 c += 0xff61-0xa1;
 260                                 break;
 261                         }
 262                         l = 2;
 263                         if (*inb < 2) goto starved;
 264                         d = *((unsigned char *)*in + 1);
 265                         if (c-129 <= 159-129) c -= 129;
 266                         else if (c-224 <= 239-224) c -= 193;
 267                         else goto ilseq;
 268                         c *= 2;
 269                         if (d-64 <= 158-64) {
 270                                 if (d==127) goto ilseq;
 271                                 if (d>127) d--;
 272                                 d -= 64;
 273                         } else if (d-159 <= 252-159) {
 274                                 c++;
 275                                 d -= 159;
 276                         }
 277                         c = jis0208[c][d];
 278                         if (!c) goto ilseq;
 279                         break;
 280                 case EUC_JP:
 281                         if (c < 128) break;
 282                         l = 2;
 283                         if (*inb < 2) goto starved;
 284                         d = *((unsigned char *)*in + 1);
 285                         if (c==0x8e) {
 286                                 c = d;
 287                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 288                                 c += 0xff61 - 0xa1;
 289                                 break;
 290                         }
 291                         c -= 0xa1;
 292                         d -= 0xa1;
 293                         if (c >= 84 || d >= 94) goto ilseq;
 294                         c = jis0208[c][d];
 295                         if (!c) goto ilseq;
 296                         break;
 297                 case GB2312:
 298                         if (c < 128) break;
 299                         if (c < 0xa1) goto ilseq;
 300                 case GBK:
 301                 case GB18030:
 302                         if (c < 128) break;
 303                         c -= 0x81;
 304                         if (c >= 126) goto ilseq;
 305                         l = 2;
 306                         if (*inb < 2) goto starved;
 307                         d = *((unsigned char *)*in + 1);
 308                         if (d < 0xa1 && type == GB2312) goto ilseq;
 309                         if (d-0x40>=191 || d==127) {
 310                                 if (d-'0'>9 || type != GB18030)
 311                                         goto ilseq;
 312                                 l = 4;
 313                                 if (*inb < 4) goto starved;
 314                                 c = (10*c + d-'0') * 1260;
 315                                 d = *((unsigned char *)*in + 2);
 316                                 if (d-0x81>126) goto ilseq;
 317                                 c += 10*(d-0x81);
 318                                 d = *((unsigned char *)*in + 3);
 319                                 if (d-'0'>9) goto ilseq;
 320                                 c += d-'0';
 321                                 c += 128;
 322                                 for (d=0; d<=c; ) {
 323                                         k = 0;
 324                                         for (int i=0; i<126; i++)
 325                                                 for (int j=0; j<190; j++)
 326                                                         if (gb18030[i][j]-d <= c-d)
 327                                                                 k++;
 328                                         d = c+1;
 329                                         c += k;
 330                                 }
 331                                 break;
 332                         }
 333                         d -= 0x40;
 334                         if (d>63) d--;
 335                         c = gb18030[c][d];
 336                         break;
 337                 case BIG5:
 338                         if (c < 128) break;
 339                         l = 2;
 340                         if (*inb < 2) goto starved;
 341                         d = *((unsigned char *)*in + 1);
 342                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 343                         d -= 0x40;
 344                         if (d > 0x3e) d -= 0x22;
 345                         if (c-0xa1>=0xfa-0xa1) {
 346                                 if (c-0x87>=0xff-0x87) goto ilseq;
 347                                 if (c < 0xa1) c -= 0x87;
 348                                 else c -= 0x87 + (0xfa-0xa1);
 349                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 350                                         | hkscs[c*157+d];
 351                                 /* A few HKSCS characters map to pairs of UCS
 352                                  * characters. These are mapped to surrogate
 353                                  * range in the hkscs table then hard-coded
 354                                  * here. Ugly, yes. */
 355                                 if (c/256 == 0xdc) {
 356                                         if (totype-0300U > 8) k = 2;
 357                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 358                                         if (k > *outb) goto toobig;
 359                                         x += iconv(combine_to_from(to, 0),
 360                                                 &(char *){"\303\212\314\204"
 361                                                 "\303\212\314\214"
 362                                                 "\303\252\314\204"
 363                                                 "\303\252\314\214"
 364                                                 +c%256}, &(size_t){4},
 365                                                 out, outb);
 366                                         continue;
 367                                 }
 368                                 if (!c) goto ilseq;
 369                                 break;
 370                         }
 371                         c -= 0xa1;
 372                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 373                         if (!c) goto ilseq;
 374                         break;
 375                 case EUC_KR:
 376                         if (c < 128) break;
 377                         l = 2;
 378                         if (*inb < 2) goto starved;
 379                         d = *((unsigned char *)*in + 1);
 380                         c -= 0xa1;
 381                         d -= 0xa1;
 382                         if (c >= 93 || d >= 94) {
 383                                 c += (0xa1-0x81);
 384                                 d += 0xa1;
 385                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 386                                         goto ilseq;
 387                                 if (d-'A'<26) d = d-'A';
 388                                 else if (d-'a'<26) d = d-'a'+26;
 389                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 390                                 else goto ilseq;
 391                                 if (c < 0x20) c = 178*c + d;
 392                                 else c = 178*0x20 + 84*(c-0x20) + d;
 393                                 c += 0xac00;
 394                                 for (d=0xac00; d<=c; ) {
 395                                         k = 0;
 396                                         for (int i=0; i<93; i++)
 397                                                 for (int j=0; j<94; j++)
 398                                                         if (ksc[i][j]-d <= c-d)
 399                                                                 k++;
 400                                         d = c+1;
 401                                         c += k;
 402                                 }
 403                                 break;
 404                         }
 405                         c = ksc[c][d];
 406                         if (!c) goto ilseq;
 407                         break;
 408                 default:
 409                         if (c < 128+type) break;
 410                         c = legacy_map(map, c);
 411                         if (c==1) goto ilseq;
 412                 }
 413
 414                 switch (totype) {
 415                 case WCHAR_T:
 416                         if (*outb < sizeof(wchar_t)) goto toobig;
 417                         *(wchar_t *)*out = c;
 418                         *out += sizeof(wchar_t);
 419                         *outb -= sizeof(wchar_t);
 420                         break;
 421                 case UTF_8:
 422                         if (*outb < 4) {
 423                                 char tmp[4];
 424                                 k = wctomb_utf8(tmp, c);
 425                                 if (*outb < k) goto toobig;
 426                                 memcpy(*out, tmp, k);
 427                         } else k = wctomb_utf8(*out, c);
 428                         *out += k;
 429                         *outb -= k;
 430                         break;
 431                 case US_ASCII:
 432                         if (c > 0x7f) subst: x++, c='*';
 433                 default:
 434                         if (*outb < 1) goto toobig;
 435                         if (c < 128+totype || (c<256 && c==legacy_map(tomap, c))) {
 436                         revout:
 437                                 *(*out)++ = c;
 438                                 *outb -= 1;
 439                                 break;
 440                         }
 441                         d = c;
 442                         for (c=128+totype; c<256; c++) {
 443                                 if (d == legacy_map(tomap, c)) {
 444                                         goto revout;
 445                                 }
 446                         }
 447                         goto subst;
 448                 case UCS2BE:
 449                 case UCS2LE:
 450                 case UTF_16BE:
 451                 case UTF_16LE:
 452                         if (c < 0x10000 || type-UCS2BE < 2U) {
 453                                 if (c >= 0x10000) c = 0xFFFD;
 454                                 if (*outb < 2) goto toobig;
 455                                 put_16((void *)*out, c, totype);
 456                                 *out += 2;
 457                                 *outb -= 2;
 458                                 break;
 459                         }
 460                         if (*outb < 4) goto toobig;
 461                         c -= 0x10000;
 462                         put_16((void *)*out, (c>>10)|0xd800, totype);
 463                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 464                         *out += 4;
 465                         *outb -= 4;
 466                         break;
 467                 case UTF_32BE:
 468                 case UTF_32LE:
 469                         if (*outb < 4) goto toobig;
 470                         put_32((void *)*out, c, totype);
 471                         *out += 4;
 472                         *outb -= 4;
 473                         break;
 474                 }
 475         }
 476         *ploc = loc;
 477         return x;
 478 ilseq:
 479         err = EILSEQ;
 480         x = -1;
 481         goto end;
 482 toobig:
 483         err = E2BIG;
 484         x = -1;
 485         goto end;
 486 starved:
 487         err = EINVAL;
 488         x = -1;
 489 end:
 490         errno = err;
 491         *ploc = loc;
 492         return x;
 493 }