1 /* 2 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 3 * Copyright 2012 Garrett D'Amore <garrett@damore.org> All rights reserved. 4 * Copyright 2015 John Marino <draco@marino.st> 5 * 6 * This source code is derived from the illumos localedef command, and 7 * provided under BSD-style license terms by Nexenta Systems, Inc. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 20 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 23 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * The functions in this file convert from the standard multibyte forms 34 * to the wide character forms used internally by libc. Unfortunately, 35 * this approach means that we need a method for each and every encoding. 36 */ 37 38 #include <stdlib.h> 39 #include <wchar.h> 40 #include <string.h> 41 #include <sys/types.h> 42 #include "localedef.h" 43 44 static int towide_none(wchar_t *, const char *, unsigned); 45 static int towide_utf8(wchar_t *, const char *, unsigned); 46 static int towide_big5(wchar_t *, const char *, unsigned); 47 static int towide_gbk(wchar_t *, const char *, unsigned); 48 static int towide_gb2312(wchar_t *, const char *, unsigned); 49 static int towide_gb18030(wchar_t *, const char *, unsigned); 50 static int towide_mskanji(wchar_t *, const char *, unsigned); 51 static int towide_euccn(wchar_t *, const char *, unsigned); 52 static int towide_eucjp(wchar_t *, const char *, unsigned); 53 static int towide_euckr(wchar_t *, const char *, unsigned); 54 static int towide_euctw(wchar_t *, const char *, unsigned); 55 56 static int tomb_none(char *, wchar_t); 57 static int tomb_utf8(char *, wchar_t); 58 static int tomb_mbs(char *, wchar_t); 59 60 static int (*_towide)(wchar_t *, const char *, unsigned) = towide_none; 61 static int (*_tomb)(char *, wchar_t) = tomb_none; 62 static const char *_encoding = "NONE"; 63 static int _nbits = 7; 64 65 /* 66 * Table of supported encodings. We only bother to list the multibyte 67 * encodings here, because single byte locales are handed by "NONE". 68 */ 69 static struct { 70 const char *name; 71 /* the name that the underlying libc implemenation uses */ 72 const char *cname; 73 /* the maximum number of bits required for priorities */ 74 int nbits; 75 int (*towide)(wchar_t *, const char *, unsigned); 76 int (*tomb)(char *, wchar_t); 77 } mb_encodings[] = { 78 /* 79 * UTF8 values max out at 0x1fffff (although in theory there could 80 * be later extensions, but it won't happen.) This means we only need 81 * 21 bits to be able to encode the entire range of priorities. 82 */ 83 { "UTF-8", "UTF-8", 21, towide_utf8, tomb_utf8 }, 84 { "UTF8", "UTF-8", 21, towide_utf8, tomb_utf8 }, 85 { "utf8", "UTF-8", 21, towide_utf8, tomb_utf8 }, 86 { "utf-8", "UTF-8", 21, towide_utf8, tomb_utf8 }, 87 88 { "EUC-CN", "EUC-CN", 16, towide_euccn, tomb_mbs }, 89 { "eucCN", "EUC-CN", 16, towide_euccn, tomb_mbs }, 90 /* 91 * Becuase the 3-byte form of EUC-JP use the same leading byte, 92 * only 17 bits required to provide unique priorities. (The low 93 * bit of that first byte is set.) By setting this value low, 94 * we can get by with only 3 bytes in the strxfrm expansion. 95 */ 96 { "EUC-JP", "EUC-JP", 17, towide_eucjp, tomb_mbs }, 97 { "eucJP", "EUC-JP", 17, towide_eucjp, tomb_mbs }, 98 99 { "EUC-KR", "EUC-KR", 16, towide_euckr, tomb_mbs }, 100 { "eucKR", "EUC-KR", 16, towide_euckr, tomb_mbs }, 101 /* 102 * EUC-TW uses 2 bytes most of the time, but 4 bytes if the 103 * high order byte is 0x8E. However, with 4 byte encodings, 104 * the third byte will be A0-B0. So we only need to consider 105 * the lower order 24 bits for collation. 106 */ 107 { "EUC-TW", "EUC-TW", 24, towide_euctw, tomb_mbs }, 108 { "eucTW", "EUC-TW", 24, towide_euctw, tomb_mbs }, 109 110 { "MS_Kanji", "MSKanji", 16, towide_mskanji, tomb_mbs }, 111 { "MSKanji", "MSKanji", 16, towide_mskanji, tomb_mbs }, 112 { "PCK", "MSKanji", 16, towide_mskanji, tomb_mbs }, 113 { "SJIS", "MSKanji", 16, towide_mskanji, tomb_mbs }, 114 { "Shift_JIS", "MSKanji", 16, towide_mskanji, tomb_mbs }, 115 116 { "BIG5", "BIG5", 16, towide_big5, tomb_mbs }, 117 { "big5", "BIG5", 16, towide_big5, tomb_mbs }, 118 { "Big5", "BIG5", 16, towide_big5, tomb_mbs }, 119 120 { "GBK", "GBK", 16, towide_gbk, tomb_mbs }, 121 122 /* 123 * GB18030 can get away with just 31 bits. This is because the 124 * high order bit is always set for 4 byte values, and the 125 * at least one of the other bits in that 4 byte value will 126 * be non-zero. 127 */ 128 { "GB18030", "GB18030", 31, towide_gb18030, tomb_mbs }, 129 130 /* 131 * This should probably be an aliase for euc-cn, or vice versa. 132 */ 133 { "GB2312", "GB2312", 16, towide_gb2312, tomb_mbs }, 134 135 { NULL, NULL, 0, 0, 0 }, 136 }; 137 138 static char * 139 show_mb(const char *mb) 140 { 141 static char buf[64]; 142 143 /* ASCII stuff we just print */ 144 if (isascii(*mb) && isgraph(*mb)) { 145 buf[0] = *mb; 146 buf[1] = 0; 147 return (buf); 148 } 149 buf[0] = 0; 150 while (*mb != 0) { 151 char scr[8]; 152 (void) snprintf(scr, sizeof (scr), "\\x%02x", *mb); 153 (void) strlcat(buf, scr, sizeof (buf)); 154 mb++; 155 } 156 return (buf); 157 } 158 159 static char *widemsg; 160 161 void 162 werr(const char *fmt, ...) 163 { 164 char *msg; 165 166 va_list va; 167 va_start(va, fmt); 168 (void) vasprintf(&msg, fmt, va); 169 va_end(va); 170 171 free(widemsg); 172 widemsg = msg; 173 } 174 175 /* 176 * This is used for 8-bit encodings. 177 */ 178 int 179 towide_none(wchar_t *c, const char *mb, unsigned n __unused) 180 { 181 if (mb_cur_max != 1) { 182 werr("invalid or unsupported multibyte locale"); 183 return (-1); 184 } 185 *c = (uint8_t)*mb; 186 return (1); 187 } 188 189 int 190 tomb_none(char *mb, wchar_t wc) 191 { 192 if (mb_cur_max != 1) { 193 werr("invalid or unsupported multibyte locale"); 194 return (-1); 195 } 196 *(uint8_t *)mb = (wc & 0xff); 197 mb[1] = 0; 198 return (1); 199 } 200 201 /* 202 * UTF-8 stores wide characters in UTF-32 form. 203 */ 204 int 205 towide_utf8(wchar_t *wc, const char *mb, unsigned n) 206 { 207 wchar_t c; 208 int nb; 209 int lv; /* lowest legal value */ 210 int i; 211 const uint8_t *s = (const uint8_t *)mb; 212 213 c = *s; 214 215 if ((c & 0x80) == 0) { 216 /* 7-bit ASCII */ 217 *wc = c; 218 return (1); 219 } else if ((c & 0xe0) == 0xc0) { 220 /* u80-u7ff - two bytes encoded */ 221 nb = 2; 222 lv = 0x80; 223 c &= ~0xe0; 224 } else if ((c & 0xf0) == 0xe0) { 225 /* u800-uffff - three bytes encoded */ 226 nb = 3; 227 lv = 0x800; 228 c &= ~0xf0; 229 } else if ((c & 0xf8) == 0xf0) { 230 /* u1000-u1fffff - four bytes encoded */ 231 nb = 4; 232 lv = 0x1000; 233 c &= ~0xf8; 234 } else { 235 /* 5 and 6 byte encodings are not legal unicode */ 236 werr("utf8 encoding too large (%s)", show_mb(mb)); 237 return (-1); 238 } 239 if (nb > (int)n) { 240 werr("incomplete utf8 sequence (%s)", show_mb(mb)); 241 return (-1); 242 } 243 244 for (i = 1; i < nb; i++) { 245 if (((s[i]) & 0xc0) != 0x80) { 246 werr("illegal utf8 byte (%x)", s[i]); 247 return (-1); 248 } 249 c <<= 6; 250 c |= (s[i] & 0x3f); 251 } 252 253 if (c < lv) { 254 werr("illegal redundant utf8 encoding (%s)", show_mb(mb)); 255 return (-1); 256 } 257 *wc = c; 258 return (nb); 259 } 260 261 int 262 tomb_utf8(char *mb, wchar_t wc) 263 { 264 uint8_t *s = (uint8_t *)mb; 265 uint8_t msk; 266 int cnt; 267 int i; 268 269 if (wc <= 0x7f) { 270 s[0] = wc & 0x7f; 271 s[1] = 0; 272 return (1); 273 } 274 if (wc <= 0x7ff) { 275 cnt = 2; 276 msk = 0xc0; 277 } else if (wc <= 0xffff) { 278 cnt = 3; 279 msk = 0xe0; 280 } else if (wc <= 0x1fffff) { 281 cnt = 4; 282 msk = 0xf0; 283 } else { 284 werr("illegal uf8 char (%x)", wc); 285 return (-1); 286 } 287 for (i = cnt - 1; i; i--) { 288 s[i] = (wc & 0x3f) | 0x80; 289 wc >>= 6; 290 } 291 s[0] = (msk) | wc; 292 s[cnt] = 0; 293 return (cnt); 294 } 295 296 /* 297 * Several encodings share a simplistic dual byte encoding. In these 298 * forms, they all indicate that a two byte sequence is to be used if 299 * the first byte has its high bit set. They all store this simple 300 * encoding as a 16-bit value, although a great many of the possible 301 * code points are not used in most character sets. This gives a possible 302 * set of just over 32,000 valid code points. 303 * 304 * 0x00 - 0x7f - 1 byte encoding 305 * 0x80 - 0x7fff - illegal 306 * 0x8000 - 0xffff - 2 byte encoding 307 */ 308 309 #pragma GCC diagnostic push 310 #pragma GCC diagnostic ignored "-Wcast-qual" 311 312 static int 313 towide_dbcs(wchar_t *wc, const char *mb, unsigned n) 314 { 315 wchar_t c; 316 317 c = *(uint8_t *)mb; 318 319 if ((c & 0x80) == 0) { 320 /* 7-bit */ 321 *wc = c; 322 return (1); 323 } 324 if (n < 2) { 325 werr("incomplete character sequence (%s)", show_mb(mb)); 326 return (-1); 327 } 328 329 /* Store both bytes as a single 16-bit wide. */ 330 c <<= 8; 331 c |= (uint8_t)(mb[1]); 332 *wc = c; 333 return (2); 334 } 335 336 /* 337 * Most multibyte locales just convert the wide character to the multibyte 338 * form by stripping leading null bytes, and writing the 32-bit quantity 339 * in big-endian order. 340 */ 341 int 342 tomb_mbs(char *mb, wchar_t wc) 343 { 344 uint8_t *s = (uint8_t *)mb; 345 int n = 0, c; 346 347 if ((wc & 0xff000000U) != 0) { 348 n = 4; 349 } else if ((wc & 0x00ff0000U) != 0) { 350 n = 3; 351 } else if ((wc & 0x0000ff00U) != 0) { 352 n = 2; 353 } else { 354 n = 1; 355 } 356 c = n; 357 while (n) { 358 n--; 359 s[n] = wc & 0xff; 360 wc >>= 8; 361 } 362 /* ensure null termination */ 363 s[c] = 0; 364 return (c); 365 } 366 367 368 /* 369 * big5 is a simple dual byte character set. 370 */ 371 int 372 towide_big5(wchar_t *wc, const char *mb, unsigned n) 373 { 374 return (towide_dbcs(wc, mb, n)); 375 } 376 377 /* 378 * GBK encodes wides in the same way that big5 does, the high order 379 * bit of the first byte indicates a double byte character. 380 */ 381 int 382 towide_gbk(wchar_t *wc, const char *mb, unsigned n) 383 { 384 return (towide_dbcs(wc, mb, n)); 385 } 386 387 /* 388 * GB2312 is another DBCS. Its cleaner than others in that the second 389 * byte does not encode ASCII, but it supports characters. 390 */ 391 int 392 towide_gb2312(wchar_t *wc, const char *mb, unsigned n) 393 { 394 return (towide_dbcs(wc, mb, n)); 395 } 396 397 /* 398 * GB18030. This encodes as 8, 16, or 32-bits. 399 * 7-bit values are in 1 byte, 4 byte sequences are used when 400 * the second byte encodes 0x30-39 and all other sequences are 2 bytes. 401 */ 402 int 403 towide_gb18030(wchar_t *wc, const char *mb, unsigned n) 404 { 405 wchar_t c; 406 407 c = *(uint8_t *)mb; 408 409 if ((c & 0x80) == 0) { 410 /* 7-bit */ 411 *wc = c; 412 return (1); 413 } 414 if (n < 2) { 415 werr("incomplete character sequence (%s)", show_mb(mb)); 416 return (-1); 417 } 418 419 /* pull in the second byte */ 420 c <<= 8; 421 c |= (uint8_t)(mb[1]); 422 423 if (((c & 0xff) >= 0x30) && ((c & 0xff) <= 0x39)) { 424 if (n < 4) { 425 werr("incomplete 4-byte character sequence (%s)", 426 show_mb(mb)); 427 return (-1); 428 } 429 c <<= 8; 430 c |= (uint8_t)(mb[2]); 431 c <<= 8; 432 c |= (uint8_t)(mb[3]); 433 *wc = c; 434 return (4); 435 } 436 437 *wc = c; 438 return (2); 439 } 440 441 /* 442 * MS-Kanji (aka SJIS) is almost a clean DBCS like the others, but it 443 * also has a range of single byte characters above 0x80. (0xa1-0xdf). 444 */ 445 int 446 towide_mskanji(wchar_t *wc, const char *mb, unsigned n) 447 { 448 wchar_t c; 449 450 c = *(uint8_t *)mb; 451 452 if ((c < 0x80) || ((c > 0xa0) && (c < 0xe0))) { 453 /* 7-bit */ 454 *wc = c; 455 return (1); 456 } 457 458 if (n < 2) { 459 werr("incomplete character sequence (%s)", show_mb(mb)); 460 return (-1); 461 } 462 463 /* Store both bytes as a single 16-bit wide. */ 464 c <<= 8; 465 c |= (uint8_t)(mb[1]); 466 *wc = c; 467 return (2); 468 } 469 470 /* 471 * EUC forms. EUC encodings are "variable". FreeBSD carries some additional 472 * variable data to encode these, but we're going to treat each as independent 473 * instead. Its the only way we can sensibly move forward. 474 * 475 * Note that the way in which the different EUC forms vary is how wide 476 * CS2 and CS3 are and what the first byte of them is. 477 */ 478 static int 479 towide_euc_impl(wchar_t *wc, const char *mb, unsigned n, 480 uint8_t cs2, uint8_t cs2width, uint8_t cs3, uint8_t cs3width) 481 { 482 int i; 483 int width = 2; 484 wchar_t c; 485 486 c = *(uint8_t *)mb; 487 488 /* 489 * All variations of EUC encode 7-bit ASCII as one byte, and use 490 * additional bytes for more than that. 491 */ 492 if ((c & 0x80) == 0) { 493 /* 7-bit */ 494 *wc = c; 495 return (1); 496 } 497 498 /* 499 * All EUC variants reserve 0xa1-0xff to identify CS1, which 500 * is always two bytes wide. Note that unused CS will be zero, 501 * and that cannot be true because we know that the high order 502 * bit must be set. 503 */ 504 if (c >= 0xa1) { 505 width = 2; 506 } else if (c == cs2) { 507 width = cs2width; 508 } else if (c == cs3) { 509 width = cs3width; 510 } 511 512 if ((int)n < width) { 513 werr("incomplete character sequence (%s)", show_mb(mb)); 514 return (-1); 515 } 516 517 for (i = 1; i < width; i++) { 518 /* pull in the next byte */ 519 c <<= 8; 520 c |= (uint8_t)(mb[i]); 521 } 522 523 *wc = c; 524 return (width); 525 } 526 527 #pragma GCC diagnostic pop 528 529 /* 530 * EUC-CN encodes as follows: 531 * 532 * Code set 0 (ASCII): 0x21-0x7E 533 * Code set 1 (CNS 11643-1992 Plane 1): 0xA1A1-0xFEFE 534 * Code set 2: unused 535 * Code set 3: unused 536 */ 537 int 538 towide_euccn(wchar_t *wc, const char *mb, unsigned n) 539 { 540 return (towide_euc_impl(wc, mb, n, 0x8e, 4, 0, 0)); 541 } 542 543 /* 544 * EUC-JP encodes as follows: 545 * 546 * Code set 0 (ASCII or JIS X 0201-1976 Roman): 0x21-0x7E 547 * Code set 1 (JIS X 0208): 0xA1A1-0xFEFE 548 * Code set 2 (half-width katakana): 0x8EA1-0x8EDF 549 * Code set 3 (JIS X 0212-1990): 0x8FA1A1-0x8FFEFE 550 */ 551 int 552 towide_eucjp(wchar_t *wc, const char *mb, unsigned n) 553 { 554 return (towide_euc_impl(wc, mb, n, 0x8e, 2, 0x8f, 3)); 555 } 556 557 /* 558 * EUC-KR encodes as follows: 559 * 560 * Code set 0 (ASCII or KS C 5636-1993): 0x21-0x7E 561 * Code set 1 (KS C 5601-1992): 0xA1A1-0xFEFE 562 * Code set 2: unused 563 * Code set 3: unused 564 */ 565 int 566 towide_euckr(wchar_t *wc, const char *mb, unsigned n) 567 { 568 return (towide_euc_impl(wc, mb, n, 0, 0, 0, 0)); 569 } 570 571 /* 572 * EUC-TW encodes as follows: 573 * 574 * Code set 0 (ASCII): 0x21-0x7E 575 * Code set 1 (CNS 11643-1992 Plane 1): 0xA1A1-0xFEFE 576 * Code set 2 (CNS 11643-1992 Planes 1-16): 0x8EA1A1A1-0x8EB0FEFE 577 * Code set 3: unused 578 */ 579 int 580 towide_euctw(wchar_t *wc, const char *mb, unsigned n) 581 { 582 return (towide_euc_impl(wc, mb, n, 0x8e, 4, 0, 0)); 583 } 584 585 /* 586 * Public entry points. 587 */ 588 589 int 590 to_wide(wchar_t *wc, const char *mb) 591 { 592 /* this won't fail hard */ 593 return (_towide(wc, mb, strlen(mb))); 594 } 595 596 int 597 to_mb(char *mb, wchar_t wc) 598 { 599 int rv; 600 601 if ((rv = _tomb(mb, wc)) < 0) { 602 errf(widemsg); 603 free(widemsg); 604 widemsg = NULL; 605 } 606 return (rv); 607 } 608 609 char * 610 to_mb_string(const wchar_t *wcs) 611 { 612 char *mbs; 613 char *ptr; 614 int len; 615 616 mbs = malloc((wcslen(wcs) * mb_cur_max) + 1); 617 if (mbs == NULL) { 618 errf("out of memory"); 619 return (NULL); 620 } 621 ptr = mbs; 622 while (*wcs) { 623 if ((len = to_mb(ptr, *wcs)) < 0) { 624 INTERR; 625 free(mbs); 626 return (NULL); 627 } 628 wcs++; 629 ptr += len; 630 } 631 *ptr = 0; 632 return (mbs); 633 } 634 635 void 636 set_wide_encoding(const char *encoding) 637 { 638 int i; 639 640 _towide = towide_none; 641 _tomb = tomb_none; 642 _encoding = "NONE"; 643 _nbits = 8; 644 645 for (i = 0; mb_encodings[i].name; i++) { 646 if (strcasecmp(encoding, mb_encodings[i].name) == 0) { 647 _towide = mb_encodings[i].towide; 648 _tomb = mb_encodings[i].tomb; 649 _encoding = mb_encodings[i].cname; 650 _nbits = mb_encodings[i].nbits; 651 break; 652 } 653 } 654 } 655 656 const char * 657 get_wide_encoding(void) 658 { 659 return (_encoding); 660 } 661 662 int 663 max_wide(void) 664 { 665 return ((int)((1U << _nbits) - 1)); 666 } 667