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 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 size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 155 {
 156         size_t x=0;
 157         unsigned long cd = (unsigned long)cd0;
 158         unsigned to = cd & 0xffff;
 159         unsigned from = cd >> 16;
 160         const unsigned char *map = charmaps+from+1;
 161         const unsigned char *tomap = charmaps+to+1;
 162         mbstate_t st = {0};
 163         wchar_t wc;
 164         unsigned c, d;
 165         size_t k, l;
 166         int err;
 167         unsigned char type = map[-1];
 168         unsigned char totype = tomap[-1];
 169         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 170
 171         if (!in || !*in || !*inb) return 0;
 172
 173         *ploc = UTF8_LOCALE;
 174
 175         for (; *inb; *in+=l, *inb-=l) {
 176                 c = *(unsigned char *)*in;
 177                 l = 1;
 178
 179                 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
 180                 case UTF_8:
 181                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 182                         if (!l) l++;
 183                         else if (l == (size_t)-1) goto ilseq;
 184                         else if (l == (size_t)-2) goto starved;
 185                         c = wc;
 186                         break;
 187                 case US_ASCII:
 188                         goto ilseq;
 189                 case WCHAR_T:
 190                         l = sizeof(wchar_t);
 191                         if (*inb < l) goto starved;
 192                         c = *(wchar_t *)*in;
 193                         if (0) {
 194                 case UTF_32BE:
 195                 case UTF_32LE:
 196                         l = 4;
 197                         if (*inb < 4) goto starved;
 198                         c = get_32((void *)*in, type);
 199                         }
 200                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 201                         break;
 202                 case UCS2BE:
 203                 case UCS2LE:
 204                 case UTF_16BE:
 205                 case UTF_16LE:
 206                         l = 2;
 207                         if (*inb < 2) goto starved;
 208                         c = get_16((void *)*in, type);
 209                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 210                         if ((unsigned)(c-0xd800) < 0x400) {
 211                                 if (type-UCS2BE < 2U) goto ilseq;
 212                                 l = 4;
 213                                 if (*inb < 4) goto starved;
 214                                 d = get_16((void *)(*in + 2), type);
 215                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 216                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 217                         }
 218                         break;
 219                 case SHIFT_JIS:
 220                         if (c-0xa1 <= 0xdf-0xa1) {
 221                                 c += 0xff61-0xa1;
 222                                 break;
 223                         }
 224                         l = 2;
 225                         if (*inb < 2) goto starved;
 226                         d = *((unsigned char *)*in + 1);
 227                         if (c-129 <= 159-129) c -= 129;
 228                         else if (c-224 <= 239-224) c -= 193;
 229                         else goto ilseq;
 230                         c *= 2;
 231                         if (d-64 <= 158-64) {
 232                                 if (d==127) goto ilseq;
 233                                 if (d>127) d--;
 234                                 d -= 64;
 235                         } else if (d-159 <= 252-159) {
 236                                 c++;
 237                                 d -= 159;
 238                         }
 239                         c = jis0208[c][d];
 240                         if (!c) goto ilseq;
 241                         break;
 242                 case EUC_JP:
 243                         l = 2;
 244                         if (*inb < 2) goto starved;
 245                         d = *((unsigned char *)*in + 1);
 246                         if (c==0x8e) {
 247                                 c = d;
 248                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 249                                 c += 0xff61 - 0xa1;
 250                                 break;
 251                         }
 252                         c -= 0xa1;
 253                         d -= 0xa1;
 254                         if (c >= 84 || d >= 94) goto ilseq;
 255                         c = jis0208[c][d];
 256                         if (!c) goto ilseq;
 257                         break;
 258                 case GB2312:
 259                         if (c < 0xa1) goto ilseq;
 260                 case GBK:
 261                 case GB18030:
 262                         c -= 0x81;
 263                         if (c >= 126) goto ilseq;
 264                         l = 2;
 265                         if (*inb < 2) goto starved;
 266                         d = *((unsigned char *)*in + 1);
 267                         if (d < 0xa1 && type == GB2312) goto ilseq;
 268                         if (d-0x40>=191 || d==127) {
 269                                 if (d-'0'>9 || type != GB18030)
 270                                         goto ilseq;
 271                                 l = 4;
 272                                 if (*inb < 4) goto starved;
 273                                 c = (10*c + d-'0') * 1260;
 274                                 d = *((unsigned char *)*in + 2);
 275                                 if (d-0x81>126) goto ilseq;
 276                                 c += 10*(d-0x81);
 277                                 d = *((unsigned char *)*in + 3);
 278                                 if (d-'0'>9) goto ilseq;
 279                                 c += d-'0';
 280                                 c += 128;
 281                                 for (d=0; d<=c; ) {
 282                                         k = 0;
 283                                         for (int i=0; i<126; i++)
 284                                                 for (int j=0; j<190; j++)
 285                                                         if (gb18030[i][j]-d <= c-d)
 286                                                                 k++;
 287                                         d = c+1;
 288                                         c += k;
 289                                 }
 290                                 break;
 291                         }
 292                         d -= 0x40;
 293                         if (d>63) d--;
 294                         c = gb18030[c][d];
 295                         break;
 296                 case BIG5:
 297                         l = 2;
 298                         if (*inb < 2) goto starved;
 299                         d = *((unsigned char *)*in + 1);
 300                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 301                         d -= 0x40;
 302                         if (d > 0x3e) d -= 0x22;
 303                         if (c-0xa1>=0xfa-0xa1) {
 304                                 if (c-0x87>=0xff-0x87) goto ilseq;
 305                                 if (c < 0xa1) c -= 0x87;
 306                                 else c -= 0x87 + (0xfa-0xa1);
 307                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 308                                         | hkscs[c*157+d];
 309                                 /* A few HKSCS characters map to pairs of UCS
 310                                  * characters. These are mapped to surrogate
 311                                  * range in the hkscs table then hard-coded
 312                                  * here. Ugly, yes. */
 313                                 if (c/256 == 0xdc) {
 314                                         if (totype-0300U > 8) k = 2;
 315                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 316                                         if (k > *outb) goto toobig;
 317                                         x += iconv((iconv_t)(uintptr_t)to,
 318                                                 &(char *){"\303\212\314\204"
 319                                                 "\303\212\314\214"
 320                                                 "\303\252\314\204"
 321                                                 "\303\252\314\214"
 322                                                 +c%256}, &(size_t){4},
 323                                                 out, outb);
 324                                         continue;
 325                                 }
 326                                 if (!c) goto ilseq;
 327                                 break;
 328                         }
 329                         c -= 0xa1;
 330                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 331                         if (!c) goto ilseq;
 332                         break;
 333                 case EUC_KR:
 334                         l = 2;
 335                         if (*inb < 2) goto starved;
 336                         d = *((unsigned char *)*in + 1);
 337                         c -= 0xa1;
 338                         d -= 0xa1;
 339                         if (c >= 93 || d >= 94) {
 340                                 c += (0xa1-0x81);
 341                                 d += 0xa1;
 342                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 343                                         goto ilseq;
 344                                 if (d-'A'<26) d = d-'A';
 345                                 else if (d-'a'<26) d = d-'a'+26;
 346                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 347                                 else goto ilseq;
 348                                 if (c < 0x20) c = 178*c + d;
 349                                 else c = 178*0x20 + 84*(c-0x20) + d;
 350                                 c += 0xac00;
 351                                 for (d=0xac00; d<=c; ) {
 352                                         k = 0;
 353                                         for (int i=0; i<93; i++)
 354                                                 for (int j=0; j<94; j++)
 355                                                         if (ksc[i][j]-d <= c-d)
 356                                                                 k++;
 357                                         d = c+1;
 358                                         c += k;
 359                                 }
 360                                 break;
 361                         }
 362                         c = ksc[c][d];
 363                         if (!c) goto ilseq;
 364                         break;
 365                 default:
 366                         if (c < 128+type) break;
 367                         c -= 128+type;
 368                         c = legacy_chars[ map[c*5/4]>>2*c%8 |
 369                                 map[c*5/4+1]<<8-2*c%8 & 1023 ];
 370                         if (!c) c = *(unsigned char *)*in;
 371                         if (c==1) goto ilseq;
 372                 }
 373
 374                 switch (totype) {
 375                 case WCHAR_T:
 376                         if (*outb < sizeof(wchar_t)) goto toobig;
 377                         *(wchar_t *)*out = c;
 378                         *out += sizeof(wchar_t);
 379                         *outb -= sizeof(wchar_t);
 380                         break;
 381                 case UTF_8:
 382                         if (*outb < 4) {
 383                                 char tmp[4];
 384                                 k = wctomb_utf8(tmp, c);
 385                                 if (*outb < k) goto toobig;
 386                                 memcpy(*out, tmp, k);
 387                         } else k = wctomb_utf8(*out, c);
 388                         *out += k;
 389                         *outb -= k;
 390                         break;
 391                 case US_ASCII:
 392                         if (c > 0x7f) subst: x++, c='*';
 393                 default:
 394                         if (*outb < 1) goto toobig;
 395                         if (c < 128+totype) {
 396                         revout:
 397                                 *(*out)++ = c;
 398                                 *outb -= 1;
 399                                 break;
 400                         }
 401                         d = c;
 402                         for (c=0; c<128-totype; c++) {
 403                                 if (d == legacy_chars[ tomap[c*5/4]>>2*c%8 |
 404                                         tomap[c*5/4+1]<<8-2*c%8 & 1023 ]) {
 405                                         c += 128;
 406                                         goto revout;
 407                                 }
 408                         }
 409                         goto subst;
 410                 case UCS2BE:
 411                 case UCS2LE:
 412                 case UTF_16BE:
 413                 case UTF_16LE:
 414                         if (c < 0x10000 || type-UCS2BE < 2U) {
 415                                 if (c >= 0x10000) c = 0xFFFD;
 416                                 if (*outb < 2) goto toobig;
 417                                 put_16((void *)*out, c, totype);
 418                                 *out += 2;
 419                                 *outb -= 2;
 420                                 break;
 421                         }
 422                         if (*outb < 4) goto toobig;
 423                         c -= 0x10000;
 424                         put_16((void *)*out, (c>>10)|0xd800, totype);
 425                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 426                         *out += 4;
 427                         *outb -= 4;
 428                         break;
 429                 case UTF_32BE:
 430                 case UTF_32LE:
 431                         if (*outb < 4) goto toobig;
 432                         put_32((void *)*out, c, totype);
 433                         *out += 4;
 434                         *outb -= 4;
 435                         break;
 436                 }
 437         }
 438         *ploc = loc;
 439         return x;
 440 ilseq:
 441         err = EILSEQ;
 442         x = -1;
 443         goto end;
 444 toobig:
 445         err = E2BIG;
 446         x = -1;
 447         goto end;
 448 starved:
 449         err = EINVAL;
 450         x = -1;
 451 end:
 452         errno = err;
 453         *ploc = loc;
 454         return x;
 455 }