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 #undef max 79 #define max(a, b) ((a)>(b)?(a):(b)) 80 81 struct archive_string_conv { 82 struct archive_string_conv *next; 83 char *from_charset; 84 char *to_charset; 85 unsigned from_cp; 86 unsigned to_cp; 87 /* Set 1 if from_charset and to_charset are the same. */ 88 int same; 89 int flag; 90 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified 91 * charset. */ 92 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from 93 * specified charset. */ 94 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */ 95 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting 96 * MBS. */ 97 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive 98 * 2.x in the wrong assumption. */ 99 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C. 100 * Before UTF-8 characters are actually 101 * processed. */ 102 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D. 103 * Before UTF-8 characters are actually 104 * processed. 105 * Currently this only for MAC OS X. */ 106 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */ 107 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */ 108 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */ 109 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */ 110 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */ 111 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */ 112 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE) 113 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE) 114 115 #if HAVE_ICONV 116 iconv_t cd; 117 iconv_t cd_w;/* Use at archive_mstring on 118 * Windows. */ 119 #endif 120 /* A temporary buffer for normalization. */ 121 struct archive_string utftmp; 122 int (*converter[2])(struct archive_string *, const void *, size_t, 123 struct archive_string_conv *); 124 int nconverter; 125 }; 126 127 #define CP_C_LOCALE 0 /* "C" locale only for this file. */ 128 #define CP_UTF16LE 1200 129 #define CP_UTF16BE 1201 130 131 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF) 132 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF) 133 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF) 134 #define UNICODE_MAX 0x10FFFF 135 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */ 136 /* Set U+FFFD(Replacement character) in UTF-8. */ 137 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd}; 138 139 static struct archive_string_conv *find_sconv_object(struct archive *, 140 const char *, const char *); 141 static void add_sconv_object(struct archive *, struct archive_string_conv *); 142 static struct archive_string_conv *create_sconv_object(const char *, 143 const char *, unsigned, int); 144 static void free_sconv_object(struct archive_string_conv *); 145 static struct archive_string_conv *get_sconv_object(struct archive *, 146 const char *, const char *, int); 147 static unsigned make_codepage_from_charset(const char *); 148 static unsigned get_current_codepage(void); 149 static unsigned get_current_oemcp(void); 150 static size_t mbsnbytes(const void *, size_t); 151 static size_t utf16nbytes(const void *, size_t); 152 #if defined(_WIN32) && !defined(__CYGWIN__) 153 static int archive_wstring_append_from_mbs_in_codepage( 154 struct archive_wstring *, const char *, size_t, 155 struct archive_string_conv *); 156 static int archive_string_append_from_wcs_in_codepage(struct archive_string *, 157 const wchar_t *, size_t, struct archive_string_conv *); 158 static int is_big_endian(void); 159 static int strncat_in_codepage(struct archive_string *, const void *, 160 size_t, struct archive_string_conv *); 161 static int win_strncat_from_utf16be(struct archive_string *, const void *, 162 size_t, struct archive_string_conv *); 163 static int win_strncat_from_utf16le(struct archive_string *, const void *, 164 size_t, struct archive_string_conv *); 165 static int win_strncat_to_utf16be(struct archive_string *, const void *, 166 size_t, struct archive_string_conv *); 167 static int win_strncat_to_utf16le(struct archive_string *, const void *, 168 size_t, struct archive_string_conv *); 169 #endif 170 static int best_effort_strncat_from_utf16be(struct archive_string *, 171 const void *, size_t, struct archive_string_conv *); 172 static int best_effort_strncat_from_utf16le(struct archive_string *, 173 const void *, size_t, struct archive_string_conv *); 174 static int best_effort_strncat_to_utf16be(struct archive_string *, 175 const void *, size_t, struct archive_string_conv *); 176 static int best_effort_strncat_to_utf16le(struct archive_string *, 177 const void *, size_t, struct archive_string_conv *); 178 #if defined(HAVE_ICONV) 179 static int iconv_strncat_in_locale(struct archive_string *, const void *, 180 size_t, struct archive_string_conv *); 181 #endif 182 static int best_effort_strncat_in_locale(struct archive_string *, 183 const void *, size_t, struct archive_string_conv *); 184 static int _utf8_to_unicode(uint32_t *, const char *, size_t); 185 static int utf8_to_unicode(uint32_t *, const char *, size_t); 186 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t); 187 static int cesu8_to_unicode(uint32_t *, const char *, size_t); 188 static size_t unicode_to_utf8(char *, size_t, uint32_t); 189 static int utf16_to_unicode(uint32_t *, const char *, size_t, int); 190 static size_t unicode_to_utf16be(char *, size_t, uint32_t); 191 static size_t unicode_to_utf16le(char *, size_t, uint32_t); 192 static int strncat_from_utf8_libarchive2(struct archive_string *, 193 const void *, size_t, struct archive_string_conv *); 194 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *, 195 size_t, struct archive_string_conv *); 196 static int archive_string_normalize_C(struct archive_string *, const void *, 197 size_t, struct archive_string_conv *); 198 static int archive_string_normalize_D(struct archive_string *, const void *, 199 size_t, struct archive_string_conv *); 200 static int archive_string_append_unicode(struct archive_string *, 201 const void *, size_t, struct archive_string_conv *); 202 203 static struct archive_string * 204 archive_string_append(struct archive_string *as, const char *p, size_t s) 205 { 206 if (archive_string_ensure(as, as->length + s + 1) == NULL) 207 return (NULL); 208 if (s) 209 memmove(as->s + as->length, p, s); 210 as->length += s; 211 as->s[as->length] = 0; 212 return (as); 213 } 214 215 static struct archive_wstring * 216 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s) 217 { 218 if (archive_wstring_ensure(as, as->length + s + 1) == NULL) 219 return (NULL); 220 if (s) 221 wmemmove(as->s + as->length, p, s); 222 as->length += s; 223 as->s[as->length] = 0; 224 return (as); 225 } 226 227 struct archive_string * 228 archive_array_append(struct archive_string *as, const char *p, size_t s) 229 { 230 return archive_string_append(as, p, s); 231 } 232 233 void 234 archive_string_concat(struct archive_string *dest, struct archive_string *src) 235 { 236 if (archive_string_append(dest, src->s, src->length) == NULL) 237 __archive_errx(1, "Out of memory"); 238 } 239 240 void 241 archive_wstring_concat(struct archive_wstring *dest, 242 struct archive_wstring *src) 243 { 244 if (archive_wstring_append(dest, src->s, src->length) == NULL) 245 __archive_errx(1, "Out of memory"); 246 } 247 248 void 249 archive_string_free(struct archive_string *as) 250 { 251 as->length = 0; 252 as->buffer_length = 0; 253 free(as->s); 254 as->s = NULL; 255 } 256 257 void 258 archive_wstring_free(struct archive_wstring *as) 259 { 260 as->length = 0; 261 as->buffer_length = 0; 262 free(as->s); 263 as->s = NULL; 264 } 265 266 struct archive_wstring * 267 archive_wstring_ensure(struct archive_wstring *as, size_t s) 268 { 269 return (struct archive_wstring *) 270 archive_string_ensure((struct archive_string *)as, 271 s * sizeof(wchar_t)); 272 } 273 274 /* Returns NULL on any allocation failure. */ 275 struct archive_string * 276 archive_string_ensure(struct archive_string *as, size_t s) 277 { 278 char *p; 279 size_t new_length; 280 281 /* If buffer is already big enough, don't reallocate. */ 282 if (as->s && (s <= as->buffer_length)) 283 return (as); 284 285 /* 286 * Growing the buffer at least exponentially ensures that 287 * append operations are always linear in the number of 288 * characters appended. Using a smaller growth rate for 289 * larger buffers reduces memory waste somewhat at the cost of 290 * a larger constant factor. 291 */ 292 if (as->buffer_length < 32) 293 /* Start with a minimum 32-character buffer. */ 294 new_length = 32; 295 else if (as->buffer_length < 8192) 296 /* Buffers under 8k are doubled for speed. */ 297 new_length = as->buffer_length + as->buffer_length; 298 else { 299 /* Buffers 8k and over grow by at least 25% each time. */ 300 new_length = as->buffer_length + as->buffer_length / 4; 301 /* Be safe: If size wraps, fail. */ 302 if (new_length < as->buffer_length) { 303 /* On failure, wipe the string and return NULL. */ 304 archive_string_free(as); 305 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 306 return (NULL); 307 } 308 } 309 /* 310 * The computation above is a lower limit to how much we'll 311 * grow the buffer. In any case, we have to grow it enough to 312 * hold the request. 313 */ 314 if (new_length < s) 315 new_length = s; 316 /* Now we can reallocate the buffer. */ 317 p = (char *)realloc(as->s, new_length); 318 if (p == NULL) { 319 /* On failure, wipe the string and return NULL. */ 320 archive_string_free(as); 321 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 322 return (NULL); 323 } 324 325 as->s = p; 326 as->buffer_length = new_length; 327 return (as); 328 } 329 330 /* 331 * TODO: See if there's a way to avoid scanning 332 * the source string twice. Then test to see 333 * if it actually helps (remember that we're almost 334 * always called with pretty short arguments, so 335 * such an optimization might not help). 336 */ 337 struct archive_string * 338 archive_strncat(struct archive_string *as, const void *_p, size_t n) 339 { 340 size_t s; 341 const char *p, *pp; 342 343 p = (const char *)_p; 344 345 /* Like strlen(p), except won't examine positions beyond p[n]. */ 346 s = 0; 347 pp = p; 348 while (s < n && *pp) { 349 pp++; 350 s++; 351 } 352 if ((as = archive_string_append(as, p, s)) == NULL) 353 __archive_errx(1, "Out of memory"); 354 return (as); 355 } 356 357 struct archive_wstring * 358 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n) 359 { 360 size_t s; 361 const wchar_t *pp; 362 363 /* Like strlen(p), except won't examine positions beyond p[n]. */ 364 s = 0; 365 pp = p; 366 while (s < n && *pp) { 367 pp++; 368 s++; 369 } 370 if ((as = archive_wstring_append(as, p, s)) == NULL) 371 __archive_errx(1, "Out of memory"); 372 return (as); 373 } 374 375 struct archive_string * 376 archive_strcat(struct archive_string *as, const void *p) 377 { 378 /* strcat is just strncat without an effective limit. 379 * Assert that we'll never get called with a source 380 * string over 16MB. 381 * TODO: Review all uses of strcat in the source 382 * and try to replace them with strncat(). 383 */ 384 return archive_strncat(as, p, 0x1000000); 385 } 386 387 struct archive_wstring * 388 archive_wstrcat(struct archive_wstring *as, const wchar_t *p) 389 { 390 /* Ditto. */ 391 return archive_wstrncat(as, p, 0x1000000); 392 } 393 394 struct archive_string * 395 archive_strappend_char(struct archive_string *as, char c) 396 { 397 if ((as = archive_string_append(as, &c, 1)) == NULL) 398 __archive_errx(1, "Out of memory"); 399 return (as); 400 } 401 402 struct archive_wstring * 403 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c) 404 { 405 if ((as = archive_wstring_append(as, &c, 1)) == NULL) 406 __archive_errx(1, "Out of memory"); 407 return (as); 408 } 409 410 /* 411 * Get the "current character set" name to use with iconv. 412 * On FreeBSD, the empty character set name "" chooses 413 * the correct character encoding for the current locale, 414 * so this isn't necessary. 415 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly; 416 * on that system, we have to explicitly call nl_langinfo() 417 * to get the right name. Not sure about other platforms. 418 * 419 * NOTE: GNU libiconv does not recognize the character-set name 420 * which some platform nl_langinfo(CODESET) returns, so we should 421 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv. 422 */ 423 static const char * 424 default_iconv_charset(const char *charset) { 425 if (charset != NULL && charset[0] != '\0') 426 return charset; 427 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__) 428 /* locale_charset() is broken on Mac OS */ 429 return locale_charset(); 430 #elif HAVE_NL_LANGINFO 431 return nl_langinfo(CODESET); 432 #else 433 return ""; 434 #endif 435 } 436 437 #if defined(_WIN32) && !defined(__CYGWIN__) 438 439 /* 440 * Convert MBS to WCS. 441 * Note: returns -1 if conversion fails. 442 */ 443 int 444 archive_wstring_append_from_mbs(struct archive_wstring *dest, 445 const char *p, size_t len) 446 { 447 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL); 448 } 449 450 static int 451 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest, 452 const char *s, size_t length, struct archive_string_conv *sc) 453 { 454 int count, ret = 0; 455 UINT from_cp; 456 457 if (sc != NULL) 458 from_cp = sc->from_cp; 459 else 460 from_cp = get_current_codepage(); 461 462 if (from_cp == CP_C_LOCALE) { 463 /* 464 * "C" locale special processing. 465 */ 466 wchar_t *ws; 467 const unsigned char *mp; 468 469 if (NULL == archive_wstring_ensure(dest, 470 dest->length + length + 1)) 471 return (-1); 472 473 ws = dest->s + dest->length; 474 mp = (const unsigned char *)s; 475 count = 0; 476 while (count < (int)length && *mp) { 477 *ws++ = (wchar_t)*mp++; 478 count++; 479 } 480 } else if (sc != NULL && 481 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) { 482 /* 483 * Normalize UTF-8 and UTF-16BE and convert it directly 484 * to UTF-16 as wchar_t. 485 */ 486 struct archive_string u16; 487 int saved_flag = sc->flag;/* save current flag. */ 488 489 if (is_big_endian()) 490 sc->flag |= SCONV_TO_UTF16BE; 491 else 492 sc->flag |= SCONV_TO_UTF16LE; 493 494 if (sc->flag & SCONV_FROM_UTF16) { 495 /* 496 * UTF-16BE/LE NFD ===> UTF-16 NFC 497 * UTF-16BE/LE NFC ===> UTF-16 NFD 498 */ 499 count = (int)utf16nbytes(s, length); 500 } else { 501 /* 502 * UTF-8 NFD ===> UTF-16 NFC 503 * UTF-8 NFC ===> UTF-16 NFD 504 */ 505 count = (int)mbsnbytes(s, length); 506 } 507 u16.s = (char *)dest->s; 508 u16.length = dest->length << 1;; 509 u16.buffer_length = dest->buffer_length; 510 if (sc->flag & SCONV_NORMALIZATION_C) 511 ret = archive_string_normalize_C(&u16, s, count, sc); 512 else 513 ret = archive_string_normalize_D(&u16, s, count, sc); 514 dest->s = (wchar_t *)u16.s; 515 dest->length = u16.length >> 1; 516 dest->buffer_length = u16.buffer_length; 517 sc->flag = saved_flag;/* restore the saved flag. */ 518 return (ret); 519 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) { 520 count = (int)utf16nbytes(s, length); 521 count >>= 1; /* to be WCS length */ 522 /* Allocate memory for WCS. */ 523 if (NULL == archive_wstring_ensure(dest, 524 dest->length + count + 1)) 525 return (-1); 526 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count); 527 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) { 528 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 529 int b; 530 for (b = 0; b < count; b++) { 531 uint16_t val = archive_le16dec(u16+b); 532 archive_be16enc(u16+b, val); 533 } 534 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) { 535 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 536 int b; 537 for (b = 0; b < count; b++) { 538 uint16_t val = archive_be16dec(u16+b); 539 archive_le16enc(u16+b, val); 540 } 541 } 542 } else { 543 DWORD mbflag; 544 size_t buffsize; 545 546 if (sc == NULL) 547 mbflag = 0; 548 else if (sc->flag & SCONV_FROM_CHARSET) { 549 /* Do not trust the length which comes from 550 * an archive file. */ 551 length = mbsnbytes(s, length); 552 mbflag = 0; 553 } else 554 mbflag = MB_PRECOMPOSED; 555 556 buffsize = dest->length + length + 1; 557 do { 558 /* Allocate memory for WCS. */ 559 if (NULL == archive_wstring_ensure(dest, buffsize)) 560 return (-1); 561 /* Convert MBS to WCS. */ 562 count = MultiByteToWideChar(from_cp, 563 mbflag, s, (int)length, dest->s + dest->length, 564 (int)(dest->buffer_length >> 1) -1); 565 if (count == 0 && 566 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 567 /* Expand the WCS buffer. */ 568 buffsize = dest->buffer_length << 1; 569 continue; 570 } 571 if (count == 0 && length != 0) 572 ret = -1; 573 break; 574 } while (1); 575 } 576 dest->length += count; 577 dest->s[dest->length] = L'\0'; 578 return (ret); 579 } 580 581 #else 582 583 /* 584 * Convert MBS to WCS. 585 * Note: returns -1 if conversion fails. 586 */ 587 int 588 archive_wstring_append_from_mbs(struct archive_wstring *dest, 589 const char *p, size_t len) 590 { 591 size_t r; 592 int ret_val = 0; 593 /* 594 * No single byte will be more than one wide character, 595 * so this length estimate will always be big enough. 596 */ 597 // size_t wcs_length = len; 598 size_t mbs_length = len; 599 const char *mbs = p; 600 wchar_t *wcs; 601 #if HAVE_MBRTOWC 602 mbstate_t shift_state; 603 604 memset(&shift_state, 0, sizeof(shift_state)); 605 #endif 606 /* 607 * As we decided to have wcs_length == mbs_length == len 608 * we can use len here instead of wcs_length 609 */ 610 if (NULL == archive_wstring_ensure(dest, dest->length + len + 1)) 611 return (-1); 612 wcs = dest->s + dest->length; 613 /* 614 * We cannot use mbsrtowcs/mbstowcs here because those may convert 615 * extra MBS when strlen(p) > len and one wide character consists of 616 * multi bytes. 617 */ 618 while (*mbs && mbs_length > 0) { 619 /* 620 * The buffer we allocated is always big enough. 621 * Keep this code path in a comment if we decide to choose 622 * smaller wcs_length in the future 623 */ 624 /* 625 if (wcs_length == 0) { 626 dest->length = wcs - dest->s; 627 dest->s[dest->length] = L'\0'; 628 wcs_length = mbs_length; 629 if (NULL == archive_wstring_ensure(dest, 630 dest->length + wcs_length + 1)) 631 return (-1); 632 wcs = dest->s + dest->length; 633 } 634 */ 635 #if HAVE_MBRTOWC 636 r = mbrtowc(wcs, mbs, mbs_length, &shift_state); 637 #else 638 r = mbtowc(wcs, mbs, mbs_length); 639 #endif 640 if (r == (size_t)-1 || r == (size_t)-2) { 641 ret_val = -1; 642 break; 643 } 644 if (r == 0 || r > mbs_length) 645 break; 646 wcs++; 647 // wcs_length--; 648 mbs += r; 649 mbs_length -= r; 650 } 651 dest->length = wcs - dest->s; 652 dest->s[dest->length] = L'\0'; 653 return (ret_val); 654 } 655 656 #endif 657 658 #if defined(_WIN32) && !defined(__CYGWIN__) 659 660 /* 661 * WCS ==> MBS. 662 * Note: returns -1 if conversion fails. 663 * 664 * Win32 builds use WideCharToMultiByte from the Windows API. 665 * (Maybe Cygwin should too? WideCharToMultiByte will know a 666 * lot more about local character encodings than the wcrtomb() 667 * wrapper is going to know.) 668 */ 669 int 670 archive_string_append_from_wcs(struct archive_string *as, 671 const wchar_t *w, size_t len) 672 { 673 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL); 674 } 675 676 static int 677 archive_string_append_from_wcs_in_codepage(struct archive_string *as, 678 const wchar_t *ws, size_t len, struct archive_string_conv *sc) 679 { 680 BOOL defchar_used, *dp; 681 int count, ret = 0; 682 UINT to_cp; 683 int wslen = (int)len; 684 685 if (sc != NULL) 686 to_cp = sc->to_cp; 687 else 688 to_cp = get_current_codepage(); 689 690 if (to_cp == CP_C_LOCALE) { 691 /* 692 * "C" locale special processing. 693 */ 694 const wchar_t *wp = ws; 695 char *p; 696 697 if (NULL == archive_string_ensure(as, 698 as->length + wslen +1)) 699 return (-1); 700 p = as->s + as->length; 701 count = 0; 702 defchar_used = 0; 703 while (count < wslen && *wp) { 704 if (*wp > 255) { 705 *p++ = '?'; 706 wp++; 707 defchar_used = 1; 708 } else 709 *p++ = (char)*wp++; 710 count++; 711 } 712 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) { 713 uint16_t *u16; 714 715 if (NULL == 716 archive_string_ensure(as, as->length + len * 2 + 2)) 717 return (-1); 718 u16 = (uint16_t *)(as->s + as->length); 719 count = 0; 720 defchar_used = 0; 721 if (sc->flag & SCONV_TO_UTF16BE) { 722 while (count < (int)len && *ws) { 723 archive_be16enc(u16+count, *ws); 724 ws++; 725 count++; 726 } 727 } else { 728 while (count < (int)len && *ws) { 729 archive_le16enc(u16+count, *ws); 730 ws++; 731 count++; 732 } 733 } 734 count <<= 1; /* to be byte size */ 735 } else { 736 /* Make sure the MBS buffer has plenty to set. */ 737 if (NULL == 738 archive_string_ensure(as, as->length + len * 2 + 1)) 739 return (-1); 740 do { 741 defchar_used = 0; 742 if (to_cp == CP_UTF8 || sc == NULL) 743 dp = NULL; 744 else 745 dp = &defchar_used; 746 count = WideCharToMultiByte(to_cp, 0, ws, wslen, 747 as->s + as->length, (int)as->buffer_length-1, NULL, dp); 748 if (count == 0 && 749 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 750 /* Expand the MBS buffer and retry. */ 751 if (NULL == archive_string_ensure(as, 752 as->buffer_length + len)) 753 return (-1); 754 continue; 755 } 756 if (count == 0) 757 ret = -1; 758 break; 759 } while (1); 760 } 761 as->length += count; 762 as->s[as->length] = '\0'; 763 return (defchar_used?-1:ret); 764 } 765 766 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) 767 768 /* 769 * Translates a wide character string into current locale character set 770 * and appends to the archive_string. Note: returns -1 if conversion 771 * fails. 772 */ 773 int 774 archive_string_append_from_wcs(struct archive_string *as, 775 const wchar_t *w, size_t len) 776 { 777 /* We cannot use the standard wcstombs() here because it 778 * cannot tell us how big the output buffer should be. So 779 * I've built a loop around wcrtomb() or wctomb() that 780 * converts a character at a time and resizes the string as 781 * needed. We prefer wcrtomb() when it's available because 782 * it's thread-safe. */ 783 int n, ret_val = 0; 784 char *p; 785 char *end; 786 #if HAVE_WCRTOMB 787 mbstate_t shift_state; 788 789 memset(&shift_state, 0, sizeof(shift_state)); 790 #else 791 /* Clear the shift state before starting. */ 792 wctomb(NULL, L'\0'); 793 #endif 794 /* 795 * Allocate buffer for MBS. 796 * We need this allocation here since it is possible that 797 * as->s is still NULL. 798 */ 799 if (archive_string_ensure(as, as->length + len + 1) == NULL) 800 return (-1); 801 802 p = as->s + as->length; 803 end = as->s + as->buffer_length - MB_CUR_MAX -1; 804 while (*w != L'\0' && len > 0) { 805 if (p >= end) { 806 as->length = p - as->s; 807 as->s[as->length] = '\0'; 808 /* Re-allocate buffer for MBS. */ 809 if (archive_string_ensure(as, 810 as->length + max(len * 2, 811 (size_t)MB_CUR_MAX) + 1) == NULL) 812 return (-1); 813 p = as->s + as->length; 814 end = as->s + as->buffer_length - MB_CUR_MAX -1; 815 } 816 #if HAVE_WCRTOMB 817 n = wcrtomb(p, *w++, &shift_state); 818 #else 819 n = wctomb(p, *w++); 820 #endif 821 if (n == -1) { 822 if (errno == EILSEQ) { 823 /* Skip an illegal wide char. */ 824 *p++ = '?'; 825 ret_val = -1; 826 } else { 827 ret_val = -1; 828 break; 829 } 830 } else 831 p += n; 832 len--; 833 } 834 as->length = p - as->s; 835 as->s[as->length] = '\0'; 836 return (ret_val); 837 } 838 839 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */ 840 841 /* 842 * TODO: Test if __STDC_ISO_10646__ is defined. 843 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion 844 * one character at a time. If a non-Windows platform doesn't have 845 * either of these, fall back to the built-in UTF8 conversion. 846 */ 847 int 848 archive_string_append_from_wcs(struct archive_string *as, 849 const wchar_t *w, size_t len) 850 { 851 (void)as;/* UNUSED */ 852 (void)w;/* UNUSED */ 853 (void)len;/* UNUSED */ 854 errno = ENOSYS; 855 return (-1); 856 } 857 858 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ 859 860 /* 861 * Find a string conversion object by a pair of 'from' charset name 862 * and 'to' charset name from an archive object. 863 * Return NULL if not found. 864 */ 865 static struct archive_string_conv * 866 find_sconv_object(struct archive *a, const char *fc, const char *tc) 867 { 868 struct archive_string_conv *sc; 869 870 if (a == NULL) 871 return (NULL); 872 873 for (sc = a->sconv; sc != NULL; sc = sc->next) { 874 if (strcmp(sc->from_charset, fc) == 0 && 875 strcmp(sc->to_charset, tc) == 0) 876 break; 877 } 878 return (sc); 879 } 880 881 /* 882 * Register a string object to an archive object. 883 */ 884 static void 885 add_sconv_object(struct archive *a, struct archive_string_conv *sc) 886 { 887 struct archive_string_conv **psc; 888 889 /* Add a new sconv to sconv list. */ 890 psc = &(a->sconv); 891 while (*psc != NULL) 892 psc = &((*psc)->next); 893 *psc = sc; 894 } 895 896 static void 897 add_converter(struct archive_string_conv *sc, int (*converter) 898 (struct archive_string *, const void *, size_t, 899 struct archive_string_conv *)) 900 { 901 if (sc == NULL || sc->nconverter >= 2) 902 __archive_errx(1, "Programming error"); 903 sc->converter[sc->nconverter++] = converter; 904 } 905 906 static void 907 setup_converter(struct archive_string_conv *sc) 908 { 909 910 /* Reset. */ 911 sc->nconverter = 0; 912 913 /* 914 * Perform special sequence for the incorrect UTF-8 filenames 915 * made by libarchive2.x. 916 */ 917 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { 918 add_converter(sc, strncat_from_utf8_libarchive2); 919 return; 920 } 921 922 /* 923 * Convert a string to UTF-16BE/LE. 924 */ 925 if (sc->flag & SCONV_TO_UTF16) { 926 /* 927 * If the current locale is UTF-8, we can translate 928 * a UTF-8 string into a UTF-16BE string. 929 */ 930 if (sc->flag & SCONV_FROM_UTF8) { 931 add_converter(sc, archive_string_append_unicode); 932 return; 933 } 934 935 #if defined(_WIN32) && !defined(__CYGWIN__) 936 if (sc->flag & SCONV_WIN_CP) { 937 if (sc->flag & SCONV_TO_UTF16BE) 938 add_converter(sc, win_strncat_to_utf16be); 939 else 940 add_converter(sc, win_strncat_to_utf16le); 941 return; 942 } 943 #endif 944 945 #if defined(HAVE_ICONV) 946 if (sc->cd != (iconv_t)-1) { 947 add_converter(sc, iconv_strncat_in_locale); 948 return; 949 } 950 #endif 951 952 if (sc->flag & SCONV_BEST_EFFORT) { 953 if (sc->flag & SCONV_TO_UTF16BE) 954 add_converter(sc, 955 best_effort_strncat_to_utf16be); 956 else 957 add_converter(sc, 958 best_effort_strncat_to_utf16le); 959 } else 960 /* Make sure we have no converter. */ 961 sc->nconverter = 0; 962 return; 963 } 964 965 /* 966 * Convert a string from UTF-16BE/LE. 967 */ 968 if (sc->flag & SCONV_FROM_UTF16) { 969 /* 970 * At least we should normalize a UTF-16BE string. 971 */ 972 if (sc->flag & SCONV_NORMALIZATION_D) 973 add_converter(sc,archive_string_normalize_D); 974 else if (sc->flag & SCONV_NORMALIZATION_C) 975 add_converter(sc, archive_string_normalize_C); 976 977 if (sc->flag & SCONV_TO_UTF8) { 978 /* 979 * If the current locale is UTF-8, we can translate 980 * a UTF-16BE/LE string into a UTF-8 string directly. 981 */ 982 if (!(sc->flag & 983 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 984 add_converter(sc, 985 archive_string_append_unicode); 986 return; 987 } 988 989 #if defined(_WIN32) && !defined(__CYGWIN__) 990 if (sc->flag & SCONV_WIN_CP) { 991 if (sc->flag & SCONV_FROM_UTF16BE) 992 add_converter(sc, win_strncat_from_utf16be); 993 else 994 add_converter(sc, win_strncat_from_utf16le); 995 return; 996 } 997 #endif 998 999 #if defined(HAVE_ICONV) 1000 if (sc->cd != (iconv_t)-1) { 1001 add_converter(sc, iconv_strncat_in_locale); 1002 return; 1003 } 1004 #endif 1005 1006 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 1007 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 1008 add_converter(sc, best_effort_strncat_from_utf16be); 1009 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 1010 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 1011 add_converter(sc, best_effort_strncat_from_utf16le); 1012 else 1013 /* Make sure we have no converter. */ 1014 sc->nconverter = 0; 1015 return; 1016 } 1017 1018 if (sc->flag & SCONV_FROM_UTF8) { 1019 /* 1020 * At least we should normalize a UTF-8 string. 1021 */ 1022 if (sc->flag & SCONV_NORMALIZATION_D) 1023 add_converter(sc,archive_string_normalize_D); 1024 else if (sc->flag & SCONV_NORMALIZATION_C) 1025 add_converter(sc, archive_string_normalize_C); 1026 1027 /* 1028 * Copy UTF-8 string with a check of CESU-8. 1029 * Apparently, iconv does not check surrogate pairs in UTF-8 1030 * when both from-charset and to-charset are UTF-8, and then 1031 * we use our UTF-8 copy code. 1032 */ 1033 if (sc->flag & SCONV_TO_UTF8) { 1034 /* 1035 * If the current locale is UTF-8, we can translate 1036 * a UTF-16BE string into a UTF-8 string directly. 1037 */ 1038 if (!(sc->flag & 1039 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 1040 add_converter(sc, strncat_from_utf8_to_utf8); 1041 return; 1042 } 1043 } 1044 1045 #if defined(_WIN32) && !defined(__CYGWIN__) 1046 /* 1047 * On Windows we can use Windows API for a string conversion. 1048 */ 1049 if (sc->flag & SCONV_WIN_CP) { 1050 add_converter(sc, strncat_in_codepage); 1051 return; 1052 } 1053 #endif 1054 1055 #if HAVE_ICONV 1056 if (sc->cd != (iconv_t)-1) { 1057 add_converter(sc, iconv_strncat_in_locale); 1058 /* 1059 * iconv generally does not support UTF-8-MAC and so 1060 * we have to the output of iconv from NFC to NFD if 1061 * need. 1062 */ 1063 if ((sc->flag & SCONV_FROM_CHARSET) && 1064 (sc->flag & SCONV_TO_UTF8)) { 1065 if (sc->flag & SCONV_NORMALIZATION_D) 1066 add_converter(sc, archive_string_normalize_D); 1067 } 1068 return; 1069 } 1070 #endif 1071 1072 /* 1073 * Try conversion in the best effort or no conversion. 1074 */ 1075 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) 1076 add_converter(sc, best_effort_strncat_in_locale); 1077 else 1078 /* Make sure we have no converter. */ 1079 sc->nconverter = 0; 1080 } 1081 1082 /* 1083 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE 1084 * and CP932 which are referenced in create_sconv_object(). 1085 */ 1086 static const char * 1087 canonical_charset_name(const char *charset) 1088 { 1089 char cs[16]; 1090 char *p; 1091 const char *s; 1092 1093 if (charset == NULL || charset[0] == '\0' 1094 || strlen(charset) > 15) 1095 return (charset); 1096 1097 /* Copy name to uppercase. */ 1098 p = cs; 1099 s = charset; 1100 while (*s) { 1101 char c = *s++; 1102 if (c >= 'a' && c <= 'z') 1103 c -= 'a' - 'A'; 1104 *p++ = c; 1105 } 1106 *p++ = '\0'; 1107 1108 if (strcmp(cs, "UTF-8") == 0 || 1109 strcmp(cs, "UTF8") == 0) 1110 return ("UTF-8"); 1111 if (strcmp(cs, "UTF-16BE") == 0 || 1112 strcmp(cs, "UTF16BE") == 0) 1113 return ("UTF-16BE"); 1114 if (strcmp(cs, "UTF-16LE") == 0 || 1115 strcmp(cs, "UTF16LE") == 0) 1116 return ("UTF-16LE"); 1117 if (strcmp(cs, "CP932") == 0) 1118 return ("CP932"); 1119 return (charset); 1120 } 1121 1122 /* 1123 * Create a string conversion object. 1124 */ 1125 static struct archive_string_conv * 1126 create_sconv_object(const char *fc, const char *tc, 1127 unsigned current_codepage, int flag) 1128 { 1129 struct archive_string_conv *sc; 1130 1131 sc = calloc(1, sizeof(*sc)); 1132 if (sc == NULL) 1133 return (NULL); 1134 sc->next = NULL; 1135 sc->from_charset = strdup(fc); 1136 if (sc->from_charset == NULL) { 1137 free(sc); 1138 return (NULL); 1139 } 1140 sc->to_charset = strdup(tc); 1141 if (sc->to_charset == NULL) { 1142 free(sc->from_charset); 1143 free(sc); 1144 return (NULL); 1145 } 1146 archive_string_init(&sc->utftmp); 1147 1148 if (flag & SCONV_TO_CHARSET) { 1149 /* 1150 * Convert characters from the current locale charset to 1151 * a specified charset. 1152 */ 1153 sc->from_cp = current_codepage; 1154 sc->to_cp = make_codepage_from_charset(tc); 1155 #if defined(_WIN32) && !defined(__CYGWIN__) 1156 if (IsValidCodePage(sc->to_cp)) 1157 flag |= SCONV_WIN_CP; 1158 #endif 1159 } else if (flag & SCONV_FROM_CHARSET) { 1160 /* 1161 * Convert characters from a specified charset to 1162 * the current locale charset. 1163 */ 1164 sc->to_cp = current_codepage; 1165 sc->from_cp = make_codepage_from_charset(fc); 1166 #if defined(_WIN32) && !defined(__CYGWIN__) 1167 if (IsValidCodePage(sc->from_cp)) 1168 flag |= SCONV_WIN_CP; 1169 #endif 1170 } 1171 1172 /* 1173 * Check if "from charset" and "to charset" are the same. 1174 */ 1175 if (strcmp(fc, tc) == 0 || 1176 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) 1177 sc->same = 1; 1178 else 1179 sc->same = 0; 1180 1181 /* 1182 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. 1183 */ 1184 if (strcmp(tc, "UTF-8") == 0) 1185 flag |= SCONV_TO_UTF8; 1186 else if (strcmp(tc, "UTF-16BE") == 0) 1187 flag |= SCONV_TO_UTF16BE; 1188 else if (strcmp(tc, "UTF-16LE") == 0) 1189 flag |= SCONV_TO_UTF16LE; 1190 if (strcmp(fc, "UTF-8") == 0) 1191 flag |= SCONV_FROM_UTF8; 1192 else if (strcmp(fc, "UTF-16BE") == 0) 1193 flag |= SCONV_FROM_UTF16BE; 1194 else if (strcmp(fc, "UTF-16LE") == 0) 1195 flag |= SCONV_FROM_UTF16LE; 1196 #if defined(_WIN32) && !defined(__CYGWIN__) 1197 if (sc->to_cp == CP_UTF8) 1198 flag |= SCONV_TO_UTF8; 1199 else if (sc->to_cp == CP_UTF16BE) 1200 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; 1201 else if (sc->to_cp == CP_UTF16LE) 1202 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; 1203 if (sc->from_cp == CP_UTF8) 1204 flag |= SCONV_FROM_UTF8; 1205 else if (sc->from_cp == CP_UTF16BE) 1206 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; 1207 else if (sc->from_cp == CP_UTF16LE) 1208 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; 1209 #endif 1210 1211 /* 1212 * Set a flag for Unicode NFD. Usually iconv cannot correctly 1213 * handle it. So we have to translate NFD characters to NFC ones 1214 * ourselves before iconv handles. Another reason is to prevent 1215 * that the same sight of two filenames, one is NFC and other 1216 * is NFD, would be in its directory. 1217 * On Mac OS X, although its filesystem layer automatically 1218 * convert filenames to NFD, it would be useful for filename 1219 * comparing to find out the same filenames that we normalize 1220 * that to be NFD ourselves. 1221 */ 1222 if ((flag & SCONV_FROM_CHARSET) && 1223 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { 1224 #if defined(__APPLE__) 1225 if (flag & SCONV_TO_UTF8) 1226 flag |= SCONV_NORMALIZATION_D; 1227 else 1228 #endif 1229 flag |= SCONV_NORMALIZATION_C; 1230 } 1231 #if defined(__APPLE__) 1232 /* 1233 * In case writing an archive file, make sure that a filename 1234 * going to be passed to iconv is a Unicode NFC string since 1235 * a filename in HFS Plus filesystem is a Unicode NFD one and 1236 * iconv cannot handle it with "UTF-8" charset. It is simpler 1237 * than a use of "UTF-8-MAC" charset. 1238 */ 1239 if ((flag & SCONV_TO_CHARSET) && 1240 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1241 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1242 flag |= SCONV_NORMALIZATION_C; 1243 /* 1244 * In case reading an archive file. make sure that a filename 1245 * will be passed to users is a Unicode NFD string in order to 1246 * correctly compare the filename with other one which comes 1247 * from HFS Plus filesystem. 1248 */ 1249 if ((flag & SCONV_FROM_CHARSET) && 1250 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1251 (flag & SCONV_TO_UTF8)) 1252 flag |= SCONV_NORMALIZATION_D; 1253 #endif 1254 1255 #if defined(HAVE_ICONV) 1256 sc->cd_w = (iconv_t)-1; 1257 /* 1258 * Create an iconv object. 1259 */ 1260 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && 1261 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || 1262 (flag & SCONV_WIN_CP)) { 1263 /* This case we won't use iconv. */ 1264 sc->cd = (iconv_t)-1; 1265 } else { 1266 sc->cd = iconv_open(tc, fc); 1267 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { 1268 /* 1269 * Unfortunately, all of iconv implements do support 1270 * "CP932" character-set, so we should use "SJIS" 1271 * instead if iconv_open failed. 1272 */ 1273 if (strcmp(tc, "CP932") == 0) 1274 sc->cd = iconv_open("SJIS", fc); 1275 else if (strcmp(fc, "CP932") == 0) 1276 sc->cd = iconv_open(tc, "SJIS"); 1277 } 1278 #if defined(_WIN32) && !defined(__CYGWIN__) 1279 /* 1280 * archive_mstring on Windows directly convert multi-bytes 1281 * into archive_wstring in order not to depend on locale 1282 * so that you can do a I18N programming. This will be 1283 * used only in archive_mstring_copy_mbs_len_l so far. 1284 */ 1285 if (flag & SCONV_FROM_CHARSET) { 1286 sc->cd_w = iconv_open("UTF-8", fc); 1287 if (sc->cd_w == (iconv_t)-1 && 1288 (sc->flag & SCONV_BEST_EFFORT)) { 1289 if (strcmp(fc, "CP932") == 0) 1290 sc->cd_w = iconv_open("UTF-8", "SJIS"); 1291 } 1292 } 1293 #endif /* _WIN32 && !__CYGWIN__ */ 1294 } 1295 #endif /* HAVE_ICONV */ 1296 1297 sc->flag = flag; 1298 1299 /* 1300 * Set up converters. 1301 */ 1302 setup_converter(sc); 1303 1304 return (sc); 1305 } 1306 1307 /* 1308 * Free a string conversion object. 1309 */ 1310 static void 1311 free_sconv_object(struct archive_string_conv *sc) 1312 { 1313 free(sc->from_charset); 1314 free(sc->to_charset); 1315 archive_string_free(&sc->utftmp); 1316 #if HAVE_ICONV 1317 if (sc->cd != (iconv_t)-1) 1318 iconv_close(sc->cd); 1319 if (sc->cd_w != (iconv_t)-1) 1320 iconv_close(sc->cd_w); 1321 #endif 1322 free(sc); 1323 } 1324 1325 #if defined(_WIN32) && !defined(__CYGWIN__) 1326 static unsigned 1327 my_atoi(const char *p) 1328 { 1329 unsigned cp; 1330 1331 cp = 0; 1332 while (*p) { 1333 if (*p >= '0' && *p <= '9') 1334 cp = cp * 10 + (*p - '0'); 1335 else 1336 return (-1); 1337 p++; 1338 } 1339 return (cp); 1340 } 1341 1342 /* 1343 * Translate Charset name (as used by iconv) into CodePage (as used by Windows) 1344 * Return -1 if failed. 1345 * 1346 * Note: This translation code may be insufficient. 1347 */ 1348 static struct charset { 1349 const char *name; 1350 unsigned cp; 1351 } charsets[] = { 1352 /* MUST BE SORTED! */ 1353 {"ASCII", 1252}, 1354 {"ASMO-708", 708}, 1355 {"BIG5", 950}, 1356 {"CHINESE", 936}, 1357 {"CP367", 1252}, 1358 {"CP819", 1252}, 1359 {"CP1025", 21025}, 1360 {"DOS-720", 720}, 1361 {"DOS-862", 862}, 1362 {"EUC-CN", 51936}, 1363 {"EUC-JP", 51932}, 1364 {"EUC-KR", 949}, 1365 {"EUCCN", 51936}, 1366 {"EUCJP", 51932}, 1367 {"EUCKR", 949}, 1368 {"GB18030", 54936}, 1369 {"GB2312", 936}, 1370 {"HEBREW", 1255}, 1371 {"HZ-GB-2312", 52936}, 1372 {"IBM273", 20273}, 1373 {"IBM277", 20277}, 1374 {"IBM278", 20278}, 1375 {"IBM280", 20280}, 1376 {"IBM284", 20284}, 1377 {"IBM285", 20285}, 1378 {"IBM290", 20290}, 1379 {"IBM297", 20297}, 1380 {"IBM367", 1252}, 1381 {"IBM420", 20420}, 1382 {"IBM423", 20423}, 1383 {"IBM424", 20424}, 1384 {"IBM819", 1252}, 1385 {"IBM871", 20871}, 1386 {"IBM880", 20880}, 1387 {"IBM905", 20905}, 1388 {"IBM924", 20924}, 1389 {"ISO-8859-1", 28591}, 1390 {"ISO-8859-13", 28603}, 1391 {"ISO-8859-15", 28605}, 1392 {"ISO-8859-2", 28592}, 1393 {"ISO-8859-3", 28593}, 1394 {"ISO-8859-4", 28594}, 1395 {"ISO-8859-5", 28595}, 1396 {"ISO-8859-6", 28596}, 1397 {"ISO-8859-7", 28597}, 1398 {"ISO-8859-8", 28598}, 1399 {"ISO-8859-9", 28599}, 1400 {"ISO8859-1", 28591}, 1401 {"ISO8859-13", 28603}, 1402 {"ISO8859-15", 28605}, 1403 {"ISO8859-2", 28592}, 1404 {"ISO8859-3", 28593}, 1405 {"ISO8859-4", 28594}, 1406 {"ISO8859-5", 28595}, 1407 {"ISO8859-6", 28596}, 1408 {"ISO8859-7", 28597}, 1409 {"ISO8859-8", 28598}, 1410 {"ISO8859-9", 28599}, 1411 {"JOHAB", 1361}, 1412 {"KOI8-R", 20866}, 1413 {"KOI8-U", 21866}, 1414 {"KS_C_5601-1987", 949}, 1415 {"LATIN1", 1252}, 1416 {"LATIN2", 28592}, 1417 {"MACINTOSH", 10000}, 1418 {"SHIFT-JIS", 932}, 1419 {"SHIFT_JIS", 932}, 1420 {"SJIS", 932}, 1421 {"US", 1252}, 1422 {"US-ASCII", 1252}, 1423 {"UTF-16", 1200}, 1424 {"UTF-16BE", 1201}, 1425 {"UTF-16LE", 1200}, 1426 {"UTF-8", CP_UTF8}, 1427 {"X-EUROPA", 29001}, 1428 {"X-MAC-ARABIC", 10004}, 1429 {"X-MAC-CE", 10029}, 1430 {"X-MAC-CHINESEIMP", 10008}, 1431 {"X-MAC-CHINESETRAD", 10002}, 1432 {"X-MAC-CROATIAN", 10082}, 1433 {"X-MAC-CYRILLIC", 10007}, 1434 {"X-MAC-GREEK", 10006}, 1435 {"X-MAC-HEBREW", 10005}, 1436 {"X-MAC-ICELANDIC", 10079}, 1437 {"X-MAC-JAPANESE", 10001}, 1438 {"X-MAC-KOREAN", 10003}, 1439 {"X-MAC-ROMANIAN", 10010}, 1440 {"X-MAC-THAI", 10021}, 1441 {"X-MAC-TURKISH", 10081}, 1442 {"X-MAC-UKRAINIAN", 10017}, 1443 }; 1444 static unsigned 1445 make_codepage_from_charset(const char *charset) 1446 { 1447 char cs[16]; 1448 char *p; 1449 unsigned cp; 1450 int a, b; 1451 1452 if (charset == NULL || strlen(charset) > 15) 1453 return -1; 1454 1455 /* Copy name to uppercase. */ 1456 p = cs; 1457 while (*charset) { 1458 char c = *charset++; 1459 if (c >= 'a' && c <= 'z') 1460 c -= 'a' - 'A'; 1461 *p++ = c; 1462 } 1463 *p++ = '\0'; 1464 cp = -1; 1465 1466 /* Look it up in the table first, so that we can easily 1467 * override CP367, which we map to 1252 instead of 367. */ 1468 a = 0; 1469 b = sizeof(charsets)/sizeof(charsets[0]); 1470 while (b > a) { 1471 int c = (b + a) / 2; 1472 int r = strcmp(charsets[c].name, cs); 1473 if (r < 0) 1474 a = c + 1; 1475 else if (r > 0) 1476 b = c; 1477 else 1478 return charsets[c].cp; 1479 } 1480 1481 /* If it's not in the table, try to parse it. */ 1482 switch (*cs) { 1483 case 'C': 1484 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { 1485 cp = my_atoi(cs + 2); 1486 } else if (strcmp(cs, "CP_ACP") == 0) 1487 cp = get_current_codepage(); 1488 else if (strcmp(cs, "CP_OEMCP") == 0) 1489 cp = get_current_oemcp(); 1490 break; 1491 case 'I': 1492 if (cs[1] == 'B' && cs[2] == 'M' && 1493 cs[3] >= '0' && cs[3] <= '9') { 1494 cp = my_atoi(cs + 3); 1495 } 1496 break; 1497 case 'W': 1498 if (strncmp(cs, "WINDOWS-", 8) == 0) { 1499 cp = my_atoi(cs + 8); 1500 if (cp != 874 && (cp < 1250 || cp > 1258)) 1501 cp = -1;/* This may invalid code. */ 1502 } 1503 break; 1504 } 1505 return (cp); 1506 } 1507 1508 /* 1509 * Return ANSI Code Page of current locale set by setlocale(). 1510 */ 1511 static unsigned 1512 get_current_codepage(void) 1513 { 1514 char *locale, *p; 1515 unsigned cp; 1516 1517 locale = setlocale(LC_CTYPE, NULL); 1518 if (locale == NULL) 1519 return (GetACP()); 1520 if (locale[0] == 'C' && locale[1] == '\0') 1521 return (CP_C_LOCALE); 1522 p = strrchr(locale, '.'); 1523 if (p == NULL) 1524 return (GetACP()); 1525 if (strcmp(p+1, "utf8") == 0) 1526 return CP_UTF8; 1527 cp = my_atoi(p+1); 1528 if ((int)cp <= 0) 1529 return (GetACP()); 1530 return (cp); 1531 } 1532 1533 /* 1534 * Translation table between Locale Name and ACP/OEMCP. 1535 */ 1536 static struct { 1537 unsigned acp; 1538 unsigned ocp; 1539 const char *locale; 1540 } acp_ocp_map[] = { 1541 { 950, 950, "Chinese_Taiwan" }, 1542 { 936, 936, "Chinese_People's Republic of China" }, 1543 { 950, 950, "Chinese_Taiwan" }, 1544 { 1250, 852, "Czech_Czech Republic" }, 1545 { 1252, 850, "Danish_Denmark" }, 1546 { 1252, 850, "Dutch_Netherlands" }, 1547 { 1252, 850, "Dutch_Belgium" }, 1548 { 1252, 437, "English_United States" }, 1549 { 1252, 850, "English_Australia" }, 1550 { 1252, 850, "English_Canada" }, 1551 { 1252, 850, "English_New Zealand" }, 1552 { 1252, 850, "English_United Kingdom" }, 1553 { 1252, 437, "English_United States" }, 1554 { 1252, 850, "Finnish_Finland" }, 1555 { 1252, 850, "French_France" }, 1556 { 1252, 850, "French_Belgium" }, 1557 { 1252, 850, "French_Canada" }, 1558 { 1252, 850, "French_Switzerland" }, 1559 { 1252, 850, "German_Germany" }, 1560 { 1252, 850, "German_Austria" }, 1561 { 1252, 850, "German_Switzerland" }, 1562 { 1253, 737, "Greek_Greece" }, 1563 { 1250, 852, "Hungarian_Hungary" }, 1564 { 1252, 850, "Icelandic_Iceland" }, 1565 { 1252, 850, "Italian_Italy" }, 1566 { 1252, 850, "Italian_Switzerland" }, 1567 { 932, 932, "Japanese_Japan" }, 1568 { 949, 949, "Korean_Korea" }, 1569 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1570 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1571 { 1252, 850, "Norwegian-Nynorsk_Norway" }, 1572 { 1250, 852, "Polish_Poland" }, 1573 { 1252, 850, "Portuguese_Portugal" }, 1574 { 1252, 850, "Portuguese_Brazil" }, 1575 { 1251, 866, "Russian_Russia" }, 1576 { 1250, 852, "Slovak_Slovakia" }, 1577 { 1252, 850, "Spanish_Spain" }, 1578 { 1252, 850, "Spanish_Mexico" }, 1579 { 1252, 850, "Spanish_Spain" }, 1580 { 1252, 850, "Swedish_Sweden" }, 1581 { 1254, 857, "Turkish_Turkey" }, 1582 { 0, 0, NULL} 1583 }; 1584 1585 /* 1586 * Return OEM Code Page of current locale set by setlocale(). 1587 */ 1588 static unsigned 1589 get_current_oemcp(void) 1590 { 1591 int i; 1592 char *locale, *p; 1593 size_t len; 1594 1595 locale = setlocale(LC_CTYPE, NULL); 1596 if (locale == NULL) 1597 return (GetOEMCP()); 1598 if (locale[0] == 'C' && locale[1] == '\0') 1599 return (CP_C_LOCALE); 1600 1601 p = strrchr(locale, '.'); 1602 if (p == NULL) 1603 return (GetOEMCP()); 1604 len = p - locale; 1605 for (i = 0; acp_ocp_map[i].acp; i++) { 1606 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) 1607 return (acp_ocp_map[i].ocp); 1608 } 1609 return (GetOEMCP()); 1610 } 1611 #else 1612 1613 /* 1614 * POSIX platform does not use CodePage. 1615 */ 1616 1617 static unsigned 1618 get_current_codepage(void) 1619 { 1620 return (-1);/* Unknown */ 1621 } 1622 static unsigned 1623 make_codepage_from_charset(const char *charset) 1624 { 1625 (void)charset; /* UNUSED */ 1626 return (-1);/* Unknown */ 1627 } 1628 static unsigned 1629 get_current_oemcp(void) 1630 { 1631 return (-1);/* Unknown */ 1632 } 1633 1634 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 1635 1636 /* 1637 * Return a string conversion object. 1638 */ 1639 static struct archive_string_conv * 1640 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) 1641 { 1642 struct archive_string_conv *sc; 1643 unsigned current_codepage; 1644 1645 /* Check if we have made the sconv object. */ 1646 sc = find_sconv_object(a, fc, tc); 1647 if (sc != NULL) 1648 return (sc); 1649 1650 if (a == NULL) 1651 current_codepage = get_current_codepage(); 1652 else 1653 current_codepage = a->current_codepage; 1654 1655 sc = create_sconv_object(canonical_charset_name(fc), 1656 canonical_charset_name(tc), current_codepage, flag); 1657 if (sc == NULL) { 1658 if (a != NULL) 1659 archive_set_error(a, ENOMEM, 1660 "Could not allocate memory for " 1661 "a string conversion object"); 1662 return (NULL); 1663 } 1664 1665 /* 1666 * If there is no converter for current string conversion object, 1667 * we cannot handle this conversion. 1668 */ 1669 if (sc->nconverter == 0) { 1670 if (a != NULL) { 1671 #if HAVE_ICONV 1672 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1673 "iconv_open failed : Cannot handle ``%s''", 1674 (flag & SCONV_TO_CHARSET)?tc:fc); 1675 #else 1676 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1677 "A character-set conversion not fully supported " 1678 "on this platform"); 1679 #endif 1680 } 1681 /* Failed; free a sconv object. */ 1682 free_sconv_object(sc); 1683 return (NULL); 1684 } 1685 1686 /* 1687 * Success! 1688 */ 1689 if (a != NULL) 1690 add_sconv_object(a, sc); 1691 return (sc); 1692 } 1693 1694 static const char * 1695 get_current_charset(struct archive *a) 1696 { 1697 const char *cur_charset; 1698 1699 if (a == NULL) 1700 cur_charset = default_iconv_charset(""); 1701 else { 1702 cur_charset = default_iconv_charset(a->current_code); 1703 if (a->current_code == NULL) { 1704 a->current_code = strdup(cur_charset); 1705 a->current_codepage = get_current_codepage(); 1706 a->current_oemcp = get_current_oemcp(); 1707 } 1708 } 1709 return (cur_charset); 1710 } 1711 1712 /* 1713 * Make and Return a string conversion object. 1714 * Return NULL if the platform does not support the specified conversion 1715 * and best_effort is 0. 1716 * If best_effort is set, A string conversion object must be returned 1717 * unless memory allocation for the object fails, but the conversion 1718 * might fail when non-ASCII code is found. 1719 */ 1720 struct archive_string_conv * 1721 archive_string_conversion_to_charset(struct archive *a, const char *charset, 1722 int best_effort) 1723 { 1724 int flag = SCONV_TO_CHARSET; 1725 1726 if (best_effort) 1727 flag |= SCONV_BEST_EFFORT; 1728 return (get_sconv_object(a, get_current_charset(a), charset, flag)); 1729 } 1730 1731 struct archive_string_conv * 1732 archive_string_conversion_from_charset(struct archive *a, const char *charset, 1733 int best_effort) 1734 { 1735 int flag = SCONV_FROM_CHARSET; 1736 1737 if (best_effort) 1738 flag |= SCONV_BEST_EFFORT; 1739 return (get_sconv_object(a, charset, get_current_charset(a), flag)); 1740 } 1741 1742 /* 1743 * archive_string_default_conversion_*_archive() are provided for Windows 1744 * platform because other archiver application use CP_OEMCP for 1745 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames 1746 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP 1747 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). 1748 * So we should make a string conversion between CP_ACP and CP_OEMCP 1749 * for compatibility. 1750 */ 1751 #if defined(_WIN32) && !defined(__CYGWIN__) 1752 struct archive_string_conv * 1753 archive_string_default_conversion_for_read(struct archive *a) 1754 { 1755 const char *cur_charset = get_current_charset(a); 1756 char oemcp[16]; 1757 1758 /* NOTE: a check of cur_charset is unneeded but we need 1759 * that get_current_charset() has been surely called at 1760 * this time whatever C compiler optimized. */ 1761 if (cur_charset != NULL && 1762 (a->current_codepage == CP_C_LOCALE || 1763 a->current_codepage == a->current_oemcp)) 1764 return (NULL);/* no conversion. */ 1765 1766 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1767 /* Make sure a null termination must be set. */ 1768 oemcp[sizeof(oemcp)-1] = '\0'; 1769 return (get_sconv_object(a, oemcp, cur_charset, 1770 SCONV_FROM_CHARSET)); 1771 } 1772 1773 struct archive_string_conv * 1774 archive_string_default_conversion_for_write(struct archive *a) 1775 { 1776 const char *cur_charset = get_current_charset(a); 1777 char oemcp[16]; 1778 1779 /* NOTE: a check of cur_charset is unneeded but we need 1780 * that get_current_charset() has been surely called at 1781 * this time whatever C compiler optimized. */ 1782 if (cur_charset != NULL && 1783 (a->current_codepage == CP_C_LOCALE || 1784 a->current_codepage == a->current_oemcp)) 1785 return (NULL);/* no conversion. */ 1786 1787 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1788 /* Make sure a null termination must be set. */ 1789 oemcp[sizeof(oemcp)-1] = '\0'; 1790 return (get_sconv_object(a, cur_charset, oemcp, 1791 SCONV_TO_CHARSET)); 1792 } 1793 #else 1794 struct archive_string_conv * 1795 archive_string_default_conversion_for_read(struct archive *a) 1796 { 1797 (void)a; /* UNUSED */ 1798 return (NULL); 1799 } 1800 1801 struct archive_string_conv * 1802 archive_string_default_conversion_for_write(struct archive *a) 1803 { 1804 (void)a; /* UNUSED */ 1805 return (NULL); 1806 } 1807 #endif 1808 1809 /* 1810 * Dispose of all character conversion objects in the archive object. 1811 */ 1812 void 1813 archive_string_conversion_free(struct archive *a) 1814 { 1815 struct archive_string_conv *sc; 1816 struct archive_string_conv *sc_next; 1817 1818 for (sc = a->sconv; sc != NULL; sc = sc_next) { 1819 sc_next = sc->next; 1820 free_sconv_object(sc); 1821 } 1822 a->sconv = NULL; 1823 free(a->current_code); 1824 a->current_code = NULL; 1825 } 1826 1827 /* 1828 * Return a conversion charset name. 1829 */ 1830 const char * 1831 archive_string_conversion_charset_name(struct archive_string_conv *sc) 1832 { 1833 if (sc->flag & SCONV_TO_CHARSET) 1834 return (sc->to_charset); 1835 else 1836 return (sc->from_charset); 1837 } 1838 1839 /* 1840 * Change the behavior of a string conversion. 1841 */ 1842 void 1843 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) 1844 { 1845 switch (opt) { 1846 /* 1847 * A filename in UTF-8 was made with libarchive 2.x in a wrong 1848 * assumption that wchar_t was Unicode. 1849 * This option enables simulating the assumption in order to read 1850 * that filename correctly. 1851 */ 1852 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: 1853 #if (defined(_WIN32) && !defined(__CYGWIN__)) \ 1854 || defined(__STDC_ISO_10646__) || defined(__APPLE__) 1855 /* 1856 * Nothing to do for it since wchar_t on these platforms 1857 * is really Unicode. 1858 */ 1859 (void)sc; /* UNUSED */ 1860 #else 1861 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { 1862 sc->flag |= SCONV_UTF8_LIBARCHIVE_2; 1863 /* Set up string converters. */ 1864 setup_converter(sc); 1865 } 1866 #endif 1867 break; 1868 case SCONV_SET_OPT_NORMALIZATION_C: 1869 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) { 1870 sc->flag |= SCONV_NORMALIZATION_C; 1871 sc->flag &= ~SCONV_NORMALIZATION_D; 1872 /* Set up string converters. */ 1873 setup_converter(sc); 1874 } 1875 break; 1876 case SCONV_SET_OPT_NORMALIZATION_D: 1877 #if defined(HAVE_ICONV) 1878 /* 1879 * If iconv will take the string, do not change the 1880 * setting of the normalization. 1881 */ 1882 if (!(sc->flag & SCONV_WIN_CP) && 1883 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1884 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1885 break; 1886 #endif 1887 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) { 1888 sc->flag |= SCONV_NORMALIZATION_D; 1889 sc->flag &= ~SCONV_NORMALIZATION_C; 1890 /* Set up string converters. */ 1891 setup_converter(sc); 1892 } 1893 break; 1894 default: 1895 break; 1896 } 1897 } 1898 1899 /* 1900 * 1901 * Copy one archive_string to another in locale conversion. 1902 * 1903 * archive_strncat_l(); 1904 * archive_strncpy_l(); 1905 * 1906 */ 1907 1908 static size_t 1909 mbsnbytes(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 while (s < n && *pp) { 1922 pp++; 1923 s++; 1924 } 1925 return (s); 1926 } 1927 1928 static size_t 1929 utf16nbytes(const void *_p, size_t n) 1930 { 1931 size_t s; 1932 const char *p, *pp; 1933 1934 if (_p == NULL) 1935 return (0); 1936 p = (const char *)_p; 1937 1938 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1939 s = 0; 1940 pp = p; 1941 n >>= 1; 1942 while (s < n && (pp[0] || pp[1])) { 1943 pp += 2; 1944 s++; 1945 } 1946 return (s<<1); 1947 } 1948 1949 int 1950 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n, 1951 struct archive_string_conv *sc) 1952 { 1953 as->length = 0; 1954 return (archive_strncat_l(as, _p, n, sc)); 1955 } 1956 1957 int 1958 archive_strncat_l(struct archive_string *as, const void *_p, size_t n, 1959 struct archive_string_conv *sc) 1960 { 1961 const void *s; 1962 size_t length = 0; 1963 int i, r = 0, r2; 1964 1965 if (_p != NULL && n > 0) { 1966 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) 1967 length = utf16nbytes(_p, n); 1968 else 1969 length = mbsnbytes(_p, n); 1970 } 1971 1972 /* We must allocate memory even if there is no data for conversion 1973 * or copy. This simulates archive_string_append behavior. */ 1974 if (length == 0) { 1975 int tn = 1; 1976 if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) 1977 tn = 2; 1978 if (archive_string_ensure(as, as->length + tn) == NULL) 1979 return (-1); 1980 as->s[as->length] = 0; 1981 if (tn == 2) 1982 as->s[as->length+1] = 0; 1983 return (0); 1984 } 1985 1986 /* 1987 * If sc is NULL, we just make a copy. 1988 */ 1989 if (sc == NULL) { 1990 if (archive_string_append(as, _p, length) == NULL) 1991 return (-1);/* No memory */ 1992 return (0); 1993 } 1994 1995 s = _p; 1996 i = 0; 1997 if (sc->nconverter > 1) { 1998 sc->utftmp.length = 0; 1999 r2 = sc->converter[0](&(sc->utftmp), s, length, sc); 2000 if (r2 != 0 && errno == ENOMEM) 2001 return (r2); 2002 if (r > r2) 2003 r = r2; 2004 s = sc->utftmp.s; 2005 length = sc->utftmp.length; 2006 ++i; 2007 } 2008 r2 = sc->converter[i](as, s, length, sc); 2009 if (r > r2) 2010 r = r2; 2011 return (r); 2012 } 2013 2014 #if HAVE_ICONV 2015 2016 /* 2017 * Return -1 if conversion fails. 2018 */ 2019 static int 2020 iconv_strncat_in_locale(struct archive_string *as, const void *_p, 2021 size_t length, struct archive_string_conv *sc) 2022 { 2023 ICONV_CONST char *itp; 2024 size_t remaining; 2025 iconv_t cd; 2026 char *outp; 2027 size_t avail, bs; 2028 int return_value = 0; /* success */ 2029 int to_size, from_size; 2030 2031 if (sc->flag & SCONV_TO_UTF16) 2032 to_size = 2; 2033 else 2034 to_size = 1; 2035 if (sc->flag & SCONV_FROM_UTF16) 2036 from_size = 2; 2037 else 2038 from_size = 1; 2039 2040 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) 2041 return (-1); 2042 2043 cd = sc->cd; 2044 itp = (char *)(uintptr_t)_p; 2045 remaining = length; 2046 outp = as->s + as->length; 2047 avail = as->buffer_length - as->length - to_size; 2048 while (remaining >= (size_t)from_size) { 2049 size_t result = iconv(cd, &itp, &remaining, &outp, &avail); 2050 2051 if (result != (size_t)-1) 2052 break; /* Conversion completed. */ 2053 2054 if (errno == EILSEQ || errno == EINVAL) { 2055 /* 2056 * If an output charset is UTF-8 or UTF-16BE/LE, 2057 * unknown character should be U+FFFD 2058 * (replacement character). 2059 */ 2060 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { 2061 size_t rbytes; 2062 if (sc->flag & SCONV_TO_UTF8) 2063 rbytes = sizeof(utf8_replacement_char); 2064 else 2065 rbytes = 2; 2066 2067 if (avail < rbytes) { 2068 as->length = outp - as->s; 2069 bs = as->buffer_length + 2070 (remaining * to_size) + rbytes; 2071 if (NULL == 2072 archive_string_ensure(as, bs)) 2073 return (-1); 2074 outp = as->s + as->length; 2075 avail = as->buffer_length 2076 - as->length - to_size; 2077 } 2078 if (sc->flag & SCONV_TO_UTF8) 2079 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); 2080 else if (sc->flag & SCONV_TO_UTF16BE) 2081 archive_be16enc(outp, UNICODE_R_CHAR); 2082 else 2083 archive_le16enc(outp, UNICODE_R_CHAR); 2084 outp += rbytes; 2085 avail -= rbytes; 2086 } else { 2087 /* Skip the illegal input bytes. */ 2088 *outp++ = '?'; 2089 avail--; 2090 } 2091 itp += from_size; 2092 remaining -= from_size; 2093 return_value = -1; /* failure */ 2094 } else { 2095 /* E2BIG no output buffer, 2096 * Increase an output buffer. */ 2097 as->length = outp - as->s; 2098 bs = as->buffer_length + remaining * 2; 2099 if (NULL == archive_string_ensure(as, bs)) 2100 return (-1); 2101 outp = as->s + as->length; 2102 avail = as->buffer_length - as->length - to_size; 2103 } 2104 } 2105 as->length = outp - as->s; 2106 as->s[as->length] = 0; 2107 if (to_size == 2) 2108 as->s[as->length+1] = 0; 2109 return (return_value); 2110 } 2111 2112 #endif /* HAVE_ICONV */ 2113 2114 2115 #if defined(_WIN32) && !defined(__CYGWIN__) 2116 2117 /* 2118 * Translate a string from a some CodePage to an another CodePage by 2119 * Windows APIs, and copy the result. Return -1 if conversion fails. 2120 */ 2121 static int 2122 strncat_in_codepage(struct archive_string *as, 2123 const void *_p, size_t length, struct archive_string_conv *sc) 2124 { 2125 const char *s = (const char *)_p; 2126 struct archive_wstring aws; 2127 size_t l; 2128 int r, saved_flag; 2129 2130 archive_string_init(&aws); 2131 saved_flag = sc->flag; 2132 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); 2133 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); 2134 sc->flag = saved_flag; 2135 if (r != 0) { 2136 archive_wstring_free(&aws); 2137 if (errno != ENOMEM) 2138 archive_string_append(as, s, length); 2139 return (-1); 2140 } 2141 2142 l = as->length; 2143 r = archive_string_append_from_wcs_in_codepage( 2144 as, aws.s, aws.length, sc); 2145 if (r != 0 && errno != ENOMEM && l == as->length) 2146 archive_string_append(as, s, length); 2147 archive_wstring_free(&aws); 2148 return (r); 2149 } 2150 2151 /* 2152 * Test whether MBS ==> WCS is okay. 2153 */ 2154 static int 2155 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2156 { 2157 const char *p = (const char *)_p; 2158 unsigned codepage; 2159 DWORD mbflag = MB_ERR_INVALID_CHARS; 2160 2161 if (sc->flag & SCONV_FROM_CHARSET) 2162 codepage = sc->to_cp; 2163 else 2164 codepage = sc->from_cp; 2165 2166 if (codepage == CP_C_LOCALE) 2167 return (0); 2168 if (codepage != CP_UTF8) 2169 mbflag |= MB_PRECOMPOSED; 2170 2171 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) 2172 return (-1); /* Invalid */ 2173 return (0); /* Okay */ 2174 } 2175 2176 #else 2177 2178 /* 2179 * Test whether MBS ==> WCS is okay. 2180 */ 2181 static int 2182 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2183 { 2184 const char *p = (const char *)_p; 2185 size_t r; 2186 2187 #if HAVE_MBRTOWC 2188 mbstate_t shift_state; 2189 2190 memset(&shift_state, 0, sizeof(shift_state)); 2191 #else 2192 /* Clear the shift state before starting. */ 2193 mbtowc(NULL, NULL, 0); 2194 #endif 2195 while (n) { 2196 wchar_t wc; 2197 2198 #if HAVE_MBRTOWC 2199 r = mbrtowc(&wc, p, n, &shift_state); 2200 #else 2201 r = mbtowc(&wc, p, n); 2202 #endif 2203 if (r == (size_t)-1 || r == (size_t)-2) 2204 return (-1);/* Invalid. */ 2205 if (r == 0) 2206 break; 2207 p += r; 2208 n -= r; 2209 } 2210 (void)sc; /* UNUSED */ 2211 return (0); /* All Okey. */ 2212 } 2213 2214 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 2215 2216 /* 2217 * Basically returns -1 because we cannot make a conversion of charset 2218 * without iconv but in some cases this would return 0. 2219 * Returns 0 if all copied characters are ASCII. 2220 * Returns 0 if both from-locale and to-locale are the same and those 2221 * can be WCS with no error. 2222 */ 2223 static int 2224 best_effort_strncat_in_locale(struct archive_string *as, const void *_p, 2225 size_t length, struct archive_string_conv *sc) 2226 { 2227 size_t remaining; 2228 const uint8_t *itp; 2229 int return_value = 0; /* success */ 2230 2231 /* 2232 * If both from-locale and to-locale is the same, this makes a copy. 2233 * And then this checks all copied MBS can be WCS if so returns 0. 2234 */ 2235 if (sc->same) { 2236 if (archive_string_append(as, _p, length) == NULL) 2237 return (-1);/* No memory */ 2238 return (invalid_mbs(_p, length, sc)); 2239 } 2240 2241 /* 2242 * If a character is ASCII, this just copies it. If not, this 2243 * assigns '?' character instead but in UTF-8 locale this assigns 2244 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, 2245 * a Replacement Character in Unicode. 2246 */ 2247 2248 remaining = length; 2249 itp = (const uint8_t *)_p; 2250 while (*itp && remaining > 0) { 2251 if (*itp > 127) { 2252 // Non-ASCII: Substitute with suitable replacement 2253 if (sc->flag & SCONV_TO_UTF8) { 2254 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { 2255 __archive_errx(1, "Out of memory"); 2256 } 2257 } else { 2258 archive_strappend_char(as, '?'); 2259 } 2260 return_value = -1; 2261 } else { 2262 archive_strappend_char(as, *itp); 2263 } 2264 ++itp; 2265 } 2266 return (return_value); 2267 } 2268 2269 2270 /* 2271 * Unicode conversion functions. 2272 * - UTF-8 <===> UTF-8 in removing surrogate pairs. 2273 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. 2274 * - UTF-8 made by libarchive 2.x ===> UTF-8. 2275 * - UTF-16BE <===> UTF-8. 2276 * 2277 */ 2278 2279 /* 2280 * Utility to convert a single UTF-8 sequence. 2281 * 2282 * Usually return used bytes, return used byte in negative value when 2283 * a unicode character is replaced with U+FFFD. 2284 * See also http://unicode.org/review/pr-121.html Public Review Issue #121 2285 * Recommended Practice for Replacement Characters. 2286 */ 2287 static int 2288 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2289 { 2290 static const char utf8_count[256] = { 2291 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 2292 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 2293 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 2294 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 2295 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 2296 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 2299 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 2300 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 2301 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 2302 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 2303 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2304 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 2305 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 2306 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ 2307 }; 2308 int ch, i; 2309 int cnt; 2310 uint32_t wc; 2311 2312 /* Sanity check. */ 2313 if (n == 0) 2314 return (0); 2315 /* 2316 * Decode 1-4 bytes depending on the value of the first byte. 2317 */ 2318 ch = (unsigned char)*s; 2319 if (ch == 0) 2320 return (0); /* Standard: return 0 for end-of-string. */ 2321 cnt = utf8_count[ch]; 2322 2323 /* Invalid sequence or there are not plenty bytes. */ 2324 if ((int)n < cnt) { 2325 cnt = (int)n; 2326 for (i = 1; i < cnt; i++) { 2327 if ((s[i] & 0xc0) != 0x80) { 2328 cnt = i; 2329 break; 2330 } 2331 } 2332 goto invalid_sequence; 2333 } 2334 2335 /* Make a Unicode code point from a single UTF-8 sequence. */ 2336 switch (cnt) { 2337 case 1: /* 1 byte sequence. */ 2338 *pwc = ch & 0x7f; 2339 return (cnt); 2340 case 2: /* 2 bytes sequence. */ 2341 if ((s[1] & 0xc0) != 0x80) { 2342 cnt = 1; 2343 goto invalid_sequence; 2344 } 2345 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); 2346 return (cnt); 2347 case 3: /* 3 bytes sequence. */ 2348 if ((s[1] & 0xc0) != 0x80) { 2349 cnt = 1; 2350 goto invalid_sequence; 2351 } 2352 if ((s[2] & 0xc0) != 0x80) { 2353 cnt = 2; 2354 goto invalid_sequence; 2355 } 2356 wc = ((ch & 0x0f) << 12) 2357 | ((s[1] & 0x3f) << 6) 2358 | (s[2] & 0x3f); 2359 if (wc < 0x800) 2360 goto invalid_sequence;/* Overlong sequence. */ 2361 break; 2362 case 4: /* 4 bytes sequence. */ 2363 if ((s[1] & 0xc0) != 0x80) { 2364 cnt = 1; 2365 goto invalid_sequence; 2366 } 2367 if ((s[2] & 0xc0) != 0x80) { 2368 cnt = 2; 2369 goto invalid_sequence; 2370 } 2371 if ((s[3] & 0xc0) != 0x80) { 2372 cnt = 3; 2373 goto invalid_sequence; 2374 } 2375 wc = ((ch & 0x07) << 18) 2376 | ((s[1] & 0x3f) << 12) 2377 | ((s[2] & 0x3f) << 6) 2378 | (s[3] & 0x3f); 2379 if (wc < 0x10000) 2380 goto invalid_sequence;/* Overlong sequence. */ 2381 break; 2382 default: /* Others are all invalid sequence. */ 2383 if (ch == 0xc0 || ch == 0xc1) 2384 cnt = 2; 2385 else if (ch >= 0xf5 && ch <= 0xf7) 2386 cnt = 4; 2387 else if (ch >= 0xf8 && ch <= 0xfb) 2388 cnt = 5; 2389 else if (ch == 0xfc || ch == 0xfd) 2390 cnt = 6; 2391 else 2392 cnt = 1; 2393 if ((int)n < cnt) 2394 cnt = (int)n; 2395 for (i = 1; i < cnt; i++) { 2396 if ((s[i] & 0xc0) != 0x80) { 2397 cnt = i; 2398 break; 2399 } 2400 } 2401 goto invalid_sequence; 2402 } 2403 2404 /* The code point larger than 0x10FFFF is not legal 2405 * Unicode values. */ 2406 if (wc > UNICODE_MAX) 2407 goto invalid_sequence; 2408 /* Correctly gets a Unicode, returns used bytes. */ 2409 *pwc = wc; 2410 return (cnt); 2411 invalid_sequence: 2412 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2413 return (cnt * -1); 2414 } 2415 2416 static int 2417 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2418 { 2419 int cnt; 2420 2421 cnt = _utf8_to_unicode(pwc, s, n); 2422 /* Any of Surrogate pair is not legal Unicode values. */ 2423 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) 2424 return (-3); 2425 return (cnt); 2426 } 2427 2428 static inline uint32_t 2429 combine_surrogate_pair(uint32_t uc, uint32_t uc2) 2430 { 2431 uc -= 0xD800; 2432 uc *= 0x400; 2433 uc += uc2 - 0xDC00; 2434 uc += 0x10000; 2435 return (uc); 2436 } 2437 2438 /* 2439 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in 2440 * removing surrogate pairs. 2441 * 2442 * CESU-8: The Compatibility Encoding Scheme for UTF-16. 2443 * 2444 * Usually return used bytes, return used byte in negative value when 2445 * a unicode character is replaced with U+FFFD. 2446 */ 2447 static int 2448 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2449 { 2450 uint32_t wc = 0; 2451 int cnt; 2452 2453 cnt = _utf8_to_unicode(&wc, s, n); 2454 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { 2455 uint32_t wc2 = 0; 2456 if (n - 3 < 3) { 2457 /* Invalid byte sequence. */ 2458 goto invalid_sequence; 2459 } 2460 cnt = _utf8_to_unicode(&wc2, s+3, n-3); 2461 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { 2462 /* Invalid byte sequence. */ 2463 goto invalid_sequence; 2464 } 2465 wc = combine_surrogate_pair(wc, wc2); 2466 cnt = 6; 2467 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { 2468 /* Invalid byte sequence. */ 2469 goto invalid_sequence; 2470 } 2471 *pwc = wc; 2472 return (cnt); 2473 invalid_sequence: 2474 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2475 if (cnt > 0) 2476 cnt *= -1; 2477 return (cnt); 2478 } 2479 2480 /* 2481 * Convert a Unicode code point to a single UTF-8 sequence. 2482 * 2483 * NOTE:This function does not check if the Unicode is legal or not. 2484 * Please you definitely check it before calling this. 2485 */ 2486 static size_t 2487 unicode_to_utf8(char *p, size_t remaining, uint32_t uc) 2488 { 2489 char *_p = p; 2490 2491 /* Invalid Unicode char maps to Replacement character */ 2492 if (uc > UNICODE_MAX) 2493 uc = UNICODE_R_CHAR; 2494 /* Translate code point to UTF8 */ 2495 if (uc <= 0x7f) { 2496 if (remaining == 0) 2497 return (0); 2498 *p++ = (char)uc; 2499 } else if (uc <= 0x7ff) { 2500 if (remaining < 2) 2501 return (0); 2502 *p++ = 0xc0 | ((uc >> 6) & 0x1f); 2503 *p++ = 0x80 | (uc & 0x3f); 2504 } else if (uc <= 0xffff) { 2505 if (remaining < 3) 2506 return (0); 2507 *p++ = 0xe0 | ((uc >> 12) & 0x0f); 2508 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2509 *p++ = 0x80 | (uc & 0x3f); 2510 } else { 2511 if (remaining < 4) 2512 return (0); 2513 *p++ = 0xf0 | ((uc >> 18) & 0x07); 2514 *p++ = 0x80 | ((uc >> 12) & 0x3f); 2515 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2516 *p++ = 0x80 | (uc & 0x3f); 2517 } 2518 return (p - _p); 2519 } 2520 2521 static int 2522 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) 2523 { 2524 return (utf16_to_unicode(pwc, s, n, 1)); 2525 } 2526 2527 static int 2528 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) 2529 { 2530 return (utf16_to_unicode(pwc, s, n, 0)); 2531 } 2532 2533 static int 2534 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) 2535 { 2536 const char *utf16 = s; 2537 unsigned uc; 2538 2539 if (n == 0) 2540 return (0); 2541 if (n == 1) { 2542 /* set the Replacement Character instead. */ 2543 *pwc = UNICODE_R_CHAR; 2544 return (-1); 2545 } 2546 2547 if (be) 2548 uc = archive_be16dec(utf16); 2549 else 2550 uc = archive_le16dec(utf16); 2551 utf16 += 2; 2552 2553 /* If this is a surrogate pair, assemble the full code point.*/ 2554 if (IS_HIGH_SURROGATE_LA(uc)) { 2555 unsigned uc2; 2556 2557 if (n >= 4) { 2558 if (be) 2559 uc2 = archive_be16dec(utf16); 2560 else 2561 uc2 = archive_le16dec(utf16); 2562 } else 2563 uc2 = 0; 2564 if (IS_LOW_SURROGATE_LA(uc2)) { 2565 uc = combine_surrogate_pair(uc, uc2); 2566 utf16 += 2; 2567 } else { 2568 /* Undescribed code point should be U+FFFD 2569 * (replacement character). */ 2570 *pwc = UNICODE_R_CHAR; 2571 return (-2); 2572 } 2573 } 2574 2575 /* 2576 * Surrogate pair values(0xd800 through 0xdfff) are only 2577 * used by UTF-16, so, after above calculation, the code 2578 * must not be surrogate values, and Unicode has no codes 2579 * larger than 0x10ffff. Thus, those are not legal Unicode 2580 * values. 2581 */ 2582 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { 2583 /* Undescribed code point should be U+FFFD 2584 * (replacement character). */ 2585 *pwc = UNICODE_R_CHAR; 2586 return (((int)(utf16 - s)) * -1); 2587 } 2588 *pwc = uc; 2589 return ((int)(utf16 - s)); 2590 } 2591 2592 static size_t 2593 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) 2594 { 2595 char *utf16 = p; 2596 2597 if (uc > 0xffff) { 2598 /* We have a code point that won't fit into a 2599 * wchar_t; convert it to a surrogate pair. */ 2600 if (remaining < 4) 2601 return (0); 2602 uc -= 0x10000; 2603 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2604 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2605 return (4); 2606 } else { 2607 if (remaining < 2) 2608 return (0); 2609 archive_be16enc(utf16, uc); 2610 return (2); 2611 } 2612 } 2613 2614 static size_t 2615 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) 2616 { 2617 char *utf16 = p; 2618 2619 if (uc > 0xffff) { 2620 /* We have a code point that won't fit into a 2621 * wchar_t; convert it to a surrogate pair. */ 2622 if (remaining < 4) 2623 return (0); 2624 uc -= 0x10000; 2625 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2626 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2627 return (4); 2628 } else { 2629 if (remaining < 2) 2630 return (0); 2631 archive_le16enc(utf16, uc); 2632 return (2); 2633 } 2634 } 2635 2636 /* 2637 * Copy UTF-8 string in checking surrogate pair. 2638 * If any surrogate pair are found, it would be canonicalized. 2639 */ 2640 static int 2641 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, 2642 size_t len, struct archive_string_conv *sc) 2643 { 2644 const char *s; 2645 char *p, *endp; 2646 int n, ret = 0; 2647 2648 (void)sc; /* UNUSED */ 2649 2650 if (archive_string_ensure(as, as->length + len + 1) == NULL) 2651 return (-1); 2652 2653 s = (const char *)_p; 2654 p = as->s + as->length; 2655 endp = as->s + as->buffer_length -1; 2656 do { 2657 uint32_t uc; 2658 const char *ss = s; 2659 size_t w; 2660 2661 /* 2662 * Forward byte sequence until a conversion of that is needed. 2663 */ 2664 while ((n = utf8_to_unicode(&uc, s, len)) > 0) { 2665 s += n; 2666 len -= n; 2667 } 2668 if (ss < s) { 2669 if (p + (s - ss) > endp) { 2670 as->length = p - as->s; 2671 if (archive_string_ensure(as, 2672 as->buffer_length + len + 1) == NULL) 2673 return (-1); 2674 p = as->s + as->length; 2675 endp = as->s + as->buffer_length -1; 2676 } 2677 2678 memcpy(p, ss, s - ss); 2679 p += s - ss; 2680 } 2681 2682 /* 2683 * If n is negative, current byte sequence needs a replacement. 2684 */ 2685 if (n < 0) { 2686 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { 2687 /* Current byte sequence may be CESU-8. */ 2688 n = cesu8_to_unicode(&uc, s, len); 2689 } 2690 if (n < 0) { 2691 ret = -1; 2692 n *= -1;/* Use a replaced unicode character. */ 2693 } 2694 2695 /* Rebuild UTF-8 byte sequence. */ 2696 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { 2697 as->length = p - as->s; 2698 if (archive_string_ensure(as, 2699 as->buffer_length + len + 1) == NULL) 2700 return (-1); 2701 p = as->s + as->length; 2702 endp = as->s + as->buffer_length -1; 2703 } 2704 p += w; 2705 s += n; 2706 len -= n; 2707 } 2708 } while (n > 0); 2709 as->length = p - as->s; 2710 as->s[as->length] = '\0'; 2711 return (ret); 2712 } 2713 2714 static int 2715 archive_string_append_unicode(struct archive_string *as, const void *_p, 2716 size_t len, struct archive_string_conv *sc) 2717 { 2718 const char *s; 2719 char *p, *endp; 2720 uint32_t uc; 2721 size_t w; 2722 int n, ret = 0, ts, tm; 2723 int (*parse)(uint32_t *, const char *, size_t); 2724 size_t (*unparse)(char *, size_t, uint32_t); 2725 2726 if (sc->flag & SCONV_TO_UTF16BE) { 2727 unparse = unicode_to_utf16be; 2728 ts = 2; 2729 } else if (sc->flag & SCONV_TO_UTF16LE) { 2730 unparse = unicode_to_utf16le; 2731 ts = 2; 2732 } else if (sc->flag & SCONV_TO_UTF8) { 2733 unparse = unicode_to_utf8; 2734 ts = 1; 2735 } else { 2736 /* 2737 * This case is going to be converted to another 2738 * character-set through iconv. 2739 */ 2740 if (sc->flag & SCONV_FROM_UTF16BE) { 2741 unparse = unicode_to_utf16be; 2742 ts = 2; 2743 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2744 unparse = unicode_to_utf16le; 2745 ts = 2; 2746 } else { 2747 unparse = unicode_to_utf8; 2748 ts = 1; 2749 } 2750 } 2751 2752 if (sc->flag & SCONV_FROM_UTF16BE) { 2753 parse = utf16be_to_unicode; 2754 tm = 1; 2755 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2756 parse = utf16le_to_unicode; 2757 tm = 1; 2758 } else { 2759 parse = cesu8_to_unicode; 2760 tm = ts; 2761 } 2762 2763 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2764 return (-1); 2765 2766 s = (const char *)_p; 2767 p = as->s + as->length; 2768 endp = as->s + as->buffer_length - ts; 2769 while ((n = parse(&uc, s, len)) != 0) { 2770 if (n < 0) { 2771 /* Use a replaced unicode character. */ 2772 n *= -1; 2773 ret = -1; 2774 } 2775 s += n; 2776 len -= n; 2777 while ((w = unparse(p, endp - p, uc)) == 0) { 2778 /* There is not enough output buffer so 2779 * we have to expand it. */ 2780 as->length = p - as->s; 2781 if (archive_string_ensure(as, 2782 as->buffer_length + len * tm + ts) == NULL) 2783 return (-1); 2784 p = as->s + as->length; 2785 endp = as->s + as->buffer_length - ts; 2786 } 2787 p += w; 2788 } 2789 as->length = p - as->s; 2790 as->s[as->length] = '\0'; 2791 if (ts == 2) 2792 as->s[as->length+1] = '\0'; 2793 return (ret); 2794 } 2795 2796 /* 2797 * Following Constants for Hangul compositions this information comes from 2798 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ 2799 */ 2800 #define HC_SBASE 0xAC00 2801 #define HC_LBASE 0x1100 2802 #define HC_VBASE 0x1161 2803 #define HC_TBASE 0x11A7 2804 #define HC_LCOUNT 19 2805 #define HC_VCOUNT 21 2806 #define HC_TCOUNT 28 2807 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) 2808 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) 2809 2810 static uint32_t 2811 get_nfc(uint32_t uc, uint32_t uc2) 2812 { 2813 int t, b; 2814 2815 t = 0; 2816 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; 2817 while (b >= t) { 2818 int m = (t + b) / 2; 2819 if (u_composition_table[m].cp1 < uc) 2820 t = m + 1; 2821 else if (u_composition_table[m].cp1 > uc) 2822 b = m - 1; 2823 else if (u_composition_table[m].cp2 < uc2) 2824 t = m + 1; 2825 else if (u_composition_table[m].cp2 > uc2) 2826 b = m - 1; 2827 else 2828 return (u_composition_table[m].nfc); 2829 } 2830 return (0); 2831 } 2832 2833 #define FDC_MAX 10 /* The maximum number of Following Decomposable 2834 * Characters. */ 2835 2836 /* 2837 * Update first code point. 2838 */ 2839 #define UPDATE_UC(new_uc) do { \ 2840 uc = new_uc; \ 2841 ucptr = NULL; \ 2842 } while (0) 2843 2844 /* 2845 * Replace first code point with second code point. 2846 */ 2847 #define REPLACE_UC_WITH_UC2() do { \ 2848 uc = uc2; \ 2849 ucptr = uc2ptr; \ 2850 n = n2; \ 2851 } while (0) 2852 2853 #define EXPAND_BUFFER() do { \ 2854 as->length = p - as->s; \ 2855 if (archive_string_ensure(as, \ 2856 as->buffer_length + len * tm + ts) == NULL)\ 2857 return (-1); \ 2858 p = as->s + as->length; \ 2859 endp = as->s + as->buffer_length - ts; \ 2860 } while (0) 2861 2862 #define UNPARSE(p, endp, uc) do { \ 2863 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ 2864 EXPAND_BUFFER(); \ 2865 } \ 2866 p += w; \ 2867 } while (0) 2868 2869 /* 2870 * Write first code point. 2871 * If the code point has not be changed from its original code, 2872 * this just copies it from its original buffer pointer. 2873 * If not, this converts it to UTF-8 byte sequence and copies it. 2874 */ 2875 #define WRITE_UC() do { \ 2876 if (ucptr) { \ 2877 if (p + n > endp) \ 2878 EXPAND_BUFFER(); \ 2879 switch (n) { \ 2880 case 4: \ 2881 *p++ = *ucptr++; \ 2882 /* FALL THROUGH */ \ 2883 case 3: \ 2884 *p++ = *ucptr++; \ 2885 /* FALL THROUGH */ \ 2886 case 2: \ 2887 *p++ = *ucptr++; \ 2888 /* FALL THROUGH */ \ 2889 case 1: \ 2890 *p++ = *ucptr; \ 2891 break; \ 2892 } \ 2893 ucptr = NULL; \ 2894 } else { \ 2895 UNPARSE(p, endp, uc); \ 2896 } \ 2897 } while (0) 2898 2899 /* 2900 * Collect following decomposable code points. 2901 */ 2902 #define COLLECT_CPS(start) do { \ 2903 int _i; \ 2904 for (_i = start; _i < FDC_MAX ; _i++) { \ 2905 nx = parse(&ucx[_i], s, len); \ 2906 if (nx <= 0) \ 2907 break; \ 2908 cx = CCC(ucx[_i]); \ 2909 if (cl >= cx && cl != 228 && cx != 228)\ 2910 break; \ 2911 s += nx; \ 2912 len -= nx; \ 2913 cl = cx; \ 2914 ccx[_i] = cx; \ 2915 } \ 2916 if (_i >= FDC_MAX) { \ 2917 ret = -1; \ 2918 ucx_size = FDC_MAX; \ 2919 } else \ 2920 ucx_size = _i; \ 2921 } while (0) 2922 2923 /* 2924 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. 2925 * 2926 * TODO: Convert composition exclusions, which are never converted 2927 * from NFC,NFD,NFKC and NFKD, to Form C. 2928 */ 2929 static int 2930 archive_string_normalize_C(struct archive_string *as, const void *_p, 2931 size_t len, struct archive_string_conv *sc) 2932 { 2933 const char *s = (const char *)_p; 2934 char *p, *endp; 2935 uint32_t uc, uc2; 2936 size_t w; 2937 int always_replace, n, n2, ret = 0, spair, ts, tm; 2938 int (*parse)(uint32_t *, const char *, size_t); 2939 size_t (*unparse)(char *, size_t, uint32_t); 2940 2941 always_replace = 1; 2942 ts = 1;/* text size. */ 2943 if (sc->flag & SCONV_TO_UTF16BE) { 2944 unparse = unicode_to_utf16be; 2945 ts = 2; 2946 if (sc->flag & SCONV_FROM_UTF16BE) 2947 always_replace = 0; 2948 } else if (sc->flag & SCONV_TO_UTF16LE) { 2949 unparse = unicode_to_utf16le; 2950 ts = 2; 2951 if (sc->flag & SCONV_FROM_UTF16LE) 2952 always_replace = 0; 2953 } else if (sc->flag & SCONV_TO_UTF8) { 2954 unparse = unicode_to_utf8; 2955 if (sc->flag & SCONV_FROM_UTF8) 2956 always_replace = 0; 2957 } else { 2958 /* 2959 * This case is going to be converted to another 2960 * character-set through iconv. 2961 */ 2962 always_replace = 0; 2963 if (sc->flag & SCONV_FROM_UTF16BE) { 2964 unparse = unicode_to_utf16be; 2965 ts = 2; 2966 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2967 unparse = unicode_to_utf16le; 2968 ts = 2; 2969 } else { 2970 unparse = unicode_to_utf8; 2971 } 2972 } 2973 2974 if (sc->flag & SCONV_FROM_UTF16BE) { 2975 parse = utf16be_to_unicode; 2976 tm = 1; 2977 spair = 4;/* surrogate pair size in UTF-16. */ 2978 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2979 parse = utf16le_to_unicode; 2980 tm = 1; 2981 spair = 4;/* surrogate pair size in UTF-16. */ 2982 } else { 2983 parse = cesu8_to_unicode; 2984 tm = ts; 2985 spair = 6;/* surrogate pair size in UTF-8. */ 2986 } 2987 2988 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2989 return (-1); 2990 2991 p = as->s + as->length; 2992 endp = as->s + as->buffer_length - ts; 2993 while ((n = parse(&uc, s, len)) != 0) { 2994 const char *ucptr, *uc2ptr; 2995 2996 if (n < 0) { 2997 /* Use a replaced unicode character. */ 2998 UNPARSE(p, endp, uc); 2999 s += n*-1; 3000 len -= n*-1; 3001 ret = -1; 3002 continue; 3003 } else if (n == spair || always_replace) 3004 /* uc is converted from a surrogate pair. 3005 * this should be treated as a changed code. */ 3006 ucptr = NULL; 3007 else 3008 ucptr = s; 3009 s += n; 3010 len -= n; 3011 3012 /* Read second code point. */ 3013 while ((n2 = parse(&uc2, s, len)) > 0) { 3014 uint32_t ucx[FDC_MAX]; 3015 int ccx[FDC_MAX]; 3016 int cl, cx, i, nx, ucx_size; 3017 int LIndex,SIndex; 3018 uint32_t nfc; 3019 3020 if (n2 == spair || always_replace) 3021 /* uc2 is converted from a surrogate pair. 3022 * this should be treated as a changed code. */ 3023 uc2ptr = NULL; 3024 else 3025 uc2ptr = s; 3026 s += n2; 3027 len -= n2; 3028 3029 /* 3030 * If current second code point is out of decomposable 3031 * code points, finding compositions is unneeded. 3032 */ 3033 if (!IS_DECOMPOSABLE_BLOCK(uc2)) { 3034 WRITE_UC(); 3035 REPLACE_UC_WITH_UC2(); 3036 continue; 3037 } 3038 3039 /* 3040 * Try to combine current code points. 3041 */ 3042 /* 3043 * We have to combine Hangul characters according to 3044 * http://uniicode.org/reports/tr15/#Hangul 3045 */ 3046 if (0 <= (LIndex = uc - HC_LBASE) && 3047 LIndex < HC_LCOUNT) { 3048 /* 3049 * Hangul Composition. 3050 * 1. Two current code points are L and V. 3051 */ 3052 int VIndex = uc2 - HC_VBASE; 3053 if (0 <= VIndex && VIndex < HC_VCOUNT) { 3054 /* Make syllable of form LV. */ 3055 UPDATE_UC(HC_SBASE + 3056 (LIndex * HC_VCOUNT + VIndex) * 3057 HC_TCOUNT); 3058 } else { 3059 WRITE_UC(); 3060 REPLACE_UC_WITH_UC2(); 3061 } 3062 continue; 3063 } else if (0 <= (SIndex = uc - HC_SBASE) && 3064 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { 3065 /* 3066 * Hangul Composition. 3067 * 2. Two current code points are LV and T. 3068 */ 3069 int TIndex = uc2 - HC_TBASE; 3070 if (0 < TIndex && TIndex < HC_TCOUNT) { 3071 /* Make syllable of form LVT. */ 3072 UPDATE_UC(uc + TIndex); 3073 } else { 3074 WRITE_UC(); 3075 REPLACE_UC_WITH_UC2(); 3076 } 3077 continue; 3078 } else if ((nfc = get_nfc(uc, uc2)) != 0) { 3079 /* A composition to current code points 3080 * is found. */ 3081 UPDATE_UC(nfc); 3082 continue; 3083 } else if ((cl = CCC(uc2)) == 0) { 3084 /* Clearly 'uc2' the second code point is not 3085 * a decomposable code. */ 3086 WRITE_UC(); 3087 REPLACE_UC_WITH_UC2(); 3088 continue; 3089 } 3090 3091 /* 3092 * Collect following decomposable code points. 3093 */ 3094 cx = 0; 3095 ucx[0] = uc2; 3096 ccx[0] = cl; 3097 COLLECT_CPS(1); 3098 3099 /* 3100 * Find a composed code in the collected code points. 3101 */ 3102 i = 1; 3103 while (i < ucx_size) { 3104 int j; 3105 3106 if ((nfc = get_nfc(uc, ucx[i])) == 0) { 3107 i++; 3108 continue; 3109 } 3110 3111 /* 3112 * nfc is composed of uc and ucx[i]. 3113 */ 3114 UPDATE_UC(nfc); 3115 3116 /* 3117 * Remove ucx[i] by shifting 3118 * following code points. 3119 */ 3120 for (j = i; j+1 < ucx_size; j++) { 3121 ucx[j] = ucx[j+1]; 3122 ccx[j] = ccx[j+1]; 3123 } 3124 ucx_size --; 3125 3126 /* 3127 * Collect following code points blocked 3128 * by ucx[i] the removed code point. 3129 */ 3130 if (ucx_size > 0 && i == ucx_size && 3131 nx > 0 && cx == cl) { 3132 cl = ccx[ucx_size-1]; 3133 COLLECT_CPS(ucx_size); 3134 } 3135 /* 3136 * Restart finding a composed code with 3137 * the updated uc from the top of the 3138 * collected code points. 3139 */ 3140 i = 0; 3141 } 3142 3143 /* 3144 * Apparently the current code points are not 3145 * decomposed characters or already composed. 3146 */ 3147 WRITE_UC(); 3148 for (i = 0; i < ucx_size; i++) 3149 UNPARSE(p, endp, ucx[i]); 3150 3151 /* 3152 * Flush out remaining canonical combining characters. 3153 */ 3154 if (nx > 0 && cx == cl && len > 0) { 3155 while ((nx = parse(&ucx[0], s, len)) 3156 > 0) { 3157 cx = CCC(ucx[0]); 3158 if (cl > cx) 3159 break; 3160 s += nx; 3161 len -= nx; 3162 cl = cx; 3163 UNPARSE(p, endp, ucx[0]); 3164 } 3165 } 3166 break; 3167 } 3168 if (n2 < 0) { 3169 WRITE_UC(); 3170 /* Use a replaced unicode character. */ 3171 UNPARSE(p, endp, uc2); 3172 s += n2*-1; 3173 len -= n2*-1; 3174 ret = -1; 3175 continue; 3176 } else if (n2 == 0) { 3177 WRITE_UC(); 3178 break; 3179 } 3180 } 3181 as->length = p - as->s; 3182 as->s[as->length] = '\0'; 3183 if (ts == 2) 3184 as->s[as->length+1] = '\0'; 3185 return (ret); 3186 } 3187 3188 static int 3189 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) 3190 { 3191 int t, b; 3192 3193 /* 3194 * These are not converted to NFD on Mac OS. 3195 */ 3196 if ((uc >= 0x2000 && uc <= 0x2FFF) || 3197 (uc >= 0xF900 && uc <= 0xFAFF) || 3198 (uc >= 0x2F800 && uc <= 0x2FAFF)) 3199 return (0); 3200 /* 3201 * Those code points are not converted to NFD on Mac OS. 3202 * I do not know the reason because it is undocumented. 3203 * NFC NFD 3204 * 1109A ==> 11099 110BA 3205 * 1109C ==> 1109B 110BA 3206 * 110AB ==> 110A5 110BA 3207 */ 3208 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) 3209 return (0); 3210 3211 t = 0; 3212 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; 3213 while (b >= t) { 3214 int m = (t + b) / 2; 3215 if (u_decomposition_table[m].nfc < uc) 3216 t = m + 1; 3217 else if (u_decomposition_table[m].nfc > uc) 3218 b = m - 1; 3219 else { 3220 *cp1 = u_decomposition_table[m].cp1; 3221 *cp2 = u_decomposition_table[m].cp2; 3222 return (1); 3223 } 3224 } 3225 return (0); 3226 } 3227 3228 #define REPLACE_UC_WITH(cp) do { \ 3229 uc = cp; \ 3230 ucptr = NULL; \ 3231 } while (0) 3232 3233 /* 3234 * Normalize UTF-8 characters to Form D and copy the result. 3235 */ 3236 static int 3237 archive_string_normalize_D(struct archive_string *as, const void *_p, 3238 size_t len, struct archive_string_conv *sc) 3239 { 3240 const char *s = (const char *)_p; 3241 char *p, *endp; 3242 uint32_t uc, uc2; 3243 size_t w; 3244 int always_replace, n, n2, ret = 0, spair, ts, tm; 3245 int (*parse)(uint32_t *, const char *, size_t); 3246 size_t (*unparse)(char *, size_t, uint32_t); 3247 3248 always_replace = 1; 3249 ts = 1;/* text size. */ 3250 if (sc->flag & SCONV_TO_UTF16BE) { 3251 unparse = unicode_to_utf16be; 3252 ts = 2; 3253 if (sc->flag & SCONV_FROM_UTF16BE) 3254 always_replace = 0; 3255 } else if (sc->flag & SCONV_TO_UTF16LE) { 3256 unparse = unicode_to_utf16le; 3257 ts = 2; 3258 if (sc->flag & SCONV_FROM_UTF16LE) 3259 always_replace = 0; 3260 } else if (sc->flag & SCONV_TO_UTF8) { 3261 unparse = unicode_to_utf8; 3262 if (sc->flag & SCONV_FROM_UTF8) 3263 always_replace = 0; 3264 } else { 3265 /* 3266 * This case is going to be converted to another 3267 * character-set through iconv. 3268 */ 3269 always_replace = 0; 3270 if (sc->flag & SCONV_FROM_UTF16BE) { 3271 unparse = unicode_to_utf16be; 3272 ts = 2; 3273 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3274 unparse = unicode_to_utf16le; 3275 ts = 2; 3276 } else { 3277 unparse = unicode_to_utf8; 3278 } 3279 } 3280 3281 if (sc->flag & SCONV_FROM_UTF16BE) { 3282 parse = utf16be_to_unicode; 3283 tm = 1; 3284 spair = 4;/* surrogate pair size in UTF-16. */ 3285 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3286 parse = utf16le_to_unicode; 3287 tm = 1; 3288 spair = 4;/* surrogate pair size in UTF-16. */ 3289 } else { 3290 parse = cesu8_to_unicode; 3291 tm = ts; 3292 spair = 6;/* surrogate pair size in UTF-8. */ 3293 } 3294 3295 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 3296 return (-1); 3297 3298 p = as->s + as->length; 3299 endp = as->s + as->buffer_length - ts; 3300 while ((n = parse(&uc, s, len)) != 0) { 3301 const char *ucptr; 3302 uint32_t cp1, cp2; 3303 int SIndex; 3304 struct { 3305 uint32_t uc; 3306 int ccc; 3307 } fdc[FDC_MAX]; 3308 int fdi, fdj; 3309 int ccc; 3310 3311 check_first_code: 3312 if (n < 0) { 3313 /* Use a replaced unicode character. */ 3314 UNPARSE(p, endp, uc); 3315 s += n*-1; 3316 len -= n*-1; 3317 ret = -1; 3318 continue; 3319 } else if (n == spair || always_replace) 3320 /* uc is converted from a surrogate pair. 3321 * this should be treated as a changed code. */ 3322 ucptr = NULL; 3323 else 3324 ucptr = s; 3325 s += n; 3326 len -= n; 3327 3328 /* Hangul Decomposition. */ 3329 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { 3330 int L = HC_LBASE + SIndex / HC_NCOUNT; 3331 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; 3332 int T = HC_TBASE + SIndex % HC_TCOUNT; 3333 3334 REPLACE_UC_WITH(L); 3335 WRITE_UC(); 3336 REPLACE_UC_WITH(V); 3337 WRITE_UC(); 3338 if (T != HC_TBASE) { 3339 REPLACE_UC_WITH(T); 3340 WRITE_UC(); 3341 } 3342 continue; 3343 } 3344 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { 3345 WRITE_UC(); 3346 continue; 3347 } 3348 3349 fdi = 0; 3350 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { 3351 int k; 3352 3353 for (k = fdi; k > 0; k--) 3354 fdc[k] = fdc[k-1]; 3355 fdc[0].ccc = CCC(cp2); 3356 fdc[0].uc = cp2; 3357 fdi++; 3358 REPLACE_UC_WITH(cp1); 3359 } 3360 3361 /* Read following code points. */ 3362 while ((n2 = parse(&uc2, s, len)) > 0 && 3363 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { 3364 int j, k; 3365 3366 s += n2; 3367 len -= n2; 3368 for (j = 0; j < fdi; j++) { 3369 if (fdc[j].ccc > ccc) 3370 break; 3371 } 3372 if (j < fdi) { 3373 for (k = fdi; k > j; k--) 3374 fdc[k] = fdc[k-1]; 3375 fdc[j].ccc = ccc; 3376 fdc[j].uc = uc2; 3377 } else { 3378 fdc[fdi].ccc = ccc; 3379 fdc[fdi].uc = uc2; 3380 } 3381 fdi++; 3382 } 3383 3384 WRITE_UC(); 3385 for (fdj = 0; fdj < fdi; fdj++) { 3386 REPLACE_UC_WITH(fdc[fdj].uc); 3387 WRITE_UC(); 3388 } 3389 3390 if (n2 == 0) 3391 break; 3392 REPLACE_UC_WITH(uc2); 3393 n = n2; 3394 goto check_first_code; 3395 } 3396 as->length = p - as->s; 3397 as->s[as->length] = '\0'; 3398 if (ts == 2) 3399 as->s[as->length+1] = '\0'; 3400 return (ret); 3401 } 3402 3403 /* 3404 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption 3405 * that WCS is Unicode. It is true for several platforms but some are false. 3406 * And then people who did not use UTF-8 locale on the non Unicode WCS 3407 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those 3408 * now cannot get right filename from libarchive 3.x and later since we 3409 * fixed the wrong assumption and it is incompatible to older its versions. 3410 * So we provide special option, "compat-2x.x", for resolving it. 3411 * That option enable the string conversion of libarchive 2.x. 3412 * 3413 * Translates the wrong UTF-8 string made by libarchive 2.x into current 3414 * locale character set and appends to the archive_string. 3415 * Note: returns -1 if conversion fails. 3416 */ 3417 static int 3418 strncat_from_utf8_libarchive2(struct archive_string *as, 3419 const void *_p, size_t len, struct archive_string_conv *sc) 3420 { 3421 const char *s; 3422 int n; 3423 char *p; 3424 char *end; 3425 uint32_t unicode; 3426 #if HAVE_WCRTOMB 3427 mbstate_t shift_state; 3428 3429 memset(&shift_state, 0, sizeof(shift_state)); 3430 #else 3431 /* Clear the shift state before starting. */ 3432 wctomb(NULL, L'\0'); 3433 #endif 3434 (void)sc; /* UNUSED */ 3435 /* 3436 * Allocate buffer for MBS. 3437 * We need this allocation here since it is possible that 3438 * as->s is still NULL. 3439 */ 3440 if (archive_string_ensure(as, as->length + len + 1) == NULL) 3441 return (-1); 3442 3443 s = (const char *)_p; 3444 p = as->s + as->length; 3445 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3446 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { 3447 wchar_t wc; 3448 3449 if (p >= end) { 3450 as->length = p - as->s; 3451 /* Re-allocate buffer for MBS. */ 3452 if (archive_string_ensure(as, 3453 as->length + max(len * 2, 3454 (size_t)MB_CUR_MAX) + 1) == NULL) 3455 return (-1); 3456 p = as->s + as->length; 3457 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3458 } 3459 3460 /* 3461 * As libarchive 2.x, translates the UTF-8 characters into 3462 * wide-characters in the assumption that WCS is Unicode. 3463 */ 3464 if (n < 0) { 3465 n *= -1; 3466 wc = L'?'; 3467 } else 3468 wc = (wchar_t)unicode; 3469 3470 s += n; 3471 len -= n; 3472 /* 3473 * Translates the wide-character into the current locale MBS. 3474 */ 3475 #if HAVE_WCRTOMB 3476 n = (int)wcrtomb(p, wc, &shift_state); 3477 #else 3478 n = (int)wctomb(p, wc); 3479 #endif 3480 if (n == -1) 3481 return (-1); 3482 p += n; 3483 } 3484 as->length = p - as->s; 3485 as->s[as->length] = '\0'; 3486 return (0); 3487 } 3488 3489 3490 /* 3491 * Conversion functions between current locale dependent MBS and UTF-16BE. 3492 * strncat_from_utf16be() : UTF-16BE --> MBS 3493 * strncat_to_utf16be() : MBS --> UTF16BE 3494 */ 3495 3496 #if defined(_WIN32) && !defined(__CYGWIN__) 3497 3498 /* 3499 * Convert a UTF-16BE/LE string to current locale and copy the result. 3500 * Return -1 if conversion fails. 3501 */ 3502 static int 3503 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, 3504 struct archive_string_conv *sc, int be) 3505 { 3506 struct archive_string tmp; 3507 const char *u16; 3508 int ll; 3509 BOOL defchar; 3510 char *mbs; 3511 size_t mbs_size, b; 3512 int ret = 0; 3513 3514 bytes &= ~1; 3515 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3516 return (-1); 3517 3518 mbs = as->s + as->length; 3519 mbs_size = as->buffer_length - as->length -1; 3520 3521 if (sc->to_cp == CP_C_LOCALE) { 3522 /* 3523 * "C" locale special process. 3524 */ 3525 u16 = _p; 3526 ll = 0; 3527 for (b = 0; b < bytes; b += 2) { 3528 uint16_t val; 3529 if (be) 3530 val = archive_be16dec(u16+b); 3531 else 3532 val = archive_le16dec(u16+b); 3533 if (val > 255) { 3534 *mbs++ = '?'; 3535 ret = -1; 3536 } else 3537 *mbs++ = (char)(val&0xff); 3538 ll++; 3539 } 3540 as->length += ll; 3541 as->s[as->length] = '\0'; 3542 return (ret); 3543 } 3544 3545 archive_string_init(&tmp); 3546 if (be) { 3547 if (is_big_endian()) { 3548 u16 = _p; 3549 } else { 3550 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3551 return (-1); 3552 memcpy(tmp.s, _p, bytes); 3553 for (b = 0; b < bytes; b += 2) { 3554 uint16_t val = archive_be16dec(tmp.s+b); 3555 archive_le16enc(tmp.s+b, val); 3556 } 3557 u16 = tmp.s; 3558 } 3559 } else { 3560 if (!is_big_endian()) { 3561 u16 = _p; 3562 } else { 3563 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3564 return (-1); 3565 memcpy(tmp.s, _p, bytes); 3566 for (b = 0; b < bytes; b += 2) { 3567 uint16_t val = archive_le16dec(tmp.s+b); 3568 archive_be16enc(tmp.s+b, val); 3569 } 3570 u16 = tmp.s; 3571 } 3572 } 3573 3574 do { 3575 defchar = 0; 3576 ll = WideCharToMultiByte(sc->to_cp, 0, 3577 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, 3578 NULL, &defchar); 3579 /* Exit loop if we succeeded */ 3580 if (ll != 0 || 3581 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3582 break; 3583 } 3584 /* Else expand buffer and loop to try again. */ 3585 ll = WideCharToMultiByte(sc->to_cp, 0, 3586 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); 3587 if (archive_string_ensure(as, ll +1) == NULL) 3588 return (-1); 3589 mbs = as->s + as->length; 3590 mbs_size = as->buffer_length - as->length -1; 3591 } while (1); 3592 archive_string_free(&tmp); 3593 as->length += ll; 3594 as->s[as->length] = '\0'; 3595 if (ll == 0 || defchar) 3596 ret = -1; 3597 return (ret); 3598 } 3599 3600 static int 3601 win_strncat_from_utf16be(struct archive_string *as, const void *_p, 3602 size_t bytes, struct archive_string_conv *sc) 3603 { 3604 return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); 3605 } 3606 3607 static int 3608 win_strncat_from_utf16le(struct archive_string *as, const void *_p, 3609 size_t bytes, struct archive_string_conv *sc) 3610 { 3611 return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); 3612 } 3613 3614 static int 3615 is_big_endian(void) 3616 { 3617 uint16_t d = 1; 3618 3619 return (archive_be16dec(&d) == 1); 3620 } 3621 3622 /* 3623 * Convert a current locale string to UTF-16BE/LE and copy the result. 3624 * Return -1 if conversion fails. 3625 */ 3626 static int 3627 win_strncat_to_utf16(struct archive_string *as16, const void *_p, 3628 size_t length, struct archive_string_conv *sc, int bigendian) 3629 { 3630 const char *s = (const char *)_p; 3631 char *u16; 3632 size_t count, avail; 3633 3634 if (archive_string_ensure(as16, 3635 as16->length + (length + 1) * 2) == NULL) 3636 return (-1); 3637 3638 u16 = as16->s + as16->length; 3639 avail = as16->buffer_length - 2; 3640 if (sc->from_cp == CP_C_LOCALE) { 3641 /* 3642 * "C" locale special process. 3643 */ 3644 count = 0; 3645 while (count < length && *s) { 3646 if (bigendian) 3647 archive_be16enc(u16, *s); 3648 else 3649 archive_le16enc(u16, *s); 3650 u16 += 2; 3651 s++; 3652 count++; 3653 } 3654 as16->length += count << 1; 3655 as16->s[as16->length] = 0; 3656 as16->s[as16->length+1] = 0; 3657 return (0); 3658 } 3659 do { 3660 count = MultiByteToWideChar(sc->from_cp, 3661 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); 3662 /* Exit loop if we succeeded */ 3663 if (count != 0 || 3664 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3665 break; 3666 } 3667 /* Expand buffer and try again */ 3668 count = MultiByteToWideChar(sc->from_cp, 3669 MB_PRECOMPOSED, s, (int)length, NULL, 0); 3670 if (archive_string_ensure(as16, (count +1) * 2) 3671 == NULL) 3672 return (-1); 3673 u16 = as16->s + as16->length; 3674 avail = as16->buffer_length - 2; 3675 } while (1); 3676 as16->length += count * 2; 3677 as16->s[as16->length] = 0; 3678 as16->s[as16->length+1] = 0; 3679 if (count == 0) 3680 return (-1); 3681 3682 if (is_big_endian()) { 3683 if (!bigendian) { 3684 while (count > 0) { 3685 uint16_t v = archive_be16dec(u16); 3686 archive_le16enc(u16, v); 3687 u16 += 2; 3688 count--; 3689 } 3690 } 3691 } else { 3692 if (bigendian) { 3693 while (count > 0) { 3694 uint16_t v = archive_le16dec(u16); 3695 archive_be16enc(u16, v); 3696 u16 += 2; 3697 count--; 3698 } 3699 } 3700 } 3701 return (0); 3702 } 3703 3704 static int 3705 win_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3706 size_t length, struct archive_string_conv *sc) 3707 { 3708 return (win_strncat_to_utf16(as16, _p, length, sc, 1)); 3709 } 3710 3711 static int 3712 win_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3713 size_t length, struct archive_string_conv *sc) 3714 { 3715 return (win_strncat_to_utf16(as16, _p, length, sc, 0)); 3716 } 3717 3718 #endif /* _WIN32 && !__CYGWIN__ */ 3719 3720 /* 3721 * Do the best effort for conversions. 3722 * We cannot handle UTF-16BE character-set without such iconv, 3723 * but there is a chance if a string consists just ASCII code or 3724 * a current locale is UTF-8. 3725 */ 3726 3727 /* 3728 * Convert a UTF-16BE string to current locale and copy the result. 3729 * Return -1 if conversion fails. 3730 */ 3731 static int 3732 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, 3733 size_t bytes, struct archive_string_conv *sc, int be) 3734 { 3735 const char *utf16 = (const char *)_p; 3736 char *mbs; 3737 uint32_t uc; 3738 int n, ret; 3739 3740 (void)sc; /* UNUSED */ 3741 /* 3742 * Other case, we should do the best effort. 3743 * If all character are ASCII(<0x7f), we can convert it. 3744 * if not , we set a alternative character and return -1. 3745 */ 3746 ret = 0; 3747 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3748 return (-1); 3749 mbs = as->s + as->length; 3750 3751 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { 3752 if (n < 0) { 3753 n *= -1; 3754 ret = -1; 3755 } 3756 bytes -= n; 3757 utf16 += n; 3758 3759 if (uc > 127) { 3760 /* We cannot handle it. */ 3761 *mbs++ = '?'; 3762 ret = -1; 3763 } else 3764 *mbs++ = (char)uc; 3765 } 3766 as->length = mbs - as->s; 3767 as->s[as->length] = '\0'; 3768 return (ret); 3769 } 3770 3771 static int 3772 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, 3773 size_t bytes, struct archive_string_conv *sc) 3774 { 3775 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); 3776 } 3777 3778 static int 3779 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, 3780 size_t bytes, struct archive_string_conv *sc) 3781 { 3782 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); 3783 } 3784 3785 /* 3786 * Convert a current locale string to UTF-16BE/LE and copy the result. 3787 * Return -1 if conversion fails. 3788 */ 3789 static int 3790 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, 3791 size_t length, struct archive_string_conv *sc, int bigendian) 3792 { 3793 const char *s = (const char *)_p; 3794 char *utf16; 3795 size_t remaining; 3796 int ret; 3797 3798 (void)sc; /* UNUSED */ 3799 /* 3800 * Other case, we should do the best effort. 3801 * If all character are ASCII(<0x7f), we can convert it. 3802 * if not , we set a alternative character and return -1. 3803 */ 3804 ret = 0; 3805 remaining = length; 3806 3807 if (archive_string_ensure(as16, 3808 as16->length + (length + 1) * 2) == NULL) 3809 return (-1); 3810 3811 utf16 = as16->s + as16->length; 3812 while (remaining--) { 3813 unsigned c = *s++; 3814 if (c > 127) { 3815 /* We cannot handle it. */ 3816 c = UNICODE_R_CHAR; 3817 ret = -1; 3818 } 3819 if (bigendian) 3820 archive_be16enc(utf16, c); 3821 else 3822 archive_le16enc(utf16, c); 3823 utf16 += 2; 3824 } 3825 as16->length = utf16 - as16->s; 3826 as16->s[as16->length] = 0; 3827 as16->s[as16->length+1] = 0; 3828 return (ret); 3829 } 3830 3831 static int 3832 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3833 size_t length, struct archive_string_conv *sc) 3834 { 3835 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); 3836 } 3837 3838 static int 3839 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3840 size_t length, struct archive_string_conv *sc) 3841 { 3842 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); 3843 } 3844 3845 3846 /* 3847 * Multistring operations. 3848 */ 3849 3850 void 3851 archive_mstring_clean(struct archive_mstring *aes) 3852 { 3853 archive_wstring_free(&(aes->aes_wcs)); 3854 archive_string_free(&(aes->aes_mbs)); 3855 archive_string_free(&(aes->aes_utf8)); 3856 archive_string_free(&(aes->aes_mbs_in_locale)); 3857 aes->aes_set = 0; 3858 } 3859 3860 void 3861 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) 3862 { 3863 dest->aes_set = src->aes_set; 3864 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); 3865 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); 3866 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); 3867 } 3868 3869 int 3870 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, 3871 const char **p) 3872 { 3873 struct archive_string_conv *sc; 3874 int r; 3875 3876 /* If we already have a UTF8 form, return that immediately. */ 3877 if (aes->aes_set & AES_SET_UTF8) { 3878 *p = aes->aes_utf8.s; 3879 return (0); 3880 } 3881 3882 *p = NULL; 3883 if (aes->aes_set & AES_SET_MBS) { 3884 sc = archive_string_conversion_to_charset(a, "UTF-8", 1); 3885 if (sc == NULL) 3886 return (-1);/* Couldn't allocate memory for sc. */ 3887 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, 3888 aes->aes_mbs.length, sc); 3889 if (a == NULL) 3890 free_sconv_object(sc); 3891 if (r == 0) { 3892 aes->aes_set |= AES_SET_UTF8; 3893 *p = aes->aes_utf8.s; 3894 return (0);/* success. */ 3895 } else 3896 return (-1);/* failure. */ 3897 } 3898 return (0);/* success. */ 3899 } 3900 3901 int 3902 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, 3903 const char **p) 3904 { 3905 int r, ret = 0; 3906 3907 (void)a; /* UNUSED */ 3908 /* If we already have an MBS form, return that immediately. */ 3909 if (aes->aes_set & AES_SET_MBS) { 3910 *p = aes->aes_mbs.s; 3911 return (ret); 3912 } 3913 3914 *p = NULL; 3915 /* If there's a WCS form, try converting with the native locale. */ 3916 if (aes->aes_set & AES_SET_WCS) { 3917 archive_string_empty(&(aes->aes_mbs)); 3918 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3919 aes->aes_wcs.s, aes->aes_wcs.length); 3920 *p = aes->aes_mbs.s; 3921 if (r == 0) { 3922 aes->aes_set |= AES_SET_MBS; 3923 return (ret); 3924 } else 3925 ret = -1; 3926 } 3927 3928 /* 3929 * Only a UTF-8 form cannot avail because its conversion already 3930 * failed at archive_mstring_update_utf8(). 3931 */ 3932 return (ret); 3933 } 3934 3935 int 3936 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, 3937 const wchar_t **wp) 3938 { 3939 int r, ret = 0; 3940 3941 (void)a;/* UNUSED */ 3942 /* Return WCS form if we already have it. */ 3943 if (aes->aes_set & AES_SET_WCS) { 3944 *wp = aes->aes_wcs.s; 3945 return (ret); 3946 } 3947 3948 *wp = NULL; 3949 /* Try converting MBS to WCS using native locale. */ 3950 if (aes->aes_set & AES_SET_MBS) { 3951 archive_wstring_empty(&(aes->aes_wcs)); 3952 r = archive_wstring_append_from_mbs(&(aes->aes_wcs), 3953 aes->aes_mbs.s, aes->aes_mbs.length); 3954 if (r == 0) { 3955 aes->aes_set |= AES_SET_WCS; 3956 *wp = aes->aes_wcs.s; 3957 } else 3958 ret = -1;/* failure. */ 3959 } 3960 return (ret); 3961 } 3962 3963 int 3964 archive_mstring_get_mbs_l(struct archive_mstring *aes, 3965 const char **p, size_t *length, struct archive_string_conv *sc) 3966 { 3967 int r, ret = 0; 3968 3969 #if defined(_WIN32) && !defined(__CYGWIN__) 3970 /* 3971 * Internationalization programming on Windows must use Wide 3972 * characters because Windows platform cannot make locale UTF-8. 3973 */ 3974 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { 3975 archive_string_empty(&(aes->aes_mbs_in_locale)); 3976 r = archive_string_append_from_wcs_in_codepage( 3977 &(aes->aes_mbs_in_locale), aes->aes_wcs.s, 3978 aes->aes_wcs.length, sc); 3979 if (r == 0) { 3980 *p = aes->aes_mbs_in_locale.s; 3981 if (length != NULL) 3982 *length = aes->aes_mbs_in_locale.length; 3983 return (0); 3984 } else if (errno == ENOMEM) 3985 return (-1); 3986 else 3987 ret = -1; 3988 } 3989 #endif 3990 3991 /* If there is not an MBS form but is a WCS form, try converting 3992 * with the native locale to be used for translating it to specified 3993 * character-set. */ 3994 if ((aes->aes_set & AES_SET_MBS) == 0 && 3995 (aes->aes_set & AES_SET_WCS) != 0) { 3996 archive_string_empty(&(aes->aes_mbs)); 3997 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3998 aes->aes_wcs.s, aes->aes_wcs.length); 3999 if (r == 0) 4000 aes->aes_set |= AES_SET_MBS; 4001 else if (errno == ENOMEM) 4002 return (-1); 4003 else 4004 ret = -1; 4005 } 4006 /* If we already have an MBS form, use it to be translated to 4007 * specified character-set. */ 4008 if (aes->aes_set & AES_SET_MBS) { 4009 if (sc == NULL) { 4010 /* Conversion is unneeded. */ 4011 *p = aes->aes_mbs.s; 4012 if (length != NULL) 4013 *length = aes->aes_mbs.length; 4014 return (0); 4015 } 4016 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), 4017 aes->aes_mbs.s, aes->aes_mbs.length, sc); 4018 *p = aes->aes_mbs_in_locale.s; 4019 if (length != NULL) 4020 *length = aes->aes_mbs_in_locale.length; 4021 } else { 4022 *p = NULL; 4023 if (length != NULL) 4024 *length = 0; 4025 } 4026 return (ret); 4027 } 4028 4029 int 4030 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) 4031 { 4032 if (mbs == NULL) { 4033 aes->aes_set = 0; 4034 return (0); 4035 } 4036 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); 4037 } 4038 4039 int 4040 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, 4041 size_t len) 4042 { 4043 if (mbs == NULL) { 4044 aes->aes_set = 0; 4045 return (0); 4046 } 4047 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4048 archive_strncpy(&(aes->aes_mbs), mbs, len); 4049 archive_string_empty(&(aes->aes_utf8)); 4050 archive_wstring_empty(&(aes->aes_wcs)); 4051 return (0); 4052 } 4053 4054 int 4055 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) 4056 { 4057 return archive_mstring_copy_wcs_len(aes, wcs, 4058 wcs == NULL ? 0 : wcslen(wcs)); 4059 } 4060 4061 int 4062 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) 4063 { 4064 if (utf8 == NULL) { 4065 aes->aes_set = 0; 4066 return (0); 4067 } 4068 aes->aes_set = AES_SET_UTF8; 4069 archive_string_empty(&(aes->aes_mbs)); 4070 archive_string_empty(&(aes->aes_wcs)); 4071 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); 4072 return (int)strlen(utf8); 4073 } 4074 4075 int 4076 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, 4077 size_t len) 4078 { 4079 if (wcs == NULL) { 4080 aes->aes_set = 0; 4081 return (0); 4082 } 4083 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ 4084 archive_string_empty(&(aes->aes_mbs)); 4085 archive_string_empty(&(aes->aes_utf8)); 4086 archive_wstrncpy(&(aes->aes_wcs), wcs, len); 4087 return (0); 4088 } 4089 4090 int 4091 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, 4092 const char *mbs, size_t len, struct archive_string_conv *sc) 4093 { 4094 int r; 4095 4096 if (mbs == NULL) { 4097 aes->aes_set = 0; 4098 return (0); 4099 } 4100 archive_string_empty(&(aes->aes_mbs)); 4101 archive_wstring_empty(&(aes->aes_wcs)); 4102 archive_string_empty(&(aes->aes_utf8)); 4103 #if defined(_WIN32) && !defined(__CYGWIN__) 4104 /* 4105 * Internationalization programming on Windows must use Wide 4106 * characters because Windows platform cannot make locale UTF-8. 4107 */ 4108 if (sc == NULL) { 4109 if (archive_string_append(&(aes->aes_mbs), 4110 mbs, mbsnbytes(mbs, len)) == NULL) { 4111 aes->aes_set = 0; 4112 r = -1; 4113 } else { 4114 aes->aes_set = AES_SET_MBS; 4115 r = 0; 4116 } 4117 #if defined(HAVE_ICONV) 4118 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { 4119 /* 4120 * This case happens only when MultiByteToWideChar() cannot 4121 * handle sc->from_cp, and we have to iconv in order to 4122 * translate character-set to wchar_t,UTF-16. 4123 */ 4124 iconv_t cd = sc->cd; 4125 unsigned from_cp; 4126 int flag; 4127 4128 /* 4129 * Translate multi-bytes from some character-set to UTF-8. 4130 */ 4131 sc->cd = sc->cd_w; 4132 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); 4133 sc->cd = cd; 4134 if (r != 0) { 4135 aes->aes_set = 0; 4136 return (r); 4137 } 4138 aes->aes_set = AES_SET_UTF8; 4139 4140 /* 4141 * Append the UTF-8 string into wstring. 4142 */ 4143 flag = sc->flag; 4144 sc->flag &= ~(SCONV_NORMALIZATION_C 4145 | SCONV_TO_UTF16| SCONV_FROM_UTF16); 4146 from_cp = sc->from_cp; 4147 sc->from_cp = CP_UTF8; 4148 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), 4149 aes->aes_utf8.s, aes->aes_utf8.length, sc); 4150 sc->flag = flag; 4151 sc->from_cp = from_cp; 4152 if (r == 0) 4153 aes->aes_set |= AES_SET_WCS; 4154 #endif 4155 } else { 4156 r = archive_wstring_append_from_mbs_in_codepage( 4157 &(aes->aes_wcs), mbs, len, sc); 4158 if (r == 0) 4159 aes->aes_set = AES_SET_WCS; 4160 else 4161 aes->aes_set = 0; 4162 } 4163 #else 4164 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); 4165 if (r == 0) 4166 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4167 else 4168 aes->aes_set = 0; 4169 #endif 4170 return (r); 4171 } 4172 4173 /* 4174 * The 'update' form tries to proactively update all forms of 4175 * this string (WCS and MBS) and returns an error if any of 4176 * them fail. This is used by the 'pax' handler, for instance, 4177 * to detect and report character-conversion failures early while 4178 * still allowing clients to get potentially useful values from 4179 * the more tolerant lazy conversions. (get_mbs and get_wcs will 4180 * strive to give the user something useful, so you can get hopefully 4181 * usable values even if some of the character conversions are failing.) 4182 */ 4183 int 4184 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, 4185 const char *utf8) 4186 { 4187 struct archive_string_conv *sc; 4188 int r; 4189 4190 if (utf8 == NULL) { 4191 aes->aes_set = 0; 4192 return (0); /* Succeeded in clearing everything. */ 4193 } 4194 4195 /* Save the UTF8 string. */ 4196 archive_strcpy(&(aes->aes_utf8), utf8); 4197 4198 /* Empty the mbs and wcs strings. */ 4199 archive_string_empty(&(aes->aes_mbs)); 4200 archive_wstring_empty(&(aes->aes_wcs)); 4201 4202 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ 4203 4204 /* Try converting UTF-8 to MBS, return false on failure. */ 4205 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 4206 if (sc == NULL) 4207 return (-1);/* Couldn't allocate memory for sc. */ 4208 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); 4209 if (a == NULL) 4210 free_sconv_object(sc); 4211 if (r != 0) 4212 return (-1); 4213 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ 4214 4215 /* Try converting MBS to WCS, return false on failure. */ 4216 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, 4217 aes->aes_mbs.length)) 4218 return (-1); 4219 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; 4220 4221 /* All conversions succeeded. */ 4222 return (0); 4223 } 4224