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 ISO2022_JP  0322
  22 #define GB18030     0330
  23 #define GBK         0331
  24 #define GB2312      0332
  25 #define BIG5        0340
  26 #define EUC_KR      0350
  27
  28 /* Definitions of charmaps. Each charmap consists of:
  29  * 1. Empty-string-terminated list of null-terminated aliases.
  30  * 2. Special type code or number of elided quads of entries.
  31  * 3. Character table (size determined by field 2), consisting
  32  *    of 5 bytes for every 4 characters, interpreted as 10-bit
  33  *    indices into the legacy_chars table. */
  34
  35 static const unsigned char charmaps[] =
  36 "utf8\0char\0\0\310"
  37 "wchart\0\0\306"
  38 "ucs2\0ucs2be\0\0\304"
  39 "ucs2le\0\0\305"
  40 "utf16\0utf16be\0\0\302"
  41 "utf16le\0\0\301"
  42 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
  43 "ucs4le\0utf32le\0\0\303"
  44 "ascii\0usascii\0iso646\0iso646us\0\0\307"
  45 "eucjp\0\0\320"
  46 "shiftjis\0sjis\0\0\321"
  47 "iso2022jp\0\0\322"
  48 "gb18030\0\0\330"
  49 "gbk\0\0\331"
  50 "gb2312\0\0\332"
  51 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
  52 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
  53 #include "codepages.h"
  54 ;
  55
  56 /* Table of characters that appear in legacy 8-bit codepages,
  57  * limited to 1024 slots (10 bit indices). The first 256 entries
  58  * are elided since those characters are obviously all included. */
  59 static const unsigned short legacy_chars[] = {
  60 #include "legacychars.h"
  61 };
  62
  63 static const unsigned short jis0208[84][94] = {
  64 #include "jis0208.h"
  65 };
  66
  67 static const unsigned short gb18030[126][190] = {
  68 #include "gb18030.h"
  69 };
  70
  71 static const unsigned short big5[89][157] = {
  72 #include "big5.h"
  73 };
  74
  75 static const unsigned short hkscs[] = {
  76 #include "hkscs.h"
  77 };
  78
  79 static const unsigned short ksc[93][94] = {
  80 #include "ksc.h"
  81 };
  82
  83 static const unsigned short rev_jis[] = {
  84 #include "revjis.h"
  85 };
  86
  87 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
  88 {
  89         for (; *a && *b; a++, b++) {
  90                 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
  91                 if ((*a|32U) != *b) return 1;
  92         }
  93         return *a != *b;
  94 }
  95
  96 static size_t find_charmap(const void *name)
  97 {
  98         const unsigned char *s;
  99         if (!*(char *)name) name=charmaps; /* "utf8" */
 100         for (s=charmaps; *s; ) {
 101                 if (!fuzzycmp(name, s)) {
 102                         for (; *s; s+=strlen((void *)s)+1);
 103                         return s+1-charmaps;
 104                 }
 105                 s += strlen((void *)s)+1;
 106                 if (!*s) {
 107                         if (s[1] > 0200) s+=2;
 108                         else s+=2+(64U-s[1])*5;
 109                 }
 110         }
 111         return -1;
 112 }
 113
 114 struct stateful_cd {
 115         iconv_t base_cd;
 116         unsigned state;
 117 };
 118
 119 static iconv_t combine_to_from(size_t t, size_t f)
 120 {
 121         return (void *)(f<<16 | t<<1 | 1);
 122 }
 123
 124 static size_t extract_from(iconv_t cd)
 125 {
 126         return (size_t)cd >> 16;
 127 }
 128
 129 static size_t extract_to(iconv_t cd)
 130 {
 131         return (size_t)cd >> 1 & 0x7fff;
 132 }
 133
 134 iconv_t iconv_open(const char *to, const char *from)
 135 {
 136         size_t f, t;
 137         struct stateful_cd *scd;
 138
 139         if ((t = find_charmap(to))==-1
 140          || (f = find_charmap(from))==-1
 141          || (charmaps[t] >= 0330)) {
 142                 errno = EINVAL;
 143                 return (iconv_t)-1;
 144         }
 145         iconv_t cd = combine_to_from(t, f);
 146
 147         switch (charmaps[f]) {
 148         case ISO2022_JP:
 149                 scd = malloc(sizeof *scd);
 150                 if (!scd) return (iconv_t)-1;
 151                 scd->base_cd = cd;
 152                 scd->state = 0;
 153                 cd = (iconv_t)scd;
 154         }
 155
 156         return cd;
 157 }
 158
 159 static unsigned get_16(const unsigned char *s, int e)
 160 {
 161         e &= 1;
 162         return s[e]<<8 | s[1-e];
 163 }
 164
 165 static void put_16(unsigned char *s, unsigned c, int e)
 166 {
 167         e &= 1;
 168         s[e] = c>>8;
 169         s[1-e] = c;
 170 }
 171
 172 static unsigned get_32(const unsigned char *s, int e)
 173 {
 174         e &= 3;
 175         return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
 176 }
 177
 178 static void put_32(unsigned char *s, unsigned c, int e)
 179 {
 180         e &= 3;
 181         s[e^0] = c>>24;
 182         s[e^1] = c>>16;
 183         s[e^2] = c>>8;
 184         s[e^3] = c;
 185 }
 186
 187 /* Adapt as needed */
 188 #define mbrtowc_utf8 mbrtowc
 189 #define wctomb_utf8 wctomb
 190
 191 static unsigned legacy_map(const unsigned char *map, unsigned c)
 192 {
 193         if (c < 4*map[-1]) return c;
 194         unsigned x = c - 4*map[-1];
 195         x = map[x*5/4]>>2*x%8 | map[x*5/4+1]<<8-2*x%8 & 1023;
 196         return x < 256 ? x : legacy_chars[x-256];
 197 }
 198
 199 static unsigned uni_to_jis(unsigned c)
 200 {
 201         unsigned nel = sizeof rev_jis / sizeof *rev_jis;
 202         unsigned d, j, i, b = 0;
 203         for (;;) {
 204                 i = nel/2;
 205                 j = rev_jis[b+i];
 206                 d = jis0208[j/256][j%256];
 207                 if (d==c) return j + 0x2121;
 208                 else if (nel == 1) return 0;
 209                 else if (c < d)
 210                         nel /= 2;
 211                 else {
 212                         b += i;
 213                         nel -= nel/2;
 214                 }
 215         }
 216 }
 217
 218 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
 219 {
 220         size_t x=0;
 221         struct stateful_cd *scd=0;
 222         if (!((size_t)cd & 1)) {
 223                 scd = (void *)cd;
 224                 cd = scd->base_cd;
 225         }
 226         unsigned to = extract_to(cd);
 227         unsigned from = extract_from(cd);
 228         const unsigned char *map = charmaps+from+1;
 229         const unsigned char *tomap = charmaps+to+1;
 230         mbstate_t st = {0};
 231         wchar_t wc;
 232         unsigned c, d;
 233         size_t k, l;
 234         int err;
 235         unsigned char type = map[-1];
 236         unsigned char totype = tomap[-1];
 237         locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 238
 239         if (!in || !*in || !*inb) return 0;
 240
 241         *ploc = UTF8_LOCALE;
 242
 243         for (; *inb; *in+=l, *inb-=l) {
 244                 c = *(unsigned char *)*in;
 245                 l = 1;
 246
 247                 switch (type) {
 248                 case UTF_8:
 249                         if (c < 128) break;
 250                         l = mbrtowc_utf8(&wc, *in, *inb, &st);
 251                         if (l == (size_t)-1) goto ilseq;
 252                         if (l == (size_t)-2) goto starved;
 253                         c = wc;
 254                         break;
 255                 case US_ASCII:
 256                         if (c >= 128) goto ilseq;
 257                         break;
 258                 case WCHAR_T:
 259                         l = sizeof(wchar_t);
 260                         if (*inb < l) goto starved;
 261                         c = *(wchar_t *)*in;
 262                         if (0) {
 263                 case UTF_32BE:
 264                 case UTF_32LE:
 265                         l = 4;
 266                         if (*inb < 4) goto starved;
 267                         c = get_32((void *)*in, type);
 268                         }
 269                         if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
 270                         break;
 271                 case UCS2BE:
 272                 case UCS2LE:
 273                 case UTF_16BE:
 274                 case UTF_16LE:
 275                         l = 2;
 276                         if (*inb < 2) goto starved;
 277                         c = get_16((void *)*in, type);
 278                         if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
 279                         if ((unsigned)(c-0xd800) < 0x400) {
 280                                 if (type-UCS2BE < 2U) goto ilseq;
 281                                 l = 4;
 282                                 if (*inb < 4) goto starved;
 283                                 d = get_16((void *)(*in + 2), type);
 284                                 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
 285                                 c = ((c-0xd7c0)<<10) + (d-0xdc00);
 286                         }
 287                         break;
 288                 case SHIFT_JIS:
 289                         if (c < 128) break;
 290                         if (c-0xa1 <= 0xdf-0xa1) {
 291                                 c += 0xff61-0xa1;
 292                                 break;
 293                         }
 294                         l = 2;
 295                         if (*inb < 2) goto starved;
 296                         d = *((unsigned char *)*in + 1);
 297                         if (c-129 <= 159-129) c -= 129;
 298                         else if (c-224 <= 239-224) c -= 193;
 299                         else goto ilseq;
 300                         c *= 2;
 301                         if (d-64 <= 158-64) {
 302                                 if (d==127) goto ilseq;
 303                                 if (d>127) d--;
 304                                 d -= 64;
 305                         } else if (d-159 <= 252-159) {
 306                                 c++;
 307                                 d -= 159;
 308                         }
 309                         c = jis0208[c][d];
 310                         if (!c) goto ilseq;
 311                         break;
 312                 case EUC_JP:
 313                         if (c < 128) break;
 314                         l = 2;
 315                         if (*inb < 2) goto starved;
 316                         d = *((unsigned char *)*in + 1);
 317                         if (c==0x8e) {
 318                                 c = d;
 319                                 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
 320                                 c += 0xff61 - 0xa1;
 321                                 break;
 322                         }
 323                         c -= 0xa1;
 324                         d -= 0xa1;
 325                         if (c >= 84 || d >= 94) goto ilseq;
 326                         c = jis0208[c][d];
 327                         if (!c) goto ilseq;
 328                         break;
 329                 case ISO2022_JP:
 330                         if (c >= 128) goto ilseq;
 331                         if (c == '\033') {
 332                                 l = 3;
 333                                 if (*inb < 3) goto starved;
 334                                 c = *((unsigned char *)*in + 1);
 335                                 d = *((unsigned char *)*in + 2);
 336                                 if (c != '(' && c != '$') goto ilseq;
 337                                 switch (128*(c=='$') + d) {
 338                                 case 'B': scd->state=0; continue;
 339                                 case 'J': scd->state=1; continue;
 340                                 case 'I': scd->state=4; continue;
 341                                 case 128+'@': scd->state=2; continue;
 342                                 case 128+'B': scd->state=3; continue;
 343                                 }
 344                                 goto ilseq;
 345                         }
 346                         switch (scd->state) {
 347                         case 1:
 348                                 if (c=='\\') c = 0xa5;
 349                                 if (c=='~') c = 0x203e;
 350                                 break;
 351                         case 2:
 352                         case 3:
 353                                 l = 2;
 354                                 if (*inb < 2) goto starved;
 355                                 d = *((unsigned char *)*in + 1);
 356                                 c -= 0x21;
 357                                 d -= 0x21;
 358                                 if (c >= 84 || d >= 94) goto ilseq;
 359                                 c = jis0208[c][d];
 360                                 if (!c) goto ilseq;
 361                                 break;
 362                         case 4:
 363                                 if (c-0x60 < 0x1f) goto ilseq;
 364                                 if (c-0x21 < 0x5e) c += 0xff61-0x21;
 365                                 break;
 366                         }
 367                         break;
 368                 case GB2312:
 369                         if (c < 128) break;
 370                         if (c < 0xa1) goto ilseq;
 371                 case GBK:
 372                 case GB18030:
 373                         if (c < 128) break;
 374                         c -= 0x81;
 375                         if (c >= 126) goto ilseq;
 376                         l = 2;
 377                         if (*inb < 2) goto starved;
 378                         d = *((unsigned char *)*in + 1);
 379                         if (d < 0xa1 && type == GB2312) goto ilseq;
 380                         if (d-0x40>=191 || d==127) {
 381                                 if (d-'0'>9 || type != GB18030)
 382                                         goto ilseq;
 383                                 l = 4;
 384                                 if (*inb < 4) goto starved;
 385                                 c = (10*c + d-'0') * 1260;
 386                                 d = *((unsigned char *)*in + 2);
 387                                 if (d-0x81>126) goto ilseq;
 388                                 c += 10*(d-0x81);
 389                                 d = *((unsigned char *)*in + 3);
 390                                 if (d-'0'>9) goto ilseq;
 391                                 c += d-'0';
 392                                 c += 128;
 393                                 for (d=0; d<=c; ) {
 394                                         k = 0;
 395                                         for (int i=0; i<126; i++)
 396                                                 for (int j=0; j<190; j++)
 397                                                         if (gb18030[i][j]-d <= c-d)
 398                                                                 k++;
 399                                         d = c+1;
 400                                         c += k;
 401                                 }
 402                                 break;
 403                         }
 404                         d -= 0x40;
 405                         if (d>63) d--;
 406                         c = gb18030[c][d];
 407                         break;
 408                 case BIG5:
 409                         if (c < 128) break;
 410                         l = 2;
 411                         if (*inb < 2) goto starved;
 412                         d = *((unsigned char *)*in + 1);
 413                         if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
 414                         d -= 0x40;
 415                         if (d > 0x3e) d -= 0x22;
 416                         if (c-0xa1>=0xfa-0xa1) {
 417                                 if (c-0x87>=0xff-0x87) goto ilseq;
 418                                 if (c < 0xa1) c -= 0x87;
 419                                 else c -= 0x87 + (0xfa-0xa1);
 420                                 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
 421                                         | hkscs[c*157+d];
 422                                 /* A few HKSCS characters map to pairs of UCS
 423                                  * characters. These are mapped to surrogate
 424                                  * range in the hkscs table then hard-coded
 425                                  * here. Ugly, yes. */
 426                                 if (c/256 == 0xdc) {
 427                                         if (totype-0300U > 8) k = 2;
 428                                         else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
 429                                         if (k > *outb) goto toobig;
 430                                         x += iconv(combine_to_from(to, 0),
 431                                                 &(char *){"\303\212\314\204"
 432                                                 "\303\212\314\214"
 433                                                 "\303\252\314\204"
 434                                                 "\303\252\314\214"
 435                                                 +c%256}, &(size_t){4},
 436                                                 out, outb);
 437                                         continue;
 438                                 }
 439                                 if (!c) goto ilseq;
 440                                 break;
 441                         }
 442                         c -= 0xa1;
 443                         c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
 444                         if (!c) goto ilseq;
 445                         break;
 446                 case EUC_KR:
 447                         if (c < 128) break;
 448                         l = 2;
 449                         if (*inb < 2) goto starved;
 450                         d = *((unsigned char *)*in + 1);
 451                         c -= 0xa1;
 452                         d -= 0xa1;
 453                         if (c >= 93 || d >= 94) {
 454                                 c += (0xa1-0x81);
 455                                 d += 0xa1;
 456                                 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
 457                                         goto ilseq;
 458                                 if (d-'A'<26) d = d-'A';
 459                                 else if (d-'a'<26) d = d-'a'+26;
 460                                 else if (d-0x81<0xff-0x81) d = d-0x81+52;
 461                                 else goto ilseq;
 462                                 if (c < 0x20) c = 178*c + d;
 463                                 else c = 178*0x20 + 84*(c-0x20) + d;
 464                                 c += 0xac00;
 465                                 for (d=0xac00; d<=c; ) {
 466                                         k = 0;
 467                                         for (int i=0; i<93; i++)
 468                                                 for (int j=0; j<94; j++)
 469                                                         if (ksc[i][j]-d <= c-d)
 470                                                                 k++;
 471                                         d = c+1;
 472                                         c += k;
 473                                 }
 474                                 break;
 475                         }
 476                         c = ksc[c][d];
 477                         if (!c) goto ilseq;
 478                         break;
 479                 default:
 480                         if (!c) break;
 481                         c = legacy_map(map, c);
 482                         if (!c) goto ilseq;
 483                 }
 484
 485                 switch (totype) {
 486                 case WCHAR_T:
 487                         if (*outb < sizeof(wchar_t)) goto toobig;
 488                         *(wchar_t *)*out = c;
 489                         *out += sizeof(wchar_t);
 490                         *outb -= sizeof(wchar_t);
 491                         break;
 492                 case UTF_8:
 493                         if (*outb < 4) {
 494                                 char tmp[4];
 495                                 k = wctomb_utf8(tmp, c);
 496                                 if (*outb < k) goto toobig;
 497                                 memcpy(*out, tmp, k);
 498                         } else k = wctomb_utf8(*out, c);
 499                         *out += k;
 500                         *outb -= k;
 501                         break;
 502                 case US_ASCII:
 503                         if (c > 0x7f) subst: x++, c='*';
 504                 default:
 505                         if (*outb < 1) goto toobig;
 506                         if (c<256 && c==legacy_map(tomap, c)) {
 507                         revout:
 508                                 *(*out)++ = c;
 509                                 *outb -= 1;
 510                                 break;
 511                         }
 512                         d = c;
 513                         for (c=4*totype; c<256; c++) {
 514                                 if (d == legacy_map(tomap, c)) {
 515                                         goto revout;
 516                                 }
 517                         }
 518                         goto subst;
 519                 case SHIFT_JIS:
 520                         if (c < 128) goto revout;
 521                         if (c == 0xa5) {
 522                                 x++;
 523                                 c = '\\';
 524                                 goto revout;
 525                         }
 526                         if (c == 0x203e) {
 527                                 x++;
 528                                 c = '~';
 529                                 goto revout;
 530                         }
 531                         if (c-0xff61 <= 0xdf-0xa1) {
 532                                 c += 0xa1 - 0xff61;
 533                                 goto revout;
 534                         }
 535                         c = uni_to_jis(c);
 536                         if (!c) goto subst;
 537                         if (*outb < 2) goto toobig;
 538                         d = c%256;
 539                         c = c/256;
 540                         *(*out)++ = (c+1)/2 + (c<95 ? 112 : 176);
 541                         *(*out)++ = c%2 ? d + 31 + d/96 : d + 126;
 542                         *outb -= 2;
 543                         break;
 544                 case EUC_JP:
 545                         if (c < 128) goto revout;
 546                         if (c-0xff61 <= 0xdf-0xa1) {
 547                                 c += 0x0e00 + 0x21 - 0xff61;
 548                         } else {
 549                                 c = uni_to_jis(c);
 550                         }
 551                         if (!c) goto subst;
 552                         if (*outb < 2) goto toobig;
 553                         *(*out)++ = c/256 + 0x80;
 554                         *(*out)++ = c%256 + 0x80;
 555                         *outb -= 2;
 556                         break;
 557                 case ISO2022_JP:
 558                         if (c < 128) goto revout;
 559                         if (c-0xff61 <= 0xdf-0xa1 || c==0xa5 || c==0x203e) {
 560                                 if (*outb < 7) goto toobig;
 561                                 *(*out)++ = '\033';
 562                                 *(*out)++ = '(';
 563                                 if (c==0xa5) {
 564                                         *(*out)++ = 'J';
 565                                         *(*out)++ = '\\';
 566                                 } else if (c==0x203e) {
 567                                         *(*out)++ = 'J';
 568                                         *(*out)++ = '~';
 569                                 } else {
 570                                         *(*out)++ = 'I';
 571                                         *(*out)++ = c-0xff61+0x21;
 572                                 }
 573                                 *(*out)++ = '\033';
 574                                 *(*out)++ = '(';
 575                                 *(*out)++ = 'B';
 576                                 *outb -= 7;
 577                                 break;
 578                         }
 579                         c = uni_to_jis(c);
 580                         if (!c) goto subst;
 581                         if (*outb < 8) goto toobig;
 582                         *(*out)++ = '\033';
 583                         *(*out)++ = '$';
 584                         *(*out)++ = 'B';
 585                         *(*out)++ = c/256;
 586                         *(*out)++ = c%256;
 587                         *(*out)++ = '\033';
 588                         *(*out)++ = '(';
 589                         *(*out)++ = 'B';
 590                         *outb -= 8;
 591                         break;
 592                 case UCS2BE:
 593                 case UCS2LE:
 594                 case UTF_16BE:
 595                 case UTF_16LE:
 596                         if (c < 0x10000 || type-UCS2BE < 2U) {
 597                                 if (c >= 0x10000) c = 0xFFFD;
 598                                 if (*outb < 2) goto toobig;
 599                                 put_16((void *)*out, c, totype);
 600                                 *out += 2;
 601                                 *outb -= 2;
 602                                 break;
 603                         }
 604                         if (*outb < 4) goto toobig;
 605                         c -= 0x10000;
 606                         put_16((void *)*out, (c>>10)|0xd800, totype);
 607                         put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
 608                         *out += 4;
 609                         *outb -= 4;
 610                         break;
 611                 case UTF_32BE:
 612                 case UTF_32LE:
 613                         if (*outb < 4) goto toobig;
 614                         put_32((void *)*out, c, totype);
 615                         *out += 4;
 616                         *outb -= 4;
 617                         break;
 618                 }
 619         }
 620         *ploc = loc;
 621         return x;
 622 ilseq:
 623         err = EILSEQ;
 624         x = -1;
 625         goto end;
 626 toobig:
 627         err = E2BIG;
 628         x = -1;
 629         goto end;
 630 starved:
 631         err = EINVAL;
 632         x = -1;
 633 end:
 634         errno = err;
 635         *ploc = loc;
 636         return x;
 637 }