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