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 static iconv_t combine_to_from(size_t t, size_t f)
 104 {
 105         return (void *)(f<<16 | t);
 106 }
 107
 108 static size_t extract_from(iconv_t cd)
 109 {
 110         return (size_t)cd >> 16;
 111 }
 112
 113 static size_t extract_to(iconv_t cd)
 114 {
 115         return (size_t)cd & 0xffff;
 116 }
 117
 118 iconv_t iconv_open(const char *to, const char *from)
 119 {
 120         size_t f, t;
 121
 122         if ((t = find_charmap(to))==-1
 123          || (f = find_charmap(from))==-1
 124          || (charmaps[t] >= 0320)) {
 125                 errno = EINVAL;
 126                 return (iconv_t)-1;
 127         }
 128
 129         return combine_to_from(t, f);
 130 }
 131
 132 int iconv_close(iconv_t cd)
 133 {
 134         return 0;
 135 }
 136
 137 static unsigned get_16(const unsigned char *s, int e)
 138 {
 139         e &= 1;
 140         return s[e]<<8 | s[1-e];
 141 }
 142
 143 static void put_16(unsigned char *s, unsigned c, int e)
 144 {
 145         e &= 1;
 146         s[e] = c>>8;
 147         s[1-e] = c;
 148 }
 149
 150 static unsigned get_32(const unsigned char *s, int e)
 151 {
 152         e &= 3;
 153         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 154 }
 155
 156 static void put_32(unsigned char *s, unsigned c, int e)
 157 {
 158         e &= 3;
 159         s[e^0] = c>>24;
 160         s[e^1] = c>>16;
 161         s[e^2] = c>>8;
 162         s[e^3] = c;
 163 }
 164
 165 /* Adapt as needed */
 166 #define mbrtowc_utf8 mbrtowc
 167 #define wctomb_utf8 wctomb
 168
 169 static unsigned legacy_map(const unsigned char *map, unsigned c)
 170 {
 171         unsigned x = c - 128 - map[-1];
 172         x = legacy_chars[ map[x*5/4]>>2*x%8 |
 173                 map[x*5/4+1]<<8-2*x%8 & 1023 ];
 174         return x ? x : c;
 175 }
 176
 177 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 178 {
 179         size_t x=0;
 180         unsigned to = extract_to(cd);
 181         unsigned from = extract_from(cd);
 182         const unsigned char *map = charmaps+from+1;
 183         const unsigned char *tomap = charmaps+to+1;
 184         mbstate_t st = {0};
 185         wchar_t wc;
 186         unsigned c, d;
 187         size_t k, l;
 188         int err;
 189         unsigned char type = map[-1];
 190         unsigned char totype = tomap[-1];
 191         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 192
 193         if (!in || !*in || !*inb) return 0;
 194
 195         *ploc = UTF8_LOCALE;
 196
 197         for (; *inb; *in+=l, *inb-=l) {
 198                 c = *(unsigned char *)*in;
 199                 l = 1;
 200
 201                 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
 202                 case UTF_8:
 203                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 204                         if (!l) l++;
 205                         else if (l == (size_t)-1) goto ilseq;
 206                         else if (l == (size_t)-2) goto starved;
 207                         c = wc;
 208                         break;
 209                 case US_ASCII:
 210                         goto ilseq;
 211                 case WCHAR_T:
 212                         l = sizeof(wchar_t);
 213                         if (*inb < l) goto starved;
 214                         c = *(wchar_t *)*in;
 215                         if (0) {
 216                 case UTF_32BE:
 217                 case UTF_32LE:
 218                         l = 4;
 219                         if (*inb < 4) goto starved;
 220                         c = get_32((void *)*in, type);
 221                         }
 222                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 223                         break;
 224                 case UCS2BE:
 225                 case UCS2LE:
 226                 case UTF_16BE:
 227                 case UTF_16LE:
 228                         l = 2;
 229                         if (*inb < 2) goto starved;
 230                         c = get_16((void *)*in, type);
 231                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 232                         if ((unsigned)(c-0xd800) < 0x400) {
 233                                 if (type-UCS2BE < 2U) goto ilseq;
 234                                 l = 4;
 235                                 if (*inb < 4) goto starved;
 236                                 d = get_16((void *)(*in + 2), type);
 237                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 238                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 239                         }
 240                         break;
 241                 case SHIFT_JIS:
 242                         if (c-0xa1 <= 0xdf-0xa1) {
 243                                 c += 0xff61-0xa1;
 244                                 break;
 245                         }
 246                         l = 2;
 247                         if (*inb < 2) goto starved;
 248                         d = *((unsigned char *)*in + 1);
 249                         if (c-129 <= 159-129) c -= 129;
 250                         else if (c-224 <= 239-224) c -= 193;
 251                         else goto ilseq;
 252                         c *= 2;
 253                         if (d-64 <= 158-64) {
 254                                 if (d==127) goto ilseq;
 255                                 if (d>127) d--;
 256                                 d -= 64;
 257                         } else if (d-159 <= 252-159) {
 258                                 c++;
 259                                 d -= 159;
 260                         }
 261                         c = jis0208[c][d];
 262                         if (!c) goto ilseq;
 263                         break;
 264                 case EUC_JP:
 265                         l = 2;
 266                         if (*inb < 2) goto starved;
 267                         d = *((unsigned char *)*in + 1);
 268                         if (c==0x8e) {
 269                                 c = d;
 270                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 271                                 c += 0xff61 - 0xa1;
 272                                 break;
 273                         }
 274                         c -= 0xa1;
 275                         d -= 0xa1;
 276                         if (c >= 84 || d >= 94) goto ilseq;
 277                         c = jis0208[c][d];
 278                         if (!c) goto ilseq;
 279                         break;
 280                 case GB2312:
 281                         if (c < 0xa1) goto ilseq;
 282                 case GBK:
 283                 case GB18030:
 284                         c -= 0x81;
 285                         if (c >= 126) goto ilseq;
 286                         l = 2;
 287                         if (*inb < 2) goto starved;
 288                         d = *((unsigned char *)*in + 1);
 289                         if (d < 0xa1 && type == GB2312) goto ilseq;
 290                         if (d-0x40>=191 || d==127) {
 291                                 if (d-'0'>9 || type != GB18030)
 292                                         goto ilseq;
 293                                 l = 4;
 294                                 if (*inb < 4) goto starved;
 295                                 c = (10*c + d-'0') * 1260;
 296                                 d = *((unsigned char *)*in + 2);
 297                                 if (d-0x81>126) goto ilseq;
 298                                 c += 10*(d-0x81);
 299                                 d = *((unsigned char *)*in + 3);
 300                                 if (d-'0'>9) goto ilseq;
 301                                 c += d-'0';
 302                                 c += 128;
 303                                 for (d=0; d<=c; ) {
 304                                         k = 0;
 305                                         for (int i=0; i<126; i++)
 306                                                 for (int j=0; j<190; j++)
 307                                                         if (gb18030[i][j]-d <= c-d)
 308                                                                 k++;
 309                                         d = c+1;
 310                                         c += k;
 311                                 }
 312                                 break;
 313                         }
 314                         d -= 0x40;
 315                         if (d>63) d--;
 316                         c = gb18030[c][d];
 317                         break;
 318                 case BIG5:
 319                         l = 2;
 320                         if (*inb < 2) goto starved;
 321                         d = *((unsigned char *)*in + 1);
 322                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 323                         d -= 0x40;
 324                         if (d > 0x3e) d -= 0x22;
 325                         if (c-0xa1>=0xfa-0xa1) {
 326                                 if (c-0x87>=0xff-0x87) goto ilseq;
 327                                 if (c < 0xa1) c -= 0x87;
 328                                 else c -= 0x87 + (0xfa-0xa1);
 329                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 330                                         | hkscs[c*157+d];
 331                                 /* A few HKSCS characters map to pairs of UCS
 332                                  * characters. These are mapped to surrogate
 333                                  * range in the hkscs table then hard-coded
 334                                  * here. Ugly, yes. */
 335                                 if (c/256 == 0xdc) {
 336                                         if (totype-0300U > 8) k = 2;
 337                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 338                                         if (k > *outb) goto toobig;
 339                                         x += iconv(combine_to_from(to, 0),
 340                                                 &(char *){"\303\212\314\204"
 341                                                 "\303\212\314\214"
 342                                                 "\303\252\314\204"
 343                                                 "\303\252\314\214"
 344                                                 +c%256}, &(size_t){4},
 345                                                 out, outb);
 346                                         continue;
 347                                 }
 348                                 if (!c) goto ilseq;
 349                                 break;
 350                         }
 351                         c -= 0xa1;
 352                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 353                         if (!c) goto ilseq;
 354                         break;
 355                 case EUC_KR:
 356                         l = 2;
 357                         if (*inb < 2) goto starved;
 358                         d = *((unsigned char *)*in + 1);
 359                         c -= 0xa1;
 360                         d -= 0xa1;
 361                         if (c >= 93 || d >= 94) {
 362                                 c += (0xa1-0x81);
 363                                 d += 0xa1;
 364                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 365                                         goto ilseq;
 366                                 if (d-'A'<26) d = d-'A';
 367                                 else if (d-'a'<26) d = d-'a'+26;
 368                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 369                                 else goto ilseq;
 370                                 if (c < 0x20) c = 178*c + d;
 371                                 else c = 178*0x20 + 84*(c-0x20) + d;
 372                                 c += 0xac00;
 373                                 for (d=0xac00; d<=c; ) {
 374                                         k = 0;
 375                                         for (int i=0; i<93; i++)
 376                                                 for (int j=0; j<94; j++)
 377                                                         if (ksc[i][j]-d <= c-d)
 378                                                                 k++;
 379                                         d = c+1;
 380                                         c += k;
 381                                 }
 382                                 break;
 383                         }
 384                         c = ksc[c][d];
 385                         if (!c) goto ilseq;
 386                         break;
 387                 default:
 388                         if (c < 128+type) break;
 389                         c = legacy_map(map, c);
 390                         if (c==1) goto ilseq;
 391                 }
 392
 393                 switch (totype) {
 394                 case WCHAR_T:
 395                         if (*outb < sizeof(wchar_t)) goto toobig;
 396                         *(wchar_t *)*out = c;
 397                         *out += sizeof(wchar_t);
 398                         *outb -= sizeof(wchar_t);
 399                         break;
 400                 case UTF_8:
 401                         if (*outb < 4) {
 402                                 char tmp[4];
 403                                 k = wctomb_utf8(tmp, c);
 404                                 if (*outb < k) goto toobig;
 405                                 memcpy(*out, tmp, k);
 406                         } else k = wctomb_utf8(*out, c);
 407                         *out += k;
 408                         *outb -= k;
 409                         break;
 410                 case US_ASCII:
 411                         if (c > 0x7f) subst: x++, c='*';
 412                 default:
 413                         if (*outb < 1) goto toobig;
 414                         if (c < 128+totype || (c<256 && c==legacy_map(tomap, c))) {
 415                         revout:
 416                                 *(*out)++ = c;
 417                                 *outb -= 1;
 418                                 break;
 419                         }
 420                         d = c;
 421                         for (c=128+totype; c<256; c++) {
 422                                 if (d == legacy_map(tomap, c)) {
 423                                         goto revout;
 424                                 }
 425                         }
 426                         goto subst;
 427                 case UCS2BE:
 428                 case UCS2LE:
 429                 case UTF_16BE:
 430                 case UTF_16LE:
 431                         if (c < 0x10000 || type-UCS2BE < 2U) {
 432                                 if (c >= 0x10000) c = 0xFFFD;
 433                                 if (*outb < 2) goto toobig;
 434                                 put_16((void *)*out, c, totype);
 435                                 *out += 2;
 436                                 *outb -= 2;
 437                                 break;
 438                         }
 439                         if (*outb < 4) goto toobig;
 440                         c -= 0x10000;
 441                         put_16((void *)*out, (c>>10)|0xd800, totype);
 442                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 443                         *out += 4;
 444                         *outb -= 4;
 445                         break;
 446                 case UTF_32BE:
 447                 case UTF_32LE:
 448                         if (*outb < 4) goto toobig;
 449                         put_32((void *)*out, c, totype);
 450                         *out += 4;
 451                         *outb -= 4;
 452                         break;
 453                 }
 454         }
 455         *ploc = loc;
 456         return x;
 457 ilseq:
 458         err = EILSEQ;
 459         x = -1;
 460         goto end;
 461 toobig:
 462         err = E2BIG;
 463         x = -1;
 464         goto end;
 465 starved:
 466         err = EINVAL;
 467         x = -1;
 468 end:
 469         errno = err;
 470         *ploc = loc;
 471         return x;
 472 }