1 /*- 2 * Copyright (c) 2003-2011 Tim Kientzle 3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include "archive_platform.h" 28 __FBSDID("$FreeBSD$"); 29 30 /* 31 * Basic resizable string support, to simplify manipulating arbitrary-sized 32 * strings while minimizing heap activity. 33 * 34 * In particular, the buffer used by a string object is only grown, it 35 * never shrinks, so you can clear and reuse the same string object 36 * without incurring additional memory allocations. 37 */ 38 39 #ifdef HAVE_ERRNO_H 40 #include <errno.h> 41 #endif 42 #ifdef HAVE_ICONV_H 43 #include <iconv.h> 44 #endif 45 #ifdef HAVE_LANGINFO_H 46 #include <langinfo.h> 47 #endif 48 #ifdef HAVE_LOCALCHARSET_H 49 #include <localcharset.h> 50 #endif 51 #ifdef HAVE_STDLIB_H 52 #include <stdlib.h> 53 #endif 54 #ifdef HAVE_STRING_H 55 #include <string.h> 56 #endif 57 #ifdef HAVE_WCHAR_H 58 #include <wchar.h> 59 #endif 60 #if defined(_WIN32) && !defined(__CYGWIN__) 61 #include <windows.h> 62 #include <locale.h> 63 #endif 64 65 #include "archive_endian.h" 66 #include "archive_private.h" 67 #include "archive_string.h" 68 #include "archive_string_composition.h" 69 70 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy) 71 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t)) 72 #endif 73 74 #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove) 75 #define wmemmove(a,b,i) (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t)) 76 #endif 77 78 struct archive_string_conv { 79 struct archive_string_conv *next; 80 char *from_charset; 81 char *to_charset; 82 unsigned from_cp; 83 unsigned to_cp; 84 /* Set 1 if from_charset and to_charset are the same. */ 85 int same; 86 int flag; 87 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified 88 * charset. */ 89 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from 90 * specified charset. */ 91 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */ 92 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting 93 * MBS. */ 94 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive 95 * 2.x in the wrong assumption. */ 96 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C. 97 * Before UTF-8 characters are actually 98 * processed. */ 99 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D. 100 * Before UTF-8 characters are actually 101 * processed. 102 * Currently this only for MAC OS X. */ 103 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */ 104 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */ 105 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */ 106 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */ 107 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */ 108 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */ 109 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE) 110 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE) 111 112 #if HAVE_ICONV 113 iconv_t cd; 114 iconv_t cd_w;/* Use at archive_mstring on 115 * Windows. */ 116 #endif 117 /* A temporary buffer for normalization. */ 118 struct archive_string utftmp; 119 int (*converter[2])(struct archive_string *, const void *, size_t, 120 struct archive_string_conv *); 121 int nconverter; 122 }; 123 124 #define CP_C_LOCALE 0 /* "C" locale only for this file. */ 125 #define CP_UTF16LE 1200 126 #define CP_UTF16BE 1201 127 128 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF) 129 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF) 130 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF) 131 #define UNICODE_MAX 0x10FFFF 132 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */ 133 /* Set U+FFFD(Replacement character) in UTF-8. */ 134 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd}; 135 136 static struct archive_string_conv *find_sconv_object(struct archive *, 137 const char *, const char *); 138 static void add_sconv_object(struct archive *, struct archive_string_conv *); 139 static struct archive_string_conv *create_sconv_object(const char *, 140 const char *, unsigned, int); 141 static void free_sconv_object(struct archive_string_conv *); 142 static struct archive_string_conv *get_sconv_object(struct archive *, 143 const char *, const char *, int); 144 static unsigned make_codepage_from_charset(const char *); 145 static unsigned get_current_codepage(void); 146 static unsigned get_current_oemcp(void); 147 static size_t mbsnbytes(const void *, size_t); 148 static size_t utf16nbytes(const void *, size_t); 149 #if defined(_WIN32) && !defined(__CYGWIN__) 150 static int archive_wstring_append_from_mbs_in_codepage( 151 struct archive_wstring *, const char *, size_t, 152 struct archive_string_conv *); 153 static int archive_string_append_from_wcs_in_codepage(struct archive_string *, 154 const wchar_t *, size_t, struct archive_string_conv *); 155 static int is_big_endian(void); 156 static int strncat_in_codepage(struct archive_string *, const void *, 157 size_t, struct archive_string_conv *); 158 static int win_strncat_from_utf16be(struct archive_string *, const void *, 159 size_t, struct archive_string_conv *); 160 static int win_strncat_from_utf16le(struct archive_string *, const void *, 161 size_t, struct archive_string_conv *); 162 static int win_strncat_to_utf16be(struct archive_string *, const void *, 163 size_t, struct archive_string_conv *); 164 static int win_strncat_to_utf16le(struct archive_string *, const void *, 165 size_t, struct archive_string_conv *); 166 #endif 167 static int best_effort_strncat_from_utf16be(struct archive_string *, 168 const void *, size_t, struct archive_string_conv *); 169 static int best_effort_strncat_from_utf16le(struct archive_string *, 170 const void *, size_t, struct archive_string_conv *); 171 static int best_effort_strncat_to_utf16be(struct archive_string *, 172 const void *, size_t, struct archive_string_conv *); 173 static int best_effort_strncat_to_utf16le(struct archive_string *, 174 const void *, size_t, struct archive_string_conv *); 175 #if defined(HAVE_ICONV) 176 static int iconv_strncat_in_locale(struct archive_string *, const void *, 177 size_t, struct archive_string_conv *); 178 #endif 179 static int best_effort_strncat_in_locale(struct archive_string *, 180 const void *, size_t, struct archive_string_conv *); 181 static int _utf8_to_unicode(uint32_t *, const char *, size_t); 182 static int utf8_to_unicode(uint32_t *, const char *, size_t); 183 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t); 184 static int cesu8_to_unicode(uint32_t *, const char *, size_t); 185 static size_t unicode_to_utf8(char *, size_t, uint32_t); 186 static int utf16_to_unicode(uint32_t *, const char *, size_t, int); 187 static size_t unicode_to_utf16be(char *, size_t, uint32_t); 188 static size_t unicode_to_utf16le(char *, size_t, uint32_t); 189 static int strncat_from_utf8_libarchive2(struct archive_string *, 190 const void *, size_t, struct archive_string_conv *); 191 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *, 192 size_t, struct archive_string_conv *); 193 static int archive_string_normalize_C(struct archive_string *, const void *, 194 size_t, struct archive_string_conv *); 195 static int archive_string_normalize_D(struct archive_string *, const void *, 196 size_t, struct archive_string_conv *); 197 static int archive_string_append_unicode(struct archive_string *, 198 const void *, size_t, struct archive_string_conv *); 199 200 static struct archive_string * 201 archive_string_append(struct archive_string *as, const char *p, size_t s) 202 { 203 if (archive_string_ensure(as, as->length + s + 1) == NULL) 204 return (NULL); 205 memmove(as->s + as->length, p, s); 206 as->length += s; 207 as->s[as->length] = 0; 208 return (as); 209 } 210 211 static struct archive_wstring * 212 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s) 213 { 214 if (archive_wstring_ensure(as, as->length + s + 1) == NULL) 215 return (NULL); 216 wmemmove(as->s + as->length, p, s); 217 as->length += s; 218 as->s[as->length] = 0; 219 return (as); 220 } 221 222 void 223 archive_string_concat(struct archive_string *dest, struct archive_string *src) 224 { 225 if (archive_string_append(dest, src->s, src->length) == NULL) 226 __archive_errx(1, "Out of memory"); 227 } 228 229 void 230 archive_wstring_concat(struct archive_wstring *dest, 231 struct archive_wstring *src) 232 { 233 if (archive_wstring_append(dest, src->s, src->length) == NULL) 234 __archive_errx(1, "Out of memory"); 235 } 236 237 void 238 archive_string_free(struct archive_string *as) 239 { 240 as->length = 0; 241 as->buffer_length = 0; 242 free(as->s); 243 as->s = NULL; 244 } 245 246 void 247 archive_wstring_free(struct archive_wstring *as) 248 { 249 as->length = 0; 250 as->buffer_length = 0; 251 free(as->s); 252 as->s = NULL; 253 } 254 255 struct archive_wstring * 256 archive_wstring_ensure(struct archive_wstring *as, size_t s) 257 { 258 return (struct archive_wstring *) 259 archive_string_ensure((struct archive_string *)as, 260 s * sizeof(wchar_t)); 261 } 262 263 /* Returns NULL on any allocation failure. */ 264 struct archive_string * 265 archive_string_ensure(struct archive_string *as, size_t s) 266 { 267 char *p; 268 size_t new_length; 269 270 /* If buffer is already big enough, don't reallocate. */ 271 if (as->s && (s <= as->buffer_length)) 272 return (as); 273 274 /* 275 * Growing the buffer at least exponentially ensures that 276 * append operations are always linear in the number of 277 * characters appended. Using a smaller growth rate for 278 * larger buffers reduces memory waste somewhat at the cost of 279 * a larger constant factor. 280 */ 281 if (as->buffer_length < 32) 282 /* Start with a minimum 32-character buffer. */ 283 new_length = 32; 284 else if (as->buffer_length < 8192) 285 /* Buffers under 8k are doubled for speed. */ 286 new_length = as->buffer_length + as->buffer_length; 287 else { 288 /* Buffers 8k and over grow by at least 25% each time. */ 289 new_length = as->buffer_length + as->buffer_length / 4; 290 /* Be safe: If size wraps, fail. */ 291 if (new_length < as->buffer_length) { 292 /* On failure, wipe the string and return NULL. */ 293 archive_string_free(as); 294 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 295 return (NULL); 296 } 297 } 298 /* 299 * The computation above is a lower limit to how much we'll 300 * grow the buffer. In any case, we have to grow it enough to 301 * hold the request. 302 */ 303 if (new_length < s) 304 new_length = s; 305 /* Now we can reallocate the buffer. */ 306 p = (char *)realloc(as->s, new_length); 307 if (p == NULL) { 308 /* On failure, wipe the string and return NULL. */ 309 archive_string_free(as); 310 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 311 return (NULL); 312 } 313 314 as->s = p; 315 as->buffer_length = new_length; 316 return (as); 317 } 318 319 /* 320 * TODO: See if there's a way to avoid scanning 321 * the source string twice. Then test to see 322 * if it actually helps (remember that we're almost 323 * always called with pretty short arguments, so 324 * such an optimization might not help). 325 */ 326 struct archive_string * 327 archive_strncat(struct archive_string *as, const void *_p, size_t n) 328 { 329 size_t s; 330 const char *p, *pp; 331 332 p = (const char *)_p; 333 334 /* Like strlen(p), except won't examine positions beyond p[n]. */ 335 s = 0; 336 pp = p; 337 while (s < n && *pp) { 338 pp++; 339 s++; 340 } 341 if ((as = archive_string_append(as, p, s)) == NULL) 342 __archive_errx(1, "Out of memory"); 343 return (as); 344 } 345 346 struct archive_wstring * 347 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n) 348 { 349 size_t s; 350 const wchar_t *pp; 351 352 /* Like strlen(p), except won't examine positions beyond p[n]. */ 353 s = 0; 354 pp = p; 355 while (s < n && *pp) { 356 pp++; 357 s++; 358 } 359 if ((as = archive_wstring_append(as, p, s)) == NULL) 360 __archive_errx(1, "Out of memory"); 361 return (as); 362 } 363 364 struct archive_string * 365 archive_strcat(struct archive_string *as, const void *p) 366 { 367 /* strcat is just strncat without an effective limit. 368 * Assert that we'll never get called with a source 369 * string over 16MB. 370 * TODO: Review all uses of strcat in the source 371 * and try to replace them with strncat(). 372 */ 373 return archive_strncat(as, p, 0x1000000); 374 } 375 376 struct archive_wstring * 377 archive_wstrcat(struct archive_wstring *as, const wchar_t *p) 378 { 379 /* Ditto. */ 380 return archive_wstrncat(as, p, 0x1000000); 381 } 382 383 struct archive_string * 384 archive_strappend_char(struct archive_string *as, char c) 385 { 386 if ((as = archive_string_append(as, &c, 1)) == NULL) 387 __archive_errx(1, "Out of memory"); 388 return (as); 389 } 390 391 struct archive_wstring * 392 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c) 393 { 394 if ((as = archive_wstring_append(as, &c, 1)) == NULL) 395 __archive_errx(1, "Out of memory"); 396 return (as); 397 } 398 399 /* 400 * Get the "current character set" name to use with iconv. 401 * On FreeBSD, the empty character set name "" chooses 402 * the correct character encoding for the current locale, 403 * so this isn't necessary. 404 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly; 405 * on that system, we have to explicitly call nl_langinfo() 406 * to get the right name. Not sure about other platforms. 407 * 408 * NOTE: GNU libiconv does not recognize the character-set name 409 * which some platform nl_langinfo(CODESET) returns, so we should 410 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv. 411 */ 412 static const char * 413 default_iconv_charset(const char *charset) { 414 if (charset != NULL && charset[0] != '\0') 415 return charset; 416 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__) 417 /* locale_charset() is broken on Mac OS */ 418 return locale_charset(); 419 #elif HAVE_NL_LANGINFO 420 return nl_langinfo(CODESET); 421 #else 422 return ""; 423 #endif 424 } 425 426 #if defined(_WIN32) && !defined(__CYGWIN__) 427 428 /* 429 * Convert MBS to WCS. 430 * Note: returns -1 if conversion fails. 431 */ 432 int 433 archive_wstring_append_from_mbs(struct archive_wstring *dest, 434 const char *p, size_t len) 435 { 436 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL); 437 } 438 439 static int 440 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest, 441 const char *s, size_t length, struct archive_string_conv *sc) 442 { 443 int count, ret = 0; 444 UINT from_cp; 445 446 if (sc != NULL) 447 from_cp = sc->from_cp; 448 else 449 from_cp = get_current_codepage(); 450 451 if (from_cp == CP_C_LOCALE) { 452 /* 453 * "C" locale special process. 454 */ 455 wchar_t *ws; 456 const unsigned char *mp; 457 458 if (NULL == archive_wstring_ensure(dest, 459 dest->length + length + 1)) 460 return (-1); 461 462 ws = dest->s + dest->length; 463 mp = (const unsigned char *)s; 464 count = 0; 465 while (count < (int)length && *mp) { 466 *ws++ = (wchar_t)*mp++; 467 count++; 468 } 469 } else if (sc != NULL && 470 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) { 471 /* 472 * Normalize UTF-8 and UTF-16BE and convert it directly 473 * to UTF-16 as wchar_t. 474 */ 475 struct archive_string u16; 476 int saved_flag = sc->flag;/* save current flag. */ 477 478 if (is_big_endian()) 479 sc->flag |= SCONV_TO_UTF16BE; 480 else 481 sc->flag |= SCONV_TO_UTF16LE; 482 483 if (sc->flag & SCONV_FROM_UTF16) { 484 /* 485 * UTF-16BE/LE NFD ===> UTF-16 NFC 486 * UTF-16BE/LE NFC ===> UTF-16 NFD 487 */ 488 count = (int)utf16nbytes(s, length); 489 } else { 490 /* 491 * UTF-8 NFD ===> UTF-16 NFC 492 * UTF-8 NFC ===> UTF-16 NFD 493 */ 494 count = (int)mbsnbytes(s, length); 495 } 496 u16.s = (char *)dest->s; 497 u16.length = dest->length << 1;; 498 u16.buffer_length = dest->buffer_length; 499 if (sc->flag & SCONV_NORMALIZATION_C) 500 ret = archive_string_normalize_C(&u16, s, count, sc); 501 else 502 ret = archive_string_normalize_D(&u16, s, count, sc); 503 dest->s = (wchar_t *)u16.s; 504 dest->length = u16.length >> 1; 505 dest->buffer_length = u16.buffer_length; 506 sc->flag = saved_flag;/* restore the saved flag. */ 507 return (ret); 508 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) { 509 count = (int)utf16nbytes(s, length); 510 count >>= 1; /* to be WCS length */ 511 /* Allocate memory for WCS. */ 512 if (NULL == archive_wstring_ensure(dest, 513 dest->length + count + 1)) 514 return (-1); 515 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count); 516 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) { 517 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 518 int b; 519 for (b = 0; b < count; b++) { 520 uint16_t val = archive_le16dec(u16+b); 521 archive_be16enc(u16+b, val); 522 } 523 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) { 524 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 525 int b; 526 for (b = 0; b < count; b++) { 527 uint16_t val = archive_be16dec(u16+b); 528 archive_le16enc(u16+b, val); 529 } 530 } 531 } else { 532 DWORD mbflag; 533 size_t buffsize; 534 535 if (sc == NULL) 536 mbflag = 0; 537 else if (sc->flag & SCONV_FROM_CHARSET) { 538 /* Do not trust the length which comes from 539 * an archive file. */ 540 length = mbsnbytes(s, length); 541 mbflag = 0; 542 } else 543 mbflag = MB_PRECOMPOSED; 544 545 buffsize = dest->length + length + 1; 546 do { 547 /* Allocate memory for WCS. */ 548 if (NULL == archive_wstring_ensure(dest, buffsize)) 549 return (-1); 550 /* Convert MBS to WCS. */ 551 count = MultiByteToWideChar(from_cp, 552 mbflag, s, (int)length, dest->s + dest->length, 553 (int)(dest->buffer_length >> 1) -1); 554 if (count == 0 && 555 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 556 /* Expand the WCS buffer. */ 557 buffsize = dest->buffer_length << 1; 558 continue; 559 } 560 if (count == 0 && length != 0) 561 ret = -1; 562 break; 563 } while (1); 564 } 565 dest->length += count; 566 dest->s[dest->length] = L'\0'; 567 return (ret); 568 } 569 570 #else 571 572 /* 573 * Convert MBS to WCS. 574 * Note: returns -1 if conversion fails. 575 */ 576 int 577 archive_wstring_append_from_mbs(struct archive_wstring *dest, 578 const char *p, size_t len) 579 { 580 size_t r; 581 int ret_val = 0; 582 /* 583 * No single byte will be more than one wide character, 584 * so this length estimate will always be big enough. 585 */ 586 size_t wcs_length = len; 587 size_t mbs_length = len; 588 const char *mbs = p; 589 wchar_t *wcs; 590 #if HAVE_MBRTOWC 591 mbstate_t shift_state; 592 593 memset(&shift_state, 0, sizeof(shift_state)); 594 #endif 595 if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) 596 return (-1); 597 wcs = dest->s + dest->length; 598 /* 599 * We cannot use mbsrtowcs/mbstowcs here because those may convert 600 * extra MBS when strlen(p) > len and one wide character consis of 601 * multi bytes. 602 */ 603 while (*mbs && mbs_length > 0) { 604 if (wcs_length == 0) { 605 dest->length = wcs - dest->s; 606 dest->s[dest->length] = L'\0'; 607 wcs_length = mbs_length; 608 if (NULL == archive_wstring_ensure(dest, 609 dest->length + wcs_length + 1)) 610 return (-1); 611 wcs = dest->s + dest->length; 612 } 613 #if HAVE_MBRTOWC 614 r = mbrtowc(wcs, mbs, wcs_length, &shift_state); 615 #else 616 r = mbtowc(wcs, mbs, wcs_length); 617 #endif 618 if (r == (size_t)-1 || r == (size_t)-2) { 619 ret_val = -1; 620 if (errno == EILSEQ) { 621 ++mbs; 622 --mbs_length; 623 continue; 624 } else 625 break; 626 } 627 if (r == 0 || r > mbs_length) 628 break; 629 wcs++; 630 wcs_length--; 631 mbs += r; 632 mbs_length -= r; 633 } 634 dest->length = wcs - dest->s; 635 dest->s[dest->length] = L'\0'; 636 return (ret_val); 637 } 638 639 #endif 640 641 #if defined(_WIN32) && !defined(__CYGWIN__) 642 643 /* 644 * WCS ==> MBS. 645 * Note: returns -1 if conversion fails. 646 * 647 * Win32 builds use WideCharToMultiByte from the Windows API. 648 * (Maybe Cygwin should too? WideCharToMultiByte will know a 649 * lot more about local character encodings than the wcrtomb() 650 * wrapper is going to know.) 651 */ 652 int 653 archive_string_append_from_wcs(struct archive_string *as, 654 const wchar_t *w, size_t len) 655 { 656 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL); 657 } 658 659 static int 660 archive_string_append_from_wcs_in_codepage(struct archive_string *as, 661 const wchar_t *ws, size_t len, struct archive_string_conv *sc) 662 { 663 BOOL defchar_used, *dp; 664 int count, ret = 0; 665 UINT to_cp; 666 int wslen = (int)len; 667 668 if (sc != NULL) 669 to_cp = sc->to_cp; 670 else 671 to_cp = get_current_codepage(); 672 673 if (to_cp == CP_C_LOCALE) { 674 /* 675 * "C" locale special process. 676 */ 677 const wchar_t *wp = ws; 678 char *p; 679 680 if (NULL == archive_string_ensure(as, 681 as->length + wslen +1)) 682 return (-1); 683 p = as->s + as->length; 684 count = 0; 685 defchar_used = 0; 686 while (count < wslen && *wp) { 687 if (*wp > 255) { 688 *p++ = '?'; 689 wp++; 690 defchar_used = 1; 691 } else 692 *p++ = (char)*wp++; 693 count++; 694 } 695 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) { 696 uint16_t *u16; 697 698 if (NULL == 699 archive_string_ensure(as, as->length + len * 2 + 2)) 700 return (-1); 701 u16 = (uint16_t *)(as->s + as->length); 702 count = 0; 703 defchar_used = 0; 704 if (sc->flag & SCONV_TO_UTF16BE) { 705 while (count < (int)len && *ws) { 706 archive_be16enc(u16+count, *ws); 707 ws++; 708 count++; 709 } 710 } else { 711 while (count < (int)len && *ws) { 712 archive_le16enc(u16+count, *ws); 713 ws++; 714 count++; 715 } 716 } 717 count <<= 1; /* to be byte size */ 718 } else { 719 /* Make sure the MBS buffer has plenty to set. */ 720 if (NULL == 721 archive_string_ensure(as, as->length + len * 2 + 1)) 722 return (-1); 723 do { 724 defchar_used = 0; 725 if (to_cp == CP_UTF8 || sc == NULL) 726 dp = NULL; 727 else 728 dp = &defchar_used; 729 count = WideCharToMultiByte(to_cp, 0, ws, wslen, 730 as->s + as->length, (int)as->buffer_length-1, NULL, dp); 731 if (count == 0 && 732 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 733 /* Expand the MBS buffer and retry. */ 734 if (NULL == archive_string_ensure(as, 735 as->buffer_length + len)) 736 return (-1); 737 continue; 738 } 739 if (count == 0) 740 ret = -1; 741 break; 742 } while (1); 743 } 744 as->length += count; 745 as->s[as->length] = '\0'; 746 return (defchar_used?-1:ret); 747 } 748 749 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) 750 751 /* 752 * Translates a wide character string into current locale character set 753 * and appends to the archive_string. Note: returns -1 if conversion 754 * fails. 755 */ 756 int 757 archive_string_append_from_wcs(struct archive_string *as, 758 const wchar_t *w, size_t len) 759 { 760 /* We cannot use the standard wcstombs() here because it 761 * cannot tell us how big the output buffer should be. So 762 * I've built a loop around wcrtomb() or wctomb() that 763 * converts a character at a time and resizes the string as 764 * needed. We prefer wcrtomb() when it's available because 765 * it's thread-safe. */ 766 int n, ret_val = 0; 767 char *p; 768 char *end; 769 #if HAVE_WCRTOMB 770 mbstate_t shift_state; 771 772 memset(&shift_state, 0, sizeof(shift_state)); 773 #else 774 /* Clear the shift state before starting. */ 775 wctomb(NULL, L'\0'); 776 #endif 777 /* 778 * Allocate buffer for MBS. 779 * We need this allocation here since it is possible that 780 * as->s is still NULL. 781 */ 782 if (archive_string_ensure(as, as->length + len + 1) == NULL) 783 return (-1); 784 785 p = as->s + as->length; 786 end = as->s + as->buffer_length - MB_CUR_MAX -1; 787 while (*w != L'\0' && len > 0) { 788 if (p >= end) { 789 as->length = p - as->s; 790 as->s[as->length] = '\0'; 791 /* Re-allocate buffer for MBS. */ 792 if (archive_string_ensure(as, 793 as->length + len * 2 + 1) == NULL) 794 return (-1); 795 p = as->s + as->length; 796 end = as->s + as->buffer_length - MB_CUR_MAX -1; 797 } 798 #if HAVE_WCRTOMB 799 n = wcrtomb(p, *w++, &shift_state); 800 #else 801 n = wctomb(p, *w++); 802 #endif 803 if (n == -1) { 804 if (errno == EILSEQ) { 805 /* Skip an illegal wide char. */ 806 *p++ = '?'; 807 ret_val = -1; 808 } else { 809 ret_val = -1; 810 break; 811 } 812 } else 813 p += n; 814 len--; 815 } 816 as->length = p - as->s; 817 as->s[as->length] = '\0'; 818 return (ret_val); 819 } 820 821 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */ 822 823 /* 824 * TODO: Test if __STDC_ISO_10646__ is defined. 825 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion 826 * one character at a time. If a non-Windows platform doesn't have 827 * either of these, fall back to the built-in UTF8 conversion. 828 */ 829 int 830 archive_string_append_from_wcs(struct archive_string *as, 831 const wchar_t *w, size_t len) 832 { 833 (void)as;/* UNUSED */ 834 (void)w;/* UNUSED */ 835 (void)len;/* UNUSED */ 836 errno = ENOSYS; 837 return (-1); 838 } 839 840 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ 841 842 /* 843 * Find a string conversion object by a pair of 'from' charset name 844 * and 'to' charset name from an archive object. 845 * Return NULL if not found. 846 */ 847 static struct archive_string_conv * 848 find_sconv_object(struct archive *a, const char *fc, const char *tc) 849 { 850 struct archive_string_conv *sc; 851 852 if (a == NULL) 853 return (NULL); 854 855 for (sc = a->sconv; sc != NULL; sc = sc->next) { 856 if (strcmp(sc->from_charset, fc) == 0 && 857 strcmp(sc->to_charset, tc) == 0) 858 break; 859 } 860 return (sc); 861 } 862 863 /* 864 * Register a string object to an archive object. 865 */ 866 static void 867 add_sconv_object(struct archive *a, struct archive_string_conv *sc) 868 { 869 struct archive_string_conv **psc; 870 871 /* Add a new sconv to sconv list. */ 872 psc = &(a->sconv); 873 while (*psc != NULL) 874 psc = &((*psc)->next); 875 *psc = sc; 876 } 877 878 static void 879 add_converter(struct archive_string_conv *sc, int (*converter) 880 (struct archive_string *, const void *, size_t, 881 struct archive_string_conv *)) 882 { 883 if (sc == NULL || sc->nconverter >= 2) 884 __archive_errx(1, "Programing error"); 885 sc->converter[sc->nconverter++] = converter; 886 } 887 888 static void 889 setup_converter(struct archive_string_conv *sc) 890 { 891 892 /* Reset. */ 893 sc->nconverter = 0; 894 895 /* 896 * Perform special sequence for the incorrect UTF-8 filenames 897 * made by libarchive2.x. 898 */ 899 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { 900 add_converter(sc, strncat_from_utf8_libarchive2); 901 return; 902 } 903 904 /* 905 * Convert a string to UTF-16BE/LE. 906 */ 907 if (sc->flag & SCONV_TO_UTF16) { 908 /* 909 * If the current locale is UTF-8, we can translate 910 * a UTF-8 string into a UTF-16BE string. 911 */ 912 if (sc->flag & SCONV_FROM_UTF8) { 913 add_converter(sc, archive_string_append_unicode); 914 return; 915 } 916 917 #if defined(_WIN32) && !defined(__CYGWIN__) 918 if (sc->flag & SCONV_WIN_CP) { 919 if (sc->flag & SCONV_TO_UTF16BE) 920 add_converter(sc, win_strncat_to_utf16be); 921 else 922 add_converter(sc, win_strncat_to_utf16le); 923 return; 924 } 925 #endif 926 927 #if defined(HAVE_ICONV) 928 if (sc->cd != (iconv_t)-1) { 929 add_converter(sc, iconv_strncat_in_locale); 930 return; 931 } 932 #endif 933 934 if (sc->flag & SCONV_BEST_EFFORT) { 935 if (sc->flag & SCONV_TO_UTF16BE) 936 add_converter(sc, 937 best_effort_strncat_to_utf16be); 938 else 939 add_converter(sc, 940 best_effort_strncat_to_utf16le); 941 } else 942 /* Make sure we have no converter. */ 943 sc->nconverter = 0; 944 return; 945 } 946 947 /* 948 * Convert a string from UTF-16BE/LE. 949 */ 950 if (sc->flag & SCONV_FROM_UTF16) { 951 /* 952 * At least we should normalize a UTF-16BE string. 953 */ 954 if (sc->flag & SCONV_NORMALIZATION_D) 955 add_converter(sc,archive_string_normalize_D); 956 else if (sc->flag & SCONV_NORMALIZATION_C) 957 add_converter(sc, archive_string_normalize_C); 958 959 if (sc->flag & SCONV_TO_UTF8) { 960 /* 961 * If the current locale is UTF-8, we can translate 962 * a UTF-16BE/LE string into a UTF-8 string directly. 963 */ 964 if (!(sc->flag & 965 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 966 add_converter(sc, 967 archive_string_append_unicode); 968 return; 969 } 970 971 #if defined(_WIN32) && !defined(__CYGWIN__) 972 if (sc->flag & SCONV_WIN_CP) { 973 if (sc->flag & SCONV_FROM_UTF16BE) 974 add_converter(sc, win_strncat_from_utf16be); 975 else 976 add_converter(sc, win_strncat_from_utf16le); 977 return; 978 } 979 #endif 980 981 #if defined(HAVE_ICONV) 982 if (sc->cd != (iconv_t)-1) { 983 add_converter(sc, iconv_strncat_in_locale); 984 return; 985 } 986 #endif 987 988 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 989 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 990 add_converter(sc, best_effort_strncat_from_utf16be); 991 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 992 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 993 add_converter(sc, best_effort_strncat_from_utf16le); 994 else 995 /* Make sure we have no converter. */ 996 sc->nconverter = 0; 997 return; 998 } 999 1000 if (sc->flag & SCONV_FROM_UTF8) { 1001 /* 1002 * At least we should normalize a UTF-8 string. 1003 */ 1004 if (sc->flag & SCONV_NORMALIZATION_D) 1005 add_converter(sc,archive_string_normalize_D); 1006 else if (sc->flag & SCONV_NORMALIZATION_C) 1007 add_converter(sc, archive_string_normalize_C); 1008 1009 /* 1010 * Copy UTF-8 string with a check of CESU-8. 1011 * Apparently, iconv does not check surrogate pairs in UTF-8 1012 * when both from-charset and to-charset are UTF-8, and then 1013 * we use our UTF-8 copy code. 1014 */ 1015 if (sc->flag & SCONV_TO_UTF8) { 1016 /* 1017 * If the current locale is UTF-8, we can translate 1018 * a UTF-16BE string into a UTF-8 string directly. 1019 */ 1020 if (!(sc->flag & 1021 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 1022 add_converter(sc, strncat_from_utf8_to_utf8); 1023 return; 1024 } 1025 } 1026 1027 #if defined(_WIN32) && !defined(__CYGWIN__) 1028 /* 1029 * On Windows we can use Windows API for a string conversion. 1030 */ 1031 if (sc->flag & SCONV_WIN_CP) { 1032 add_converter(sc, strncat_in_codepage); 1033 return; 1034 } 1035 #endif 1036 1037 #if HAVE_ICONV 1038 if (sc->cd != (iconv_t)-1) { 1039 add_converter(sc, iconv_strncat_in_locale); 1040 /* 1041 * iconv generally does not support UTF-8-MAC and so 1042 * we have to the output of iconv from NFC to NFD if 1043 * need. 1044 */ 1045 if ((sc->flag & SCONV_FROM_CHARSET) && 1046 (sc->flag & SCONV_TO_UTF8)) { 1047 if (sc->flag & SCONV_NORMALIZATION_D) 1048 add_converter(sc, archive_string_normalize_D); 1049 } 1050 return; 1051 } 1052 #endif 1053 1054 /* 1055 * Try conversion in the best effort or no conversion. 1056 */ 1057 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) 1058 add_converter(sc, best_effort_strncat_in_locale); 1059 else 1060 /* Make sure we have no converter. */ 1061 sc->nconverter = 0; 1062 } 1063 1064 /* 1065 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE 1066 * and CP932 which are referenced in create_sconv_object(). 1067 */ 1068 static const char * 1069 canonical_charset_name(const char *charset) 1070 { 1071 char cs[16]; 1072 char *p; 1073 const char *s; 1074 1075 if (charset == NULL || charset[0] == '\0' 1076 || strlen(charset) > 15) 1077 return (charset); 1078 1079 /* Copy name to uppercase. */ 1080 p = cs; 1081 s = charset; 1082 while (*s) { 1083 char c = *s++; 1084 if (c >= 'a' && c <= 'z') 1085 c -= 'a' - 'A'; 1086 *p++ = c; 1087 } 1088 *p++ = '\0'; 1089 1090 if (strcmp(cs, "UTF-8") == 0 || 1091 strcmp(cs, "UTF8") == 0) 1092 return ("UTF-8"); 1093 if (strcmp(cs, "UTF-16BE") == 0 || 1094 strcmp(cs, "UTF16BE") == 0) 1095 return ("UTF-16BE"); 1096 if (strcmp(cs, "UTF-16LE") == 0 || 1097 strcmp(cs, "UTF16LE") == 0) 1098 return ("UTF-16LE"); 1099 if (strcmp(cs, "CP932") == 0) 1100 return ("CP932"); 1101 return (charset); 1102 } 1103 1104 /* 1105 * Create a string conversion object. 1106 */ 1107 static struct archive_string_conv * 1108 create_sconv_object(const char *fc, const char *tc, 1109 unsigned current_codepage, int flag) 1110 { 1111 struct archive_string_conv *sc; 1112 1113 sc = calloc(1, sizeof(*sc)); 1114 if (sc == NULL) 1115 return (NULL); 1116 sc->next = NULL; 1117 sc->from_charset = strdup(fc); 1118 if (sc->from_charset == NULL) { 1119 free(sc); 1120 return (NULL); 1121 } 1122 sc->to_charset = strdup(tc); 1123 if (sc->to_charset == NULL) { 1124 free(sc->from_charset); 1125 free(sc); 1126 return (NULL); 1127 } 1128 archive_string_init(&sc->utftmp); 1129 1130 if (flag & SCONV_TO_CHARSET) { 1131 /* 1132 * Convert characters from the current locale charset to 1133 * a specified charset. 1134 */ 1135 sc->from_cp = current_codepage; 1136 sc->to_cp = make_codepage_from_charset(tc); 1137 #if defined(_WIN32) && !defined(__CYGWIN__) 1138 if (IsValidCodePage(sc->to_cp)) 1139 flag |= SCONV_WIN_CP; 1140 #endif 1141 } else if (flag & SCONV_FROM_CHARSET) { 1142 /* 1143 * Convert characters from a specified charset to 1144 * the current locale charset. 1145 */ 1146 sc->to_cp = current_codepage; 1147 sc->from_cp = make_codepage_from_charset(fc); 1148 #if defined(_WIN32) && !defined(__CYGWIN__) 1149 if (IsValidCodePage(sc->from_cp)) 1150 flag |= SCONV_WIN_CP; 1151 #endif 1152 } 1153 1154 /* 1155 * Check if "from charset" and "to charset" are the same. 1156 */ 1157 if (strcmp(fc, tc) == 0 || 1158 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) 1159 sc->same = 1; 1160 else 1161 sc->same = 0; 1162 1163 /* 1164 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. 1165 */ 1166 if (strcmp(tc, "UTF-8") == 0) 1167 flag |= SCONV_TO_UTF8; 1168 else if (strcmp(tc, "UTF-16BE") == 0) 1169 flag |= SCONV_TO_UTF16BE; 1170 else if (strcmp(tc, "UTF-16LE") == 0) 1171 flag |= SCONV_TO_UTF16LE; 1172 if (strcmp(fc, "UTF-8") == 0) 1173 flag |= SCONV_FROM_UTF8; 1174 else if (strcmp(fc, "UTF-16BE") == 0) 1175 flag |= SCONV_FROM_UTF16BE; 1176 else if (strcmp(fc, "UTF-16LE") == 0) 1177 flag |= SCONV_FROM_UTF16LE; 1178 #if defined(_WIN32) && !defined(__CYGWIN__) 1179 if (sc->to_cp == CP_UTF8) 1180 flag |= SCONV_TO_UTF8; 1181 else if (sc->to_cp == CP_UTF16BE) 1182 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; 1183 else if (sc->to_cp == CP_UTF16LE) 1184 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; 1185 if (sc->from_cp == CP_UTF8) 1186 flag |= SCONV_FROM_UTF8; 1187 else if (sc->from_cp == CP_UTF16BE) 1188 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; 1189 else if (sc->from_cp == CP_UTF16LE) 1190 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; 1191 #endif 1192 1193 /* 1194 * Set a flag for Unicode NFD. Usually iconv cannot correctly 1195 * handle it. So we have to translate NFD characters to NFC ones 1196 * ourselves before iconv handles. Another reason is to prevent 1197 * that the same sight of two filenames, one is NFC and other 1198 * is NFD, would be in its directory. 1199 * On Mac OS X, although its filesystem layer automatically 1200 * convert filenames to NFD, it would be useful for filename 1201 * comparing to find out the same filenames that we normalize 1202 * that to be NFD ourselves. 1203 */ 1204 if ((flag & SCONV_FROM_CHARSET) && 1205 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { 1206 #if defined(__APPLE__) 1207 if (flag & SCONV_TO_UTF8) 1208 flag |= SCONV_NORMALIZATION_D; 1209 else 1210 #endif 1211 flag |= SCONV_NORMALIZATION_C; 1212 } 1213 #if defined(__APPLE__) 1214 /* 1215 * In case writing an archive file, make sure that a filename 1216 * going to be passed to iconv is a Unicode NFC string since 1217 * a filename in HFS Plus filesystem is a Unicode NFD one and 1218 * iconv cannot handle it with "UTF-8" charset. It is simpler 1219 * than a use of "UTF-8-MAC" charset. 1220 */ 1221 if ((flag & SCONV_TO_CHARSET) && 1222 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1223 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1224 flag |= SCONV_NORMALIZATION_C; 1225 /* 1226 * In case reading an archive file. make sure that a filename 1227 * will be passed to users is a Unicode NFD string in order to 1228 * correctly compare the filename with other one which comes 1229 * from HFS Plus filesystem. 1230 */ 1231 if ((flag & SCONV_FROM_CHARSET) && 1232 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1233 (flag & SCONV_TO_UTF8)) 1234 flag |= SCONV_NORMALIZATION_D; 1235 #endif 1236 1237 #if defined(HAVE_ICONV) 1238 sc->cd_w = (iconv_t)-1; 1239 /* 1240 * Create an iconv object. 1241 */ 1242 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && 1243 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || 1244 (flag & SCONV_WIN_CP)) { 1245 /* This case we won't use iconv. */ 1246 sc->cd = (iconv_t)-1; 1247 } else { 1248 sc->cd = iconv_open(tc, fc); 1249 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { 1250 /* 1251 * Unfortunaly, all of iconv implements do support 1252 * "CP932" character-set, so we should use "SJIS" 1253 * instead if iconv_open failed. 1254 */ 1255 if (strcmp(tc, "CP932") == 0) 1256 sc->cd = iconv_open("SJIS", fc); 1257 else if (strcmp(fc, "CP932") == 0) 1258 sc->cd = iconv_open(tc, "SJIS"); 1259 } 1260 #if defined(_WIN32) && !defined(__CYGWIN__) 1261 /* 1262 * archive_mstring on Windows directly convert multi-bytes 1263 * into archive_wstring in order not to depend on locale 1264 * so that you can do a I18N programing. This will be 1265 * used only in archive_mstring_copy_mbs_len_l so far. 1266 */ 1267 if (flag & SCONV_FROM_CHARSET) { 1268 sc->cd_w = iconv_open("UTF-8", fc); 1269 if (sc->cd_w == (iconv_t)-1 && 1270 (sc->flag & SCONV_BEST_EFFORT)) { 1271 if (strcmp(fc, "CP932") == 0) 1272 sc->cd_w = iconv_open("UTF-8", "SJIS"); 1273 } 1274 } 1275 #endif /* _WIN32 && !__CYGWIN__ */ 1276 } 1277 #endif /* HAVE_ICONV */ 1278 1279 sc->flag = flag; 1280 1281 /* 1282 * Set up converters. 1283 */ 1284 setup_converter(sc); 1285 1286 return (sc); 1287 } 1288 1289 /* 1290 * Free a string conversion object. 1291 */ 1292 static void 1293 free_sconv_object(struct archive_string_conv *sc) 1294 { 1295 free(sc->from_charset); 1296 free(sc->to_charset); 1297 archive_string_free(&sc->utftmp); 1298 #if HAVE_ICONV 1299 if (sc->cd != (iconv_t)-1) 1300 iconv_close(sc->cd); 1301 if (sc->cd_w != (iconv_t)-1) 1302 iconv_close(sc->cd_w); 1303 #endif 1304 free(sc); 1305 } 1306 1307 #if defined(_WIN32) && !defined(__CYGWIN__) 1308 static unsigned 1309 my_atoi(const char *p) 1310 { 1311 unsigned cp; 1312 1313 cp = 0; 1314 while (*p) { 1315 if (*p >= '0' && *p <= '9') 1316 cp = cp * 10 + (*p - '0'); 1317 else 1318 return (-1); 1319 p++; 1320 } 1321 return (cp); 1322 } 1323 1324 /* 1325 * Translate Charset name (as used by iconv) into CodePage (as used by Windows) 1326 * Return -1 if failed. 1327 * 1328 * Note: This translation code may be insufficient. 1329 */ 1330 static struct charset { 1331 const char *name; 1332 unsigned cp; 1333 } charsets[] = { 1334 /* MUST BE SORTED! */ 1335 {"ASCII", 1252}, 1336 {"ASMO-708", 708}, 1337 {"BIG5", 950}, 1338 {"CHINESE", 936}, 1339 {"CP367", 1252}, 1340 {"CP819", 1252}, 1341 {"CP1025", 21025}, 1342 {"DOS-720", 720}, 1343 {"DOS-862", 862}, 1344 {"EUC-CN", 51936}, 1345 {"EUC-JP", 51932}, 1346 {"EUC-KR", 949}, 1347 {"EUCCN", 51936}, 1348 {"EUCJP", 51932}, 1349 {"EUCKR", 949}, 1350 {"GB18030", 54936}, 1351 {"GB2312", 936}, 1352 {"HEBREW", 1255}, 1353 {"HZ-GB-2312", 52936}, 1354 {"IBM273", 20273}, 1355 {"IBM277", 20277}, 1356 {"IBM278", 20278}, 1357 {"IBM280", 20280}, 1358 {"IBM284", 20284}, 1359 {"IBM285", 20285}, 1360 {"IBM290", 20290}, 1361 {"IBM297", 20297}, 1362 {"IBM367", 1252}, 1363 {"IBM420", 20420}, 1364 {"IBM423", 20423}, 1365 {"IBM424", 20424}, 1366 {"IBM819", 1252}, 1367 {"IBM871", 20871}, 1368 {"IBM880", 20880}, 1369 {"IBM905", 20905}, 1370 {"IBM924", 20924}, 1371 {"ISO-8859-1", 28591}, 1372 {"ISO-8859-13", 28603}, 1373 {"ISO-8859-15", 28605}, 1374 {"ISO-8859-2", 28592}, 1375 {"ISO-8859-3", 28593}, 1376 {"ISO-8859-4", 28594}, 1377 {"ISO-8859-5", 28595}, 1378 {"ISO-8859-6", 28596}, 1379 {"ISO-8859-7", 28597}, 1380 {"ISO-8859-8", 28598}, 1381 {"ISO-8859-9", 28599}, 1382 {"ISO8859-1", 28591}, 1383 {"ISO8859-13", 28603}, 1384 {"ISO8859-15", 28605}, 1385 {"ISO8859-2", 28592}, 1386 {"ISO8859-3", 28593}, 1387 {"ISO8859-4", 28594}, 1388 {"ISO8859-5", 28595}, 1389 {"ISO8859-6", 28596}, 1390 {"ISO8859-7", 28597}, 1391 {"ISO8859-8", 28598}, 1392 {"ISO8859-9", 28599}, 1393 {"JOHAB", 1361}, 1394 {"KOI8-R", 20866}, 1395 {"KOI8-U", 21866}, 1396 {"KS_C_5601-1987", 949}, 1397 {"LATIN1", 1252}, 1398 {"LATIN2", 28592}, 1399 {"MACINTOSH", 10000}, 1400 {"SHIFT-JIS", 932}, 1401 {"SHIFT_JIS", 932}, 1402 {"SJIS", 932}, 1403 {"US", 1252}, 1404 {"US-ASCII", 1252}, 1405 {"UTF-16", 1200}, 1406 {"UTF-16BE", 1201}, 1407 {"UTF-16LE", 1200}, 1408 {"UTF-8", CP_UTF8}, 1409 {"X-EUROPA", 29001}, 1410 {"X-MAC-ARABIC", 10004}, 1411 {"X-MAC-CE", 10029}, 1412 {"X-MAC-CHINESEIMP", 10008}, 1413 {"X-MAC-CHINESETRAD", 10002}, 1414 {"X-MAC-CROATIAN", 10082}, 1415 {"X-MAC-CYRILLIC", 10007}, 1416 {"X-MAC-GREEK", 10006}, 1417 {"X-MAC-HEBREW", 10005}, 1418 {"X-MAC-ICELANDIC", 10079}, 1419 {"X-MAC-JAPANESE", 10001}, 1420 {"X-MAC-KOREAN", 10003}, 1421 {"X-MAC-ROMANIAN", 10010}, 1422 {"X-MAC-THAI", 10021}, 1423 {"X-MAC-TURKISH", 10081}, 1424 {"X-MAC-UKRAINIAN", 10017}, 1425 }; 1426 static unsigned 1427 make_codepage_from_charset(const char *charset) 1428 { 1429 char cs[16]; 1430 char *p; 1431 unsigned cp; 1432 int a, b; 1433 1434 if (charset == NULL || strlen(charset) > 15) 1435 return -1; 1436 1437 /* Copy name to uppercase. */ 1438 p = cs; 1439 while (*charset) { 1440 char c = *charset++; 1441 if (c >= 'a' && c <= 'z') 1442 c -= 'a' - 'A'; 1443 *p++ = c; 1444 } 1445 *p++ = '\0'; 1446 cp = -1; 1447 1448 /* Look it up in the table first, so that we can easily 1449 * override CP367, which we map to 1252 instead of 367. */ 1450 a = 0; 1451 b = sizeof(charsets)/sizeof(charsets[0]); 1452 while (b > a) { 1453 int c = (b + a) / 2; 1454 int r = strcmp(charsets[c].name, cs); 1455 if (r < 0) 1456 a = c + 1; 1457 else if (r > 0) 1458 b = c; 1459 else 1460 return charsets[c].cp; 1461 } 1462 1463 /* If it's not in the table, try to parse it. */ 1464 switch (*cs) { 1465 case 'C': 1466 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { 1467 cp = my_atoi(cs + 2); 1468 } else if (strcmp(cs, "CP_ACP") == 0) 1469 cp = get_current_codepage(); 1470 else if (strcmp(cs, "CP_OEMCP") == 0) 1471 cp = get_current_oemcp(); 1472 break; 1473 case 'I': 1474 if (cs[1] == 'B' && cs[2] == 'M' && 1475 cs[3] >= '0' && cs[3] <= '9') { 1476 cp = my_atoi(cs + 3); 1477 } 1478 break; 1479 case 'W': 1480 if (strncmp(cs, "WINDOWS-", 8) == 0) { 1481 cp = my_atoi(cs + 8); 1482 if (cp != 874 && (cp < 1250 || cp > 1258)) 1483 cp = -1;/* This may invalid code. */ 1484 } 1485 break; 1486 } 1487 return (cp); 1488 } 1489 1490 /* 1491 * Return ANSI Code Page of current locale set by setlocale(). 1492 */ 1493 static unsigned 1494 get_current_codepage(void) 1495 { 1496 char *locale, *p; 1497 unsigned cp; 1498 1499 locale = setlocale(LC_CTYPE, NULL); 1500 if (locale == NULL) 1501 return (GetACP()); 1502 if (locale[0] == 'C' && locale[1] == '\0') 1503 return (CP_C_LOCALE); 1504 p = strrchr(locale, '.'); 1505 if (p == NULL) 1506 return (GetACP()); 1507 cp = my_atoi(p+1); 1508 if (cp <= 0) 1509 return (GetACP()); 1510 return (cp); 1511 } 1512 1513 /* 1514 * Translation table between Locale Name and ACP/OEMCP. 1515 */ 1516 static struct { 1517 unsigned acp; 1518 unsigned ocp; 1519 const char *locale; 1520 } acp_ocp_map[] = { 1521 { 950, 950, "Chinese_Taiwan" }, 1522 { 936, 936, "Chinese_People's Republic of China" }, 1523 { 950, 950, "Chinese_Taiwan" }, 1524 { 1250, 852, "Czech_Czech Republic" }, 1525 { 1252, 850, "Danish_Denmark" }, 1526 { 1252, 850, "Dutch_Netherlands" }, 1527 { 1252, 850, "Dutch_Belgium" }, 1528 { 1252, 437, "English_United States" }, 1529 { 1252, 850, "English_Australia" }, 1530 { 1252, 850, "English_Canada" }, 1531 { 1252, 850, "English_New Zealand" }, 1532 { 1252, 850, "English_United Kingdom" }, 1533 { 1252, 437, "English_United States" }, 1534 { 1252, 850, "Finnish_Finland" }, 1535 { 1252, 850, "French_France" }, 1536 { 1252, 850, "French_Belgium" }, 1537 { 1252, 850, "French_Canada" }, 1538 { 1252, 850, "French_Switzerland" }, 1539 { 1252, 850, "German_Germany" }, 1540 { 1252, 850, "German_Austria" }, 1541 { 1252, 850, "German_Switzerland" }, 1542 { 1253, 737, "Greek_Greece" }, 1543 { 1250, 852, "Hungarian_Hungary" }, 1544 { 1252, 850, "Icelandic_Iceland" }, 1545 { 1252, 850, "Italian_Italy" }, 1546 { 1252, 850, "Italian_Switzerland" }, 1547 { 932, 932, "Japanese_Japan" }, 1548 { 949, 949, "Korean_Korea" }, 1549 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1550 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1551 { 1252, 850, "Norwegian-Nynorsk_Norway" }, 1552 { 1250, 852, "Polish_Poland" }, 1553 { 1252, 850, "Portuguese_Portugal" }, 1554 { 1252, 850, "Portuguese_Brazil" }, 1555 { 1251, 866, "Russian_Russia" }, 1556 { 1250, 852, "Slovak_Slovakia" }, 1557 { 1252, 850, "Spanish_Spain" }, 1558 { 1252, 850, "Spanish_Mexico" }, 1559 { 1252, 850, "Spanish_Spain" }, 1560 { 1252, 850, "Swedish_Sweden" }, 1561 { 1254, 857, "Turkish_Turkey" }, 1562 { 0, 0, NULL} 1563 }; 1564 1565 /* 1566 * Return OEM Code Page of current locale set by setlocale(). 1567 */ 1568 static unsigned 1569 get_current_oemcp(void) 1570 { 1571 int i; 1572 char *locale, *p; 1573 size_t len; 1574 1575 locale = setlocale(LC_CTYPE, NULL); 1576 if (locale == NULL) 1577 return (GetOEMCP()); 1578 if (locale[0] == 'C' && locale[1] == '\0') 1579 return (CP_C_LOCALE); 1580 1581 p = strrchr(locale, '.'); 1582 if (p == NULL) 1583 return (GetOEMCP()); 1584 len = p - locale; 1585 for (i = 0; acp_ocp_map[i].acp; i++) { 1586 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) 1587 return (acp_ocp_map[i].ocp); 1588 } 1589 return (GetOEMCP()); 1590 } 1591 #else 1592 1593 /* 1594 * POSIX platform does not use CodePage. 1595 */ 1596 1597 static unsigned 1598 get_current_codepage(void) 1599 { 1600 return (-1);/* Unknown */ 1601 } 1602 static unsigned 1603 make_codepage_from_charset(const char *charset) 1604 { 1605 (void)charset; /* UNUSED */ 1606 return (-1);/* Unknown */ 1607 } 1608 static unsigned 1609 get_current_oemcp(void) 1610 { 1611 return (-1);/* Unknown */ 1612 } 1613 1614 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 1615 1616 /* 1617 * Return a string conversion object. 1618 */ 1619 static struct archive_string_conv * 1620 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) 1621 { 1622 struct archive_string_conv *sc; 1623 unsigned current_codepage; 1624 1625 /* Check if we have made the sconv object. */ 1626 sc = find_sconv_object(a, fc, tc); 1627 if (sc != NULL) 1628 return (sc); 1629 1630 if (a == NULL) 1631 current_codepage = get_current_codepage(); 1632 else 1633 current_codepage = a->current_codepage; 1634 1635 sc = create_sconv_object(canonical_charset_name(fc), 1636 canonical_charset_name(tc), current_codepage, flag); 1637 if (sc == NULL) { 1638 if (a != NULL) 1639 archive_set_error(a, ENOMEM, 1640 "Could not allocate memory for " 1641 "a string conversion object"); 1642 return (NULL); 1643 } 1644 1645 /* 1646 * If there is no converter for current string conversion object, 1647 * we cannot handle this conversion. 1648 */ 1649 if (sc->nconverter == 0) { 1650 if (a != NULL) { 1651 #if HAVE_ICONV 1652 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1653 "iconv_open failed : Cannot handle ``%s''", 1654 (flag & SCONV_TO_CHARSET)?tc:fc); 1655 #else 1656 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1657 "A character-set conversion not fully supported " 1658 "on this platform"); 1659 #endif 1660 } 1661 /* Failed; free a sconv object. */ 1662 free_sconv_object(sc); 1663 return (NULL); 1664 } 1665 1666 /* 1667 * Success! 1668 */ 1669 if (a != NULL) 1670 add_sconv_object(a, sc); 1671 return (sc); 1672 } 1673 1674 static const char * 1675 get_current_charset(struct archive *a) 1676 { 1677 const char *cur_charset; 1678 1679 if (a == NULL) 1680 cur_charset = default_iconv_charset(""); 1681 else { 1682 cur_charset = default_iconv_charset(a->current_code); 1683 if (a->current_code == NULL) { 1684 a->current_code = strdup(cur_charset); 1685 a->current_codepage = get_current_codepage(); 1686 a->current_oemcp = get_current_oemcp(); 1687 } 1688 } 1689 return (cur_charset); 1690 } 1691 1692 /* 1693 * Make and Return a string conversion object. 1694 * Return NULL if the platform does not support the specified conversion 1695 * and best_effort is 0. 1696 * If best_effort is set, A string conversion object must be returned 1697 * unless memory allocation for the object fails, but the conversion 1698 * might fail when non-ASCII code is found. 1699 */ 1700 struct archive_string_conv * 1701 archive_string_conversion_to_charset(struct archive *a, const char *charset, 1702 int best_effort) 1703 { 1704 int flag = SCONV_TO_CHARSET; 1705 1706 if (best_effort) 1707 flag |= SCONV_BEST_EFFORT; 1708 return (get_sconv_object(a, get_current_charset(a), charset, flag)); 1709 } 1710 1711 struct archive_string_conv * 1712 archive_string_conversion_from_charset(struct archive *a, const char *charset, 1713 int best_effort) 1714 { 1715 int flag = SCONV_FROM_CHARSET; 1716 1717 if (best_effort) 1718 flag |= SCONV_BEST_EFFORT; 1719 return (get_sconv_object(a, charset, get_current_charset(a), flag)); 1720 } 1721 1722 /* 1723 * archive_string_default_conversion_*_archive() are provided for Windows 1724 * platform because other archiver application use CP_OEMCP for 1725 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames 1726 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP 1727 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). 1728 * So we should make a string conversion between CP_ACP and CP_OEMCP 1729 * for compatibillty. 1730 */ 1731 #if defined(_WIN32) && !defined(__CYGWIN__) 1732 struct archive_string_conv * 1733 archive_string_default_conversion_for_read(struct archive *a) 1734 { 1735 const char *cur_charset = get_current_charset(a); 1736 char oemcp[16]; 1737 1738 /* NOTE: a check of cur_charset is unneeded but we need 1739 * that get_current_charset() has been surely called at 1740 * this time whatever C compiler optimized. */ 1741 if (cur_charset != NULL && 1742 (a->current_codepage == CP_C_LOCALE || 1743 a->current_codepage == a->current_oemcp)) 1744 return (NULL);/* no conversion. */ 1745 1746 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1747 /* Make sure a null termination must be set. */ 1748 oemcp[sizeof(oemcp)-1] = '\0'; 1749 return (get_sconv_object(a, oemcp, cur_charset, 1750 SCONV_FROM_CHARSET)); 1751 } 1752 1753 struct archive_string_conv * 1754 archive_string_default_conversion_for_write(struct archive *a) 1755 { 1756 const char *cur_charset = get_current_charset(a); 1757 char oemcp[16]; 1758 1759 /* NOTE: a check of cur_charset is unneeded but we need 1760 * that get_current_charset() has been surely called at 1761 * this time whatever C compiler optimized. */ 1762 if (cur_charset != NULL && 1763 (a->current_codepage == CP_C_LOCALE || 1764 a->current_codepage == a->current_oemcp)) 1765 return (NULL);/* no conversion. */ 1766 1767 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1768 /* Make sure a null termination must be set. */ 1769 oemcp[sizeof(oemcp)-1] = '\0'; 1770 return (get_sconv_object(a, cur_charset, oemcp, 1771 SCONV_TO_CHARSET)); 1772 } 1773 #else 1774 struct archive_string_conv * 1775 archive_string_default_conversion_for_read(struct archive *a) 1776 { 1777 (void)a; /* UNUSED */ 1778 return (NULL); 1779 } 1780 1781 struct archive_string_conv * 1782 archive_string_default_conversion_for_write(struct archive *a) 1783 { 1784 (void)a; /* UNUSED */ 1785 return (NULL); 1786 } 1787 #endif 1788 1789 /* 1790 * Dispose of all character conversion objects in the archive object. 1791 */ 1792 void 1793 archive_string_conversion_free(struct archive *a) 1794 { 1795 struct archive_string_conv *sc; 1796 struct archive_string_conv *sc_next; 1797 1798 for (sc = a->sconv; sc != NULL; sc = sc_next) { 1799 sc_next = sc->next; 1800 free_sconv_object(sc); 1801 } 1802 a->sconv = NULL; 1803 free(a->current_code); 1804 a->current_code = NULL; 1805 } 1806 1807 /* 1808 * Return a conversion charset name. 1809 */ 1810 const char * 1811 archive_string_conversion_charset_name(struct archive_string_conv *sc) 1812 { 1813 if (sc->flag & SCONV_TO_CHARSET) 1814 return (sc->to_charset); 1815 else 1816 return (sc->from_charset); 1817 } 1818 1819 /* 1820 * Change the behavior of a string conversion. 1821 */ 1822 void 1823 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) 1824 { 1825 switch (opt) { 1826 /* 1827 * A filename in UTF-8 was made with libarchive 2.x in a wrong 1828 * assumption that wchar_t was Unicode. 1829 * This option enables simulating the assumption in order to read 1830 * that filname correctly. 1831 */ 1832 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: 1833 #if (defined(_WIN32) && !defined(__CYGWIN__)) \ 1834 || defined(__STDC_ISO_10646__) || defined(__APPLE__) 1835 /* 1836 * Nothing to do for it since wchar_t on these platforms 1837 * is really Unicode. 1838 */ 1839 (void)sc; /* UNUSED */ 1840 #else 1841 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { 1842 sc->flag |= SCONV_UTF8_LIBARCHIVE_2; 1843 /* Set up string converters. */ 1844 setup_converter(sc); 1845 } 1846 #endif 1847 break; 1848 case SCONV_SET_OPT_NORMALIZATION_C: 1849 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) { 1850 sc->flag |= SCONV_NORMALIZATION_C; 1851 sc->flag &= ~SCONV_NORMALIZATION_D; 1852 /* Set up string converters. */ 1853 setup_converter(sc); 1854 } 1855 break; 1856 case SCONV_SET_OPT_NORMALIZATION_D: 1857 #if defined(HAVE_ICONV) 1858 /* 1859 * If iconv will take the string, do not change the 1860 * setting of the normalization. 1861 */ 1862 if (!(sc->flag & SCONV_WIN_CP) && 1863 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1864 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1865 break; 1866 #endif 1867 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) { 1868 sc->flag |= SCONV_NORMALIZATION_D; 1869 sc->flag &= ~SCONV_NORMALIZATION_C; 1870 /* Set up string converters. */ 1871 setup_converter(sc); 1872 } 1873 break; 1874 default: 1875 break; 1876 } 1877 } 1878 1879 /* 1880 * 1881 * Copy one archive_string to another in locale conversion. 1882 * 1883 * archive_strncat_l(); 1884 * archive_strncpy_l(); 1885 * 1886 */ 1887 1888 static size_t 1889 mbsnbytes(const void *_p, size_t n) 1890 { 1891 size_t s; 1892 const char *p, *pp; 1893 1894 if (_p == NULL) 1895 return (0); 1896 p = (const char *)_p; 1897 1898 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1899 s = 0; 1900 pp = p; 1901 while (s < n && *pp) { 1902 pp++; 1903 s++; 1904 } 1905 return (s); 1906 } 1907 1908 static size_t 1909 utf16nbytes(const void *_p, size_t n) 1910 { 1911 size_t s; 1912 const char *p, *pp; 1913 1914 if (_p == NULL) 1915 return (0); 1916 p = (const char *)_p; 1917 1918 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1919 s = 0; 1920 pp = p; 1921 n >>= 1; 1922 while (s < n && (pp[0] || pp[1])) { 1923 pp += 2; 1924 s++; 1925 } 1926 return (s<<1); 1927 } 1928 1929 int 1930 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n, 1931 struct archive_string_conv *sc) 1932 { 1933 as->length = 0; 1934 return (archive_strncat_l(as, _p, n, sc)); 1935 } 1936 1937 int 1938 archive_strncat_l(struct archive_string *as, const void *_p, size_t n, 1939 struct archive_string_conv *sc) 1940 { 1941 const void *s; 1942 size_t length; 1943 int i, r = 0, r2; 1944 1945 /* We must allocate memory even if there is no data for conversion 1946 * or copy. This simulates archive_string_append behavior. */ 1947 if (_p == NULL || n == 0) { 1948 int tn = 1; 1949 if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) 1950 tn = 2; 1951 if (archive_string_ensure(as, as->length + tn) == NULL) 1952 return (-1); 1953 as->s[as->length] = 0; 1954 if (tn == 2) 1955 as->s[as->length+1] = 0; 1956 return (0); 1957 } 1958 1959 /* 1960 * If sc is NULL, we just make a copy. 1961 */ 1962 if (sc == NULL) { 1963 length = mbsnbytes(_p, n); 1964 if (archive_string_append(as, _p, length) == NULL) 1965 return (-1);/* No memory */ 1966 return (0); 1967 } 1968 1969 if (sc->flag & SCONV_FROM_UTF16) 1970 length = utf16nbytes(_p, n); 1971 else 1972 length = mbsnbytes(_p, n); 1973 s = _p; 1974 i = 0; 1975 if (sc->nconverter > 1) { 1976 sc->utftmp.length = 0; 1977 r2 = sc->converter[0](&(sc->utftmp), s, length, sc); 1978 if (r2 != 0 && errno == ENOMEM) 1979 return (r2); 1980 if (r > r2) 1981 r = r2; 1982 s = sc->utftmp.s; 1983 length = sc->utftmp.length; 1984 ++i; 1985 } 1986 r2 = sc->converter[i](as, s, length, sc); 1987 if (r > r2) 1988 r = r2; 1989 return (r); 1990 } 1991 1992 #if HAVE_ICONV 1993 1994 /* 1995 * Return -1 if conversion failes. 1996 */ 1997 static int 1998 iconv_strncat_in_locale(struct archive_string *as, const void *_p, 1999 size_t length, struct archive_string_conv *sc) 2000 { 2001 ICONV_CONST char *itp; 2002 size_t remaining; 2003 iconv_t cd; 2004 char *outp; 2005 size_t avail, bs; 2006 int return_value = 0; /* success */ 2007 int to_size, from_size; 2008 2009 if (sc->flag & SCONV_TO_UTF16) 2010 to_size = 2; 2011 else 2012 to_size = 1; 2013 if (sc->flag & SCONV_FROM_UTF16) 2014 from_size = 2; 2015 else 2016 from_size = 1; 2017 2018 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) 2019 return (-1); 2020 2021 cd = sc->cd; 2022 itp = (char *)(uintptr_t)_p; 2023 remaining = length; 2024 outp = as->s + as->length; 2025 avail = as->buffer_length - as->length - to_size; 2026 while (remaining >= (size_t)from_size) { 2027 size_t result = iconv(cd, &itp, &remaining, &outp, &avail); 2028 2029 if (result != (size_t)-1) 2030 break; /* Conversion completed. */ 2031 2032 if (errno == EILSEQ || errno == EINVAL) { 2033 /* 2034 * If an output charset is UTF-8 or UTF-16BE/LE, 2035 * unknown character should be U+FFFD 2036 * (replacement character). 2037 */ 2038 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { 2039 size_t rbytes; 2040 if (sc->flag & SCONV_TO_UTF8) 2041 rbytes = sizeof(utf8_replacement_char); 2042 else 2043 rbytes = 2; 2044 2045 if (avail < rbytes) { 2046 as->length = outp - as->s; 2047 bs = as->buffer_length + 2048 (remaining * to_size) + rbytes; 2049 if (NULL == 2050 archive_string_ensure(as, bs)) 2051 return (-1); 2052 outp = as->s + as->length; 2053 avail = as->buffer_length 2054 - as->length - to_size; 2055 } 2056 if (sc->flag & SCONV_TO_UTF8) 2057 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); 2058 else if (sc->flag & SCONV_TO_UTF16BE) 2059 archive_be16enc(outp, UNICODE_R_CHAR); 2060 else 2061 archive_le16enc(outp, UNICODE_R_CHAR); 2062 outp += rbytes; 2063 avail -= rbytes; 2064 } else { 2065 /* Skip the illegal input bytes. */ 2066 *outp++ = '?'; 2067 avail--; 2068 } 2069 itp += from_size; 2070 remaining -= from_size; 2071 return_value = -1; /* failure */ 2072 } else { 2073 /* E2BIG no output buffer, 2074 * Increase an output buffer. */ 2075 as->length = outp - as->s; 2076 bs = as->buffer_length + remaining * 2; 2077 if (NULL == archive_string_ensure(as, bs)) 2078 return (-1); 2079 outp = as->s + as->length; 2080 avail = as->buffer_length - as->length - to_size; 2081 } 2082 } 2083 as->length = outp - as->s; 2084 as->s[as->length] = 0; 2085 if (to_size == 2) 2086 as->s[as->length+1] = 0; 2087 return (return_value); 2088 } 2089 2090 #endif /* HAVE_ICONV */ 2091 2092 2093 #if defined(_WIN32) && !defined(__CYGWIN__) 2094 2095 /* 2096 * Translate a string from a some CodePage to an another CodePage by 2097 * Windows APIs, and copy the result. Return -1 if conversion failes. 2098 */ 2099 static int 2100 strncat_in_codepage(struct archive_string *as, 2101 const void *_p, size_t length, struct archive_string_conv *sc) 2102 { 2103 const char *s = (const char *)_p; 2104 struct archive_wstring aws; 2105 size_t l; 2106 int r, saved_flag; 2107 2108 archive_string_init(&aws); 2109 saved_flag = sc->flag; 2110 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); 2111 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); 2112 sc->flag = saved_flag; 2113 if (r != 0) { 2114 archive_wstring_free(&aws); 2115 if (errno != ENOMEM) 2116 archive_string_append(as, s, length); 2117 return (-1); 2118 } 2119 2120 l = as->length; 2121 r = archive_string_append_from_wcs_in_codepage( 2122 as, aws.s, aws.length, sc); 2123 if (r != 0 && errno != ENOMEM && l == as->length) 2124 archive_string_append(as, s, length); 2125 archive_wstring_free(&aws); 2126 return (r); 2127 } 2128 2129 /* 2130 * Test whether MBS ==> WCS is okay. 2131 */ 2132 static int 2133 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2134 { 2135 const char *p = (const char *)_p; 2136 unsigned codepage; 2137 DWORD mbflag = MB_ERR_INVALID_CHARS; 2138 2139 if (sc->flag & SCONV_FROM_CHARSET) 2140 codepage = sc->to_cp; 2141 else 2142 codepage = sc->from_cp; 2143 2144 if (codepage == CP_C_LOCALE) 2145 return (0); 2146 if (codepage != CP_UTF8) 2147 mbflag |= MB_PRECOMPOSED; 2148 2149 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) 2150 return (-1); /* Invalid */ 2151 return (0); /* Okay */ 2152 } 2153 2154 #else 2155 2156 /* 2157 * Test whether MBS ==> WCS is okay. 2158 */ 2159 static int 2160 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2161 { 2162 const char *p = (const char *)_p; 2163 size_t r; 2164 2165 #if HAVE_MBRTOWC 2166 mbstate_t shift_state; 2167 2168 memset(&shift_state, 0, sizeof(shift_state)); 2169 #else 2170 /* Clear the shift state before starting. */ 2171 mbtowc(NULL, NULL, 0); 2172 #endif 2173 while (n) { 2174 wchar_t wc; 2175 2176 #if HAVE_MBRTOWC 2177 r = mbrtowc(&wc, p, n, &shift_state); 2178 #else 2179 r = mbtowc(&wc, p, n); 2180 #endif 2181 if (r == (size_t)-1 || r == (size_t)-2) 2182 return (-1);/* Invalid. */ 2183 if (r == 0) 2184 break; 2185 p += r; 2186 n -= r; 2187 } 2188 (void)sc; /* UNUSED */ 2189 return (0); /* All Okey. */ 2190 } 2191 2192 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 2193 2194 /* 2195 * Basically returns -1 because we cannot make a conversion of charset 2196 * without iconv but in some cases this would return 0. 2197 * Returns 0 if all copied characters are ASCII. 2198 * Returns 0 if both from-locale and to-locale are the same and those 2199 * can be WCS with no error. 2200 */ 2201 static int 2202 best_effort_strncat_in_locale(struct archive_string *as, const void *_p, 2203 size_t length, struct archive_string_conv *sc) 2204 { 2205 size_t remaining; 2206 const uint8_t *itp; 2207 int return_value = 0; /* success */ 2208 2209 /* 2210 * If both from-locale and to-locale is the same, this makes a copy. 2211 * And then this checks all copied MBS can be WCS if so returns 0. 2212 */ 2213 if (sc->same) { 2214 if (archive_string_append(as, _p, length) == NULL) 2215 return (-1);/* No memory */ 2216 return (invalid_mbs(_p, length, sc)); 2217 } 2218 2219 /* 2220 * If a character is ASCII, this just copies it. If not, this 2221 * assigns '?' charater instead but in UTF-8 locale this assigns 2222 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, 2223 * a Replacement Character in Unicode. 2224 */ 2225 2226 remaining = length; 2227 itp = (const uint8_t *)_p; 2228 while (*itp && remaining > 0) { 2229 if (*itp > 127) { 2230 // Non-ASCII: Substitute with suitable replacement 2231 if (sc->flag & SCONV_TO_UTF8) { 2232 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { 2233 __archive_errx(1, "Out of memory"); 2234 } 2235 } else { 2236 archive_strappend_char(as, '?'); 2237 } 2238 return_value = -1; 2239 } else { 2240 archive_strappend_char(as, *itp); 2241 } 2242 ++itp; 2243 } 2244 return (return_value); 2245 } 2246 2247 2248 /* 2249 * Unicode conversion functions. 2250 * - UTF-8 <===> UTF-8 in removing surrogate pairs. 2251 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. 2252 * - UTF-8 made by libarchive 2.x ===> UTF-8. 2253 * - UTF-16BE <===> UTF-8. 2254 * 2255 */ 2256 2257 /* 2258 * Utility to convert a single UTF-8 sequence. 2259 * 2260 * Usually return used bytes, return used byte in negative value when 2261 * a unicode character is replaced with U+FFFD. 2262 * See also http://unicode.org/review/pr-121.html Public Review Issue #121 2263 * Recommended Practice for Replacement Characters. 2264 */ 2265 static int 2266 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2267 { 2268 static const char utf8_count[256] = { 2269 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 2270 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 2271 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 2272 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 2273 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 2274 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 2275 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 2276 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 2277 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 2278 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 2279 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 2280 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 2281 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2282 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 2283 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 2284 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ 2285 }; 2286 int ch, i; 2287 int cnt; 2288 uint32_t wc; 2289 2290 /* Sanity check. */ 2291 if (n == 0) 2292 return (0); 2293 /* 2294 * Decode 1-4 bytes depending on the value of the first byte. 2295 */ 2296 ch = (unsigned char)*s; 2297 if (ch == 0) 2298 return (0); /* Standard: return 0 for end-of-string. */ 2299 cnt = utf8_count[ch]; 2300 2301 /* Invalid sequence or there are not plenty bytes. */ 2302 if ((int)n < cnt) { 2303 cnt = (int)n; 2304 for (i = 1; i < cnt; i++) { 2305 if ((s[i] & 0xc0) != 0x80) { 2306 cnt = i; 2307 break; 2308 } 2309 } 2310 goto invalid_sequence; 2311 } 2312 2313 /* Make a Unicode code point from a single UTF-8 sequence. */ 2314 switch (cnt) { 2315 case 1: /* 1 byte sequence. */ 2316 *pwc = ch & 0x7f; 2317 return (cnt); 2318 case 2: /* 2 bytes sequence. */ 2319 if ((s[1] & 0xc0) != 0x80) { 2320 cnt = 1; 2321 goto invalid_sequence; 2322 } 2323 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); 2324 return (cnt); 2325 case 3: /* 3 bytes sequence. */ 2326 if ((s[1] & 0xc0) != 0x80) { 2327 cnt = 1; 2328 goto invalid_sequence; 2329 } 2330 if ((s[2] & 0xc0) != 0x80) { 2331 cnt = 2; 2332 goto invalid_sequence; 2333 } 2334 wc = ((ch & 0x0f) << 12) 2335 | ((s[1] & 0x3f) << 6) 2336 | (s[2] & 0x3f); 2337 if (wc < 0x800) 2338 goto invalid_sequence;/* Overlong sequence. */ 2339 break; 2340 case 4: /* 4 bytes sequence. */ 2341 if ((s[1] & 0xc0) != 0x80) { 2342 cnt = 1; 2343 goto invalid_sequence; 2344 } 2345 if ((s[2] & 0xc0) != 0x80) { 2346 cnt = 2; 2347 goto invalid_sequence; 2348 } 2349 if ((s[3] & 0xc0) != 0x80) { 2350 cnt = 3; 2351 goto invalid_sequence; 2352 } 2353 wc = ((ch & 0x07) << 18) 2354 | ((s[1] & 0x3f) << 12) 2355 | ((s[2] & 0x3f) << 6) 2356 | (s[3] & 0x3f); 2357 if (wc < 0x10000) 2358 goto invalid_sequence;/* Overlong sequence. */ 2359 break; 2360 default: /* Others are all invalid sequence. */ 2361 if (ch == 0xc0 || ch == 0xc1) 2362 cnt = 2; 2363 else if (ch >= 0xf5 && ch <= 0xf7) 2364 cnt = 4; 2365 else if (ch >= 0xf8 && ch <= 0xfb) 2366 cnt = 5; 2367 else if (ch == 0xfc || ch == 0xfd) 2368 cnt = 6; 2369 else 2370 cnt = 1; 2371 if ((int)n < cnt) 2372 cnt = (int)n; 2373 for (i = 1; i < cnt; i++) { 2374 if ((s[i] & 0xc0) != 0x80) { 2375 cnt = i; 2376 break; 2377 } 2378 } 2379 goto invalid_sequence; 2380 } 2381 2382 /* The code point larger than 0x10FFFF is not legal 2383 * Unicode values. */ 2384 if (wc > UNICODE_MAX) 2385 goto invalid_sequence; 2386 /* Correctly gets a Unicode, returns used bytes. */ 2387 *pwc = wc; 2388 return (cnt); 2389 invalid_sequence: 2390 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2391 return (cnt * -1); 2392 } 2393 2394 static int 2395 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2396 { 2397 int cnt; 2398 2399 cnt = _utf8_to_unicode(pwc, s, n); 2400 /* Any of Surrogate pair is not legal Unicode values. */ 2401 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) 2402 return (-3); 2403 return (cnt); 2404 } 2405 2406 static inline uint32_t 2407 combine_surrogate_pair(uint32_t uc, uint32_t uc2) 2408 { 2409 uc -= 0xD800; 2410 uc *= 0x400; 2411 uc += uc2 - 0xDC00; 2412 uc += 0x10000; 2413 return (uc); 2414 } 2415 2416 /* 2417 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in 2418 * removing surrogate pairs. 2419 * 2420 * CESU-8: The Compatibility Encoding Scheme for UTF-16. 2421 * 2422 * Usually return used bytes, return used byte in negative value when 2423 * a unicode character is replaced with U+FFFD. 2424 */ 2425 static int 2426 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2427 { 2428 uint32_t wc = 0; 2429 int cnt; 2430 2431 cnt = _utf8_to_unicode(&wc, s, n); 2432 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { 2433 uint32_t wc2 = 0; 2434 if (n - 3 < 3) { 2435 /* Invalid byte sequence. */ 2436 goto invalid_sequence; 2437 } 2438 cnt = _utf8_to_unicode(&wc2, s+3, n-3); 2439 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { 2440 /* Invalid byte sequence. */ 2441 goto invalid_sequence; 2442 } 2443 wc = combine_surrogate_pair(wc, wc2); 2444 cnt = 6; 2445 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { 2446 /* Invalid byte sequence. */ 2447 goto invalid_sequence; 2448 } 2449 *pwc = wc; 2450 return (cnt); 2451 invalid_sequence: 2452 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2453 if (cnt > 0) 2454 cnt *= -1; 2455 return (cnt); 2456 } 2457 2458 /* 2459 * Convert a Unicode code point to a single UTF-8 sequence. 2460 * 2461 * NOTE:This function does not check if the Unicode is legal or not. 2462 * Please you definitely check it before calling this. 2463 */ 2464 static size_t 2465 unicode_to_utf8(char *p, size_t remaining, uint32_t uc) 2466 { 2467 char *_p = p; 2468 2469 /* Invalid Unicode char maps to Replacement character */ 2470 if (uc > UNICODE_MAX) 2471 uc = UNICODE_R_CHAR; 2472 /* Translate code point to UTF8 */ 2473 if (uc <= 0x7f) { 2474 if (remaining == 0) 2475 return (0); 2476 *p++ = (char)uc; 2477 } else if (uc <= 0x7ff) { 2478 if (remaining < 2) 2479 return (0); 2480 *p++ = 0xc0 | ((uc >> 6) & 0x1f); 2481 *p++ = 0x80 | (uc & 0x3f); 2482 } else if (uc <= 0xffff) { 2483 if (remaining < 3) 2484 return (0); 2485 *p++ = 0xe0 | ((uc >> 12) & 0x0f); 2486 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2487 *p++ = 0x80 | (uc & 0x3f); 2488 } else { 2489 if (remaining < 4) 2490 return (0); 2491 *p++ = 0xf0 | ((uc >> 18) & 0x07); 2492 *p++ = 0x80 | ((uc >> 12) & 0x3f); 2493 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2494 *p++ = 0x80 | (uc & 0x3f); 2495 } 2496 return (p - _p); 2497 } 2498 2499 static int 2500 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) 2501 { 2502 return (utf16_to_unicode(pwc, s, n, 1)); 2503 } 2504 2505 static int 2506 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) 2507 { 2508 return (utf16_to_unicode(pwc, s, n, 0)); 2509 } 2510 2511 static int 2512 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) 2513 { 2514 const char *utf16 = s; 2515 unsigned uc; 2516 2517 if (n == 0) 2518 return (0); 2519 if (n == 1) { 2520 /* set the Replacement Character instead. */ 2521 *pwc = UNICODE_R_CHAR; 2522 return (-1); 2523 } 2524 2525 if (be) 2526 uc = archive_be16dec(utf16); 2527 else 2528 uc = archive_le16dec(utf16); 2529 utf16 += 2; 2530 2531 /* If this is a surrogate pair, assemble the full code point.*/ 2532 if (IS_HIGH_SURROGATE_LA(uc)) { 2533 unsigned uc2; 2534 2535 if (n >= 4) { 2536 if (be) 2537 uc2 = archive_be16dec(utf16); 2538 else 2539 uc2 = archive_le16dec(utf16); 2540 } else 2541 uc2 = 0; 2542 if (IS_LOW_SURROGATE_LA(uc2)) { 2543 uc = combine_surrogate_pair(uc, uc2); 2544 utf16 += 2; 2545 } else { 2546 /* Undescribed code point should be U+FFFD 2547 * (replacement character). */ 2548 *pwc = UNICODE_R_CHAR; 2549 return (-2); 2550 } 2551 } 2552 2553 /* 2554 * Surrogate pair values(0xd800 through 0xdfff) are only 2555 * used by UTF-16, so, after above culculation, the code 2556 * must not be surrogate values, and Unicode has no codes 2557 * larger than 0x10ffff. Thus, those are not legal Unicode 2558 * values. 2559 */ 2560 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { 2561 /* Undescribed code point should be U+FFFD 2562 * (replacement character). */ 2563 *pwc = UNICODE_R_CHAR; 2564 return (((int)(utf16 - s)) * -1); 2565 } 2566 *pwc = uc; 2567 return ((int)(utf16 - s)); 2568 } 2569 2570 static size_t 2571 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) 2572 { 2573 char *utf16 = p; 2574 2575 if (uc > 0xffff) { 2576 /* We have a code point that won't fit into a 2577 * wchar_t; convert it to a surrogate pair. */ 2578 if (remaining < 4) 2579 return (0); 2580 uc -= 0x10000; 2581 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2582 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2583 return (4); 2584 } else { 2585 if (remaining < 2) 2586 return (0); 2587 archive_be16enc(utf16, uc); 2588 return (2); 2589 } 2590 } 2591 2592 static size_t 2593 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) 2594 { 2595 char *utf16 = p; 2596 2597 if (uc > 0xffff) { 2598 /* We have a code point that won't fit into a 2599 * wchar_t; convert it to a surrogate pair. */ 2600 if (remaining < 4) 2601 return (0); 2602 uc -= 0x10000; 2603 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2604 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2605 return (4); 2606 } else { 2607 if (remaining < 2) 2608 return (0); 2609 archive_le16enc(utf16, uc); 2610 return (2); 2611 } 2612 } 2613 2614 /* 2615 * Copy UTF-8 string in checking surrogate pair. 2616 * If any surrogate pair are found, it would be canonicalized. 2617 */ 2618 static int 2619 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, 2620 size_t len, struct archive_string_conv *sc) 2621 { 2622 const char *s; 2623 char *p, *endp; 2624 int n, ret = 0; 2625 2626 (void)sc; /* UNUSED */ 2627 2628 if (archive_string_ensure(as, as->length + len + 1) == NULL) 2629 return (-1); 2630 2631 s = (const char *)_p; 2632 p = as->s + as->length; 2633 endp = as->s + as->buffer_length -1; 2634 do { 2635 uint32_t uc; 2636 const char *ss = s; 2637 size_t w; 2638 2639 /* 2640 * Forward byte sequence until a conversion of that is needed. 2641 */ 2642 while ((n = utf8_to_unicode(&uc, s, len)) > 0) { 2643 s += n; 2644 len -= n; 2645 } 2646 if (ss < s) { 2647 if (p + (s - ss) > endp) { 2648 as->length = p - as->s; 2649 if (archive_string_ensure(as, 2650 as->buffer_length + len + 1) == NULL) 2651 return (-1); 2652 p = as->s + as->length; 2653 endp = as->s + as->buffer_length -1; 2654 } 2655 2656 memcpy(p, ss, s - ss); 2657 p += s - ss; 2658 } 2659 2660 /* 2661 * If n is negative, current byte sequence needs a replacement. 2662 */ 2663 if (n < 0) { 2664 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { 2665 /* Current byte sequence may be CESU-8. */ 2666 n = cesu8_to_unicode(&uc, s, len); 2667 } 2668 if (n < 0) { 2669 ret = -1; 2670 n *= -1;/* Use a replaced unicode character. */ 2671 } 2672 2673 /* Rebuild UTF-8 byte sequence. */ 2674 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { 2675 as->length = p - as->s; 2676 if (archive_string_ensure(as, 2677 as->buffer_length + len + 1) == NULL) 2678 return (-1); 2679 p = as->s + as->length; 2680 endp = as->s + as->buffer_length -1; 2681 } 2682 p += w; 2683 s += n; 2684 len -= n; 2685 } 2686 } while (n > 0); 2687 as->length = p - as->s; 2688 as->s[as->length] = '\0'; 2689 return (ret); 2690 } 2691 2692 static int 2693 archive_string_append_unicode(struct archive_string *as, const void *_p, 2694 size_t len, struct archive_string_conv *sc) 2695 { 2696 const char *s; 2697 char *p, *endp; 2698 uint32_t uc; 2699 size_t w; 2700 int n, ret = 0, ts, tm; 2701 int (*parse)(uint32_t *, const char *, size_t); 2702 size_t (*unparse)(char *, size_t, uint32_t); 2703 2704 if (sc->flag & SCONV_TO_UTF16BE) { 2705 unparse = unicode_to_utf16be; 2706 ts = 2; 2707 } else if (sc->flag & SCONV_TO_UTF16LE) { 2708 unparse = unicode_to_utf16le; 2709 ts = 2; 2710 } else if (sc->flag & SCONV_TO_UTF8) { 2711 unparse = unicode_to_utf8; 2712 ts = 1; 2713 } else { 2714 /* 2715 * This case is going to be converted to another 2716 * character-set through iconv. 2717 */ 2718 if (sc->flag & SCONV_FROM_UTF16BE) { 2719 unparse = unicode_to_utf16be; 2720 ts = 2; 2721 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2722 unparse = unicode_to_utf16le; 2723 ts = 2; 2724 } else { 2725 unparse = unicode_to_utf8; 2726 ts = 1; 2727 } 2728 } 2729 2730 if (sc->flag & SCONV_FROM_UTF16BE) { 2731 parse = utf16be_to_unicode; 2732 tm = 1; 2733 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2734 parse = utf16le_to_unicode; 2735 tm = 1; 2736 } else { 2737 parse = cesu8_to_unicode; 2738 tm = ts; 2739 } 2740 2741 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2742 return (-1); 2743 2744 s = (const char *)_p; 2745 p = as->s + as->length; 2746 endp = as->s + as->buffer_length - ts; 2747 while ((n = parse(&uc, s, len)) != 0) { 2748 if (n < 0) { 2749 /* Use a replaced unicode character. */ 2750 n *= -1; 2751 ret = -1; 2752 } 2753 s += n; 2754 len -= n; 2755 while ((w = unparse(p, endp - p, uc)) == 0) { 2756 /* There is not enough output buffer so 2757 * we have to expand it. */ 2758 as->length = p - as->s; 2759 if (archive_string_ensure(as, 2760 as->buffer_length + len * tm + ts) == NULL) 2761 return (-1); 2762 p = as->s + as->length; 2763 endp = as->s + as->buffer_length - ts; 2764 } 2765 p += w; 2766 } 2767 as->length = p - as->s; 2768 as->s[as->length] = '\0'; 2769 if (ts == 2) 2770 as->s[as->length+1] = '\0'; 2771 return (ret); 2772 } 2773 2774 /* 2775 * Following Constants for Hangul compositions this information comes from 2776 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ 2777 */ 2778 #define HC_SBASE 0xAC00 2779 #define HC_LBASE 0x1100 2780 #define HC_VBASE 0x1161 2781 #define HC_TBASE 0x11A7 2782 #define HC_LCOUNT 19 2783 #define HC_VCOUNT 21 2784 #define HC_TCOUNT 28 2785 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) 2786 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) 2787 2788 static uint32_t 2789 get_nfc(uint32_t uc, uint32_t uc2) 2790 { 2791 int t, b; 2792 2793 t = 0; 2794 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; 2795 while (b >= t) { 2796 int m = (t + b) / 2; 2797 if (u_composition_table[m].cp1 < uc) 2798 t = m + 1; 2799 else if (u_composition_table[m].cp1 > uc) 2800 b = m - 1; 2801 else if (u_composition_table[m].cp2 < uc2) 2802 t = m + 1; 2803 else if (u_composition_table[m].cp2 > uc2) 2804 b = m - 1; 2805 else 2806 return (u_composition_table[m].nfc); 2807 } 2808 return (0); 2809 } 2810 2811 #define FDC_MAX 10 /* The maximum number of Following Decomposable 2812 * Characters. */ 2813 2814 /* 2815 * Update first code point. 2816 */ 2817 #define UPDATE_UC(new_uc) do { \ 2818 uc = new_uc; \ 2819 ucptr = NULL; \ 2820 } while (0) 2821 2822 /* 2823 * Replace first code point with second code point. 2824 */ 2825 #define REPLACE_UC_WITH_UC2() do { \ 2826 uc = uc2; \ 2827 ucptr = uc2ptr; \ 2828 n = n2; \ 2829 } while (0) 2830 2831 #define EXPAND_BUFFER() do { \ 2832 as->length = p - as->s; \ 2833 if (archive_string_ensure(as, \ 2834 as->buffer_length + len * tm + ts) == NULL)\ 2835 return (-1); \ 2836 p = as->s + as->length; \ 2837 endp = as->s + as->buffer_length - ts; \ 2838 } while (0) 2839 2840 #define UNPARSE(p, endp, uc) do { \ 2841 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ 2842 EXPAND_BUFFER(); \ 2843 } \ 2844 p += w; \ 2845 } while (0) 2846 2847 /* 2848 * Write first code point. 2849 * If the code point has not be changed from its original code, 2850 * this just copies it from its original buffer pointer. 2851 * If not, this converts it to UTF-8 byte sequence and copies it. 2852 */ 2853 #define WRITE_UC() do { \ 2854 if (ucptr) { \ 2855 if (p + n > endp) \ 2856 EXPAND_BUFFER(); \ 2857 switch (n) { \ 2858 case 4: \ 2859 *p++ = *ucptr++; \ 2860 /* FALL THROUGH */ \ 2861 case 3: \ 2862 *p++ = *ucptr++; \ 2863 /* FALL THROUGH */ \ 2864 case 2: \ 2865 *p++ = *ucptr++; \ 2866 /* FALL THROUGH */ \ 2867 case 1: \ 2868 *p++ = *ucptr; \ 2869 break; \ 2870 } \ 2871 ucptr = NULL; \ 2872 } else { \ 2873 UNPARSE(p, endp, uc); \ 2874 } \ 2875 } while (0) 2876 2877 /* 2878 * Collect following decomposable code points. 2879 */ 2880 #define COLLECT_CPS(start) do { \ 2881 int _i; \ 2882 for (_i = start; _i < FDC_MAX ; _i++) { \ 2883 nx = parse(&ucx[_i], s, len); \ 2884 if (nx <= 0) \ 2885 break; \ 2886 cx = CCC(ucx[_i]); \ 2887 if (cl >= cx && cl != 228 && cx != 228)\ 2888 break; \ 2889 s += nx; \ 2890 len -= nx; \ 2891 cl = cx; \ 2892 ccx[_i] = cx; \ 2893 } \ 2894 if (_i >= FDC_MAX) { \ 2895 ret = -1; \ 2896 ucx_size = FDC_MAX; \ 2897 } else \ 2898 ucx_size = _i; \ 2899 } while (0) 2900 2901 /* 2902 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. 2903 * 2904 * TODO: Convert composition exclusions,which are never converted 2905 * from NFC,NFD,NFKC and NFKD, to Form C. 2906 */ 2907 static int 2908 archive_string_normalize_C(struct archive_string *as, const void *_p, 2909 size_t len, struct archive_string_conv *sc) 2910 { 2911 const char *s = (const char *)_p; 2912 char *p, *endp; 2913 uint32_t uc, uc2; 2914 size_t w; 2915 int always_replace, n, n2, ret = 0, spair, ts, tm; 2916 int (*parse)(uint32_t *, const char *, size_t); 2917 size_t (*unparse)(char *, size_t, uint32_t); 2918 2919 always_replace = 1; 2920 ts = 1;/* text size. */ 2921 if (sc->flag & SCONV_TO_UTF16BE) { 2922 unparse = unicode_to_utf16be; 2923 ts = 2; 2924 if (sc->flag & SCONV_FROM_UTF16BE) 2925 always_replace = 0; 2926 } else if (sc->flag & SCONV_TO_UTF16LE) { 2927 unparse = unicode_to_utf16le; 2928 ts = 2; 2929 if (sc->flag & SCONV_FROM_UTF16LE) 2930 always_replace = 0; 2931 } else if (sc->flag & SCONV_TO_UTF8) { 2932 unparse = unicode_to_utf8; 2933 if (sc->flag & SCONV_FROM_UTF8) 2934 always_replace = 0; 2935 } else { 2936 /* 2937 * This case is going to be converted to another 2938 * character-set through iconv. 2939 */ 2940 always_replace = 0; 2941 if (sc->flag & SCONV_FROM_UTF16BE) { 2942 unparse = unicode_to_utf16be; 2943 ts = 2; 2944 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2945 unparse = unicode_to_utf16le; 2946 ts = 2; 2947 } else { 2948 unparse = unicode_to_utf8; 2949 } 2950 } 2951 2952 if (sc->flag & SCONV_FROM_UTF16BE) { 2953 parse = utf16be_to_unicode; 2954 tm = 1; 2955 spair = 4;/* surrogate pair size in UTF-16. */ 2956 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2957 parse = utf16le_to_unicode; 2958 tm = 1; 2959 spair = 4;/* surrogate pair size in UTF-16. */ 2960 } else { 2961 parse = cesu8_to_unicode; 2962 tm = ts; 2963 spair = 6;/* surrogate pair size in UTF-8. */ 2964 } 2965 2966 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2967 return (-1); 2968 2969 p = as->s + as->length; 2970 endp = as->s + as->buffer_length - ts; 2971 while ((n = parse(&uc, s, len)) != 0) { 2972 const char *ucptr, *uc2ptr; 2973 2974 if (n < 0) { 2975 /* Use a replaced unicode character. */ 2976 UNPARSE(p, endp, uc); 2977 s += n*-1; 2978 len -= n*-1; 2979 ret = -1; 2980 continue; 2981 } else if (n == spair || always_replace) 2982 /* uc is converted from a surrogate pair. 2983 * this should be treated as a changed code. */ 2984 ucptr = NULL; 2985 else 2986 ucptr = s; 2987 s += n; 2988 len -= n; 2989 2990 /* Read second code point. */ 2991 while ((n2 = parse(&uc2, s, len)) > 0) { 2992 uint32_t ucx[FDC_MAX]; 2993 int ccx[FDC_MAX]; 2994 int cl, cx, i, nx, ucx_size; 2995 int LIndex,SIndex; 2996 uint32_t nfc; 2997 2998 if (n2 == spair || always_replace) 2999 /* uc2 is converted from a surrogate pair. 3000 * this should be treated as a changed code. */ 3001 uc2ptr = NULL; 3002 else 3003 uc2ptr = s; 3004 s += n2; 3005 len -= n2; 3006 3007 /* 3008 * If current second code point is out of decomposable 3009 * code points, finding compositions is unneeded. 3010 */ 3011 if (!IS_DECOMPOSABLE_BLOCK(uc2)) { 3012 WRITE_UC(); 3013 REPLACE_UC_WITH_UC2(); 3014 continue; 3015 } 3016 3017 /* 3018 * Try to combine current code points. 3019 */ 3020 /* 3021 * We have to combine Hangul characters according to 3022 * http://uniicode.org/reports/tr15/#Hangul 3023 */ 3024 if (0 <= (LIndex = uc - HC_LBASE) && 3025 LIndex < HC_LCOUNT) { 3026 /* 3027 * Hangul Composition. 3028 * 1. Two current code points are L and V. 3029 */ 3030 int VIndex = uc2 - HC_VBASE; 3031 if (0 <= VIndex && VIndex < HC_VCOUNT) { 3032 /* Make syllable of form LV. */ 3033 UPDATE_UC(HC_SBASE + 3034 (LIndex * HC_VCOUNT + VIndex) * 3035 HC_TCOUNT); 3036 } else { 3037 WRITE_UC(); 3038 REPLACE_UC_WITH_UC2(); 3039 } 3040 continue; 3041 } else if (0 <= (SIndex = uc - HC_SBASE) && 3042 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { 3043 /* 3044 * Hangul Composition. 3045 * 2. Two current code points are LV and T. 3046 */ 3047 int TIndex = uc2 - HC_TBASE; 3048 if (0 < TIndex && TIndex < HC_TCOUNT) { 3049 /* Make syllable of form LVT. */ 3050 UPDATE_UC(uc + TIndex); 3051 } else { 3052 WRITE_UC(); 3053 REPLACE_UC_WITH_UC2(); 3054 } 3055 continue; 3056 } else if ((nfc = get_nfc(uc, uc2)) != 0) { 3057 /* A composition to current code points 3058 * is found. */ 3059 UPDATE_UC(nfc); 3060 continue; 3061 } else if ((cl = CCC(uc2)) == 0) { 3062 /* Clearly 'uc2' the second code point is not 3063 * a decomposable code. */ 3064 WRITE_UC(); 3065 REPLACE_UC_WITH_UC2(); 3066 continue; 3067 } 3068 3069 /* 3070 * Collect following decomposable code points. 3071 */ 3072 cx = 0; 3073 ucx[0] = uc2; 3074 ccx[0] = cl; 3075 COLLECT_CPS(1); 3076 3077 /* 3078 * Find a composed code in the collected code points. 3079 */ 3080 i = 1; 3081 while (i < ucx_size) { 3082 int j; 3083 3084 if ((nfc = get_nfc(uc, ucx[i])) == 0) { 3085 i++; 3086 continue; 3087 } 3088 3089 /* 3090 * nfc is composed of uc and ucx[i]. 3091 */ 3092 UPDATE_UC(nfc); 3093 3094 /* 3095 * Remove ucx[i] by shifting 3096 * following code points. 3097 */ 3098 for (j = i; j+1 < ucx_size; j++) { 3099 ucx[j] = ucx[j+1]; 3100 ccx[j] = ccx[j+1]; 3101 } 3102 ucx_size --; 3103 3104 /* 3105 * Collect following code points blocked 3106 * by ucx[i] the removed code point. 3107 */ 3108 if (ucx_size > 0 && i == ucx_size && 3109 nx > 0 && cx == cl) { 3110 cl = ccx[ucx_size-1]; 3111 COLLECT_CPS(ucx_size); 3112 } 3113 /* 3114 * Restart finding a composed code with 3115 * the updated uc from the top of the 3116 * collected code points. 3117 */ 3118 i = 0; 3119 } 3120 3121 /* 3122 * Apparently the current code points are not 3123 * decomposed characters or already composed. 3124 */ 3125 WRITE_UC(); 3126 for (i = 0; i < ucx_size; i++) 3127 UNPARSE(p, endp, ucx[i]); 3128 3129 /* 3130 * Flush out remaining canonical combining characters. 3131 */ 3132 if (nx > 0 && cx == cl && len > 0) { 3133 while ((nx = parse(&ucx[0], s, len)) 3134 > 0) { 3135 cx = CCC(ucx[0]); 3136 if (cl > cx) 3137 break; 3138 s += nx; 3139 len -= nx; 3140 cl = cx; 3141 UNPARSE(p, endp, ucx[0]); 3142 } 3143 } 3144 break; 3145 } 3146 if (n2 < 0) { 3147 WRITE_UC(); 3148 /* Use a replaced unicode character. */ 3149 UNPARSE(p, endp, uc2); 3150 s += n2*-1; 3151 len -= n2*-1; 3152 ret = -1; 3153 continue; 3154 } else if (n2 == 0) { 3155 WRITE_UC(); 3156 break; 3157 } 3158 } 3159 as->length = p - as->s; 3160 as->s[as->length] = '\0'; 3161 if (ts == 2) 3162 as->s[as->length+1] = '\0'; 3163 return (ret); 3164 } 3165 3166 static int 3167 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) 3168 { 3169 int t, b; 3170 3171 /* 3172 * These are not converted to NFD on Mac OS. 3173 */ 3174 if ((uc >= 0x2000 && uc <= 0x2FFF) || 3175 (uc >= 0xF900 && uc <= 0xFAFF) || 3176 (uc >= 0x2F800 && uc <= 0x2FAFF)) 3177 return (0); 3178 /* 3179 * Those code points are not converted to NFD on Mac OS. 3180 * I do not know the reason because it is undocumented. 3181 * NFC NFD 3182 * 1109A ==> 11099 110BA 3183 * 1109C ==> 1109B 110BA 3184 * 110AB ==> 110A5 110BA 3185 */ 3186 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) 3187 return (0); 3188 3189 t = 0; 3190 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; 3191 while (b >= t) { 3192 int m = (t + b) / 2; 3193 if (u_decomposition_table[m].nfc < uc) 3194 t = m + 1; 3195 else if (u_decomposition_table[m].nfc > uc) 3196 b = m - 1; 3197 else { 3198 *cp1 = u_decomposition_table[m].cp1; 3199 *cp2 = u_decomposition_table[m].cp2; 3200 return (1); 3201 } 3202 } 3203 return (0); 3204 } 3205 3206 #define REPLACE_UC_WITH(cp) do { \ 3207 uc = cp; \ 3208 ucptr = NULL; \ 3209 } while (0) 3210 3211 /* 3212 * Normalize UTF-8 characters to Form D and copy the result. 3213 */ 3214 static int 3215 archive_string_normalize_D(struct archive_string *as, const void *_p, 3216 size_t len, struct archive_string_conv *sc) 3217 { 3218 const char *s = (const char *)_p; 3219 char *p, *endp; 3220 uint32_t uc, uc2; 3221 size_t w; 3222 int always_replace, n, n2, ret = 0, spair, ts, tm; 3223 int (*parse)(uint32_t *, const char *, size_t); 3224 size_t (*unparse)(char *, size_t, uint32_t); 3225 3226 always_replace = 1; 3227 ts = 1;/* text size. */ 3228 if (sc->flag & SCONV_TO_UTF16BE) { 3229 unparse = unicode_to_utf16be; 3230 ts = 2; 3231 if (sc->flag & SCONV_FROM_UTF16BE) 3232 always_replace = 0; 3233 } else if (sc->flag & SCONV_TO_UTF16LE) { 3234 unparse = unicode_to_utf16le; 3235 ts = 2; 3236 if (sc->flag & SCONV_FROM_UTF16LE) 3237 always_replace = 0; 3238 } else if (sc->flag & SCONV_TO_UTF8) { 3239 unparse = unicode_to_utf8; 3240 if (sc->flag & SCONV_FROM_UTF8) 3241 always_replace = 0; 3242 } else { 3243 /* 3244 * This case is going to be converted to another 3245 * character-set through iconv. 3246 */ 3247 always_replace = 0; 3248 if (sc->flag & SCONV_FROM_UTF16BE) { 3249 unparse = unicode_to_utf16be; 3250 ts = 2; 3251 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3252 unparse = unicode_to_utf16le; 3253 ts = 2; 3254 } else { 3255 unparse = unicode_to_utf8; 3256 } 3257 } 3258 3259 if (sc->flag & SCONV_FROM_UTF16BE) { 3260 parse = utf16be_to_unicode; 3261 tm = 1; 3262 spair = 4;/* surrogate pair size in UTF-16. */ 3263 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3264 parse = utf16le_to_unicode; 3265 tm = 1; 3266 spair = 4;/* surrogate pair size in UTF-16. */ 3267 } else { 3268 parse = cesu8_to_unicode; 3269 tm = ts; 3270 spair = 6;/* surrogate pair size in UTF-8. */ 3271 } 3272 3273 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 3274 return (-1); 3275 3276 p = as->s + as->length; 3277 endp = as->s + as->buffer_length - ts; 3278 while ((n = parse(&uc, s, len)) != 0) { 3279 const char *ucptr; 3280 uint32_t cp1, cp2; 3281 int SIndex; 3282 struct { 3283 uint32_t uc; 3284 int ccc; 3285 } fdc[FDC_MAX]; 3286 int fdi, fdj; 3287 int ccc; 3288 3289 check_first_code: 3290 if (n < 0) { 3291 /* Use a replaced unicode character. */ 3292 UNPARSE(p, endp, uc); 3293 s += n*-1; 3294 len -= n*-1; 3295 ret = -1; 3296 continue; 3297 } else if (n == spair || always_replace) 3298 /* uc is converted from a surrogate pair. 3299 * this should be treated as a changed code. */ 3300 ucptr = NULL; 3301 else 3302 ucptr = s; 3303 s += n; 3304 len -= n; 3305 3306 /* Hangul Decomposition. */ 3307 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { 3308 int L = HC_LBASE + SIndex / HC_NCOUNT; 3309 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; 3310 int T = HC_TBASE + SIndex % HC_TCOUNT; 3311 3312 REPLACE_UC_WITH(L); 3313 WRITE_UC(); 3314 REPLACE_UC_WITH(V); 3315 WRITE_UC(); 3316 if (T != HC_TBASE) { 3317 REPLACE_UC_WITH(T); 3318 WRITE_UC(); 3319 } 3320 continue; 3321 } 3322 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { 3323 WRITE_UC(); 3324 continue; 3325 } 3326 3327 fdi = 0; 3328 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { 3329 int k; 3330 3331 for (k = fdi; k > 0; k--) 3332 fdc[k] = fdc[k-1]; 3333 fdc[0].ccc = CCC(cp2); 3334 fdc[0].uc = cp2; 3335 fdi++; 3336 REPLACE_UC_WITH(cp1); 3337 } 3338 3339 /* Read following code points. */ 3340 while ((n2 = parse(&uc2, s, len)) > 0 && 3341 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { 3342 int j, k; 3343 3344 s += n2; 3345 len -= n2; 3346 for (j = 0; j < fdi; j++) { 3347 if (fdc[j].ccc > ccc) 3348 break; 3349 } 3350 if (j < fdi) { 3351 for (k = fdi; k > j; k--) 3352 fdc[k] = fdc[k-1]; 3353 fdc[j].ccc = ccc; 3354 fdc[j].uc = uc2; 3355 } else { 3356 fdc[fdi].ccc = ccc; 3357 fdc[fdi].uc = uc2; 3358 } 3359 fdi++; 3360 } 3361 3362 WRITE_UC(); 3363 for (fdj = 0; fdj < fdi; fdj++) { 3364 REPLACE_UC_WITH(fdc[fdj].uc); 3365 WRITE_UC(); 3366 } 3367 3368 if (n2 == 0) 3369 break; 3370 REPLACE_UC_WITH(uc2); 3371 n = n2; 3372 goto check_first_code; 3373 } 3374 as->length = p - as->s; 3375 as->s[as->length] = '\0'; 3376 if (ts == 2) 3377 as->s[as->length+1] = '\0'; 3378 return (ret); 3379 } 3380 3381 /* 3382 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption 3383 * that WCS is Unicode. It is true for several platforms but some are false. 3384 * And then people who did not use UTF-8 locale on the non Unicode WCS 3385 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those 3386 * now cannot get right filename from libarchive 3.x and later since we 3387 * fixed the wrong assumption and it is incompatible to older its versions. 3388 * So we provide special option, "compat-2x.x", for resolving it. 3389 * That option enable the string conversion of libarchive 2.x. 3390 * 3391 * Translates the wrong UTF-8 string made by libarchive 2.x into current 3392 * locale character set and appends to the archive_string. 3393 * Note: returns -1 if conversion fails. 3394 */ 3395 static int 3396 strncat_from_utf8_libarchive2(struct archive_string *as, 3397 const void *_p, size_t len, struct archive_string_conv *sc) 3398 { 3399 const char *s; 3400 int n; 3401 char *p; 3402 char *end; 3403 uint32_t unicode; 3404 #if HAVE_WCRTOMB 3405 mbstate_t shift_state; 3406 3407 memset(&shift_state, 0, sizeof(shift_state)); 3408 #else 3409 /* Clear the shift state before starting. */ 3410 wctomb(NULL, L'\0'); 3411 #endif 3412 (void)sc; /* UNUSED */ 3413 /* 3414 * Allocate buffer for MBS. 3415 * We need this allocation here since it is possible that 3416 * as->s is still NULL. 3417 */ 3418 if (archive_string_ensure(as, as->length + len + 1) == NULL) 3419 return (-1); 3420 3421 s = (const char *)_p; 3422 p = as->s + as->length; 3423 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3424 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { 3425 wchar_t wc; 3426 3427 if (p >= end) { 3428 as->length = p - as->s; 3429 /* Re-allocate buffer for MBS. */ 3430 if (archive_string_ensure(as, 3431 as->length + len * 2 + 1) == NULL) 3432 return (-1); 3433 p = as->s + as->length; 3434 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3435 } 3436 3437 /* 3438 * As libarchie 2.x, translates the UTF-8 characters into 3439 * wide-characters in the assumption that WCS is Unicode. 3440 */ 3441 if (n < 0) { 3442 n *= -1; 3443 wc = L'?'; 3444 } else 3445 wc = (wchar_t)unicode; 3446 3447 s += n; 3448 len -= n; 3449 /* 3450 * Translates the wide-character into the current locale MBS. 3451 */ 3452 #if HAVE_WCRTOMB 3453 n = (int)wcrtomb(p, wc, &shift_state); 3454 #else 3455 n = (int)wctomb(p, wc); 3456 #endif 3457 if (n == -1) 3458 return (-1); 3459 p += n; 3460 } 3461 as->length = p - as->s; 3462 as->s[as->length] = '\0'; 3463 return (0); 3464 } 3465 3466 3467 /* 3468 * Conversion functions between current locale dependent MBS and UTF-16BE. 3469 * strncat_from_utf16be() : UTF-16BE --> MBS 3470 * strncat_to_utf16be() : MBS --> UTF16BE 3471 */ 3472 3473 #if defined(_WIN32) && !defined(__CYGWIN__) 3474 3475 /* 3476 * Convert a UTF-16BE/LE string to current locale and copy the result. 3477 * Return -1 if conversion failes. 3478 */ 3479 static int 3480 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, 3481 struct archive_string_conv *sc, int be) 3482 { 3483 struct archive_string tmp; 3484 const char *u16; 3485 int ll; 3486 BOOL defchar; 3487 char *mbs; 3488 size_t mbs_size, b; 3489 int ret = 0; 3490 3491 bytes &= ~1; 3492 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3493 return (-1); 3494 3495 mbs = as->s + as->length; 3496 mbs_size = as->buffer_length - as->length -1; 3497 3498 if (sc->to_cp == CP_C_LOCALE) { 3499 /* 3500 * "C" locale special process. 3501 */ 3502 u16 = _p; 3503 ll = 0; 3504 for (b = 0; b < bytes; b += 2) { 3505 uint16_t val; 3506 if (be) 3507 val = archive_be16dec(u16+b); 3508 else 3509 val = archive_le16dec(u16+b); 3510 if (val > 255) { 3511 *mbs++ = '?'; 3512 ret = -1; 3513 } else 3514 *mbs++ = (char)(val&0xff); 3515 ll++; 3516 } 3517 as->length += ll; 3518 as->s[as->length] = '\0'; 3519 return (ret); 3520 } 3521 3522 archive_string_init(&tmp); 3523 if (be) { 3524 if (is_big_endian()) { 3525 u16 = _p; 3526 } else { 3527 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3528 return (-1); 3529 memcpy(tmp.s, _p, bytes); 3530 for (b = 0; b < bytes; b += 2) { 3531 uint16_t val = archive_be16dec(tmp.s+b); 3532 archive_le16enc(tmp.s+b, val); 3533 } 3534 u16 = tmp.s; 3535 } 3536 } else { 3537 if (!is_big_endian()) { 3538 u16 = _p; 3539 } else { 3540 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3541 return (-1); 3542 memcpy(tmp.s, _p, bytes); 3543 for (b = 0; b < bytes; b += 2) { 3544 uint16_t val = archive_le16dec(tmp.s+b); 3545 archive_be16enc(tmp.s+b, val); 3546 } 3547 u16 = tmp.s; 3548 } 3549 } 3550 3551 do { 3552 defchar = 0; 3553 ll = WideCharToMultiByte(sc->to_cp, 0, 3554 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, 3555 NULL, &defchar); 3556 /* Exit loop if we succeeded */ 3557 if (ll != 0 || 3558 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3559 break; 3560 } 3561 /* Else expand buffer and loop to try again. */ 3562 ll = WideCharToMultiByte(sc->to_cp, 0, 3563 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); 3564 if (archive_string_ensure(as, ll +1) == NULL) 3565 return (-1); 3566 mbs = as->s + as->length; 3567 mbs_size = as->buffer_length - as->length -1; 3568 } while (1); 3569 archive_string_free(&tmp); 3570 as->length += ll; 3571 as->s[as->length] = '\0'; 3572 if (ll == 0 || defchar) 3573 ret = -1; 3574 return (ret); 3575 } 3576 3577 static int 3578 win_strncat_from_utf16be(struct archive_string *as, const void *_p, 3579 size_t bytes, struct archive_string_conv *sc) 3580 { 3581 return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); 3582 } 3583 3584 static int 3585 win_strncat_from_utf16le(struct archive_string *as, const void *_p, 3586 size_t bytes, struct archive_string_conv *sc) 3587 { 3588 return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); 3589 } 3590 3591 static int 3592 is_big_endian(void) 3593 { 3594 uint16_t d = 1; 3595 3596 return (archive_be16dec(&d) == 1); 3597 } 3598 3599 /* 3600 * Convert a current locale string to UTF-16BE/LE and copy the result. 3601 * Return -1 if conversion failes. 3602 */ 3603 static int 3604 win_strncat_to_utf16(struct archive_string *as16, const void *_p, 3605 size_t length, struct archive_string_conv *sc, int bigendian) 3606 { 3607 const char *s = (const char *)_p; 3608 char *u16; 3609 size_t count, avail; 3610 3611 if (archive_string_ensure(as16, 3612 as16->length + (length + 1) * 2) == NULL) 3613 return (-1); 3614 3615 u16 = as16->s + as16->length; 3616 avail = as16->buffer_length - 2; 3617 if (sc->from_cp == CP_C_LOCALE) { 3618 /* 3619 * "C" locale special process. 3620 */ 3621 count = 0; 3622 while (count < length && *s) { 3623 if (bigendian) 3624 archive_be16enc(u16, *s); 3625 else 3626 archive_le16enc(u16, *s); 3627 u16 += 2; 3628 s++; 3629 count++; 3630 } 3631 as16->length += count << 1; 3632 as16->s[as16->length] = 0; 3633 as16->s[as16->length+1] = 0; 3634 return (0); 3635 } 3636 do { 3637 count = MultiByteToWideChar(sc->from_cp, 3638 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); 3639 /* Exit loop if we succeeded */ 3640 if (count != 0 || 3641 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3642 break; 3643 } 3644 /* Expand buffer and try again */ 3645 count = MultiByteToWideChar(sc->from_cp, 3646 MB_PRECOMPOSED, s, (int)length, NULL, 0); 3647 if (archive_string_ensure(as16, (count +1) * 2) 3648 == NULL) 3649 return (-1); 3650 u16 = as16->s + as16->length; 3651 avail = as16->buffer_length - 2; 3652 } while (1); 3653 as16->length += count * 2; 3654 as16->s[as16->length] = 0; 3655 as16->s[as16->length+1] = 0; 3656 if (count == 0) 3657 return (-1); 3658 3659 if (is_big_endian()) { 3660 if (!bigendian) { 3661 while (count > 0) { 3662 uint16_t v = archive_be16dec(u16); 3663 archive_le16enc(u16, v); 3664 u16 += 2; 3665 count--; 3666 } 3667 } 3668 } else { 3669 if (bigendian) { 3670 while (count > 0) { 3671 uint16_t v = archive_le16dec(u16); 3672 archive_be16enc(u16, v); 3673 u16 += 2; 3674 count--; 3675 } 3676 } 3677 } 3678 return (0); 3679 } 3680 3681 static int 3682 win_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3683 size_t length, struct archive_string_conv *sc) 3684 { 3685 return (win_strncat_to_utf16(as16, _p, length, sc, 1)); 3686 } 3687 3688 static int 3689 win_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3690 size_t length, struct archive_string_conv *sc) 3691 { 3692 return (win_strncat_to_utf16(as16, _p, length, sc, 0)); 3693 } 3694 3695 #endif /* _WIN32 && !__CYGWIN__ */ 3696 3697 /* 3698 * Do the best effort for conversions. 3699 * We cannot handle UTF-16BE character-set without such iconv, 3700 * but there is a chance if a string consists just ASCII code or 3701 * a current locale is UTF-8. 3702 */ 3703 3704 /* 3705 * Convert a UTF-16BE string to current locale and copy the result. 3706 * Return -1 if conversion failes. 3707 */ 3708 static int 3709 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, 3710 size_t bytes, struct archive_string_conv *sc, int be) 3711 { 3712 const char *utf16 = (const char *)_p; 3713 char *mbs; 3714 uint32_t uc; 3715 int n, ret; 3716 3717 (void)sc; /* UNUSED */ 3718 /* 3719 * Other case, we should do the best effort. 3720 * If all character are ASCII(<0x7f), we can convert it. 3721 * if not , we set a alternative character and return -1. 3722 */ 3723 ret = 0; 3724 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3725 return (-1); 3726 mbs = as->s + as->length; 3727 3728 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { 3729 if (n < 0) { 3730 n *= -1; 3731 ret = -1; 3732 } 3733 bytes -= n; 3734 utf16 += n; 3735 3736 if (uc > 127) { 3737 /* We cannot handle it. */ 3738 *mbs++ = '?'; 3739 ret = -1; 3740 } else 3741 *mbs++ = (char)uc; 3742 } 3743 as->length = mbs - as->s; 3744 as->s[as->length] = '\0'; 3745 return (ret); 3746 } 3747 3748 static int 3749 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, 3750 size_t bytes, struct archive_string_conv *sc) 3751 { 3752 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); 3753 } 3754 3755 static int 3756 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, 3757 size_t bytes, struct archive_string_conv *sc) 3758 { 3759 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); 3760 } 3761 3762 /* 3763 * Convert a current locale string to UTF-16BE/LE and copy the result. 3764 * Return -1 if conversion failes. 3765 */ 3766 static int 3767 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, 3768 size_t length, struct archive_string_conv *sc, int bigendian) 3769 { 3770 const char *s = (const char *)_p; 3771 char *utf16; 3772 size_t remaining; 3773 int ret; 3774 3775 (void)sc; /* UNUSED */ 3776 /* 3777 * Other case, we should do the best effort. 3778 * If all character are ASCII(<0x7f), we can convert it. 3779 * if not , we set a alternative character and return -1. 3780 */ 3781 ret = 0; 3782 remaining = length; 3783 3784 if (archive_string_ensure(as16, 3785 as16->length + (length + 1) * 2) == NULL) 3786 return (-1); 3787 3788 utf16 = as16->s + as16->length; 3789 while (remaining--) { 3790 unsigned c = *s++; 3791 if (c > 127) { 3792 /* We cannot handle it. */ 3793 c = UNICODE_R_CHAR; 3794 ret = -1; 3795 } 3796 if (bigendian) 3797 archive_be16enc(utf16, c); 3798 else 3799 archive_le16enc(utf16, c); 3800 utf16 += 2; 3801 } 3802 as16->length = utf16 - as16->s; 3803 as16->s[as16->length] = 0; 3804 as16->s[as16->length+1] = 0; 3805 return (ret); 3806 } 3807 3808 static int 3809 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3810 size_t length, struct archive_string_conv *sc) 3811 { 3812 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); 3813 } 3814 3815 static int 3816 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3817 size_t length, struct archive_string_conv *sc) 3818 { 3819 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); 3820 } 3821 3822 3823 /* 3824 * Multistring operations. 3825 */ 3826 3827 void 3828 archive_mstring_clean(struct archive_mstring *aes) 3829 { 3830 archive_wstring_free(&(aes->aes_wcs)); 3831 archive_string_free(&(aes->aes_mbs)); 3832 archive_string_free(&(aes->aes_utf8)); 3833 archive_string_free(&(aes->aes_mbs_in_locale)); 3834 aes->aes_set = 0; 3835 } 3836 3837 void 3838 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) 3839 { 3840 dest->aes_set = src->aes_set; 3841 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); 3842 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); 3843 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); 3844 } 3845 3846 int 3847 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, 3848 const char **p) 3849 { 3850 struct archive_string_conv *sc; 3851 int r; 3852 3853 /* If we already have a UTF8 form, return that immediately. */ 3854 if (aes->aes_set & AES_SET_UTF8) { 3855 *p = aes->aes_utf8.s; 3856 return (0); 3857 } 3858 3859 *p = NULL; 3860 if (aes->aes_set & AES_SET_MBS) { 3861 sc = archive_string_conversion_to_charset(a, "UTF-8", 1); 3862 if (sc == NULL) 3863 return (-1);/* Couldn't allocate memory for sc. */ 3864 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, 3865 aes->aes_mbs.length, sc); 3866 if (a == NULL) 3867 free_sconv_object(sc); 3868 if (r == 0) { 3869 aes->aes_set |= AES_SET_UTF8; 3870 *p = aes->aes_utf8.s; 3871 return (0);/* success. */ 3872 } else 3873 return (-1);/* failure. */ 3874 } 3875 return (0);/* success. */ 3876 } 3877 3878 int 3879 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, 3880 const char **p) 3881 { 3882 int r, ret = 0; 3883 3884 (void)a; /* UNUSED */ 3885 /* If we already have an MBS form, return that immediately. */ 3886 if (aes->aes_set & AES_SET_MBS) { 3887 *p = aes->aes_mbs.s; 3888 return (ret); 3889 } 3890 3891 *p = NULL; 3892 /* If there's a WCS form, try converting with the native locale. */ 3893 if (aes->aes_set & AES_SET_WCS) { 3894 archive_string_empty(&(aes->aes_mbs)); 3895 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3896 aes->aes_wcs.s, aes->aes_wcs.length); 3897 *p = aes->aes_mbs.s; 3898 if (r == 0) { 3899 aes->aes_set |= AES_SET_MBS; 3900 return (ret); 3901 } else 3902 ret = -1; 3903 } 3904 3905 /* 3906 * Only a UTF-8 form cannot avail because its conversion already 3907 * failed at archive_mstring_update_utf8(). 3908 */ 3909 return (ret); 3910 } 3911 3912 int 3913 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, 3914 const wchar_t **wp) 3915 { 3916 int r, ret = 0; 3917 3918 (void)a;/* UNUSED */ 3919 /* Return WCS form if we already have it. */ 3920 if (aes->aes_set & AES_SET_WCS) { 3921 *wp = aes->aes_wcs.s; 3922 return (ret); 3923 } 3924 3925 *wp = NULL; 3926 /* Try converting MBS to WCS using native locale. */ 3927 if (aes->aes_set & AES_SET_MBS) { 3928 archive_wstring_empty(&(aes->aes_wcs)); 3929 r = archive_wstring_append_from_mbs(&(aes->aes_wcs), 3930 aes->aes_mbs.s, aes->aes_mbs.length); 3931 if (r == 0) { 3932 aes->aes_set |= AES_SET_WCS; 3933 *wp = aes->aes_wcs.s; 3934 } else 3935 ret = -1;/* failure. */ 3936 } 3937 return (ret); 3938 } 3939 3940 int 3941 archive_mstring_get_mbs_l(struct archive_mstring *aes, 3942 const char **p, size_t *length, struct archive_string_conv *sc) 3943 { 3944 int r, ret = 0; 3945 3946 #if defined(_WIN32) && !defined(__CYGWIN__) 3947 /* 3948 * Internationalization programing on Windows must use Wide 3949 * characters because Windows platform cannot make locale UTF-8. 3950 */ 3951 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { 3952 archive_string_empty(&(aes->aes_mbs_in_locale)); 3953 r = archive_string_append_from_wcs_in_codepage( 3954 &(aes->aes_mbs_in_locale), aes->aes_wcs.s, 3955 aes->aes_wcs.length, sc); 3956 if (r == 0) { 3957 *p = aes->aes_mbs_in_locale.s; 3958 if (length != NULL) 3959 *length = aes->aes_mbs_in_locale.length; 3960 return (0); 3961 } else if (errno == ENOMEM) 3962 return (-1); 3963 else 3964 ret = -1; 3965 } 3966 #endif 3967 3968 /* If there is not an MBS form but is a WCS form, try converting 3969 * with the native locale to be used for translating it to specified 3970 * character-set. */ 3971 if ((aes->aes_set & AES_SET_MBS) == 0 && 3972 (aes->aes_set & AES_SET_WCS) != 0) { 3973 archive_string_empty(&(aes->aes_mbs)); 3974 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3975 aes->aes_wcs.s, aes->aes_wcs.length); 3976 if (r == 0) 3977 aes->aes_set |= AES_SET_MBS; 3978 else if (errno == ENOMEM) 3979 return (-1); 3980 else 3981 ret = -1; 3982 } 3983 /* If we already have an MBS form, use it to be translated to 3984 * specified character-set. */ 3985 if (aes->aes_set & AES_SET_MBS) { 3986 if (sc == NULL) { 3987 /* Conversion is unneeded. */ 3988 *p = aes->aes_mbs.s; 3989 if (length != NULL) 3990 *length = aes->aes_mbs.length; 3991 return (0); 3992 } 3993 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), 3994 aes->aes_mbs.s, aes->aes_mbs.length, sc); 3995 *p = aes->aes_mbs_in_locale.s; 3996 if (length != NULL) 3997 *length = aes->aes_mbs_in_locale.length; 3998 } else { 3999 *p = NULL; 4000 if (length != NULL) 4001 *length = 0; 4002 } 4003 return (ret); 4004 } 4005 4006 int 4007 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) 4008 { 4009 if (mbs == NULL) { 4010 aes->aes_set = 0; 4011 return (0); 4012 } 4013 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); 4014 } 4015 4016 int 4017 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, 4018 size_t len) 4019 { 4020 if (mbs == NULL) { 4021 aes->aes_set = 0; 4022 return (0); 4023 } 4024 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4025 archive_strncpy(&(aes->aes_mbs), mbs, len); 4026 archive_string_empty(&(aes->aes_utf8)); 4027 archive_wstring_empty(&(aes->aes_wcs)); 4028 return (0); 4029 } 4030 4031 int 4032 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) 4033 { 4034 return archive_mstring_copy_wcs_len(aes, wcs, 4035 wcs == NULL ? 0 : wcslen(wcs)); 4036 } 4037 4038 int 4039 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) 4040 { 4041 if (utf8 == NULL) { 4042 aes->aes_set = 0; 4043 } 4044 aes->aes_set = AES_SET_UTF8; 4045 archive_string_empty(&(aes->aes_mbs)); 4046 archive_string_empty(&(aes->aes_wcs)); 4047 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); 4048 return (int)strlen(utf8); 4049 } 4050 4051 int 4052 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, 4053 size_t len) 4054 { 4055 if (wcs == NULL) { 4056 aes->aes_set = 0; 4057 } 4058 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ 4059 archive_string_empty(&(aes->aes_mbs)); 4060 archive_string_empty(&(aes->aes_utf8)); 4061 archive_wstrncpy(&(aes->aes_wcs), wcs, len); 4062 return (0); 4063 } 4064 4065 int 4066 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, 4067 const char *mbs, size_t len, struct archive_string_conv *sc) 4068 { 4069 int r; 4070 4071 if (mbs == NULL) { 4072 aes->aes_set = 0; 4073 return (0); 4074 } 4075 archive_string_empty(&(aes->aes_mbs)); 4076 archive_wstring_empty(&(aes->aes_wcs)); 4077 archive_string_empty(&(aes->aes_utf8)); 4078 #if defined(_WIN32) && !defined(__CYGWIN__) 4079 /* 4080 * Internationalization programing on Windows must use Wide 4081 * characters because Windows platform cannot make locale UTF-8. 4082 */ 4083 if (sc == NULL) { 4084 if (archive_string_append(&(aes->aes_mbs), 4085 mbs, mbsnbytes(mbs, len)) == NULL) { 4086 aes->aes_set = 0; 4087 r = -1; 4088 } else { 4089 aes->aes_set = AES_SET_MBS; 4090 r = 0; 4091 } 4092 #if defined(HAVE_ICONV) 4093 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { 4094 /* 4095 * This case happens only when MultiByteToWideChar() cannot 4096 * handle sc->from_cp, and we have to iconv in order to 4097 * translate character-set to wchar_t,UTF-16. 4098 */ 4099 iconv_t cd = sc->cd; 4100 unsigned from_cp; 4101 int flag; 4102 4103 /* 4104 * Translate multi-bytes from some character-set to UTF-8. 4105 */ 4106 sc->cd = sc->cd_w; 4107 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); 4108 sc->cd = cd; 4109 if (r != 0) { 4110 aes->aes_set = 0; 4111 return (r); 4112 } 4113 aes->aes_set = AES_SET_UTF8; 4114 4115 /* 4116 * Append the UTF-8 string into wstring. 4117 */ 4118 flag = sc->flag; 4119 sc->flag &= ~(SCONV_NORMALIZATION_C 4120 | SCONV_TO_UTF16| SCONV_FROM_UTF16); 4121 from_cp = sc->from_cp; 4122 sc->from_cp = CP_UTF8; 4123 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), 4124 aes->aes_utf8.s, aes->aes_utf8.length, sc); 4125 sc->flag = flag; 4126 sc->from_cp = from_cp; 4127 if (r == 0) 4128 aes->aes_set |= AES_SET_WCS; 4129 #endif 4130 } else { 4131 r = archive_wstring_append_from_mbs_in_codepage( 4132 &(aes->aes_wcs), mbs, len, sc); 4133 if (r == 0) 4134 aes->aes_set = AES_SET_WCS; 4135 else 4136 aes->aes_set = 0; 4137 } 4138 #else 4139 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); 4140 if (r == 0) 4141 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4142 else 4143 aes->aes_set = 0; 4144 #endif 4145 return (r); 4146 } 4147 4148 /* 4149 * The 'update' form tries to proactively update all forms of 4150 * this string (WCS and MBS) and returns an error if any of 4151 * them fail. This is used by the 'pax' handler, for instance, 4152 * to detect and report character-conversion failures early while 4153 * still allowing clients to get potentially useful values from 4154 * the more tolerant lazy conversions. (get_mbs and get_wcs will 4155 * strive to give the user something useful, so you can get hopefully 4156 * usable values even if some of the character conversions are failing.) 4157 */ 4158 int 4159 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, 4160 const char *utf8) 4161 { 4162 struct archive_string_conv *sc; 4163 int r; 4164 4165 if (utf8 == NULL) { 4166 aes->aes_set = 0; 4167 return (0); /* Succeeded in clearing everything. */ 4168 } 4169 4170 /* Save the UTF8 string. */ 4171 archive_strcpy(&(aes->aes_utf8), utf8); 4172 4173 /* Empty the mbs and wcs strings. */ 4174 archive_string_empty(&(aes->aes_mbs)); 4175 archive_wstring_empty(&(aes->aes_wcs)); 4176 4177 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ 4178 4179 /* Try converting UTF-8 to MBS, return false on failure. */ 4180 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 4181 if (sc == NULL) 4182 return (-1);/* Couldn't allocate memory for sc. */ 4183 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); 4184 if (a == NULL) 4185 free_sconv_object(sc); 4186 if (r != 0) 4187 return (-1); 4188 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ 4189 4190 /* Try converting MBS to WCS, return false on failure. */ 4191 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, 4192 aes->aes_mbs.length)) 4193 return (-1); 4194 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; 4195 4196 /* All conversions succeeded. */ 4197 return (0); 4198 } 4199