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 filename 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 = 0; 1943 int i, r = 0, r2; 1944 1945 if (_p != NULL && n > 0) { 1946 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) 1947 length = utf16nbytes(_p, n); 1948 else 1949 length = mbsnbytes(_p, n); 1950 } 1951 1952 /* We must allocate memory even if there is no data for conversion 1953 * or copy. This simulates archive_string_append behavior. */ 1954 if (length == 0) { 1955 int tn = 1; 1956 if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) 1957 tn = 2; 1958 if (archive_string_ensure(as, as->length + tn) == NULL) 1959 return (-1); 1960 as->s[as->length] = 0; 1961 if (tn == 2) 1962 as->s[as->length+1] = 0; 1963 return (0); 1964 } 1965 1966 /* 1967 * If sc is NULL, we just make a copy. 1968 */ 1969 if (sc == NULL) { 1970 if (archive_string_append(as, _p, length) == NULL) 1971 return (-1);/* No memory */ 1972 return (0); 1973 } 1974 1975 s = _p; 1976 i = 0; 1977 if (sc->nconverter > 1) { 1978 sc->utftmp.length = 0; 1979 r2 = sc->converter[0](&(sc->utftmp), s, length, sc); 1980 if (r2 != 0 && errno == ENOMEM) 1981 return (r2); 1982 if (r > r2) 1983 r = r2; 1984 s = sc->utftmp.s; 1985 length = sc->utftmp.length; 1986 ++i; 1987 } 1988 r2 = sc->converter[i](as, s, length, sc); 1989 if (r > r2) 1990 r = r2; 1991 return (r); 1992 } 1993 1994 #if HAVE_ICONV 1995 1996 /* 1997 * Return -1 if conversion fails. 1998 */ 1999 static int 2000 iconv_strncat_in_locale(struct archive_string *as, const void *_p, 2001 size_t length, struct archive_string_conv *sc) 2002 { 2003 ICONV_CONST char *itp; 2004 size_t remaining; 2005 iconv_t cd; 2006 char *outp; 2007 size_t avail, bs; 2008 int return_value = 0; /* success */ 2009 int to_size, from_size; 2010 2011 if (sc->flag & SCONV_TO_UTF16) 2012 to_size = 2; 2013 else 2014 to_size = 1; 2015 if (sc->flag & SCONV_FROM_UTF16) 2016 from_size = 2; 2017 else 2018 from_size = 1; 2019 2020 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) 2021 return (-1); 2022 2023 cd = sc->cd; 2024 itp = (char *)(uintptr_t)_p; 2025 remaining = length; 2026 outp = as->s + as->length; 2027 avail = as->buffer_length - as->length - to_size; 2028 while (remaining >= (size_t)from_size) { 2029 size_t result = iconv(cd, &itp, &remaining, &outp, &avail); 2030 2031 if (result != (size_t)-1) 2032 break; /* Conversion completed. */ 2033 2034 if (errno == EILSEQ || errno == EINVAL) { 2035 /* 2036 * If an output charset is UTF-8 or UTF-16BE/LE, 2037 * unknown character should be U+FFFD 2038 * (replacement character). 2039 */ 2040 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { 2041 size_t rbytes; 2042 if (sc->flag & SCONV_TO_UTF8) 2043 rbytes = sizeof(utf8_replacement_char); 2044 else 2045 rbytes = 2; 2046 2047 if (avail < rbytes) { 2048 as->length = outp - as->s; 2049 bs = as->buffer_length + 2050 (remaining * to_size) + rbytes; 2051 if (NULL == 2052 archive_string_ensure(as, bs)) 2053 return (-1); 2054 outp = as->s + as->length; 2055 avail = as->buffer_length 2056 - as->length - to_size; 2057 } 2058 if (sc->flag & SCONV_TO_UTF8) 2059 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); 2060 else if (sc->flag & SCONV_TO_UTF16BE) 2061 archive_be16enc(outp, UNICODE_R_CHAR); 2062 else 2063 archive_le16enc(outp, UNICODE_R_CHAR); 2064 outp += rbytes; 2065 avail -= rbytes; 2066 } else { 2067 /* Skip the illegal input bytes. */ 2068 *outp++ = '?'; 2069 avail--; 2070 } 2071 itp += from_size; 2072 remaining -= from_size; 2073 return_value = -1; /* failure */ 2074 } else { 2075 /* E2BIG no output buffer, 2076 * Increase an output buffer. */ 2077 as->length = outp - as->s; 2078 bs = as->buffer_length + remaining * 2; 2079 if (NULL == archive_string_ensure(as, bs)) 2080 return (-1); 2081 outp = as->s + as->length; 2082 avail = as->buffer_length - as->length - to_size; 2083 } 2084 } 2085 as->length = outp - as->s; 2086 as->s[as->length] = 0; 2087 if (to_size == 2) 2088 as->s[as->length+1] = 0; 2089 return (return_value); 2090 } 2091 2092 #endif /* HAVE_ICONV */ 2093 2094 2095 #if defined(_WIN32) && !defined(__CYGWIN__) 2096 2097 /* 2098 * Translate a string from a some CodePage to an another CodePage by 2099 * Windows APIs, and copy the result. Return -1 if conversion fails. 2100 */ 2101 static int 2102 strncat_in_codepage(struct archive_string *as, 2103 const void *_p, size_t length, struct archive_string_conv *sc) 2104 { 2105 const char *s = (const char *)_p; 2106 struct archive_wstring aws; 2107 size_t l; 2108 int r, saved_flag; 2109 2110 archive_string_init(&aws); 2111 saved_flag = sc->flag; 2112 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); 2113 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); 2114 sc->flag = saved_flag; 2115 if (r != 0) { 2116 archive_wstring_free(&aws); 2117 if (errno != ENOMEM) 2118 archive_string_append(as, s, length); 2119 return (-1); 2120 } 2121 2122 l = as->length; 2123 r = archive_string_append_from_wcs_in_codepage( 2124 as, aws.s, aws.length, sc); 2125 if (r != 0 && errno != ENOMEM && l == as->length) 2126 archive_string_append(as, s, length); 2127 archive_wstring_free(&aws); 2128 return (r); 2129 } 2130 2131 /* 2132 * Test whether MBS ==> WCS is okay. 2133 */ 2134 static int 2135 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2136 { 2137 const char *p = (const char *)_p; 2138 unsigned codepage; 2139 DWORD mbflag = MB_ERR_INVALID_CHARS; 2140 2141 if (sc->flag & SCONV_FROM_CHARSET) 2142 codepage = sc->to_cp; 2143 else 2144 codepage = sc->from_cp; 2145 2146 if (codepage == CP_C_LOCALE) 2147 return (0); 2148 if (codepage != CP_UTF8) 2149 mbflag |= MB_PRECOMPOSED; 2150 2151 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) 2152 return (-1); /* Invalid */ 2153 return (0); /* Okay */ 2154 } 2155 2156 #else 2157 2158 /* 2159 * Test whether MBS ==> WCS is okay. 2160 */ 2161 static int 2162 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2163 { 2164 const char *p = (const char *)_p; 2165 size_t r; 2166 2167 #if HAVE_MBRTOWC 2168 mbstate_t shift_state; 2169 2170 memset(&shift_state, 0, sizeof(shift_state)); 2171 #else 2172 /* Clear the shift state before starting. */ 2173 mbtowc(NULL, NULL, 0); 2174 #endif 2175 while (n) { 2176 wchar_t wc; 2177 2178 #if HAVE_MBRTOWC 2179 r = mbrtowc(&wc, p, n, &shift_state); 2180 #else 2181 r = mbtowc(&wc, p, n); 2182 #endif 2183 if (r == (size_t)-1 || r == (size_t)-2) 2184 return (-1);/* Invalid. */ 2185 if (r == 0) 2186 break; 2187 p += r; 2188 n -= r; 2189 } 2190 (void)sc; /* UNUSED */ 2191 return (0); /* All Okey. */ 2192 } 2193 2194 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 2195 2196 /* 2197 * Basically returns -1 because we cannot make a conversion of charset 2198 * without iconv but in some cases this would return 0. 2199 * Returns 0 if all copied characters are ASCII. 2200 * Returns 0 if both from-locale and to-locale are the same and those 2201 * can be WCS with no error. 2202 */ 2203 static int 2204 best_effort_strncat_in_locale(struct archive_string *as, const void *_p, 2205 size_t length, struct archive_string_conv *sc) 2206 { 2207 size_t remaining; 2208 const uint8_t *itp; 2209 int return_value = 0; /* success */ 2210 2211 /* 2212 * If both from-locale and to-locale is the same, this makes a copy. 2213 * And then this checks all copied MBS can be WCS if so returns 0. 2214 */ 2215 if (sc->same) { 2216 if (archive_string_append(as, _p, length) == NULL) 2217 return (-1);/* No memory */ 2218 return (invalid_mbs(_p, length, sc)); 2219 } 2220 2221 /* 2222 * If a character is ASCII, this just copies it. If not, this 2223 * assigns '?' charater instead but in UTF-8 locale this assigns 2224 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, 2225 * a Replacement Character in Unicode. 2226 */ 2227 2228 remaining = length; 2229 itp = (const uint8_t *)_p; 2230 while (*itp && remaining > 0) { 2231 if (*itp > 127) { 2232 // Non-ASCII: Substitute with suitable replacement 2233 if (sc->flag & SCONV_TO_UTF8) { 2234 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { 2235 __archive_errx(1, "Out of memory"); 2236 } 2237 } else { 2238 archive_strappend_char(as, '?'); 2239 } 2240 return_value = -1; 2241 } else { 2242 archive_strappend_char(as, *itp); 2243 } 2244 ++itp; 2245 } 2246 return (return_value); 2247 } 2248 2249 2250 /* 2251 * Unicode conversion functions. 2252 * - UTF-8 <===> UTF-8 in removing surrogate pairs. 2253 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. 2254 * - UTF-8 made by libarchive 2.x ===> UTF-8. 2255 * - UTF-16BE <===> UTF-8. 2256 * 2257 */ 2258 2259 /* 2260 * Utility to convert a single UTF-8 sequence. 2261 * 2262 * Usually return used bytes, return used byte in negative value when 2263 * a unicode character is replaced with U+FFFD. 2264 * See also http://unicode.org/review/pr-121.html Public Review Issue #121 2265 * Recommended Practice for Replacement Characters. 2266 */ 2267 static int 2268 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2269 { 2270 static const char utf8_count[256] = { 2271 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 2272 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 2273 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 2274 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 2275 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 2276 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 2277 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 2278 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 2279 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 2280 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 2281 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 2282 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 2283 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2284 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 2285 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 2286 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ 2287 }; 2288 int ch, i; 2289 int cnt; 2290 uint32_t wc; 2291 2292 /* Sanity check. */ 2293 if (n == 0) 2294 return (0); 2295 /* 2296 * Decode 1-4 bytes depending on the value of the first byte. 2297 */ 2298 ch = (unsigned char)*s; 2299 if (ch == 0) 2300 return (0); /* Standard: return 0 for end-of-string. */ 2301 cnt = utf8_count[ch]; 2302 2303 /* Invalid sequence or there are not plenty bytes. */ 2304 if ((int)n < cnt) { 2305 cnt = (int)n; 2306 for (i = 1; i < cnt; i++) { 2307 if ((s[i] & 0xc0) != 0x80) { 2308 cnt = i; 2309 break; 2310 } 2311 } 2312 goto invalid_sequence; 2313 } 2314 2315 /* Make a Unicode code point from a single UTF-8 sequence. */ 2316 switch (cnt) { 2317 case 1: /* 1 byte sequence. */ 2318 *pwc = ch & 0x7f; 2319 return (cnt); 2320 case 2: /* 2 bytes sequence. */ 2321 if ((s[1] & 0xc0) != 0x80) { 2322 cnt = 1; 2323 goto invalid_sequence; 2324 } 2325 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); 2326 return (cnt); 2327 case 3: /* 3 bytes sequence. */ 2328 if ((s[1] & 0xc0) != 0x80) { 2329 cnt = 1; 2330 goto invalid_sequence; 2331 } 2332 if ((s[2] & 0xc0) != 0x80) { 2333 cnt = 2; 2334 goto invalid_sequence; 2335 } 2336 wc = ((ch & 0x0f) << 12) 2337 | ((s[1] & 0x3f) << 6) 2338 | (s[2] & 0x3f); 2339 if (wc < 0x800) 2340 goto invalid_sequence;/* Overlong sequence. */ 2341 break; 2342 case 4: /* 4 bytes sequence. */ 2343 if ((s[1] & 0xc0) != 0x80) { 2344 cnt = 1; 2345 goto invalid_sequence; 2346 } 2347 if ((s[2] & 0xc0) != 0x80) { 2348 cnt = 2; 2349 goto invalid_sequence; 2350 } 2351 if ((s[3] & 0xc0) != 0x80) { 2352 cnt = 3; 2353 goto invalid_sequence; 2354 } 2355 wc = ((ch & 0x07) << 18) 2356 | ((s[1] & 0x3f) << 12) 2357 | ((s[2] & 0x3f) << 6) 2358 | (s[3] & 0x3f); 2359 if (wc < 0x10000) 2360 goto invalid_sequence;/* Overlong sequence. */ 2361 break; 2362 default: /* Others are all invalid sequence. */ 2363 if (ch == 0xc0 || ch == 0xc1) 2364 cnt = 2; 2365 else if (ch >= 0xf5 && ch <= 0xf7) 2366 cnt = 4; 2367 else if (ch >= 0xf8 && ch <= 0xfb) 2368 cnt = 5; 2369 else if (ch == 0xfc || ch == 0xfd) 2370 cnt = 6; 2371 else 2372 cnt = 1; 2373 if ((int)n < cnt) 2374 cnt = (int)n; 2375 for (i = 1; i < cnt; i++) { 2376 if ((s[i] & 0xc0) != 0x80) { 2377 cnt = i; 2378 break; 2379 } 2380 } 2381 goto invalid_sequence; 2382 } 2383 2384 /* The code point larger than 0x10FFFF is not legal 2385 * Unicode values. */ 2386 if (wc > UNICODE_MAX) 2387 goto invalid_sequence; 2388 /* Correctly gets a Unicode, returns used bytes. */ 2389 *pwc = wc; 2390 return (cnt); 2391 invalid_sequence: 2392 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2393 return (cnt * -1); 2394 } 2395 2396 static int 2397 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2398 { 2399 int cnt; 2400 2401 cnt = _utf8_to_unicode(pwc, s, n); 2402 /* Any of Surrogate pair is not legal Unicode values. */ 2403 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) 2404 return (-3); 2405 return (cnt); 2406 } 2407 2408 static inline uint32_t 2409 combine_surrogate_pair(uint32_t uc, uint32_t uc2) 2410 { 2411 uc -= 0xD800; 2412 uc *= 0x400; 2413 uc += uc2 - 0xDC00; 2414 uc += 0x10000; 2415 return (uc); 2416 } 2417 2418 /* 2419 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in 2420 * removing surrogate pairs. 2421 * 2422 * CESU-8: The Compatibility Encoding Scheme for UTF-16. 2423 * 2424 * Usually return used bytes, return used byte in negative value when 2425 * a unicode character is replaced with U+FFFD. 2426 */ 2427 static int 2428 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2429 { 2430 uint32_t wc = 0; 2431 int cnt; 2432 2433 cnt = _utf8_to_unicode(&wc, s, n); 2434 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { 2435 uint32_t wc2 = 0; 2436 if (n - 3 < 3) { 2437 /* Invalid byte sequence. */ 2438 goto invalid_sequence; 2439 } 2440 cnt = _utf8_to_unicode(&wc2, s+3, n-3); 2441 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { 2442 /* Invalid byte sequence. */ 2443 goto invalid_sequence; 2444 } 2445 wc = combine_surrogate_pair(wc, wc2); 2446 cnt = 6; 2447 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { 2448 /* Invalid byte sequence. */ 2449 goto invalid_sequence; 2450 } 2451 *pwc = wc; 2452 return (cnt); 2453 invalid_sequence: 2454 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2455 if (cnt > 0) 2456 cnt *= -1; 2457 return (cnt); 2458 } 2459 2460 /* 2461 * Convert a Unicode code point to a single UTF-8 sequence. 2462 * 2463 * NOTE:This function does not check if the Unicode is legal or not. 2464 * Please you definitely check it before calling this. 2465 */ 2466 static size_t 2467 unicode_to_utf8(char *p, size_t remaining, uint32_t uc) 2468 { 2469 char *_p = p; 2470 2471 /* Invalid Unicode char maps to Replacement character */ 2472 if (uc > UNICODE_MAX) 2473 uc = UNICODE_R_CHAR; 2474 /* Translate code point to UTF8 */ 2475 if (uc <= 0x7f) { 2476 if (remaining == 0) 2477 return (0); 2478 *p++ = (char)uc; 2479 } else if (uc <= 0x7ff) { 2480 if (remaining < 2) 2481 return (0); 2482 *p++ = 0xc0 | ((uc >> 6) & 0x1f); 2483 *p++ = 0x80 | (uc & 0x3f); 2484 } else if (uc <= 0xffff) { 2485 if (remaining < 3) 2486 return (0); 2487 *p++ = 0xe0 | ((uc >> 12) & 0x0f); 2488 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2489 *p++ = 0x80 | (uc & 0x3f); 2490 } else { 2491 if (remaining < 4) 2492 return (0); 2493 *p++ = 0xf0 | ((uc >> 18) & 0x07); 2494 *p++ = 0x80 | ((uc >> 12) & 0x3f); 2495 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2496 *p++ = 0x80 | (uc & 0x3f); 2497 } 2498 return (p - _p); 2499 } 2500 2501 static int 2502 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) 2503 { 2504 return (utf16_to_unicode(pwc, s, n, 1)); 2505 } 2506 2507 static int 2508 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) 2509 { 2510 return (utf16_to_unicode(pwc, s, n, 0)); 2511 } 2512 2513 static int 2514 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) 2515 { 2516 const char *utf16 = s; 2517 unsigned uc; 2518 2519 if (n == 0) 2520 return (0); 2521 if (n == 1) { 2522 /* set the Replacement Character instead. */ 2523 *pwc = UNICODE_R_CHAR; 2524 return (-1); 2525 } 2526 2527 if (be) 2528 uc = archive_be16dec(utf16); 2529 else 2530 uc = archive_le16dec(utf16); 2531 utf16 += 2; 2532 2533 /* If this is a surrogate pair, assemble the full code point.*/ 2534 if (IS_HIGH_SURROGATE_LA(uc)) { 2535 unsigned uc2; 2536 2537 if (n >= 4) { 2538 if (be) 2539 uc2 = archive_be16dec(utf16); 2540 else 2541 uc2 = archive_le16dec(utf16); 2542 } else 2543 uc2 = 0; 2544 if (IS_LOW_SURROGATE_LA(uc2)) { 2545 uc = combine_surrogate_pair(uc, uc2); 2546 utf16 += 2; 2547 } else { 2548 /* Undescribed code point should be U+FFFD 2549 * (replacement character). */ 2550 *pwc = UNICODE_R_CHAR; 2551 return (-2); 2552 } 2553 } 2554 2555 /* 2556 * Surrogate pair values(0xd800 through 0xdfff) are only 2557 * used by UTF-16, so, after above culculation, the code 2558 * must not be surrogate values, and Unicode has no codes 2559 * larger than 0x10ffff. Thus, those are not legal Unicode 2560 * values. 2561 */ 2562 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { 2563 /* Undescribed code point should be U+FFFD 2564 * (replacement character). */ 2565 *pwc = UNICODE_R_CHAR; 2566 return (((int)(utf16 - s)) * -1); 2567 } 2568 *pwc = uc; 2569 return ((int)(utf16 - s)); 2570 } 2571 2572 static size_t 2573 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) 2574 { 2575 char *utf16 = p; 2576 2577 if (uc > 0xffff) { 2578 /* We have a code point that won't fit into a 2579 * wchar_t; convert it to a surrogate pair. */ 2580 if (remaining < 4) 2581 return (0); 2582 uc -= 0x10000; 2583 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2584 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2585 return (4); 2586 } else { 2587 if (remaining < 2) 2588 return (0); 2589 archive_be16enc(utf16, uc); 2590 return (2); 2591 } 2592 } 2593 2594 static size_t 2595 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) 2596 { 2597 char *utf16 = p; 2598 2599 if (uc > 0xffff) { 2600 /* We have a code point that won't fit into a 2601 * wchar_t; convert it to a surrogate pair. */ 2602 if (remaining < 4) 2603 return (0); 2604 uc -= 0x10000; 2605 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2606 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2607 return (4); 2608 } else { 2609 if (remaining < 2) 2610 return (0); 2611 archive_le16enc(utf16, uc); 2612 return (2); 2613 } 2614 } 2615 2616 /* 2617 * Copy UTF-8 string in checking surrogate pair. 2618 * If any surrogate pair are found, it would be canonicalized. 2619 */ 2620 static int 2621 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, 2622 size_t len, struct archive_string_conv *sc) 2623 { 2624 const char *s; 2625 char *p, *endp; 2626 int n, ret = 0; 2627 2628 (void)sc; /* UNUSED */ 2629 2630 if (archive_string_ensure(as, as->length + len + 1) == NULL) 2631 return (-1); 2632 2633 s = (const char *)_p; 2634 p = as->s + as->length; 2635 endp = as->s + as->buffer_length -1; 2636 do { 2637 uint32_t uc; 2638 const char *ss = s; 2639 size_t w; 2640 2641 /* 2642 * Forward byte sequence until a conversion of that is needed. 2643 */ 2644 while ((n = utf8_to_unicode(&uc, s, len)) > 0) { 2645 s += n; 2646 len -= n; 2647 } 2648 if (ss < s) { 2649 if (p + (s - ss) > endp) { 2650 as->length = p - as->s; 2651 if (archive_string_ensure(as, 2652 as->buffer_length + len + 1) == NULL) 2653 return (-1); 2654 p = as->s + as->length; 2655 endp = as->s + as->buffer_length -1; 2656 } 2657 2658 memcpy(p, ss, s - ss); 2659 p += s - ss; 2660 } 2661 2662 /* 2663 * If n is negative, current byte sequence needs a replacement. 2664 */ 2665 if (n < 0) { 2666 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { 2667 /* Current byte sequence may be CESU-8. */ 2668 n = cesu8_to_unicode(&uc, s, len); 2669 } 2670 if (n < 0) { 2671 ret = -1; 2672 n *= -1;/* Use a replaced unicode character. */ 2673 } 2674 2675 /* Rebuild UTF-8 byte sequence. */ 2676 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { 2677 as->length = p - as->s; 2678 if (archive_string_ensure(as, 2679 as->buffer_length + len + 1) == NULL) 2680 return (-1); 2681 p = as->s + as->length; 2682 endp = as->s + as->buffer_length -1; 2683 } 2684 p += w; 2685 s += n; 2686 len -= n; 2687 } 2688 } while (n > 0); 2689 as->length = p - as->s; 2690 as->s[as->length] = '\0'; 2691 return (ret); 2692 } 2693 2694 static int 2695 archive_string_append_unicode(struct archive_string *as, const void *_p, 2696 size_t len, struct archive_string_conv *sc) 2697 { 2698 const char *s; 2699 char *p, *endp; 2700 uint32_t uc; 2701 size_t w; 2702 int n, ret = 0, ts, tm; 2703 int (*parse)(uint32_t *, const char *, size_t); 2704 size_t (*unparse)(char *, size_t, uint32_t); 2705 2706 if (sc->flag & SCONV_TO_UTF16BE) { 2707 unparse = unicode_to_utf16be; 2708 ts = 2; 2709 } else if (sc->flag & SCONV_TO_UTF16LE) { 2710 unparse = unicode_to_utf16le; 2711 ts = 2; 2712 } else if (sc->flag & SCONV_TO_UTF8) { 2713 unparse = unicode_to_utf8; 2714 ts = 1; 2715 } else { 2716 /* 2717 * This case is going to be converted to another 2718 * character-set through iconv. 2719 */ 2720 if (sc->flag & SCONV_FROM_UTF16BE) { 2721 unparse = unicode_to_utf16be; 2722 ts = 2; 2723 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2724 unparse = unicode_to_utf16le; 2725 ts = 2; 2726 } else { 2727 unparse = unicode_to_utf8; 2728 ts = 1; 2729 } 2730 } 2731 2732 if (sc->flag & SCONV_FROM_UTF16BE) { 2733 parse = utf16be_to_unicode; 2734 tm = 1; 2735 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2736 parse = utf16le_to_unicode; 2737 tm = 1; 2738 } else { 2739 parse = cesu8_to_unicode; 2740 tm = ts; 2741 } 2742 2743 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2744 return (-1); 2745 2746 s = (const char *)_p; 2747 p = as->s + as->length; 2748 endp = as->s + as->buffer_length - ts; 2749 while ((n = parse(&uc, s, len)) != 0) { 2750 if (n < 0) { 2751 /* Use a replaced unicode character. */ 2752 n *= -1; 2753 ret = -1; 2754 } 2755 s += n; 2756 len -= n; 2757 while ((w = unparse(p, endp - p, uc)) == 0) { 2758 /* There is not enough output buffer so 2759 * we have to expand it. */ 2760 as->length = p - as->s; 2761 if (archive_string_ensure(as, 2762 as->buffer_length + len * tm + ts) == NULL) 2763 return (-1); 2764 p = as->s + as->length; 2765 endp = as->s + as->buffer_length - ts; 2766 } 2767 p += w; 2768 } 2769 as->length = p - as->s; 2770 as->s[as->length] = '\0'; 2771 if (ts == 2) 2772 as->s[as->length+1] = '\0'; 2773 return (ret); 2774 } 2775 2776 /* 2777 * Following Constants for Hangul compositions this information comes from 2778 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ 2779 */ 2780 #define HC_SBASE 0xAC00 2781 #define HC_LBASE 0x1100 2782 #define HC_VBASE 0x1161 2783 #define HC_TBASE 0x11A7 2784 #define HC_LCOUNT 19 2785 #define HC_VCOUNT 21 2786 #define HC_TCOUNT 28 2787 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) 2788 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) 2789 2790 static uint32_t 2791 get_nfc(uint32_t uc, uint32_t uc2) 2792 { 2793 int t, b; 2794 2795 t = 0; 2796 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; 2797 while (b >= t) { 2798 int m = (t + b) / 2; 2799 if (u_composition_table[m].cp1 < uc) 2800 t = m + 1; 2801 else if (u_composition_table[m].cp1 > uc) 2802 b = m - 1; 2803 else if (u_composition_table[m].cp2 < uc2) 2804 t = m + 1; 2805 else if (u_composition_table[m].cp2 > uc2) 2806 b = m - 1; 2807 else 2808 return (u_composition_table[m].nfc); 2809 } 2810 return (0); 2811 } 2812 2813 #define FDC_MAX 10 /* The maximum number of Following Decomposable 2814 * Characters. */ 2815 2816 /* 2817 * Update first code point. 2818 */ 2819 #define UPDATE_UC(new_uc) do { \ 2820 uc = new_uc; \ 2821 ucptr = NULL; \ 2822 } while (0) 2823 2824 /* 2825 * Replace first code point with second code point. 2826 */ 2827 #define REPLACE_UC_WITH_UC2() do { \ 2828 uc = uc2; \ 2829 ucptr = uc2ptr; \ 2830 n = n2; \ 2831 } while (0) 2832 2833 #define EXPAND_BUFFER() do { \ 2834 as->length = p - as->s; \ 2835 if (archive_string_ensure(as, \ 2836 as->buffer_length + len * tm + ts) == NULL)\ 2837 return (-1); \ 2838 p = as->s + as->length; \ 2839 endp = as->s + as->buffer_length - ts; \ 2840 } while (0) 2841 2842 #define UNPARSE(p, endp, uc) do { \ 2843 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ 2844 EXPAND_BUFFER(); \ 2845 } \ 2846 p += w; \ 2847 } while (0) 2848 2849 /* 2850 * Write first code point. 2851 * If the code point has not be changed from its original code, 2852 * this just copies it from its original buffer pointer. 2853 * If not, this converts it to UTF-8 byte sequence and copies it. 2854 */ 2855 #define WRITE_UC() do { \ 2856 if (ucptr) { \ 2857 if (p + n > endp) \ 2858 EXPAND_BUFFER(); \ 2859 switch (n) { \ 2860 case 4: \ 2861 *p++ = *ucptr++; \ 2862 /* FALL THROUGH */ \ 2863 case 3: \ 2864 *p++ = *ucptr++; \ 2865 /* FALL THROUGH */ \ 2866 case 2: \ 2867 *p++ = *ucptr++; \ 2868 /* FALL THROUGH */ \ 2869 case 1: \ 2870 *p++ = *ucptr; \ 2871 break; \ 2872 } \ 2873 ucptr = NULL; \ 2874 } else { \ 2875 UNPARSE(p, endp, uc); \ 2876 } \ 2877 } while (0) 2878 2879 /* 2880 * Collect following decomposable code points. 2881 */ 2882 #define COLLECT_CPS(start) do { \ 2883 int _i; \ 2884 for (_i = start; _i < FDC_MAX ; _i++) { \ 2885 nx = parse(&ucx[_i], s, len); \ 2886 if (nx <= 0) \ 2887 break; \ 2888 cx = CCC(ucx[_i]); \ 2889 if (cl >= cx && cl != 228 && cx != 228)\ 2890 break; \ 2891 s += nx; \ 2892 len -= nx; \ 2893 cl = cx; \ 2894 ccx[_i] = cx; \ 2895 } \ 2896 if (_i >= FDC_MAX) { \ 2897 ret = -1; \ 2898 ucx_size = FDC_MAX; \ 2899 } else \ 2900 ucx_size = _i; \ 2901 } while (0) 2902 2903 /* 2904 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. 2905 * 2906 * TODO: Convert composition exclusions,which are never converted 2907 * from NFC,NFD,NFKC and NFKD, to Form C. 2908 */ 2909 static int 2910 archive_string_normalize_C(struct archive_string *as, const void *_p, 2911 size_t len, struct archive_string_conv *sc) 2912 { 2913 const char *s = (const char *)_p; 2914 char *p, *endp; 2915 uint32_t uc, uc2; 2916 size_t w; 2917 int always_replace, n, n2, ret = 0, spair, ts, tm; 2918 int (*parse)(uint32_t *, const char *, size_t); 2919 size_t (*unparse)(char *, size_t, uint32_t); 2920 2921 always_replace = 1; 2922 ts = 1;/* text size. */ 2923 if (sc->flag & SCONV_TO_UTF16BE) { 2924 unparse = unicode_to_utf16be; 2925 ts = 2; 2926 if (sc->flag & SCONV_FROM_UTF16BE) 2927 always_replace = 0; 2928 } else if (sc->flag & SCONV_TO_UTF16LE) { 2929 unparse = unicode_to_utf16le; 2930 ts = 2; 2931 if (sc->flag & SCONV_FROM_UTF16LE) 2932 always_replace = 0; 2933 } else if (sc->flag & SCONV_TO_UTF8) { 2934 unparse = unicode_to_utf8; 2935 if (sc->flag & SCONV_FROM_UTF8) 2936 always_replace = 0; 2937 } else { 2938 /* 2939 * This case is going to be converted to another 2940 * character-set through iconv. 2941 */ 2942 always_replace = 0; 2943 if (sc->flag & SCONV_FROM_UTF16BE) { 2944 unparse = unicode_to_utf16be; 2945 ts = 2; 2946 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2947 unparse = unicode_to_utf16le; 2948 ts = 2; 2949 } else { 2950 unparse = unicode_to_utf8; 2951 } 2952 } 2953 2954 if (sc->flag & SCONV_FROM_UTF16BE) { 2955 parse = utf16be_to_unicode; 2956 tm = 1; 2957 spair = 4;/* surrogate pair size in UTF-16. */ 2958 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2959 parse = utf16le_to_unicode; 2960 tm = 1; 2961 spair = 4;/* surrogate pair size in UTF-16. */ 2962 } else { 2963 parse = cesu8_to_unicode; 2964 tm = ts; 2965 spair = 6;/* surrogate pair size in UTF-8. */ 2966 } 2967 2968 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2969 return (-1); 2970 2971 p = as->s + as->length; 2972 endp = as->s + as->buffer_length - ts; 2973 while ((n = parse(&uc, s, len)) != 0) { 2974 const char *ucptr, *uc2ptr; 2975 2976 if (n < 0) { 2977 /* Use a replaced unicode character. */ 2978 UNPARSE(p, endp, uc); 2979 s += n*-1; 2980 len -= n*-1; 2981 ret = -1; 2982 continue; 2983 } else if (n == spair || always_replace) 2984 /* uc is converted from a surrogate pair. 2985 * this should be treated as a changed code. */ 2986 ucptr = NULL; 2987 else 2988 ucptr = s; 2989 s += n; 2990 len -= n; 2991 2992 /* Read second code point. */ 2993 while ((n2 = parse(&uc2, s, len)) > 0) { 2994 uint32_t ucx[FDC_MAX]; 2995 int ccx[FDC_MAX]; 2996 int cl, cx, i, nx, ucx_size; 2997 int LIndex,SIndex; 2998 uint32_t nfc; 2999 3000 if (n2 == spair || always_replace) 3001 /* uc2 is converted from a surrogate pair. 3002 * this should be treated as a changed code. */ 3003 uc2ptr = NULL; 3004 else 3005 uc2ptr = s; 3006 s += n2; 3007 len -= n2; 3008 3009 /* 3010 * If current second code point is out of decomposable 3011 * code points, finding compositions is unneeded. 3012 */ 3013 if (!IS_DECOMPOSABLE_BLOCK(uc2)) { 3014 WRITE_UC(); 3015 REPLACE_UC_WITH_UC2(); 3016 continue; 3017 } 3018 3019 /* 3020 * Try to combine current code points. 3021 */ 3022 /* 3023 * We have to combine Hangul characters according to 3024 * http://uniicode.org/reports/tr15/#Hangul 3025 */ 3026 if (0 <= (LIndex = uc - HC_LBASE) && 3027 LIndex < HC_LCOUNT) { 3028 /* 3029 * Hangul Composition. 3030 * 1. Two current code points are L and V. 3031 */ 3032 int VIndex = uc2 - HC_VBASE; 3033 if (0 <= VIndex && VIndex < HC_VCOUNT) { 3034 /* Make syllable of form LV. */ 3035 UPDATE_UC(HC_SBASE + 3036 (LIndex * HC_VCOUNT + VIndex) * 3037 HC_TCOUNT); 3038 } else { 3039 WRITE_UC(); 3040 REPLACE_UC_WITH_UC2(); 3041 } 3042 continue; 3043 } else if (0 <= (SIndex = uc - HC_SBASE) && 3044 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { 3045 /* 3046 * Hangul Composition. 3047 * 2. Two current code points are LV and T. 3048 */ 3049 int TIndex = uc2 - HC_TBASE; 3050 if (0 < TIndex && TIndex < HC_TCOUNT) { 3051 /* Make syllable of form LVT. */ 3052 UPDATE_UC(uc + TIndex); 3053 } else { 3054 WRITE_UC(); 3055 REPLACE_UC_WITH_UC2(); 3056 } 3057 continue; 3058 } else if ((nfc = get_nfc(uc, uc2)) != 0) { 3059 /* A composition to current code points 3060 * is found. */ 3061 UPDATE_UC(nfc); 3062 continue; 3063 } else if ((cl = CCC(uc2)) == 0) { 3064 /* Clearly 'uc2' the second code point is not 3065 * a decomposable code. */ 3066 WRITE_UC(); 3067 REPLACE_UC_WITH_UC2(); 3068 continue; 3069 } 3070 3071 /* 3072 * Collect following decomposable code points. 3073 */ 3074 cx = 0; 3075 ucx[0] = uc2; 3076 ccx[0] = cl; 3077 COLLECT_CPS(1); 3078 3079 /* 3080 * Find a composed code in the collected code points. 3081 */ 3082 i = 1; 3083 while (i < ucx_size) { 3084 int j; 3085 3086 if ((nfc = get_nfc(uc, ucx[i])) == 0) { 3087 i++; 3088 continue; 3089 } 3090 3091 /* 3092 * nfc is composed of uc and ucx[i]. 3093 */ 3094 UPDATE_UC(nfc); 3095 3096 /* 3097 * Remove ucx[i] by shifting 3098 * following code points. 3099 */ 3100 for (j = i; j+1 < ucx_size; j++) { 3101 ucx[j] = ucx[j+1]; 3102 ccx[j] = ccx[j+1]; 3103 } 3104 ucx_size --; 3105 3106 /* 3107 * Collect following code points blocked 3108 * by ucx[i] the removed code point. 3109 */ 3110 if (ucx_size > 0 && i == ucx_size && 3111 nx > 0 && cx == cl) { 3112 cl = ccx[ucx_size-1]; 3113 COLLECT_CPS(ucx_size); 3114 } 3115 /* 3116 * Restart finding a composed code with 3117 * the updated uc from the top of the 3118 * collected code points. 3119 */ 3120 i = 0; 3121 } 3122 3123 /* 3124 * Apparently the current code points are not 3125 * decomposed characters or already composed. 3126 */ 3127 WRITE_UC(); 3128 for (i = 0; i < ucx_size; i++) 3129 UNPARSE(p, endp, ucx[i]); 3130 3131 /* 3132 * Flush out remaining canonical combining characters. 3133 */ 3134 if (nx > 0 && cx == cl && len > 0) { 3135 while ((nx = parse(&ucx[0], s, len)) 3136 > 0) { 3137 cx = CCC(ucx[0]); 3138 if (cl > cx) 3139 break; 3140 s += nx; 3141 len -= nx; 3142 cl = cx; 3143 UNPARSE(p, endp, ucx[0]); 3144 } 3145 } 3146 break; 3147 } 3148 if (n2 < 0) { 3149 WRITE_UC(); 3150 /* Use a replaced unicode character. */ 3151 UNPARSE(p, endp, uc2); 3152 s += n2*-1; 3153 len -= n2*-1; 3154 ret = -1; 3155 continue; 3156 } else if (n2 == 0) { 3157 WRITE_UC(); 3158 break; 3159 } 3160 } 3161 as->length = p - as->s; 3162 as->s[as->length] = '\0'; 3163 if (ts == 2) 3164 as->s[as->length+1] = '\0'; 3165 return (ret); 3166 } 3167 3168 static int 3169 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) 3170 { 3171 int t, b; 3172 3173 /* 3174 * These are not converted to NFD on Mac OS. 3175 */ 3176 if ((uc >= 0x2000 && uc <= 0x2FFF) || 3177 (uc >= 0xF900 && uc <= 0xFAFF) || 3178 (uc >= 0x2F800 && uc <= 0x2FAFF)) 3179 return (0); 3180 /* 3181 * Those code points are not converted to NFD on Mac OS. 3182 * I do not know the reason because it is undocumented. 3183 * NFC NFD 3184 * 1109A ==> 11099 110BA 3185 * 1109C ==> 1109B 110BA 3186 * 110AB ==> 110A5 110BA 3187 */ 3188 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) 3189 return (0); 3190 3191 t = 0; 3192 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; 3193 while (b >= t) { 3194 int m = (t + b) / 2; 3195 if (u_decomposition_table[m].nfc < uc) 3196 t = m + 1; 3197 else if (u_decomposition_table[m].nfc > uc) 3198 b = m - 1; 3199 else { 3200 *cp1 = u_decomposition_table[m].cp1; 3201 *cp2 = u_decomposition_table[m].cp2; 3202 return (1); 3203 } 3204 } 3205 return (0); 3206 } 3207 3208 #define REPLACE_UC_WITH(cp) do { \ 3209 uc = cp; \ 3210 ucptr = NULL; \ 3211 } while (0) 3212 3213 /* 3214 * Normalize UTF-8 characters to Form D and copy the result. 3215 */ 3216 static int 3217 archive_string_normalize_D(struct archive_string *as, const void *_p, 3218 size_t len, struct archive_string_conv *sc) 3219 { 3220 const char *s = (const char *)_p; 3221 char *p, *endp; 3222 uint32_t uc, uc2; 3223 size_t w; 3224 int always_replace, n, n2, ret = 0, spair, ts, tm; 3225 int (*parse)(uint32_t *, const char *, size_t); 3226 size_t (*unparse)(char *, size_t, uint32_t); 3227 3228 always_replace = 1; 3229 ts = 1;/* text size. */ 3230 if (sc->flag & SCONV_TO_UTF16BE) { 3231 unparse = unicode_to_utf16be; 3232 ts = 2; 3233 if (sc->flag & SCONV_FROM_UTF16BE) 3234 always_replace = 0; 3235 } else if (sc->flag & SCONV_TO_UTF16LE) { 3236 unparse = unicode_to_utf16le; 3237 ts = 2; 3238 if (sc->flag & SCONV_FROM_UTF16LE) 3239 always_replace = 0; 3240 } else if (sc->flag & SCONV_TO_UTF8) { 3241 unparse = unicode_to_utf8; 3242 if (sc->flag & SCONV_FROM_UTF8) 3243 always_replace = 0; 3244 } else { 3245 /* 3246 * This case is going to be converted to another 3247 * character-set through iconv. 3248 */ 3249 always_replace = 0; 3250 if (sc->flag & SCONV_FROM_UTF16BE) { 3251 unparse = unicode_to_utf16be; 3252 ts = 2; 3253 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3254 unparse = unicode_to_utf16le; 3255 ts = 2; 3256 } else { 3257 unparse = unicode_to_utf8; 3258 } 3259 } 3260 3261 if (sc->flag & SCONV_FROM_UTF16BE) { 3262 parse = utf16be_to_unicode; 3263 tm = 1; 3264 spair = 4;/* surrogate pair size in UTF-16. */ 3265 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3266 parse = utf16le_to_unicode; 3267 tm = 1; 3268 spair = 4;/* surrogate pair size in UTF-16. */ 3269 } else { 3270 parse = cesu8_to_unicode; 3271 tm = ts; 3272 spair = 6;/* surrogate pair size in UTF-8. */ 3273 } 3274 3275 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 3276 return (-1); 3277 3278 p = as->s + as->length; 3279 endp = as->s + as->buffer_length - ts; 3280 while ((n = parse(&uc, s, len)) != 0) { 3281 const char *ucptr; 3282 uint32_t cp1, cp2; 3283 int SIndex; 3284 struct { 3285 uint32_t uc; 3286 int ccc; 3287 } fdc[FDC_MAX]; 3288 int fdi, fdj; 3289 int ccc; 3290 3291 check_first_code: 3292 if (n < 0) { 3293 /* Use a replaced unicode character. */ 3294 UNPARSE(p, endp, uc); 3295 s += n*-1; 3296 len -= n*-1; 3297 ret = -1; 3298 continue; 3299 } else if (n == spair || always_replace) 3300 /* uc is converted from a surrogate pair. 3301 * this should be treated as a changed code. */ 3302 ucptr = NULL; 3303 else 3304 ucptr = s; 3305 s += n; 3306 len -= n; 3307 3308 /* Hangul Decomposition. */ 3309 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { 3310 int L = HC_LBASE + SIndex / HC_NCOUNT; 3311 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; 3312 int T = HC_TBASE + SIndex % HC_TCOUNT; 3313 3314 REPLACE_UC_WITH(L); 3315 WRITE_UC(); 3316 REPLACE_UC_WITH(V); 3317 WRITE_UC(); 3318 if (T != HC_TBASE) { 3319 REPLACE_UC_WITH(T); 3320 WRITE_UC(); 3321 } 3322 continue; 3323 } 3324 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { 3325 WRITE_UC(); 3326 continue; 3327 } 3328 3329 fdi = 0; 3330 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { 3331 int k; 3332 3333 for (k = fdi; k > 0; k--) 3334 fdc[k] = fdc[k-1]; 3335 fdc[0].ccc = CCC(cp2); 3336 fdc[0].uc = cp2; 3337 fdi++; 3338 REPLACE_UC_WITH(cp1); 3339 } 3340 3341 /* Read following code points. */ 3342 while ((n2 = parse(&uc2, s, len)) > 0 && 3343 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { 3344 int j, k; 3345 3346 s += n2; 3347 len -= n2; 3348 for (j = 0; j < fdi; j++) { 3349 if (fdc[j].ccc > ccc) 3350 break; 3351 } 3352 if (j < fdi) { 3353 for (k = fdi; k > j; k--) 3354 fdc[k] = fdc[k-1]; 3355 fdc[j].ccc = ccc; 3356 fdc[j].uc = uc2; 3357 } else { 3358 fdc[fdi].ccc = ccc; 3359 fdc[fdi].uc = uc2; 3360 } 3361 fdi++; 3362 } 3363 3364 WRITE_UC(); 3365 for (fdj = 0; fdj < fdi; fdj++) { 3366 REPLACE_UC_WITH(fdc[fdj].uc); 3367 WRITE_UC(); 3368 } 3369 3370 if (n2 == 0) 3371 break; 3372 REPLACE_UC_WITH(uc2); 3373 n = n2; 3374 goto check_first_code; 3375 } 3376 as->length = p - as->s; 3377 as->s[as->length] = '\0'; 3378 if (ts == 2) 3379 as->s[as->length+1] = '\0'; 3380 return (ret); 3381 } 3382 3383 /* 3384 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption 3385 * that WCS is Unicode. It is true for several platforms but some are false. 3386 * And then people who did not use UTF-8 locale on the non Unicode WCS 3387 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those 3388 * now cannot get right filename from libarchive 3.x and later since we 3389 * fixed the wrong assumption and it is incompatible to older its versions. 3390 * So we provide special option, "compat-2x.x", for resolving it. 3391 * That option enable the string conversion of libarchive 2.x. 3392 * 3393 * Translates the wrong UTF-8 string made by libarchive 2.x into current 3394 * locale character set and appends to the archive_string. 3395 * Note: returns -1 if conversion fails. 3396 */ 3397 static int 3398 strncat_from_utf8_libarchive2(struct archive_string *as, 3399 const void *_p, size_t len, struct archive_string_conv *sc) 3400 { 3401 const char *s; 3402 int n; 3403 char *p; 3404 char *end; 3405 uint32_t unicode; 3406 #if HAVE_WCRTOMB 3407 mbstate_t shift_state; 3408 3409 memset(&shift_state, 0, sizeof(shift_state)); 3410 #else 3411 /* Clear the shift state before starting. */ 3412 wctomb(NULL, L'\0'); 3413 #endif 3414 (void)sc; /* UNUSED */ 3415 /* 3416 * Allocate buffer for MBS. 3417 * We need this allocation here since it is possible that 3418 * as->s is still NULL. 3419 */ 3420 if (archive_string_ensure(as, as->length + len + 1) == NULL) 3421 return (-1); 3422 3423 s = (const char *)_p; 3424 p = as->s + as->length; 3425 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3426 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { 3427 wchar_t wc; 3428 3429 if (p >= end) { 3430 as->length = p - as->s; 3431 /* Re-allocate buffer for MBS. */ 3432 if (archive_string_ensure(as, 3433 as->length + len * 2 + 1) == NULL) 3434 return (-1); 3435 p = as->s + as->length; 3436 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3437 } 3438 3439 /* 3440 * As libarchie 2.x, translates the UTF-8 characters into 3441 * wide-characters in the assumption that WCS is Unicode. 3442 */ 3443 if (n < 0) { 3444 n *= -1; 3445 wc = L'?'; 3446 } else 3447 wc = (wchar_t)unicode; 3448 3449 s += n; 3450 len -= n; 3451 /* 3452 * Translates the wide-character into the current locale MBS. 3453 */ 3454 #if HAVE_WCRTOMB 3455 n = (int)wcrtomb(p, wc, &shift_state); 3456 #else 3457 n = (int)wctomb(p, wc); 3458 #endif 3459 if (n == -1) 3460 return (-1); 3461 p += n; 3462 } 3463 as->length = p - as->s; 3464 as->s[as->length] = '\0'; 3465 return (0); 3466 } 3467 3468 3469 /* 3470 * Conversion functions between current locale dependent MBS and UTF-16BE. 3471 * strncat_from_utf16be() : UTF-16BE --> MBS 3472 * strncat_to_utf16be() : MBS --> UTF16BE 3473 */ 3474 3475 #if defined(_WIN32) && !defined(__CYGWIN__) 3476 3477 /* 3478 * Convert a UTF-16BE/LE string to current locale and copy the result. 3479 * Return -1 if conversion fails. 3480 */ 3481 static int 3482 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, 3483 struct archive_string_conv *sc, int be) 3484 { 3485 struct archive_string tmp; 3486 const char *u16; 3487 int ll; 3488 BOOL defchar; 3489 char *mbs; 3490 size_t mbs_size, b; 3491 int ret = 0; 3492 3493 bytes &= ~1; 3494 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3495 return (-1); 3496 3497 mbs = as->s + as->length; 3498 mbs_size = as->buffer_length - as->length -1; 3499 3500 if (sc->to_cp == CP_C_LOCALE) { 3501 /* 3502 * "C" locale special process. 3503 */ 3504 u16 = _p; 3505 ll = 0; 3506 for (b = 0; b < bytes; b += 2) { 3507 uint16_t val; 3508 if (be) 3509 val = archive_be16dec(u16+b); 3510 else 3511 val = archive_le16dec(u16+b); 3512 if (val > 255) { 3513 *mbs++ = '?'; 3514 ret = -1; 3515 } else 3516 *mbs++ = (char)(val&0xff); 3517 ll++; 3518 } 3519 as->length += ll; 3520 as->s[as->length] = '\0'; 3521 return (ret); 3522 } 3523 3524 archive_string_init(&tmp); 3525 if (be) { 3526 if (is_big_endian()) { 3527 u16 = _p; 3528 } else { 3529 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3530 return (-1); 3531 memcpy(tmp.s, _p, bytes); 3532 for (b = 0; b < bytes; b += 2) { 3533 uint16_t val = archive_be16dec(tmp.s+b); 3534 archive_le16enc(tmp.s+b, val); 3535 } 3536 u16 = tmp.s; 3537 } 3538 } else { 3539 if (!is_big_endian()) { 3540 u16 = _p; 3541 } else { 3542 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3543 return (-1); 3544 memcpy(tmp.s, _p, bytes); 3545 for (b = 0; b < bytes; b += 2) { 3546 uint16_t val = archive_le16dec(tmp.s+b); 3547 archive_be16enc(tmp.s+b, val); 3548 } 3549 u16 = tmp.s; 3550 } 3551 } 3552 3553 do { 3554 defchar = 0; 3555 ll = WideCharToMultiByte(sc->to_cp, 0, 3556 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, 3557 NULL, &defchar); 3558 /* Exit loop if we succeeded */ 3559 if (ll != 0 || 3560 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3561 break; 3562 } 3563 /* Else expand buffer and loop to try again. */ 3564 ll = WideCharToMultiByte(sc->to_cp, 0, 3565 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); 3566 if (archive_string_ensure(as, ll +1) == NULL) 3567 return (-1); 3568 mbs = as->s + as->length; 3569 mbs_size = as->buffer_length - as->length -1; 3570 } while (1); 3571 archive_string_free(&tmp); 3572 as->length += ll; 3573 as->s[as->length] = '\0'; 3574 if (ll == 0 || defchar) 3575 ret = -1; 3576 return (ret); 3577 } 3578 3579 static int 3580 win_strncat_from_utf16be(struct archive_string *as, const void *_p, 3581 size_t bytes, struct archive_string_conv *sc) 3582 { 3583 return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); 3584 } 3585 3586 static int 3587 win_strncat_from_utf16le(struct archive_string *as, const void *_p, 3588 size_t bytes, struct archive_string_conv *sc) 3589 { 3590 return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); 3591 } 3592 3593 static int 3594 is_big_endian(void) 3595 { 3596 uint16_t d = 1; 3597 3598 return (archive_be16dec(&d) == 1); 3599 } 3600 3601 /* 3602 * Convert a current locale string to UTF-16BE/LE and copy the result. 3603 * Return -1 if conversion fails. 3604 */ 3605 static int 3606 win_strncat_to_utf16(struct archive_string *as16, const void *_p, 3607 size_t length, struct archive_string_conv *sc, int bigendian) 3608 { 3609 const char *s = (const char *)_p; 3610 char *u16; 3611 size_t count, avail; 3612 3613 if (archive_string_ensure(as16, 3614 as16->length + (length + 1) * 2) == NULL) 3615 return (-1); 3616 3617 u16 = as16->s + as16->length; 3618 avail = as16->buffer_length - 2; 3619 if (sc->from_cp == CP_C_LOCALE) { 3620 /* 3621 * "C" locale special process. 3622 */ 3623 count = 0; 3624 while (count < length && *s) { 3625 if (bigendian) 3626 archive_be16enc(u16, *s); 3627 else 3628 archive_le16enc(u16, *s); 3629 u16 += 2; 3630 s++; 3631 count++; 3632 } 3633 as16->length += count << 1; 3634 as16->s[as16->length] = 0; 3635 as16->s[as16->length+1] = 0; 3636 return (0); 3637 } 3638 do { 3639 count = MultiByteToWideChar(sc->from_cp, 3640 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); 3641 /* Exit loop if we succeeded */ 3642 if (count != 0 || 3643 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3644 break; 3645 } 3646 /* Expand buffer and try again */ 3647 count = MultiByteToWideChar(sc->from_cp, 3648 MB_PRECOMPOSED, s, (int)length, NULL, 0); 3649 if (archive_string_ensure(as16, (count +1) * 2) 3650 == NULL) 3651 return (-1); 3652 u16 = as16->s + as16->length; 3653 avail = as16->buffer_length - 2; 3654 } while (1); 3655 as16->length += count * 2; 3656 as16->s[as16->length] = 0; 3657 as16->s[as16->length+1] = 0; 3658 if (count == 0) 3659 return (-1); 3660 3661 if (is_big_endian()) { 3662 if (!bigendian) { 3663 while (count > 0) { 3664 uint16_t v = archive_be16dec(u16); 3665 archive_le16enc(u16, v); 3666 u16 += 2; 3667 count--; 3668 } 3669 } 3670 } else { 3671 if (bigendian) { 3672 while (count > 0) { 3673 uint16_t v = archive_le16dec(u16); 3674 archive_be16enc(u16, v); 3675 u16 += 2; 3676 count--; 3677 } 3678 } 3679 } 3680 return (0); 3681 } 3682 3683 static int 3684 win_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3685 size_t length, struct archive_string_conv *sc) 3686 { 3687 return (win_strncat_to_utf16(as16, _p, length, sc, 1)); 3688 } 3689 3690 static int 3691 win_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3692 size_t length, struct archive_string_conv *sc) 3693 { 3694 return (win_strncat_to_utf16(as16, _p, length, sc, 0)); 3695 } 3696 3697 #endif /* _WIN32 && !__CYGWIN__ */ 3698 3699 /* 3700 * Do the best effort for conversions. 3701 * We cannot handle UTF-16BE character-set without such iconv, 3702 * but there is a chance if a string consists just ASCII code or 3703 * a current locale is UTF-8. 3704 */ 3705 3706 /* 3707 * Convert a UTF-16BE string to current locale and copy the result. 3708 * Return -1 if conversion fails. 3709 */ 3710 static int 3711 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, 3712 size_t bytes, struct archive_string_conv *sc, int be) 3713 { 3714 const char *utf16 = (const char *)_p; 3715 char *mbs; 3716 uint32_t uc; 3717 int n, ret; 3718 3719 (void)sc; /* UNUSED */ 3720 /* 3721 * Other case, we should do the best effort. 3722 * If all character are ASCII(<0x7f), we can convert it. 3723 * if not , we set a alternative character and return -1. 3724 */ 3725 ret = 0; 3726 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3727 return (-1); 3728 mbs = as->s + as->length; 3729 3730 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { 3731 if (n < 0) { 3732 n *= -1; 3733 ret = -1; 3734 } 3735 bytes -= n; 3736 utf16 += n; 3737 3738 if (uc > 127) { 3739 /* We cannot handle it. */ 3740 *mbs++ = '?'; 3741 ret = -1; 3742 } else 3743 *mbs++ = (char)uc; 3744 } 3745 as->length = mbs - as->s; 3746 as->s[as->length] = '\0'; 3747 return (ret); 3748 } 3749 3750 static int 3751 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, 3752 size_t bytes, struct archive_string_conv *sc) 3753 { 3754 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); 3755 } 3756 3757 static int 3758 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, 3759 size_t bytes, struct archive_string_conv *sc) 3760 { 3761 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); 3762 } 3763 3764 /* 3765 * Convert a current locale string to UTF-16BE/LE and copy the result. 3766 * Return -1 if conversion fails. 3767 */ 3768 static int 3769 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, 3770 size_t length, struct archive_string_conv *sc, int bigendian) 3771 { 3772 const char *s = (const char *)_p; 3773 char *utf16; 3774 size_t remaining; 3775 int ret; 3776 3777 (void)sc; /* UNUSED */ 3778 /* 3779 * Other case, we should do the best effort. 3780 * If all character are ASCII(<0x7f), we can convert it. 3781 * if not , we set a alternative character and return -1. 3782 */ 3783 ret = 0; 3784 remaining = length; 3785 3786 if (archive_string_ensure(as16, 3787 as16->length + (length + 1) * 2) == NULL) 3788 return (-1); 3789 3790 utf16 = as16->s + as16->length; 3791 while (remaining--) { 3792 unsigned c = *s++; 3793 if (c > 127) { 3794 /* We cannot handle it. */ 3795 c = UNICODE_R_CHAR; 3796 ret = -1; 3797 } 3798 if (bigendian) 3799 archive_be16enc(utf16, c); 3800 else 3801 archive_le16enc(utf16, c); 3802 utf16 += 2; 3803 } 3804 as16->length = utf16 - as16->s; 3805 as16->s[as16->length] = 0; 3806 as16->s[as16->length+1] = 0; 3807 return (ret); 3808 } 3809 3810 static int 3811 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3812 size_t length, struct archive_string_conv *sc) 3813 { 3814 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); 3815 } 3816 3817 static int 3818 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3819 size_t length, struct archive_string_conv *sc) 3820 { 3821 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); 3822 } 3823 3824 3825 /* 3826 * Multistring operations. 3827 */ 3828 3829 void 3830 archive_mstring_clean(struct archive_mstring *aes) 3831 { 3832 archive_wstring_free(&(aes->aes_wcs)); 3833 archive_string_free(&(aes->aes_mbs)); 3834 archive_string_free(&(aes->aes_utf8)); 3835 archive_string_free(&(aes->aes_mbs_in_locale)); 3836 aes->aes_set = 0; 3837 } 3838 3839 void 3840 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) 3841 { 3842 dest->aes_set = src->aes_set; 3843 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); 3844 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); 3845 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); 3846 } 3847 3848 int 3849 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, 3850 const char **p) 3851 { 3852 struct archive_string_conv *sc; 3853 int r; 3854 3855 /* If we already have a UTF8 form, return that immediately. */ 3856 if (aes->aes_set & AES_SET_UTF8) { 3857 *p = aes->aes_utf8.s; 3858 return (0); 3859 } 3860 3861 *p = NULL; 3862 if (aes->aes_set & AES_SET_MBS) { 3863 sc = archive_string_conversion_to_charset(a, "UTF-8", 1); 3864 if (sc == NULL) 3865 return (-1);/* Couldn't allocate memory for sc. */ 3866 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, 3867 aes->aes_mbs.length, sc); 3868 if (a == NULL) 3869 free_sconv_object(sc); 3870 if (r == 0) { 3871 aes->aes_set |= AES_SET_UTF8; 3872 *p = aes->aes_utf8.s; 3873 return (0);/* success. */ 3874 } else 3875 return (-1);/* failure. */ 3876 } 3877 return (0);/* success. */ 3878 } 3879 3880 int 3881 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, 3882 const char **p) 3883 { 3884 int r, ret = 0; 3885 3886 (void)a; /* UNUSED */ 3887 /* If we already have an MBS form, return that immediately. */ 3888 if (aes->aes_set & AES_SET_MBS) { 3889 *p = aes->aes_mbs.s; 3890 return (ret); 3891 } 3892 3893 *p = NULL; 3894 /* If there's a WCS form, try converting with the native locale. */ 3895 if (aes->aes_set & AES_SET_WCS) { 3896 archive_string_empty(&(aes->aes_mbs)); 3897 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3898 aes->aes_wcs.s, aes->aes_wcs.length); 3899 *p = aes->aes_mbs.s; 3900 if (r == 0) { 3901 aes->aes_set |= AES_SET_MBS; 3902 return (ret); 3903 } else 3904 ret = -1; 3905 } 3906 3907 /* 3908 * Only a UTF-8 form cannot avail because its conversion already 3909 * failed at archive_mstring_update_utf8(). 3910 */ 3911 return (ret); 3912 } 3913 3914 int 3915 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, 3916 const wchar_t **wp) 3917 { 3918 int r, ret = 0; 3919 3920 (void)a;/* UNUSED */ 3921 /* Return WCS form if we already have it. */ 3922 if (aes->aes_set & AES_SET_WCS) { 3923 *wp = aes->aes_wcs.s; 3924 return (ret); 3925 } 3926 3927 *wp = NULL; 3928 /* Try converting MBS to WCS using native locale. */ 3929 if (aes->aes_set & AES_SET_MBS) { 3930 archive_wstring_empty(&(aes->aes_wcs)); 3931 r = archive_wstring_append_from_mbs(&(aes->aes_wcs), 3932 aes->aes_mbs.s, aes->aes_mbs.length); 3933 if (r == 0) { 3934 aes->aes_set |= AES_SET_WCS; 3935 *wp = aes->aes_wcs.s; 3936 } else 3937 ret = -1;/* failure. */ 3938 } 3939 return (ret); 3940 } 3941 3942 int 3943 archive_mstring_get_mbs_l(struct archive_mstring *aes, 3944 const char **p, size_t *length, struct archive_string_conv *sc) 3945 { 3946 int r, ret = 0; 3947 3948 #if defined(_WIN32) && !defined(__CYGWIN__) 3949 /* 3950 * Internationalization programing on Windows must use Wide 3951 * characters because Windows platform cannot make locale UTF-8. 3952 */ 3953 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { 3954 archive_string_empty(&(aes->aes_mbs_in_locale)); 3955 r = archive_string_append_from_wcs_in_codepage( 3956 &(aes->aes_mbs_in_locale), aes->aes_wcs.s, 3957 aes->aes_wcs.length, sc); 3958 if (r == 0) { 3959 *p = aes->aes_mbs_in_locale.s; 3960 if (length != NULL) 3961 *length = aes->aes_mbs_in_locale.length; 3962 return (0); 3963 } else if (errno == ENOMEM) 3964 return (-1); 3965 else 3966 ret = -1; 3967 } 3968 #endif 3969 3970 /* If there is not an MBS form but is a WCS form, try converting 3971 * with the native locale to be used for translating it to specified 3972 * character-set. */ 3973 if ((aes->aes_set & AES_SET_MBS) == 0 && 3974 (aes->aes_set & AES_SET_WCS) != 0) { 3975 archive_string_empty(&(aes->aes_mbs)); 3976 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3977 aes->aes_wcs.s, aes->aes_wcs.length); 3978 if (r == 0) 3979 aes->aes_set |= AES_SET_MBS; 3980 else if (errno == ENOMEM) 3981 return (-1); 3982 else 3983 ret = -1; 3984 } 3985 /* If we already have an MBS form, use it to be translated to 3986 * specified character-set. */ 3987 if (aes->aes_set & AES_SET_MBS) { 3988 if (sc == NULL) { 3989 /* Conversion is unneeded. */ 3990 *p = aes->aes_mbs.s; 3991 if (length != NULL) 3992 *length = aes->aes_mbs.length; 3993 return (0); 3994 } 3995 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), 3996 aes->aes_mbs.s, aes->aes_mbs.length, sc); 3997 *p = aes->aes_mbs_in_locale.s; 3998 if (length != NULL) 3999 *length = aes->aes_mbs_in_locale.length; 4000 } else { 4001 *p = NULL; 4002 if (length != NULL) 4003 *length = 0; 4004 } 4005 return (ret); 4006 } 4007 4008 int 4009 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) 4010 { 4011 if (mbs == NULL) { 4012 aes->aes_set = 0; 4013 return (0); 4014 } 4015 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); 4016 } 4017 4018 int 4019 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, 4020 size_t len) 4021 { 4022 if (mbs == NULL) { 4023 aes->aes_set = 0; 4024 return (0); 4025 } 4026 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4027 archive_strncpy(&(aes->aes_mbs), mbs, len); 4028 archive_string_empty(&(aes->aes_utf8)); 4029 archive_wstring_empty(&(aes->aes_wcs)); 4030 return (0); 4031 } 4032 4033 int 4034 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) 4035 { 4036 return archive_mstring_copy_wcs_len(aes, wcs, 4037 wcs == NULL ? 0 : wcslen(wcs)); 4038 } 4039 4040 int 4041 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) 4042 { 4043 if (utf8 == NULL) { 4044 aes->aes_set = 0; 4045 } 4046 aes->aes_set = AES_SET_UTF8; 4047 archive_string_empty(&(aes->aes_mbs)); 4048 archive_string_empty(&(aes->aes_wcs)); 4049 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); 4050 return (int)strlen(utf8); 4051 } 4052 4053 int 4054 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, 4055 size_t len) 4056 { 4057 if (wcs == NULL) { 4058 aes->aes_set = 0; 4059 } 4060 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ 4061 archive_string_empty(&(aes->aes_mbs)); 4062 archive_string_empty(&(aes->aes_utf8)); 4063 archive_wstrncpy(&(aes->aes_wcs), wcs, len); 4064 return (0); 4065 } 4066 4067 int 4068 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, 4069 const char *mbs, size_t len, struct archive_string_conv *sc) 4070 { 4071 int r; 4072 4073 if (mbs == NULL) { 4074 aes->aes_set = 0; 4075 return (0); 4076 } 4077 archive_string_empty(&(aes->aes_mbs)); 4078 archive_wstring_empty(&(aes->aes_wcs)); 4079 archive_string_empty(&(aes->aes_utf8)); 4080 #if defined(_WIN32) && !defined(__CYGWIN__) 4081 /* 4082 * Internationalization programing on Windows must use Wide 4083 * characters because Windows platform cannot make locale UTF-8. 4084 */ 4085 if (sc == NULL) { 4086 if (archive_string_append(&(aes->aes_mbs), 4087 mbs, mbsnbytes(mbs, len)) == NULL) { 4088 aes->aes_set = 0; 4089 r = -1; 4090 } else { 4091 aes->aes_set = AES_SET_MBS; 4092 r = 0; 4093 } 4094 #if defined(HAVE_ICONV) 4095 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { 4096 /* 4097 * This case happens only when MultiByteToWideChar() cannot 4098 * handle sc->from_cp, and we have to iconv in order to 4099 * translate character-set to wchar_t,UTF-16. 4100 */ 4101 iconv_t cd = sc->cd; 4102 unsigned from_cp; 4103 int flag; 4104 4105 /* 4106 * Translate multi-bytes from some character-set to UTF-8. 4107 */ 4108 sc->cd = sc->cd_w; 4109 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); 4110 sc->cd = cd; 4111 if (r != 0) { 4112 aes->aes_set = 0; 4113 return (r); 4114 } 4115 aes->aes_set = AES_SET_UTF8; 4116 4117 /* 4118 * Append the UTF-8 string into wstring. 4119 */ 4120 flag = sc->flag; 4121 sc->flag &= ~(SCONV_NORMALIZATION_C 4122 | SCONV_TO_UTF16| SCONV_FROM_UTF16); 4123 from_cp = sc->from_cp; 4124 sc->from_cp = CP_UTF8; 4125 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), 4126 aes->aes_utf8.s, aes->aes_utf8.length, sc); 4127 sc->flag = flag; 4128 sc->from_cp = from_cp; 4129 if (r == 0) 4130 aes->aes_set |= AES_SET_WCS; 4131 #endif 4132 } else { 4133 r = archive_wstring_append_from_mbs_in_codepage( 4134 &(aes->aes_wcs), mbs, len, sc); 4135 if (r == 0) 4136 aes->aes_set = AES_SET_WCS; 4137 else 4138 aes->aes_set = 0; 4139 } 4140 #else 4141 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); 4142 if (r == 0) 4143 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4144 else 4145 aes->aes_set = 0; 4146 #endif 4147 return (r); 4148 } 4149 4150 /* 4151 * The 'update' form tries to proactively update all forms of 4152 * this string (WCS and MBS) and returns an error if any of 4153 * them fail. This is used by the 'pax' handler, for instance, 4154 * to detect and report character-conversion failures early while 4155 * still allowing clients to get potentially useful values from 4156 * the more tolerant lazy conversions. (get_mbs and get_wcs will 4157 * strive to give the user something useful, so you can get hopefully 4158 * usable values even if some of the character conversions are failing.) 4159 */ 4160 int 4161 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, 4162 const char *utf8) 4163 { 4164 struct archive_string_conv *sc; 4165 int r; 4166 4167 if (utf8 == NULL) { 4168 aes->aes_set = 0; 4169 return (0); /* Succeeded in clearing everything. */ 4170 } 4171 4172 /* Save the UTF8 string. */ 4173 archive_strcpy(&(aes->aes_utf8), utf8); 4174 4175 /* Empty the mbs and wcs strings. */ 4176 archive_string_empty(&(aes->aes_mbs)); 4177 archive_wstring_empty(&(aes->aes_wcs)); 4178 4179 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ 4180 4181 /* Try converting UTF-8 to MBS, return false on failure. */ 4182 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 4183 if (sc == NULL) 4184 return (-1);/* Couldn't allocate memory for sc. */ 4185 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); 4186 if (a == NULL) 4187 free_sconv_object(sc); 4188 if (r != 0) 4189 return (-1); 4190 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ 4191 4192 /* Try converting MBS to WCS, return false on failure. */ 4193 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, 4194 aes->aes_mbs.length)) 4195 return (-1); 4196 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; 4197 4198 /* All conversions succeeded. */ 4199 return (0); 4200 } 4201