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, 748 (int)as->buffer_length - (int)as->length - 1, NULL, dp); 749 if (count == 0 && 750 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 751 /* Expand the MBS buffer and retry. */ 752 if (NULL == archive_string_ensure(as, 753 as->buffer_length + len)) 754 return (-1); 755 continue; 756 } 757 if (count == 0) 758 ret = -1; 759 break; 760 } while (1); 761 } 762 as->length += count; 763 as->s[as->length] = '\0'; 764 return (defchar_used?-1:ret); 765 } 766 767 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) 768 769 /* 770 * Translates a wide character string into current locale character set 771 * and appends to the archive_string. Note: returns -1 if conversion 772 * fails. 773 */ 774 int 775 archive_string_append_from_wcs(struct archive_string *as, 776 const wchar_t *w, size_t len) 777 { 778 /* We cannot use the standard wcstombs() here because it 779 * cannot tell us how big the output buffer should be. So 780 * I've built a loop around wcrtomb() or wctomb() that 781 * converts a character at a time and resizes the string as 782 * needed. We prefer wcrtomb() when it's available because 783 * it's thread-safe. */ 784 int n, ret_val = 0; 785 char *p; 786 char *end; 787 #if HAVE_WCRTOMB 788 mbstate_t shift_state; 789 790 memset(&shift_state, 0, sizeof(shift_state)); 791 #else 792 /* Clear the shift state before starting. */ 793 wctomb(NULL, L'\0'); 794 #endif 795 /* 796 * Allocate buffer for MBS. 797 * We need this allocation here since it is possible that 798 * as->s is still NULL. 799 */ 800 if (archive_string_ensure(as, as->length + len + 1) == NULL) 801 return (-1); 802 803 p = as->s + as->length; 804 end = as->s + as->buffer_length - MB_CUR_MAX -1; 805 while (*w != L'\0' && len > 0) { 806 if (p >= end) { 807 as->length = p - as->s; 808 as->s[as->length] = '\0'; 809 /* Re-allocate buffer for MBS. */ 810 if (archive_string_ensure(as, 811 as->length + max(len * 2, 812 (size_t)MB_CUR_MAX) + 1) == NULL) 813 return (-1); 814 p = as->s + as->length; 815 end = as->s + as->buffer_length - MB_CUR_MAX -1; 816 } 817 #if HAVE_WCRTOMB 818 n = wcrtomb(p, *w++, &shift_state); 819 #else 820 n = wctomb(p, *w++); 821 #endif 822 if (n == -1) { 823 if (errno == EILSEQ) { 824 /* Skip an illegal wide char. */ 825 *p++ = '?'; 826 ret_val = -1; 827 } else { 828 ret_val = -1; 829 break; 830 } 831 } else 832 p += n; 833 len--; 834 } 835 as->length = p - as->s; 836 as->s[as->length] = '\0'; 837 return (ret_val); 838 } 839 840 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */ 841 842 /* 843 * TODO: Test if __STDC_ISO_10646__ is defined. 844 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion 845 * one character at a time. If a non-Windows platform doesn't have 846 * either of these, fall back to the built-in UTF8 conversion. 847 */ 848 int 849 archive_string_append_from_wcs(struct archive_string *as, 850 const wchar_t *w, size_t len) 851 { 852 (void)as;/* UNUSED */ 853 (void)w;/* UNUSED */ 854 (void)len;/* UNUSED */ 855 errno = ENOSYS; 856 return (-1); 857 } 858 859 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ 860 861 /* 862 * Find a string conversion object by a pair of 'from' charset name 863 * and 'to' charset name from an archive object. 864 * Return NULL if not found. 865 */ 866 static struct archive_string_conv * 867 find_sconv_object(struct archive *a, const char *fc, const char *tc) 868 { 869 struct archive_string_conv *sc; 870 871 if (a == NULL) 872 return (NULL); 873 874 for (sc = a->sconv; sc != NULL; sc = sc->next) { 875 if (strcmp(sc->from_charset, fc) == 0 && 876 strcmp(sc->to_charset, tc) == 0) 877 break; 878 } 879 return (sc); 880 } 881 882 /* 883 * Register a string object to an archive object. 884 */ 885 static void 886 add_sconv_object(struct archive *a, struct archive_string_conv *sc) 887 { 888 struct archive_string_conv **psc; 889 890 /* Add a new sconv to sconv list. */ 891 psc = &(a->sconv); 892 while (*psc != NULL) 893 psc = &((*psc)->next); 894 *psc = sc; 895 } 896 897 static void 898 add_converter(struct archive_string_conv *sc, int (*converter) 899 (struct archive_string *, const void *, size_t, 900 struct archive_string_conv *)) 901 { 902 if (sc == NULL || sc->nconverter >= 2) 903 __archive_errx(1, "Programming error"); 904 sc->converter[sc->nconverter++] = converter; 905 } 906 907 static void 908 setup_converter(struct archive_string_conv *sc) 909 { 910 911 /* Reset. */ 912 sc->nconverter = 0; 913 914 /* 915 * Perform special sequence for the incorrect UTF-8 filenames 916 * made by libarchive2.x. 917 */ 918 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { 919 add_converter(sc, strncat_from_utf8_libarchive2); 920 return; 921 } 922 923 /* 924 * Convert a string to UTF-16BE/LE. 925 */ 926 if (sc->flag & SCONV_TO_UTF16) { 927 /* 928 * If the current locale is UTF-8, we can translate 929 * a UTF-8 string into a UTF-16BE string. 930 */ 931 if (sc->flag & SCONV_FROM_UTF8) { 932 add_converter(sc, archive_string_append_unicode); 933 return; 934 } 935 936 #if defined(_WIN32) && !defined(__CYGWIN__) 937 if (sc->flag & SCONV_WIN_CP) { 938 if (sc->flag & SCONV_TO_UTF16BE) 939 add_converter(sc, win_strncat_to_utf16be); 940 else 941 add_converter(sc, win_strncat_to_utf16le); 942 return; 943 } 944 #endif 945 946 #if defined(HAVE_ICONV) 947 if (sc->cd != (iconv_t)-1) { 948 add_converter(sc, iconv_strncat_in_locale); 949 return; 950 } 951 #endif 952 953 if (sc->flag & SCONV_BEST_EFFORT) { 954 if (sc->flag & SCONV_TO_UTF16BE) 955 add_converter(sc, 956 best_effort_strncat_to_utf16be); 957 else 958 add_converter(sc, 959 best_effort_strncat_to_utf16le); 960 } else 961 /* Make sure we have no converter. */ 962 sc->nconverter = 0; 963 return; 964 } 965 966 /* 967 * Convert a string from UTF-16BE/LE. 968 */ 969 if (sc->flag & SCONV_FROM_UTF16) { 970 /* 971 * At least we should normalize a UTF-16BE string. 972 */ 973 if (sc->flag & SCONV_NORMALIZATION_D) 974 add_converter(sc,archive_string_normalize_D); 975 else if (sc->flag & SCONV_NORMALIZATION_C) 976 add_converter(sc, archive_string_normalize_C); 977 978 if (sc->flag & SCONV_TO_UTF8) { 979 /* 980 * If the current locale is UTF-8, we can translate 981 * a UTF-16BE/LE string into a UTF-8 string directly. 982 */ 983 if (!(sc->flag & 984 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 985 add_converter(sc, 986 archive_string_append_unicode); 987 return; 988 } 989 990 #if defined(_WIN32) && !defined(__CYGWIN__) 991 if (sc->flag & SCONV_WIN_CP) { 992 if (sc->flag & SCONV_FROM_UTF16BE) 993 add_converter(sc, win_strncat_from_utf16be); 994 else 995 add_converter(sc, win_strncat_from_utf16le); 996 return; 997 } 998 #endif 999 1000 #if defined(HAVE_ICONV) 1001 if (sc->cd != (iconv_t)-1) { 1002 add_converter(sc, iconv_strncat_in_locale); 1003 return; 1004 } 1005 #endif 1006 1007 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 1008 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 1009 add_converter(sc, best_effort_strncat_from_utf16be); 1010 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 1011 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 1012 add_converter(sc, best_effort_strncat_from_utf16le); 1013 else 1014 /* Make sure we have no converter. */ 1015 sc->nconverter = 0; 1016 return; 1017 } 1018 1019 if (sc->flag & SCONV_FROM_UTF8) { 1020 /* 1021 * At least we should normalize a UTF-8 string. 1022 */ 1023 if (sc->flag & SCONV_NORMALIZATION_D) 1024 add_converter(sc,archive_string_normalize_D); 1025 else if (sc->flag & SCONV_NORMALIZATION_C) 1026 add_converter(sc, archive_string_normalize_C); 1027 1028 /* 1029 * Copy UTF-8 string with a check of CESU-8. 1030 * Apparently, iconv does not check surrogate pairs in UTF-8 1031 * when both from-charset and to-charset are UTF-8, and then 1032 * we use our UTF-8 copy code. 1033 */ 1034 if (sc->flag & SCONV_TO_UTF8) { 1035 /* 1036 * If the current locale is UTF-8, we can translate 1037 * a UTF-16BE string into a UTF-8 string directly. 1038 */ 1039 if (!(sc->flag & 1040 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 1041 add_converter(sc, strncat_from_utf8_to_utf8); 1042 return; 1043 } 1044 } 1045 1046 #if defined(_WIN32) && !defined(__CYGWIN__) 1047 /* 1048 * On Windows we can use Windows API for a string conversion. 1049 */ 1050 if (sc->flag & SCONV_WIN_CP) { 1051 add_converter(sc, strncat_in_codepage); 1052 return; 1053 } 1054 #endif 1055 1056 #if HAVE_ICONV 1057 if (sc->cd != (iconv_t)-1) { 1058 add_converter(sc, iconv_strncat_in_locale); 1059 /* 1060 * iconv generally does not support UTF-8-MAC and so 1061 * we have to the output of iconv from NFC to NFD if 1062 * need. 1063 */ 1064 if ((sc->flag & SCONV_FROM_CHARSET) && 1065 (sc->flag & SCONV_TO_UTF8)) { 1066 if (sc->flag & SCONV_NORMALIZATION_D) 1067 add_converter(sc, archive_string_normalize_D); 1068 } 1069 return; 1070 } 1071 #endif 1072 1073 /* 1074 * Try conversion in the best effort or no conversion. 1075 */ 1076 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) 1077 add_converter(sc, best_effort_strncat_in_locale); 1078 else 1079 /* Make sure we have no converter. */ 1080 sc->nconverter = 0; 1081 } 1082 1083 /* 1084 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE 1085 * and CP932 which are referenced in create_sconv_object(). 1086 */ 1087 static const char * 1088 canonical_charset_name(const char *charset) 1089 { 1090 char cs[16]; 1091 char *p; 1092 const char *s; 1093 1094 if (charset == NULL || charset[0] == '\0' 1095 || strlen(charset) > 15) 1096 return (charset); 1097 1098 /* Copy name to uppercase. */ 1099 p = cs; 1100 s = charset; 1101 while (*s) { 1102 char c = *s++; 1103 if (c >= 'a' && c <= 'z') 1104 c -= 'a' - 'A'; 1105 *p++ = c; 1106 } 1107 *p++ = '\0'; 1108 1109 if (strcmp(cs, "UTF-8") == 0 || 1110 strcmp(cs, "UTF8") == 0) 1111 return ("UTF-8"); 1112 if (strcmp(cs, "UTF-16BE") == 0 || 1113 strcmp(cs, "UTF16BE") == 0) 1114 return ("UTF-16BE"); 1115 if (strcmp(cs, "UTF-16LE") == 0 || 1116 strcmp(cs, "UTF16LE") == 0) 1117 return ("UTF-16LE"); 1118 if (strcmp(cs, "CP932") == 0) 1119 return ("CP932"); 1120 return (charset); 1121 } 1122 1123 /* 1124 * Create a string conversion object. 1125 */ 1126 static struct archive_string_conv * 1127 create_sconv_object(const char *fc, const char *tc, 1128 unsigned current_codepage, int flag) 1129 { 1130 struct archive_string_conv *sc; 1131 1132 sc = calloc(1, sizeof(*sc)); 1133 if (sc == NULL) 1134 return (NULL); 1135 sc->next = NULL; 1136 sc->from_charset = strdup(fc); 1137 if (sc->from_charset == NULL) { 1138 free(sc); 1139 return (NULL); 1140 } 1141 sc->to_charset = strdup(tc); 1142 if (sc->to_charset == NULL) { 1143 free(sc->from_charset); 1144 free(sc); 1145 return (NULL); 1146 } 1147 archive_string_init(&sc->utftmp); 1148 1149 if (flag & SCONV_TO_CHARSET) { 1150 /* 1151 * Convert characters from the current locale charset to 1152 * a specified charset. 1153 */ 1154 sc->from_cp = current_codepage; 1155 sc->to_cp = make_codepage_from_charset(tc); 1156 #if defined(_WIN32) && !defined(__CYGWIN__) 1157 if (IsValidCodePage(sc->to_cp)) 1158 flag |= SCONV_WIN_CP; 1159 #endif 1160 } else if (flag & SCONV_FROM_CHARSET) { 1161 /* 1162 * Convert characters from a specified charset to 1163 * the current locale charset. 1164 */ 1165 sc->to_cp = current_codepage; 1166 sc->from_cp = make_codepage_from_charset(fc); 1167 #if defined(_WIN32) && !defined(__CYGWIN__) 1168 if (IsValidCodePage(sc->from_cp)) 1169 flag |= SCONV_WIN_CP; 1170 #endif 1171 } 1172 1173 /* 1174 * Check if "from charset" and "to charset" are the same. 1175 */ 1176 if (strcmp(fc, tc) == 0 || 1177 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) 1178 sc->same = 1; 1179 else 1180 sc->same = 0; 1181 1182 /* 1183 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. 1184 */ 1185 if (strcmp(tc, "UTF-8") == 0) 1186 flag |= SCONV_TO_UTF8; 1187 else if (strcmp(tc, "UTF-16BE") == 0) 1188 flag |= SCONV_TO_UTF16BE; 1189 else if (strcmp(tc, "UTF-16LE") == 0) 1190 flag |= SCONV_TO_UTF16LE; 1191 if (strcmp(fc, "UTF-8") == 0) 1192 flag |= SCONV_FROM_UTF8; 1193 else if (strcmp(fc, "UTF-16BE") == 0) 1194 flag |= SCONV_FROM_UTF16BE; 1195 else if (strcmp(fc, "UTF-16LE") == 0) 1196 flag |= SCONV_FROM_UTF16LE; 1197 #if defined(_WIN32) && !defined(__CYGWIN__) 1198 if (sc->to_cp == CP_UTF8) 1199 flag |= SCONV_TO_UTF8; 1200 else if (sc->to_cp == CP_UTF16BE) 1201 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; 1202 else if (sc->to_cp == CP_UTF16LE) 1203 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; 1204 if (sc->from_cp == CP_UTF8) 1205 flag |= SCONV_FROM_UTF8; 1206 else if (sc->from_cp == CP_UTF16BE) 1207 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; 1208 else if (sc->from_cp == CP_UTF16LE) 1209 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; 1210 #endif 1211 1212 /* 1213 * Set a flag for Unicode NFD. Usually iconv cannot correctly 1214 * handle it. So we have to translate NFD characters to NFC ones 1215 * ourselves before iconv handles. Another reason is to prevent 1216 * that the same sight of two filenames, one is NFC and other 1217 * is NFD, would be in its directory. 1218 * On Mac OS X, although its filesystem layer automatically 1219 * convert filenames to NFD, it would be useful for filename 1220 * comparing to find out the same filenames that we normalize 1221 * that to be NFD ourselves. 1222 */ 1223 if ((flag & SCONV_FROM_CHARSET) && 1224 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { 1225 #if defined(__APPLE__) 1226 if (flag & SCONV_TO_UTF8) 1227 flag |= SCONV_NORMALIZATION_D; 1228 else 1229 #endif 1230 flag |= SCONV_NORMALIZATION_C; 1231 } 1232 #if defined(__APPLE__) 1233 /* 1234 * In case writing an archive file, make sure that a filename 1235 * going to be passed to iconv is a Unicode NFC string since 1236 * a filename in HFS Plus filesystem is a Unicode NFD one and 1237 * iconv cannot handle it with "UTF-8" charset. It is simpler 1238 * than a use of "UTF-8-MAC" charset. 1239 */ 1240 if ((flag & SCONV_TO_CHARSET) && 1241 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1242 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1243 flag |= SCONV_NORMALIZATION_C; 1244 /* 1245 * In case reading an archive file. make sure that a filename 1246 * will be passed to users is a Unicode NFD string in order to 1247 * correctly compare the filename with other one which comes 1248 * from HFS Plus filesystem. 1249 */ 1250 if ((flag & SCONV_FROM_CHARSET) && 1251 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1252 (flag & SCONV_TO_UTF8)) 1253 flag |= SCONV_NORMALIZATION_D; 1254 #endif 1255 1256 #if defined(HAVE_ICONV) 1257 sc->cd_w = (iconv_t)-1; 1258 /* 1259 * Create an iconv object. 1260 */ 1261 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && 1262 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || 1263 (flag & SCONV_WIN_CP)) { 1264 /* This case we won't use iconv. */ 1265 sc->cd = (iconv_t)-1; 1266 } else { 1267 sc->cd = iconv_open(tc, fc); 1268 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { 1269 /* 1270 * Unfortunately, all of iconv implements do support 1271 * "CP932" character-set, so we should use "SJIS" 1272 * instead if iconv_open failed. 1273 */ 1274 if (strcmp(tc, "CP932") == 0) 1275 sc->cd = iconv_open("SJIS", fc); 1276 else if (strcmp(fc, "CP932") == 0) 1277 sc->cd = iconv_open(tc, "SJIS"); 1278 } 1279 #if defined(_WIN32) && !defined(__CYGWIN__) 1280 /* 1281 * archive_mstring on Windows directly convert multi-bytes 1282 * into archive_wstring in order not to depend on locale 1283 * so that you can do a I18N programming. This will be 1284 * used only in archive_mstring_copy_mbs_len_l so far. 1285 */ 1286 if (flag & SCONV_FROM_CHARSET) { 1287 sc->cd_w = iconv_open("UTF-8", fc); 1288 if (sc->cd_w == (iconv_t)-1 && 1289 (sc->flag & SCONV_BEST_EFFORT)) { 1290 if (strcmp(fc, "CP932") == 0) 1291 sc->cd_w = iconv_open("UTF-8", "SJIS"); 1292 } 1293 } 1294 #endif /* _WIN32 && !__CYGWIN__ */ 1295 } 1296 #endif /* HAVE_ICONV */ 1297 1298 sc->flag = flag; 1299 1300 /* 1301 * Set up converters. 1302 */ 1303 setup_converter(sc); 1304 1305 return (sc); 1306 } 1307 1308 /* 1309 * Free a string conversion object. 1310 */ 1311 static void 1312 free_sconv_object(struct archive_string_conv *sc) 1313 { 1314 free(sc->from_charset); 1315 free(sc->to_charset); 1316 archive_string_free(&sc->utftmp); 1317 #if HAVE_ICONV 1318 if (sc->cd != (iconv_t)-1) 1319 iconv_close(sc->cd); 1320 if (sc->cd_w != (iconv_t)-1) 1321 iconv_close(sc->cd_w); 1322 #endif 1323 free(sc); 1324 } 1325 1326 #if defined(_WIN32) && !defined(__CYGWIN__) 1327 # if defined(WINAPI_FAMILY_PARTITION) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) 1328 # define GetOEMCP() CP_OEMCP 1329 # endif 1330 1331 static unsigned 1332 my_atoi(const char *p) 1333 { 1334 unsigned cp; 1335 1336 cp = 0; 1337 while (*p) { 1338 if (*p >= '0' && *p <= '9') 1339 cp = cp * 10 + (*p - '0'); 1340 else 1341 return (-1); 1342 p++; 1343 } 1344 return (cp); 1345 } 1346 1347 /* 1348 * Translate Charset name (as used by iconv) into CodePage (as used by Windows) 1349 * Return -1 if failed. 1350 * 1351 * Note: This translation code may be insufficient. 1352 */ 1353 static struct charset { 1354 const char *name; 1355 unsigned cp; 1356 } charsets[] = { 1357 /* MUST BE SORTED! */ 1358 {"ASCII", 1252}, 1359 {"ASMO-708", 708}, 1360 {"BIG5", 950}, 1361 {"CHINESE", 936}, 1362 {"CP367", 1252}, 1363 {"CP819", 1252}, 1364 {"CP1025", 21025}, 1365 {"DOS-720", 720}, 1366 {"DOS-862", 862}, 1367 {"EUC-CN", 51936}, 1368 {"EUC-JP", 51932}, 1369 {"EUC-KR", 949}, 1370 {"EUCCN", 51936}, 1371 {"EUCJP", 51932}, 1372 {"EUCKR", 949}, 1373 {"GB18030", 54936}, 1374 {"GB2312", 936}, 1375 {"HEBREW", 1255}, 1376 {"HZ-GB-2312", 52936}, 1377 {"IBM273", 20273}, 1378 {"IBM277", 20277}, 1379 {"IBM278", 20278}, 1380 {"IBM280", 20280}, 1381 {"IBM284", 20284}, 1382 {"IBM285", 20285}, 1383 {"IBM290", 20290}, 1384 {"IBM297", 20297}, 1385 {"IBM367", 1252}, 1386 {"IBM420", 20420}, 1387 {"IBM423", 20423}, 1388 {"IBM424", 20424}, 1389 {"IBM819", 1252}, 1390 {"IBM871", 20871}, 1391 {"IBM880", 20880}, 1392 {"IBM905", 20905}, 1393 {"IBM924", 20924}, 1394 {"ISO-8859-1", 28591}, 1395 {"ISO-8859-13", 28603}, 1396 {"ISO-8859-15", 28605}, 1397 {"ISO-8859-2", 28592}, 1398 {"ISO-8859-3", 28593}, 1399 {"ISO-8859-4", 28594}, 1400 {"ISO-8859-5", 28595}, 1401 {"ISO-8859-6", 28596}, 1402 {"ISO-8859-7", 28597}, 1403 {"ISO-8859-8", 28598}, 1404 {"ISO-8859-9", 28599}, 1405 {"ISO8859-1", 28591}, 1406 {"ISO8859-13", 28603}, 1407 {"ISO8859-15", 28605}, 1408 {"ISO8859-2", 28592}, 1409 {"ISO8859-3", 28593}, 1410 {"ISO8859-4", 28594}, 1411 {"ISO8859-5", 28595}, 1412 {"ISO8859-6", 28596}, 1413 {"ISO8859-7", 28597}, 1414 {"ISO8859-8", 28598}, 1415 {"ISO8859-9", 28599}, 1416 {"JOHAB", 1361}, 1417 {"KOI8-R", 20866}, 1418 {"KOI8-U", 21866}, 1419 {"KS_C_5601-1987", 949}, 1420 {"LATIN1", 1252}, 1421 {"LATIN2", 28592}, 1422 {"MACINTOSH", 10000}, 1423 {"SHIFT-JIS", 932}, 1424 {"SHIFT_JIS", 932}, 1425 {"SJIS", 932}, 1426 {"US", 1252}, 1427 {"US-ASCII", 1252}, 1428 {"UTF-16", 1200}, 1429 {"UTF-16BE", 1201}, 1430 {"UTF-16LE", 1200}, 1431 {"UTF-8", CP_UTF8}, 1432 {"X-EUROPA", 29001}, 1433 {"X-MAC-ARABIC", 10004}, 1434 {"X-MAC-CE", 10029}, 1435 {"X-MAC-CHINESEIMP", 10008}, 1436 {"X-MAC-CHINESETRAD", 10002}, 1437 {"X-MAC-CROATIAN", 10082}, 1438 {"X-MAC-CYRILLIC", 10007}, 1439 {"X-MAC-GREEK", 10006}, 1440 {"X-MAC-HEBREW", 10005}, 1441 {"X-MAC-ICELANDIC", 10079}, 1442 {"X-MAC-JAPANESE", 10001}, 1443 {"X-MAC-KOREAN", 10003}, 1444 {"X-MAC-ROMANIAN", 10010}, 1445 {"X-MAC-THAI", 10021}, 1446 {"X-MAC-TURKISH", 10081}, 1447 {"X-MAC-UKRAINIAN", 10017}, 1448 }; 1449 static unsigned 1450 make_codepage_from_charset(const char *charset) 1451 { 1452 char cs[16]; 1453 char *p; 1454 unsigned cp; 1455 int a, b; 1456 1457 if (charset == NULL || strlen(charset) > 15) 1458 return -1; 1459 1460 /* Copy name to uppercase. */ 1461 p = cs; 1462 while (*charset) { 1463 char c = *charset++; 1464 if (c >= 'a' && c <= 'z') 1465 c -= 'a' - 'A'; 1466 *p++ = c; 1467 } 1468 *p++ = '\0'; 1469 cp = -1; 1470 1471 /* Look it up in the table first, so that we can easily 1472 * override CP367, which we map to 1252 instead of 367. */ 1473 a = 0; 1474 b = sizeof(charsets)/sizeof(charsets[0]); 1475 while (b > a) { 1476 int c = (b + a) / 2; 1477 int r = strcmp(charsets[c].name, cs); 1478 if (r < 0) 1479 a = c + 1; 1480 else if (r > 0) 1481 b = c; 1482 else 1483 return charsets[c].cp; 1484 } 1485 1486 /* If it's not in the table, try to parse it. */ 1487 switch (*cs) { 1488 case 'C': 1489 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { 1490 cp = my_atoi(cs + 2); 1491 } else if (strcmp(cs, "CP_ACP") == 0) 1492 cp = get_current_codepage(); 1493 else if (strcmp(cs, "CP_OEMCP") == 0) 1494 cp = get_current_oemcp(); 1495 break; 1496 case 'I': 1497 if (cs[1] == 'B' && cs[2] == 'M' && 1498 cs[3] >= '0' && cs[3] <= '9') { 1499 cp = my_atoi(cs + 3); 1500 } 1501 break; 1502 case 'W': 1503 if (strncmp(cs, "WINDOWS-", 8) == 0) { 1504 cp = my_atoi(cs + 8); 1505 if (cp != 874 && (cp < 1250 || cp > 1258)) 1506 cp = -1;/* This may invalid code. */ 1507 } 1508 break; 1509 } 1510 return (cp); 1511 } 1512 1513 /* 1514 * Return ANSI Code Page of current locale set by setlocale(). 1515 */ 1516 static unsigned 1517 get_current_codepage(void) 1518 { 1519 char *locale, *p; 1520 unsigned cp; 1521 1522 locale = setlocale(LC_CTYPE, NULL); 1523 if (locale == NULL) 1524 return (GetACP()); 1525 if (locale[0] == 'C' && locale[1] == '\0') 1526 return (CP_C_LOCALE); 1527 p = strrchr(locale, '.'); 1528 if (p == NULL) 1529 return (GetACP()); 1530 if (strcmp(p+1, "utf8") == 0) 1531 return CP_UTF8; 1532 cp = my_atoi(p+1); 1533 if ((int)cp <= 0) 1534 return (GetACP()); 1535 return (cp); 1536 } 1537 1538 /* 1539 * Translation table between Locale Name and ACP/OEMCP. 1540 */ 1541 static struct { 1542 unsigned acp; 1543 unsigned ocp; 1544 const char *locale; 1545 } acp_ocp_map[] = { 1546 { 950, 950, "Chinese_Taiwan" }, 1547 { 936, 936, "Chinese_People's Republic of China" }, 1548 { 950, 950, "Chinese_Taiwan" }, 1549 { 1250, 852, "Czech_Czech Republic" }, 1550 { 1252, 850, "Danish_Denmark" }, 1551 { 1252, 850, "Dutch_Netherlands" }, 1552 { 1252, 850, "Dutch_Belgium" }, 1553 { 1252, 437, "English_United States" }, 1554 { 1252, 850, "English_Australia" }, 1555 { 1252, 850, "English_Canada" }, 1556 { 1252, 850, "English_New Zealand" }, 1557 { 1252, 850, "English_United Kingdom" }, 1558 { 1252, 437, "English_United States" }, 1559 { 1252, 850, "Finnish_Finland" }, 1560 { 1252, 850, "French_France" }, 1561 { 1252, 850, "French_Belgium" }, 1562 { 1252, 850, "French_Canada" }, 1563 { 1252, 850, "French_Switzerland" }, 1564 { 1252, 850, "German_Germany" }, 1565 { 1252, 850, "German_Austria" }, 1566 { 1252, 850, "German_Switzerland" }, 1567 { 1253, 737, "Greek_Greece" }, 1568 { 1250, 852, "Hungarian_Hungary" }, 1569 { 1252, 850, "Icelandic_Iceland" }, 1570 { 1252, 850, "Italian_Italy" }, 1571 { 1252, 850, "Italian_Switzerland" }, 1572 { 932, 932, "Japanese_Japan" }, 1573 { 949, 949, "Korean_Korea" }, 1574 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1575 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1576 { 1252, 850, "Norwegian-Nynorsk_Norway" }, 1577 { 1250, 852, "Polish_Poland" }, 1578 { 1252, 850, "Portuguese_Portugal" }, 1579 { 1252, 850, "Portuguese_Brazil" }, 1580 { 1251, 866, "Russian_Russia" }, 1581 { 1250, 852, "Slovak_Slovakia" }, 1582 { 1252, 850, "Spanish_Spain" }, 1583 { 1252, 850, "Spanish_Mexico" }, 1584 { 1252, 850, "Spanish_Spain" }, 1585 { 1252, 850, "Swedish_Sweden" }, 1586 { 1254, 857, "Turkish_Turkey" }, 1587 { 0, 0, NULL} 1588 }; 1589 1590 /* 1591 * Return OEM Code Page of current locale set by setlocale(). 1592 */ 1593 static unsigned 1594 get_current_oemcp(void) 1595 { 1596 int i; 1597 char *locale, *p; 1598 size_t len; 1599 1600 locale = setlocale(LC_CTYPE, NULL); 1601 if (locale == NULL) 1602 return (GetOEMCP()); 1603 if (locale[0] == 'C' && locale[1] == '\0') 1604 return (CP_C_LOCALE); 1605 1606 p = strrchr(locale, '.'); 1607 if (p == NULL) 1608 return (GetOEMCP()); 1609 len = p - locale; 1610 for (i = 0; acp_ocp_map[i].acp; i++) { 1611 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) 1612 return (acp_ocp_map[i].ocp); 1613 } 1614 return (GetOEMCP()); 1615 } 1616 #else 1617 1618 /* 1619 * POSIX platform does not use CodePage. 1620 */ 1621 1622 static unsigned 1623 get_current_codepage(void) 1624 { 1625 return (-1);/* Unknown */ 1626 } 1627 static unsigned 1628 make_codepage_from_charset(const char *charset) 1629 { 1630 (void)charset; /* UNUSED */ 1631 return (-1);/* Unknown */ 1632 } 1633 static unsigned 1634 get_current_oemcp(void) 1635 { 1636 return (-1);/* Unknown */ 1637 } 1638 1639 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 1640 1641 /* 1642 * Return a string conversion object. 1643 */ 1644 static struct archive_string_conv * 1645 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) 1646 { 1647 struct archive_string_conv *sc; 1648 unsigned current_codepage; 1649 1650 /* Check if we have made the sconv object. */ 1651 sc = find_sconv_object(a, fc, tc); 1652 if (sc != NULL) 1653 return (sc); 1654 1655 if (a == NULL) 1656 current_codepage = get_current_codepage(); 1657 else 1658 current_codepage = a->current_codepage; 1659 1660 sc = create_sconv_object(canonical_charset_name(fc), 1661 canonical_charset_name(tc), current_codepage, flag); 1662 if (sc == NULL) { 1663 if (a != NULL) 1664 archive_set_error(a, ENOMEM, 1665 "Could not allocate memory for " 1666 "a string conversion object"); 1667 return (NULL); 1668 } 1669 1670 /* 1671 * If there is no converter for current string conversion object, 1672 * we cannot handle this conversion. 1673 */ 1674 if (sc->nconverter == 0) { 1675 if (a != NULL) { 1676 #if HAVE_ICONV 1677 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1678 "iconv_open failed : Cannot handle ``%s''", 1679 (flag & SCONV_TO_CHARSET)?tc:fc); 1680 #else 1681 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1682 "A character-set conversion not fully supported " 1683 "on this platform"); 1684 #endif 1685 } 1686 /* Failed; free a sconv object. */ 1687 free_sconv_object(sc); 1688 return (NULL); 1689 } 1690 1691 /* 1692 * Success! 1693 */ 1694 if (a != NULL) 1695 add_sconv_object(a, sc); 1696 return (sc); 1697 } 1698 1699 static const char * 1700 get_current_charset(struct archive *a) 1701 { 1702 const char *cur_charset; 1703 1704 if (a == NULL) 1705 cur_charset = default_iconv_charset(""); 1706 else { 1707 cur_charset = default_iconv_charset(a->current_code); 1708 if (a->current_code == NULL) { 1709 a->current_code = strdup(cur_charset); 1710 a->current_codepage = get_current_codepage(); 1711 a->current_oemcp = get_current_oemcp(); 1712 } 1713 } 1714 return (cur_charset); 1715 } 1716 1717 /* 1718 * Make and Return a string conversion object. 1719 * Return NULL if the platform does not support the specified conversion 1720 * and best_effort is 0. 1721 * If best_effort is set, A string conversion object must be returned 1722 * unless memory allocation for the object fails, but the conversion 1723 * might fail when non-ASCII code is found. 1724 */ 1725 struct archive_string_conv * 1726 archive_string_conversion_to_charset(struct archive *a, const char *charset, 1727 int best_effort) 1728 { 1729 int flag = SCONV_TO_CHARSET; 1730 1731 if (best_effort) 1732 flag |= SCONV_BEST_EFFORT; 1733 return (get_sconv_object(a, get_current_charset(a), charset, flag)); 1734 } 1735 1736 struct archive_string_conv * 1737 archive_string_conversion_from_charset(struct archive *a, const char *charset, 1738 int best_effort) 1739 { 1740 int flag = SCONV_FROM_CHARSET; 1741 1742 if (best_effort) 1743 flag |= SCONV_BEST_EFFORT; 1744 return (get_sconv_object(a, charset, get_current_charset(a), flag)); 1745 } 1746 1747 /* 1748 * archive_string_default_conversion_*_archive() are provided for Windows 1749 * platform because other archiver application use CP_OEMCP for 1750 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames 1751 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP 1752 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). 1753 * So we should make a string conversion between CP_ACP and CP_OEMCP 1754 * for compatibility. 1755 */ 1756 #if defined(_WIN32) && !defined(__CYGWIN__) 1757 struct archive_string_conv * 1758 archive_string_default_conversion_for_read(struct archive *a) 1759 { 1760 const char *cur_charset = get_current_charset(a); 1761 char oemcp[16]; 1762 1763 /* NOTE: a check of cur_charset is unneeded but we need 1764 * that get_current_charset() has been surely called at 1765 * this time whatever C compiler optimized. */ 1766 if (cur_charset != NULL && 1767 (a->current_codepage == CP_C_LOCALE || 1768 a->current_codepage == a->current_oemcp)) 1769 return (NULL);/* no conversion. */ 1770 1771 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1772 /* Make sure a null termination must be set. */ 1773 oemcp[sizeof(oemcp)-1] = '\0'; 1774 return (get_sconv_object(a, oemcp, cur_charset, 1775 SCONV_FROM_CHARSET)); 1776 } 1777 1778 struct archive_string_conv * 1779 archive_string_default_conversion_for_write(struct archive *a) 1780 { 1781 const char *cur_charset = get_current_charset(a); 1782 char oemcp[16]; 1783 1784 /* NOTE: a check of cur_charset is unneeded but we need 1785 * that get_current_charset() has been surely called at 1786 * this time whatever C compiler optimized. */ 1787 if (cur_charset != NULL && 1788 (a->current_codepage == CP_C_LOCALE || 1789 a->current_codepage == a->current_oemcp)) 1790 return (NULL);/* no conversion. */ 1791 1792 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1793 /* Make sure a null termination must be set. */ 1794 oemcp[sizeof(oemcp)-1] = '\0'; 1795 return (get_sconv_object(a, cur_charset, oemcp, 1796 SCONV_TO_CHARSET)); 1797 } 1798 #else 1799 struct archive_string_conv * 1800 archive_string_default_conversion_for_read(struct archive *a) 1801 { 1802 (void)a; /* UNUSED */ 1803 return (NULL); 1804 } 1805 1806 struct archive_string_conv * 1807 archive_string_default_conversion_for_write(struct archive *a) 1808 { 1809 (void)a; /* UNUSED */ 1810 return (NULL); 1811 } 1812 #endif 1813 1814 /* 1815 * Dispose of all character conversion objects in the archive object. 1816 */ 1817 void 1818 archive_string_conversion_free(struct archive *a) 1819 { 1820 struct archive_string_conv *sc; 1821 struct archive_string_conv *sc_next; 1822 1823 for (sc = a->sconv; sc != NULL; sc = sc_next) { 1824 sc_next = sc->next; 1825 free_sconv_object(sc); 1826 } 1827 a->sconv = NULL; 1828 free(a->current_code); 1829 a->current_code = NULL; 1830 } 1831 1832 /* 1833 * Return a conversion charset name. 1834 */ 1835 const char * 1836 archive_string_conversion_charset_name(struct archive_string_conv *sc) 1837 { 1838 if (sc->flag & SCONV_TO_CHARSET) 1839 return (sc->to_charset); 1840 else 1841 return (sc->from_charset); 1842 } 1843 1844 /* 1845 * Change the behavior of a string conversion. 1846 */ 1847 void 1848 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) 1849 { 1850 switch (opt) { 1851 /* 1852 * A filename in UTF-8 was made with libarchive 2.x in a wrong 1853 * assumption that wchar_t was Unicode. 1854 * This option enables simulating the assumption in order to read 1855 * that filename correctly. 1856 */ 1857 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: 1858 #if (defined(_WIN32) && !defined(__CYGWIN__)) \ 1859 || defined(__STDC_ISO_10646__) || defined(__APPLE__) 1860 /* 1861 * Nothing to do for it since wchar_t on these platforms 1862 * is really Unicode. 1863 */ 1864 (void)sc; /* UNUSED */ 1865 #else 1866 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { 1867 sc->flag |= SCONV_UTF8_LIBARCHIVE_2; 1868 /* Set up string converters. */ 1869 setup_converter(sc); 1870 } 1871 #endif 1872 break; 1873 case SCONV_SET_OPT_NORMALIZATION_C: 1874 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) { 1875 sc->flag |= SCONV_NORMALIZATION_C; 1876 sc->flag &= ~SCONV_NORMALIZATION_D; 1877 /* Set up string converters. */ 1878 setup_converter(sc); 1879 } 1880 break; 1881 case SCONV_SET_OPT_NORMALIZATION_D: 1882 #if defined(HAVE_ICONV) 1883 /* 1884 * If iconv will take the string, do not change the 1885 * setting of the normalization. 1886 */ 1887 if (!(sc->flag & SCONV_WIN_CP) && 1888 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1889 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1890 break; 1891 #endif 1892 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) { 1893 sc->flag |= SCONV_NORMALIZATION_D; 1894 sc->flag &= ~SCONV_NORMALIZATION_C; 1895 /* Set up string converters. */ 1896 setup_converter(sc); 1897 } 1898 break; 1899 default: 1900 break; 1901 } 1902 } 1903 1904 /* 1905 * 1906 * Copy one archive_string to another in locale conversion. 1907 * 1908 * archive_strncat_l(); 1909 * archive_strncpy_l(); 1910 * 1911 */ 1912 1913 static size_t 1914 mbsnbytes(const void *_p, size_t n) 1915 { 1916 size_t s; 1917 const char *p, *pp; 1918 1919 if (_p == NULL) 1920 return (0); 1921 p = (const char *)_p; 1922 1923 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1924 s = 0; 1925 pp = p; 1926 while (s < n && *pp) { 1927 pp++; 1928 s++; 1929 } 1930 return (s); 1931 } 1932 1933 static size_t 1934 utf16nbytes(const void *_p, size_t n) 1935 { 1936 size_t s; 1937 const char *p, *pp; 1938 1939 if (_p == NULL) 1940 return (0); 1941 p = (const char *)_p; 1942 1943 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1944 s = 0; 1945 pp = p; 1946 n >>= 1; 1947 while (s < n && (pp[0] || pp[1])) { 1948 pp += 2; 1949 s++; 1950 } 1951 return (s<<1); 1952 } 1953 1954 int 1955 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n, 1956 struct archive_string_conv *sc) 1957 { 1958 as->length = 0; 1959 return (archive_strncat_l(as, _p, n, sc)); 1960 } 1961 1962 int 1963 archive_strncat_l(struct archive_string *as, const void *_p, size_t n, 1964 struct archive_string_conv *sc) 1965 { 1966 const void *s; 1967 size_t length = 0; 1968 int i, r = 0, r2; 1969 1970 if (_p != NULL && n > 0) { 1971 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) 1972 length = utf16nbytes(_p, n); 1973 else 1974 length = mbsnbytes(_p, n); 1975 } 1976 1977 /* We must allocate memory even if there is no data for conversion 1978 * or copy. This simulates archive_string_append behavior. */ 1979 if (length == 0) { 1980 int tn = 1; 1981 if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) 1982 tn = 2; 1983 if (archive_string_ensure(as, as->length + tn) == NULL) 1984 return (-1); 1985 as->s[as->length] = 0; 1986 if (tn == 2) 1987 as->s[as->length+1] = 0; 1988 return (0); 1989 } 1990 1991 /* 1992 * If sc is NULL, we just make a copy. 1993 */ 1994 if (sc == NULL) { 1995 if (archive_string_append(as, _p, length) == NULL) 1996 return (-1);/* No memory */ 1997 return (0); 1998 } 1999 2000 s = _p; 2001 i = 0; 2002 if (sc->nconverter > 1) { 2003 sc->utftmp.length = 0; 2004 r2 = sc->converter[0](&(sc->utftmp), s, length, sc); 2005 if (r2 != 0 && errno == ENOMEM) 2006 return (r2); 2007 if (r > r2) 2008 r = r2; 2009 s = sc->utftmp.s; 2010 length = sc->utftmp.length; 2011 ++i; 2012 } 2013 r2 = sc->converter[i](as, s, length, sc); 2014 if (r > r2) 2015 r = r2; 2016 return (r); 2017 } 2018 2019 #if HAVE_ICONV 2020 2021 /* 2022 * Return -1 if conversion fails. 2023 */ 2024 static int 2025 iconv_strncat_in_locale(struct archive_string *as, const void *_p, 2026 size_t length, struct archive_string_conv *sc) 2027 { 2028 ICONV_CONST char *itp; 2029 size_t remaining; 2030 iconv_t cd; 2031 char *outp; 2032 size_t avail, bs; 2033 int return_value = 0; /* success */ 2034 int to_size, from_size; 2035 2036 if (sc->flag & SCONV_TO_UTF16) 2037 to_size = 2; 2038 else 2039 to_size = 1; 2040 if (sc->flag & SCONV_FROM_UTF16) 2041 from_size = 2; 2042 else 2043 from_size = 1; 2044 2045 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) 2046 return (-1); 2047 2048 cd = sc->cd; 2049 itp = (char *)(uintptr_t)_p; 2050 remaining = length; 2051 outp = as->s + as->length; 2052 avail = as->buffer_length - as->length - to_size; 2053 while (remaining >= (size_t)from_size) { 2054 size_t result = iconv(cd, &itp, &remaining, &outp, &avail); 2055 2056 if (result != (size_t)-1) 2057 break; /* Conversion completed. */ 2058 2059 if (errno == EILSEQ || errno == EINVAL) { 2060 /* 2061 * If an output charset is UTF-8 or UTF-16BE/LE, 2062 * unknown character should be U+FFFD 2063 * (replacement character). 2064 */ 2065 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { 2066 size_t rbytes; 2067 if (sc->flag & SCONV_TO_UTF8) 2068 rbytes = sizeof(utf8_replacement_char); 2069 else 2070 rbytes = 2; 2071 2072 if (avail < rbytes) { 2073 as->length = outp - as->s; 2074 bs = as->buffer_length + 2075 (remaining * to_size) + rbytes; 2076 if (NULL == 2077 archive_string_ensure(as, bs)) 2078 return (-1); 2079 outp = as->s + as->length; 2080 avail = as->buffer_length 2081 - as->length - to_size; 2082 } 2083 if (sc->flag & SCONV_TO_UTF8) 2084 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); 2085 else if (sc->flag & SCONV_TO_UTF16BE) 2086 archive_be16enc(outp, UNICODE_R_CHAR); 2087 else 2088 archive_le16enc(outp, UNICODE_R_CHAR); 2089 outp += rbytes; 2090 avail -= rbytes; 2091 } else { 2092 /* Skip the illegal input bytes. */ 2093 *outp++ = '?'; 2094 avail--; 2095 } 2096 itp += from_size; 2097 remaining -= from_size; 2098 return_value = -1; /* failure */ 2099 } else { 2100 /* E2BIG no output buffer, 2101 * Increase an output buffer. */ 2102 as->length = outp - as->s; 2103 bs = as->buffer_length + remaining * 2; 2104 if (NULL == archive_string_ensure(as, bs)) 2105 return (-1); 2106 outp = as->s + as->length; 2107 avail = as->buffer_length - as->length - to_size; 2108 } 2109 } 2110 as->length = outp - as->s; 2111 as->s[as->length] = 0; 2112 if (to_size == 2) 2113 as->s[as->length+1] = 0; 2114 return (return_value); 2115 } 2116 2117 #endif /* HAVE_ICONV */ 2118 2119 2120 #if defined(_WIN32) && !defined(__CYGWIN__) 2121 2122 /* 2123 * Translate a string from a some CodePage to an another CodePage by 2124 * Windows APIs, and copy the result. Return -1 if conversion fails. 2125 */ 2126 static int 2127 strncat_in_codepage(struct archive_string *as, 2128 const void *_p, size_t length, struct archive_string_conv *sc) 2129 { 2130 const char *s = (const char *)_p; 2131 struct archive_wstring aws; 2132 size_t l; 2133 int r, saved_flag; 2134 2135 archive_string_init(&aws); 2136 saved_flag = sc->flag; 2137 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); 2138 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); 2139 sc->flag = saved_flag; 2140 if (r != 0) { 2141 archive_wstring_free(&aws); 2142 if (errno != ENOMEM) 2143 archive_string_append(as, s, length); 2144 return (-1); 2145 } 2146 2147 l = as->length; 2148 r = archive_string_append_from_wcs_in_codepage( 2149 as, aws.s, aws.length, sc); 2150 if (r != 0 && errno != ENOMEM && l == as->length) 2151 archive_string_append(as, s, length); 2152 archive_wstring_free(&aws); 2153 return (r); 2154 } 2155 2156 /* 2157 * Test whether MBS ==> WCS is okay. 2158 */ 2159 static int 2160 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2161 { 2162 const char *p = (const char *)_p; 2163 unsigned codepage; 2164 DWORD mbflag = MB_ERR_INVALID_CHARS; 2165 2166 if (sc->flag & SCONV_FROM_CHARSET) 2167 codepage = sc->to_cp; 2168 else 2169 codepage = sc->from_cp; 2170 2171 if (codepage == CP_C_LOCALE) 2172 return (0); 2173 if (codepage != CP_UTF8) 2174 mbflag |= MB_PRECOMPOSED; 2175 2176 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) 2177 return (-1); /* Invalid */ 2178 return (0); /* Okay */ 2179 } 2180 2181 #else 2182 2183 /* 2184 * Test whether MBS ==> WCS is okay. 2185 */ 2186 static int 2187 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2188 { 2189 const char *p = (const char *)_p; 2190 size_t r; 2191 2192 #if HAVE_MBRTOWC 2193 mbstate_t shift_state; 2194 2195 memset(&shift_state, 0, sizeof(shift_state)); 2196 #else 2197 /* Clear the shift state before starting. */ 2198 mbtowc(NULL, NULL, 0); 2199 #endif 2200 while (n) { 2201 wchar_t wc; 2202 2203 #if HAVE_MBRTOWC 2204 r = mbrtowc(&wc, p, n, &shift_state); 2205 #else 2206 r = mbtowc(&wc, p, n); 2207 #endif 2208 if (r == (size_t)-1 || r == (size_t)-2) 2209 return (-1);/* Invalid. */ 2210 if (r == 0) 2211 break; 2212 p += r; 2213 n -= r; 2214 } 2215 (void)sc; /* UNUSED */ 2216 return (0); /* All Okey. */ 2217 } 2218 2219 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 2220 2221 /* 2222 * Basically returns -1 because we cannot make a conversion of charset 2223 * without iconv but in some cases this would return 0. 2224 * Returns 0 if all copied characters are ASCII. 2225 * Returns 0 if both from-locale and to-locale are the same and those 2226 * can be WCS with no error. 2227 */ 2228 static int 2229 best_effort_strncat_in_locale(struct archive_string *as, const void *_p, 2230 size_t length, struct archive_string_conv *sc) 2231 { 2232 size_t remaining; 2233 const uint8_t *itp; 2234 int return_value = 0; /* success */ 2235 2236 /* 2237 * If both from-locale and to-locale is the same, this makes a copy. 2238 * And then this checks all copied MBS can be WCS if so returns 0. 2239 */ 2240 if (sc->same) { 2241 if (archive_string_append(as, _p, length) == NULL) 2242 return (-1);/* No memory */ 2243 return (invalid_mbs(_p, length, sc)); 2244 } 2245 2246 /* 2247 * If a character is ASCII, this just copies it. If not, this 2248 * assigns '?' character instead but in UTF-8 locale this assigns 2249 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, 2250 * a Replacement Character in Unicode. 2251 */ 2252 2253 remaining = length; 2254 itp = (const uint8_t *)_p; 2255 while (*itp && remaining > 0) { 2256 if (*itp > 127) { 2257 // Non-ASCII: Substitute with suitable replacement 2258 if (sc->flag & SCONV_TO_UTF8) { 2259 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { 2260 __archive_errx(1, "Out of memory"); 2261 } 2262 } else { 2263 archive_strappend_char(as, '?'); 2264 } 2265 return_value = -1; 2266 } else { 2267 archive_strappend_char(as, *itp); 2268 } 2269 ++itp; 2270 } 2271 return (return_value); 2272 } 2273 2274 2275 /* 2276 * Unicode conversion functions. 2277 * - UTF-8 <===> UTF-8 in removing surrogate pairs. 2278 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. 2279 * - UTF-8 made by libarchive 2.x ===> UTF-8. 2280 * - UTF-16BE <===> UTF-8. 2281 * 2282 */ 2283 2284 /* 2285 * Utility to convert a single UTF-8 sequence. 2286 * 2287 * Usually return used bytes, return used byte in negative value when 2288 * a unicode character is replaced with U+FFFD. 2289 * See also http://unicode.org/review/pr-121.html Public Review Issue #121 2290 * Recommended Practice for Replacement Characters. 2291 */ 2292 static int 2293 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2294 { 2295 static const char utf8_count[256] = { 2296 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 2299 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 2300 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 2301 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 2302 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 2303 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 2304 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 2305 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 2306 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 2307 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 2308 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2309 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 2310 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 2311 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ 2312 }; 2313 int ch, i; 2314 int cnt; 2315 uint32_t wc; 2316 2317 /* Sanity check. */ 2318 if (n == 0) 2319 return (0); 2320 /* 2321 * Decode 1-4 bytes depending on the value of the first byte. 2322 */ 2323 ch = (unsigned char)*s; 2324 if (ch == 0) 2325 return (0); /* Standard: return 0 for end-of-string. */ 2326 cnt = utf8_count[ch]; 2327 2328 /* Invalid sequence or there are not plenty bytes. */ 2329 if ((int)n < cnt) { 2330 cnt = (int)n; 2331 for (i = 1; i < cnt; i++) { 2332 if ((s[i] & 0xc0) != 0x80) { 2333 cnt = i; 2334 break; 2335 } 2336 } 2337 goto invalid_sequence; 2338 } 2339 2340 /* Make a Unicode code point from a single UTF-8 sequence. */ 2341 switch (cnt) { 2342 case 1: /* 1 byte sequence. */ 2343 *pwc = ch & 0x7f; 2344 return (cnt); 2345 case 2: /* 2 bytes sequence. */ 2346 if ((s[1] & 0xc0) != 0x80) { 2347 cnt = 1; 2348 goto invalid_sequence; 2349 } 2350 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); 2351 return (cnt); 2352 case 3: /* 3 bytes sequence. */ 2353 if ((s[1] & 0xc0) != 0x80) { 2354 cnt = 1; 2355 goto invalid_sequence; 2356 } 2357 if ((s[2] & 0xc0) != 0x80) { 2358 cnt = 2; 2359 goto invalid_sequence; 2360 } 2361 wc = ((ch & 0x0f) << 12) 2362 | ((s[1] & 0x3f) << 6) 2363 | (s[2] & 0x3f); 2364 if (wc < 0x800) 2365 goto invalid_sequence;/* Overlong sequence. */ 2366 break; 2367 case 4: /* 4 bytes sequence. */ 2368 if ((s[1] & 0xc0) != 0x80) { 2369 cnt = 1; 2370 goto invalid_sequence; 2371 } 2372 if ((s[2] & 0xc0) != 0x80) { 2373 cnt = 2; 2374 goto invalid_sequence; 2375 } 2376 if ((s[3] & 0xc0) != 0x80) { 2377 cnt = 3; 2378 goto invalid_sequence; 2379 } 2380 wc = ((ch & 0x07) << 18) 2381 | ((s[1] & 0x3f) << 12) 2382 | ((s[2] & 0x3f) << 6) 2383 | (s[3] & 0x3f); 2384 if (wc < 0x10000) 2385 goto invalid_sequence;/* Overlong sequence. */ 2386 break; 2387 default: /* Others are all invalid sequence. */ 2388 if (ch == 0xc0 || ch == 0xc1) 2389 cnt = 2; 2390 else if (ch >= 0xf5 && ch <= 0xf7) 2391 cnt = 4; 2392 else if (ch >= 0xf8 && ch <= 0xfb) 2393 cnt = 5; 2394 else if (ch == 0xfc || ch == 0xfd) 2395 cnt = 6; 2396 else 2397 cnt = 1; 2398 if ((int)n < cnt) 2399 cnt = (int)n; 2400 for (i = 1; i < cnt; i++) { 2401 if ((s[i] & 0xc0) != 0x80) { 2402 cnt = i; 2403 break; 2404 } 2405 } 2406 goto invalid_sequence; 2407 } 2408 2409 /* The code point larger than 0x10FFFF is not legal 2410 * Unicode values. */ 2411 if (wc > UNICODE_MAX) 2412 goto invalid_sequence; 2413 /* Correctly gets a Unicode, returns used bytes. */ 2414 *pwc = wc; 2415 return (cnt); 2416 invalid_sequence: 2417 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2418 return (cnt * -1); 2419 } 2420 2421 static int 2422 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2423 { 2424 int cnt; 2425 2426 cnt = _utf8_to_unicode(pwc, s, n); 2427 /* Any of Surrogate pair is not legal Unicode values. */ 2428 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) 2429 return (-3); 2430 return (cnt); 2431 } 2432 2433 static inline uint32_t 2434 combine_surrogate_pair(uint32_t uc, uint32_t uc2) 2435 { 2436 uc -= 0xD800; 2437 uc *= 0x400; 2438 uc += uc2 - 0xDC00; 2439 uc += 0x10000; 2440 return (uc); 2441 } 2442 2443 /* 2444 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in 2445 * removing surrogate pairs. 2446 * 2447 * CESU-8: The Compatibility Encoding Scheme for UTF-16. 2448 * 2449 * Usually return used bytes, return used byte in negative value when 2450 * a unicode character is replaced with U+FFFD. 2451 */ 2452 static int 2453 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2454 { 2455 uint32_t wc = 0; 2456 int cnt; 2457 2458 cnt = _utf8_to_unicode(&wc, s, n); 2459 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { 2460 uint32_t wc2 = 0; 2461 if (n - 3 < 3) { 2462 /* Invalid byte sequence. */ 2463 goto invalid_sequence; 2464 } 2465 cnt = _utf8_to_unicode(&wc2, s+3, n-3); 2466 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { 2467 /* Invalid byte sequence. */ 2468 goto invalid_sequence; 2469 } 2470 wc = combine_surrogate_pair(wc, wc2); 2471 cnt = 6; 2472 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { 2473 /* Invalid byte sequence. */ 2474 goto invalid_sequence; 2475 } 2476 *pwc = wc; 2477 return (cnt); 2478 invalid_sequence: 2479 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2480 if (cnt > 0) 2481 cnt *= -1; 2482 return (cnt); 2483 } 2484 2485 /* 2486 * Convert a Unicode code point to a single UTF-8 sequence. 2487 * 2488 * NOTE:This function does not check if the Unicode is legal or not. 2489 * Please you definitely check it before calling this. 2490 */ 2491 static size_t 2492 unicode_to_utf8(char *p, size_t remaining, uint32_t uc) 2493 { 2494 char *_p = p; 2495 2496 /* Invalid Unicode char maps to Replacement character */ 2497 if (uc > UNICODE_MAX) 2498 uc = UNICODE_R_CHAR; 2499 /* Translate code point to UTF8 */ 2500 if (uc <= 0x7f) { 2501 if (remaining == 0) 2502 return (0); 2503 *p++ = (char)uc; 2504 } else if (uc <= 0x7ff) { 2505 if (remaining < 2) 2506 return (0); 2507 *p++ = 0xc0 | ((uc >> 6) & 0x1f); 2508 *p++ = 0x80 | (uc & 0x3f); 2509 } else if (uc <= 0xffff) { 2510 if (remaining < 3) 2511 return (0); 2512 *p++ = 0xe0 | ((uc >> 12) & 0x0f); 2513 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2514 *p++ = 0x80 | (uc & 0x3f); 2515 } else { 2516 if (remaining < 4) 2517 return (0); 2518 *p++ = 0xf0 | ((uc >> 18) & 0x07); 2519 *p++ = 0x80 | ((uc >> 12) & 0x3f); 2520 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2521 *p++ = 0x80 | (uc & 0x3f); 2522 } 2523 return (p - _p); 2524 } 2525 2526 static int 2527 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) 2528 { 2529 return (utf16_to_unicode(pwc, s, n, 1)); 2530 } 2531 2532 static int 2533 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) 2534 { 2535 return (utf16_to_unicode(pwc, s, n, 0)); 2536 } 2537 2538 static int 2539 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) 2540 { 2541 const char *utf16 = s; 2542 unsigned uc; 2543 2544 if (n == 0) 2545 return (0); 2546 if (n == 1) { 2547 /* set the Replacement Character instead. */ 2548 *pwc = UNICODE_R_CHAR; 2549 return (-1); 2550 } 2551 2552 if (be) 2553 uc = archive_be16dec(utf16); 2554 else 2555 uc = archive_le16dec(utf16); 2556 utf16 += 2; 2557 2558 /* If this is a surrogate pair, assemble the full code point.*/ 2559 if (IS_HIGH_SURROGATE_LA(uc)) { 2560 unsigned uc2; 2561 2562 if (n >= 4) { 2563 if (be) 2564 uc2 = archive_be16dec(utf16); 2565 else 2566 uc2 = archive_le16dec(utf16); 2567 } else 2568 uc2 = 0; 2569 if (IS_LOW_SURROGATE_LA(uc2)) { 2570 uc = combine_surrogate_pair(uc, uc2); 2571 utf16 += 2; 2572 } else { 2573 /* Undescribed code point should be U+FFFD 2574 * (replacement character). */ 2575 *pwc = UNICODE_R_CHAR; 2576 return (-2); 2577 } 2578 } 2579 2580 /* 2581 * Surrogate pair values(0xd800 through 0xdfff) are only 2582 * used by UTF-16, so, after above calculation, the code 2583 * must not be surrogate values, and Unicode has no codes 2584 * larger than 0x10ffff. Thus, those are not legal Unicode 2585 * values. 2586 */ 2587 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { 2588 /* Undescribed code point should be U+FFFD 2589 * (replacement character). */ 2590 *pwc = UNICODE_R_CHAR; 2591 return (((int)(utf16 - s)) * -1); 2592 } 2593 *pwc = uc; 2594 return ((int)(utf16 - s)); 2595 } 2596 2597 static size_t 2598 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) 2599 { 2600 char *utf16 = p; 2601 2602 if (uc > 0xffff) { 2603 /* We have a code point that won't fit into a 2604 * wchar_t; convert it to a surrogate pair. */ 2605 if (remaining < 4) 2606 return (0); 2607 uc -= 0x10000; 2608 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2609 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2610 return (4); 2611 } else { 2612 if (remaining < 2) 2613 return (0); 2614 archive_be16enc(utf16, uc); 2615 return (2); 2616 } 2617 } 2618 2619 static size_t 2620 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) 2621 { 2622 char *utf16 = p; 2623 2624 if (uc > 0xffff) { 2625 /* We have a code point that won't fit into a 2626 * wchar_t; convert it to a surrogate pair. */ 2627 if (remaining < 4) 2628 return (0); 2629 uc -= 0x10000; 2630 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2631 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2632 return (4); 2633 } else { 2634 if (remaining < 2) 2635 return (0); 2636 archive_le16enc(utf16, uc); 2637 return (2); 2638 } 2639 } 2640 2641 /* 2642 * Copy UTF-8 string in checking surrogate pair. 2643 * If any surrogate pair are found, it would be canonicalized. 2644 */ 2645 static int 2646 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, 2647 size_t len, struct archive_string_conv *sc) 2648 { 2649 const char *s; 2650 char *p, *endp; 2651 int n, ret = 0; 2652 2653 (void)sc; /* UNUSED */ 2654 2655 if (archive_string_ensure(as, as->length + len + 1) == NULL) 2656 return (-1); 2657 2658 s = (const char *)_p; 2659 p = as->s + as->length; 2660 endp = as->s + as->buffer_length -1; 2661 do { 2662 uint32_t uc; 2663 const char *ss = s; 2664 size_t w; 2665 2666 /* 2667 * Forward byte sequence until a conversion of that is needed. 2668 */ 2669 while ((n = utf8_to_unicode(&uc, s, len)) > 0) { 2670 s += n; 2671 len -= n; 2672 } 2673 if (ss < s) { 2674 if (p + (s - ss) > endp) { 2675 as->length = p - as->s; 2676 if (archive_string_ensure(as, 2677 as->buffer_length + len + 1) == NULL) 2678 return (-1); 2679 p = as->s + as->length; 2680 endp = as->s + as->buffer_length -1; 2681 } 2682 2683 memcpy(p, ss, s - ss); 2684 p += s - ss; 2685 } 2686 2687 /* 2688 * If n is negative, current byte sequence needs a replacement. 2689 */ 2690 if (n < 0) { 2691 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { 2692 /* Current byte sequence may be CESU-8. */ 2693 n = cesu8_to_unicode(&uc, s, len); 2694 } 2695 if (n < 0) { 2696 ret = -1; 2697 n *= -1;/* Use a replaced unicode character. */ 2698 } 2699 2700 /* Rebuild UTF-8 byte sequence. */ 2701 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { 2702 as->length = p - as->s; 2703 if (archive_string_ensure(as, 2704 as->buffer_length + len + 1) == NULL) 2705 return (-1); 2706 p = as->s + as->length; 2707 endp = as->s + as->buffer_length -1; 2708 } 2709 p += w; 2710 s += n; 2711 len -= n; 2712 } 2713 } while (n > 0); 2714 as->length = p - as->s; 2715 as->s[as->length] = '\0'; 2716 return (ret); 2717 } 2718 2719 static int 2720 archive_string_append_unicode(struct archive_string *as, const void *_p, 2721 size_t len, struct archive_string_conv *sc) 2722 { 2723 const char *s; 2724 char *p, *endp; 2725 uint32_t uc; 2726 size_t w; 2727 int n, ret = 0, ts, tm; 2728 int (*parse)(uint32_t *, const char *, size_t); 2729 size_t (*unparse)(char *, size_t, uint32_t); 2730 2731 if (sc->flag & SCONV_TO_UTF16BE) { 2732 unparse = unicode_to_utf16be; 2733 ts = 2; 2734 } else if (sc->flag & SCONV_TO_UTF16LE) { 2735 unparse = unicode_to_utf16le; 2736 ts = 2; 2737 } else if (sc->flag & SCONV_TO_UTF8) { 2738 unparse = unicode_to_utf8; 2739 ts = 1; 2740 } else { 2741 /* 2742 * This case is going to be converted to another 2743 * character-set through iconv. 2744 */ 2745 if (sc->flag & SCONV_FROM_UTF16BE) { 2746 unparse = unicode_to_utf16be; 2747 ts = 2; 2748 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2749 unparse = unicode_to_utf16le; 2750 ts = 2; 2751 } else { 2752 unparse = unicode_to_utf8; 2753 ts = 1; 2754 } 2755 } 2756 2757 if (sc->flag & SCONV_FROM_UTF16BE) { 2758 parse = utf16be_to_unicode; 2759 tm = 1; 2760 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2761 parse = utf16le_to_unicode; 2762 tm = 1; 2763 } else { 2764 parse = cesu8_to_unicode; 2765 tm = ts; 2766 } 2767 2768 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2769 return (-1); 2770 2771 s = (const char *)_p; 2772 p = as->s + as->length; 2773 endp = as->s + as->buffer_length - ts; 2774 while ((n = parse(&uc, s, len)) != 0) { 2775 if (n < 0) { 2776 /* Use a replaced unicode character. */ 2777 n *= -1; 2778 ret = -1; 2779 } 2780 s += n; 2781 len -= n; 2782 while ((w = unparse(p, endp - p, uc)) == 0) { 2783 /* There is not enough output buffer so 2784 * we have to expand it. */ 2785 as->length = p - as->s; 2786 if (archive_string_ensure(as, 2787 as->buffer_length + len * tm + ts) == NULL) 2788 return (-1); 2789 p = as->s + as->length; 2790 endp = as->s + as->buffer_length - ts; 2791 } 2792 p += w; 2793 } 2794 as->length = p - as->s; 2795 as->s[as->length] = '\0'; 2796 if (ts == 2) 2797 as->s[as->length+1] = '\0'; 2798 return (ret); 2799 } 2800 2801 /* 2802 * Following Constants for Hangul compositions this information comes from 2803 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ 2804 */ 2805 #define HC_SBASE 0xAC00 2806 #define HC_LBASE 0x1100 2807 #define HC_VBASE 0x1161 2808 #define HC_TBASE 0x11A7 2809 #define HC_LCOUNT 19 2810 #define HC_VCOUNT 21 2811 #define HC_TCOUNT 28 2812 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) 2813 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) 2814 2815 static uint32_t 2816 get_nfc(uint32_t uc, uint32_t uc2) 2817 { 2818 int t, b; 2819 2820 t = 0; 2821 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; 2822 while (b >= t) { 2823 int m = (t + b) / 2; 2824 if (u_composition_table[m].cp1 < uc) 2825 t = m + 1; 2826 else if (u_composition_table[m].cp1 > uc) 2827 b = m - 1; 2828 else if (u_composition_table[m].cp2 < uc2) 2829 t = m + 1; 2830 else if (u_composition_table[m].cp2 > uc2) 2831 b = m - 1; 2832 else 2833 return (u_composition_table[m].nfc); 2834 } 2835 return (0); 2836 } 2837 2838 #define FDC_MAX 10 /* The maximum number of Following Decomposable 2839 * Characters. */ 2840 2841 /* 2842 * Update first code point. 2843 */ 2844 #define UPDATE_UC(new_uc) do { \ 2845 uc = new_uc; \ 2846 ucptr = NULL; \ 2847 } while (0) 2848 2849 /* 2850 * Replace first code point with second code point. 2851 */ 2852 #define REPLACE_UC_WITH_UC2() do { \ 2853 uc = uc2; \ 2854 ucptr = uc2ptr; \ 2855 n = n2; \ 2856 } while (0) 2857 2858 #define EXPAND_BUFFER() do { \ 2859 as->length = p - as->s; \ 2860 if (archive_string_ensure(as, \ 2861 as->buffer_length + len * tm + ts) == NULL)\ 2862 return (-1); \ 2863 p = as->s + as->length; \ 2864 endp = as->s + as->buffer_length - ts; \ 2865 } while (0) 2866 2867 #define UNPARSE(p, endp, uc) do { \ 2868 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ 2869 EXPAND_BUFFER(); \ 2870 } \ 2871 p += w; \ 2872 } while (0) 2873 2874 /* 2875 * Write first code point. 2876 * If the code point has not be changed from its original code, 2877 * this just copies it from its original buffer pointer. 2878 * If not, this converts it to UTF-8 byte sequence and copies it. 2879 */ 2880 #define WRITE_UC() do { \ 2881 if (ucptr) { \ 2882 if (p + n > endp) \ 2883 EXPAND_BUFFER(); \ 2884 switch (n) { \ 2885 case 4: \ 2886 *p++ = *ucptr++; \ 2887 /* FALL THROUGH */ \ 2888 case 3: \ 2889 *p++ = *ucptr++; \ 2890 /* FALL THROUGH */ \ 2891 case 2: \ 2892 *p++ = *ucptr++; \ 2893 /* FALL THROUGH */ \ 2894 case 1: \ 2895 *p++ = *ucptr; \ 2896 break; \ 2897 } \ 2898 ucptr = NULL; \ 2899 } else { \ 2900 UNPARSE(p, endp, uc); \ 2901 } \ 2902 } while (0) 2903 2904 /* 2905 * Collect following decomposable code points. 2906 */ 2907 #define COLLECT_CPS(start) do { \ 2908 int _i; \ 2909 for (_i = start; _i < FDC_MAX ; _i++) { \ 2910 nx = parse(&ucx[_i], s, len); \ 2911 if (nx <= 0) \ 2912 break; \ 2913 cx = CCC(ucx[_i]); \ 2914 if (cl >= cx && cl != 228 && cx != 228)\ 2915 break; \ 2916 s += nx; \ 2917 len -= nx; \ 2918 cl = cx; \ 2919 ccx[_i] = cx; \ 2920 } \ 2921 if (_i >= FDC_MAX) { \ 2922 ret = -1; \ 2923 ucx_size = FDC_MAX; \ 2924 } else \ 2925 ucx_size = _i; \ 2926 } while (0) 2927 2928 /* 2929 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. 2930 * 2931 * TODO: Convert composition exclusions, which are never converted 2932 * from NFC,NFD,NFKC and NFKD, to Form C. 2933 */ 2934 static int 2935 archive_string_normalize_C(struct archive_string *as, const void *_p, 2936 size_t len, struct archive_string_conv *sc) 2937 { 2938 const char *s = (const char *)_p; 2939 char *p, *endp; 2940 uint32_t uc, uc2; 2941 size_t w; 2942 int always_replace, n, n2, ret = 0, spair, ts, tm; 2943 int (*parse)(uint32_t *, const char *, size_t); 2944 size_t (*unparse)(char *, size_t, uint32_t); 2945 2946 always_replace = 1; 2947 ts = 1;/* text size. */ 2948 if (sc->flag & SCONV_TO_UTF16BE) { 2949 unparse = unicode_to_utf16be; 2950 ts = 2; 2951 if (sc->flag & SCONV_FROM_UTF16BE) 2952 always_replace = 0; 2953 } else if (sc->flag & SCONV_TO_UTF16LE) { 2954 unparse = unicode_to_utf16le; 2955 ts = 2; 2956 if (sc->flag & SCONV_FROM_UTF16LE) 2957 always_replace = 0; 2958 } else if (sc->flag & SCONV_TO_UTF8) { 2959 unparse = unicode_to_utf8; 2960 if (sc->flag & SCONV_FROM_UTF8) 2961 always_replace = 0; 2962 } else { 2963 /* 2964 * This case is going to be converted to another 2965 * character-set through iconv. 2966 */ 2967 always_replace = 0; 2968 if (sc->flag & SCONV_FROM_UTF16BE) { 2969 unparse = unicode_to_utf16be; 2970 ts = 2; 2971 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2972 unparse = unicode_to_utf16le; 2973 ts = 2; 2974 } else { 2975 unparse = unicode_to_utf8; 2976 } 2977 } 2978 2979 if (sc->flag & SCONV_FROM_UTF16BE) { 2980 parse = utf16be_to_unicode; 2981 tm = 1; 2982 spair = 4;/* surrogate pair size in UTF-16. */ 2983 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2984 parse = utf16le_to_unicode; 2985 tm = 1; 2986 spair = 4;/* surrogate pair size in UTF-16. */ 2987 } else { 2988 parse = cesu8_to_unicode; 2989 tm = ts; 2990 spair = 6;/* surrogate pair size in UTF-8. */ 2991 } 2992 2993 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2994 return (-1); 2995 2996 p = as->s + as->length; 2997 endp = as->s + as->buffer_length - ts; 2998 while ((n = parse(&uc, s, len)) != 0) { 2999 const char *ucptr, *uc2ptr; 3000 3001 if (n < 0) { 3002 /* Use a replaced unicode character. */ 3003 UNPARSE(p, endp, uc); 3004 s += n*-1; 3005 len -= n*-1; 3006 ret = -1; 3007 continue; 3008 } else if (n == spair || always_replace) 3009 /* uc is converted from a surrogate pair. 3010 * this should be treated as a changed code. */ 3011 ucptr = NULL; 3012 else 3013 ucptr = s; 3014 s += n; 3015 len -= n; 3016 3017 /* Read second code point. */ 3018 while ((n2 = parse(&uc2, s, len)) > 0) { 3019 uint32_t ucx[FDC_MAX]; 3020 int ccx[FDC_MAX]; 3021 int cl, cx, i, nx, ucx_size; 3022 int LIndex,SIndex; 3023 uint32_t nfc; 3024 3025 if (n2 == spair || always_replace) 3026 /* uc2 is converted from a surrogate pair. 3027 * this should be treated as a changed code. */ 3028 uc2ptr = NULL; 3029 else 3030 uc2ptr = s; 3031 s += n2; 3032 len -= n2; 3033 3034 /* 3035 * If current second code point is out of decomposable 3036 * code points, finding compositions is unneeded. 3037 */ 3038 if (!IS_DECOMPOSABLE_BLOCK(uc2)) { 3039 WRITE_UC(); 3040 REPLACE_UC_WITH_UC2(); 3041 continue; 3042 } 3043 3044 /* 3045 * Try to combine current code points. 3046 */ 3047 /* 3048 * We have to combine Hangul characters according to 3049 * http://uniicode.org/reports/tr15/#Hangul 3050 */ 3051 if (0 <= (LIndex = uc - HC_LBASE) && 3052 LIndex < HC_LCOUNT) { 3053 /* 3054 * Hangul Composition. 3055 * 1. Two current code points are L and V. 3056 */ 3057 int VIndex = uc2 - HC_VBASE; 3058 if (0 <= VIndex && VIndex < HC_VCOUNT) { 3059 /* Make syllable of form LV. */ 3060 UPDATE_UC(HC_SBASE + 3061 (LIndex * HC_VCOUNT + VIndex) * 3062 HC_TCOUNT); 3063 } else { 3064 WRITE_UC(); 3065 REPLACE_UC_WITH_UC2(); 3066 } 3067 continue; 3068 } else if (0 <= (SIndex = uc - HC_SBASE) && 3069 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { 3070 /* 3071 * Hangul Composition. 3072 * 2. Two current code points are LV and T. 3073 */ 3074 int TIndex = uc2 - HC_TBASE; 3075 if (0 < TIndex && TIndex < HC_TCOUNT) { 3076 /* Make syllable of form LVT. */ 3077 UPDATE_UC(uc + TIndex); 3078 } else { 3079 WRITE_UC(); 3080 REPLACE_UC_WITH_UC2(); 3081 } 3082 continue; 3083 } else if ((nfc = get_nfc(uc, uc2)) != 0) { 3084 /* A composition to current code points 3085 * is found. */ 3086 UPDATE_UC(nfc); 3087 continue; 3088 } else if ((cl = CCC(uc2)) == 0) { 3089 /* Clearly 'uc2' the second code point is not 3090 * a decomposable code. */ 3091 WRITE_UC(); 3092 REPLACE_UC_WITH_UC2(); 3093 continue; 3094 } 3095 3096 /* 3097 * Collect following decomposable code points. 3098 */ 3099 cx = 0; 3100 ucx[0] = uc2; 3101 ccx[0] = cl; 3102 COLLECT_CPS(1); 3103 3104 /* 3105 * Find a composed code in the collected code points. 3106 */ 3107 i = 1; 3108 while (i < ucx_size) { 3109 int j; 3110 3111 if ((nfc = get_nfc(uc, ucx[i])) == 0) { 3112 i++; 3113 continue; 3114 } 3115 3116 /* 3117 * nfc is composed of uc and ucx[i]. 3118 */ 3119 UPDATE_UC(nfc); 3120 3121 /* 3122 * Remove ucx[i] by shifting 3123 * following code points. 3124 */ 3125 for (j = i; j+1 < ucx_size; j++) { 3126 ucx[j] = ucx[j+1]; 3127 ccx[j] = ccx[j+1]; 3128 } 3129 ucx_size --; 3130 3131 /* 3132 * Collect following code points blocked 3133 * by ucx[i] the removed code point. 3134 */ 3135 if (ucx_size > 0 && i == ucx_size && 3136 nx > 0 && cx == cl) { 3137 cl = ccx[ucx_size-1]; 3138 COLLECT_CPS(ucx_size); 3139 } 3140 /* 3141 * Restart finding a composed code with 3142 * the updated uc from the top of the 3143 * collected code points. 3144 */ 3145 i = 0; 3146 } 3147 3148 /* 3149 * Apparently the current code points are not 3150 * decomposed characters or already composed. 3151 */ 3152 WRITE_UC(); 3153 for (i = 0; i < ucx_size; i++) 3154 UNPARSE(p, endp, ucx[i]); 3155 3156 /* 3157 * Flush out remaining canonical combining characters. 3158 */ 3159 if (nx > 0 && cx == cl && len > 0) { 3160 while ((nx = parse(&ucx[0], s, len)) 3161 > 0) { 3162 cx = CCC(ucx[0]); 3163 if (cl > cx) 3164 break; 3165 s += nx; 3166 len -= nx; 3167 cl = cx; 3168 UNPARSE(p, endp, ucx[0]); 3169 } 3170 } 3171 break; 3172 } 3173 if (n2 < 0) { 3174 WRITE_UC(); 3175 /* Use a replaced unicode character. */ 3176 UNPARSE(p, endp, uc2); 3177 s += n2*-1; 3178 len -= n2*-1; 3179 ret = -1; 3180 continue; 3181 } else if (n2 == 0) { 3182 WRITE_UC(); 3183 break; 3184 } 3185 } 3186 as->length = p - as->s; 3187 as->s[as->length] = '\0'; 3188 if (ts == 2) 3189 as->s[as->length+1] = '\0'; 3190 return (ret); 3191 } 3192 3193 static int 3194 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) 3195 { 3196 int t, b; 3197 3198 /* 3199 * These are not converted to NFD on Mac OS. 3200 */ 3201 if ((uc >= 0x2000 && uc <= 0x2FFF) || 3202 (uc >= 0xF900 && uc <= 0xFAFF) || 3203 (uc >= 0x2F800 && uc <= 0x2FAFF)) 3204 return (0); 3205 /* 3206 * Those code points are not converted to NFD on Mac OS. 3207 * I do not know the reason because it is undocumented. 3208 * NFC NFD 3209 * 1109A ==> 11099 110BA 3210 * 1109C ==> 1109B 110BA 3211 * 110AB ==> 110A5 110BA 3212 */ 3213 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) 3214 return (0); 3215 3216 t = 0; 3217 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; 3218 while (b >= t) { 3219 int m = (t + b) / 2; 3220 if (u_decomposition_table[m].nfc < uc) 3221 t = m + 1; 3222 else if (u_decomposition_table[m].nfc > uc) 3223 b = m - 1; 3224 else { 3225 *cp1 = u_decomposition_table[m].cp1; 3226 *cp2 = u_decomposition_table[m].cp2; 3227 return (1); 3228 } 3229 } 3230 return (0); 3231 } 3232 3233 #define REPLACE_UC_WITH(cp) do { \ 3234 uc = cp; \ 3235 ucptr = NULL; \ 3236 } while (0) 3237 3238 /* 3239 * Normalize UTF-8 characters to Form D and copy the result. 3240 */ 3241 static int 3242 archive_string_normalize_D(struct archive_string *as, const void *_p, 3243 size_t len, struct archive_string_conv *sc) 3244 { 3245 const char *s = (const char *)_p; 3246 char *p, *endp; 3247 uint32_t uc, uc2; 3248 size_t w; 3249 int always_replace, n, n2, ret = 0, spair, ts, tm; 3250 int (*parse)(uint32_t *, const char *, size_t); 3251 size_t (*unparse)(char *, size_t, uint32_t); 3252 3253 always_replace = 1; 3254 ts = 1;/* text size. */ 3255 if (sc->flag & SCONV_TO_UTF16BE) { 3256 unparse = unicode_to_utf16be; 3257 ts = 2; 3258 if (sc->flag & SCONV_FROM_UTF16BE) 3259 always_replace = 0; 3260 } else if (sc->flag & SCONV_TO_UTF16LE) { 3261 unparse = unicode_to_utf16le; 3262 ts = 2; 3263 if (sc->flag & SCONV_FROM_UTF16LE) 3264 always_replace = 0; 3265 } else if (sc->flag & SCONV_TO_UTF8) { 3266 unparse = unicode_to_utf8; 3267 if (sc->flag & SCONV_FROM_UTF8) 3268 always_replace = 0; 3269 } else { 3270 /* 3271 * This case is going to be converted to another 3272 * character-set through iconv. 3273 */ 3274 always_replace = 0; 3275 if (sc->flag & SCONV_FROM_UTF16BE) { 3276 unparse = unicode_to_utf16be; 3277 ts = 2; 3278 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3279 unparse = unicode_to_utf16le; 3280 ts = 2; 3281 } else { 3282 unparse = unicode_to_utf8; 3283 } 3284 } 3285 3286 if (sc->flag & SCONV_FROM_UTF16BE) { 3287 parse = utf16be_to_unicode; 3288 tm = 1; 3289 spair = 4;/* surrogate pair size in UTF-16. */ 3290 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3291 parse = utf16le_to_unicode; 3292 tm = 1; 3293 spair = 4;/* surrogate pair size in UTF-16. */ 3294 } else { 3295 parse = cesu8_to_unicode; 3296 tm = ts; 3297 spair = 6;/* surrogate pair size in UTF-8. */ 3298 } 3299 3300 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 3301 return (-1); 3302 3303 p = as->s + as->length; 3304 endp = as->s + as->buffer_length - ts; 3305 while ((n = parse(&uc, s, len)) != 0) { 3306 const char *ucptr; 3307 uint32_t cp1, cp2; 3308 int SIndex; 3309 struct { 3310 uint32_t uc; 3311 int ccc; 3312 } fdc[FDC_MAX]; 3313 int fdi, fdj; 3314 int ccc; 3315 3316 check_first_code: 3317 if (n < 0) { 3318 /* Use a replaced unicode character. */ 3319 UNPARSE(p, endp, uc); 3320 s += n*-1; 3321 len -= n*-1; 3322 ret = -1; 3323 continue; 3324 } else if (n == spair || always_replace) 3325 /* uc is converted from a surrogate pair. 3326 * this should be treated as a changed code. */ 3327 ucptr = NULL; 3328 else 3329 ucptr = s; 3330 s += n; 3331 len -= n; 3332 3333 /* Hangul Decomposition. */ 3334 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { 3335 int L = HC_LBASE + SIndex / HC_NCOUNT; 3336 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; 3337 int T = HC_TBASE + SIndex % HC_TCOUNT; 3338 3339 REPLACE_UC_WITH(L); 3340 WRITE_UC(); 3341 REPLACE_UC_WITH(V); 3342 WRITE_UC(); 3343 if (T != HC_TBASE) { 3344 REPLACE_UC_WITH(T); 3345 WRITE_UC(); 3346 } 3347 continue; 3348 } 3349 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { 3350 WRITE_UC(); 3351 continue; 3352 } 3353 3354 fdi = 0; 3355 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { 3356 int k; 3357 3358 for (k = fdi; k > 0; k--) 3359 fdc[k] = fdc[k-1]; 3360 fdc[0].ccc = CCC(cp2); 3361 fdc[0].uc = cp2; 3362 fdi++; 3363 REPLACE_UC_WITH(cp1); 3364 } 3365 3366 /* Read following code points. */ 3367 while ((n2 = parse(&uc2, s, len)) > 0 && 3368 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { 3369 int j, k; 3370 3371 s += n2; 3372 len -= n2; 3373 for (j = 0; j < fdi; j++) { 3374 if (fdc[j].ccc > ccc) 3375 break; 3376 } 3377 if (j < fdi) { 3378 for (k = fdi; k > j; k--) 3379 fdc[k] = fdc[k-1]; 3380 fdc[j].ccc = ccc; 3381 fdc[j].uc = uc2; 3382 } else { 3383 fdc[fdi].ccc = ccc; 3384 fdc[fdi].uc = uc2; 3385 } 3386 fdi++; 3387 } 3388 3389 WRITE_UC(); 3390 for (fdj = 0; fdj < fdi; fdj++) { 3391 REPLACE_UC_WITH(fdc[fdj].uc); 3392 WRITE_UC(); 3393 } 3394 3395 if (n2 == 0) 3396 break; 3397 REPLACE_UC_WITH(uc2); 3398 n = n2; 3399 goto check_first_code; 3400 } 3401 as->length = p - as->s; 3402 as->s[as->length] = '\0'; 3403 if (ts == 2) 3404 as->s[as->length+1] = '\0'; 3405 return (ret); 3406 } 3407 3408 /* 3409 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption 3410 * that WCS is Unicode. It is true for several platforms but some are false. 3411 * And then people who did not use UTF-8 locale on the non Unicode WCS 3412 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those 3413 * now cannot get right filename from libarchive 3.x and later since we 3414 * fixed the wrong assumption and it is incompatible to older its versions. 3415 * So we provide special option, "compat-2x.x", for resolving it. 3416 * That option enable the string conversion of libarchive 2.x. 3417 * 3418 * Translates the wrong UTF-8 string made by libarchive 2.x into current 3419 * locale character set and appends to the archive_string. 3420 * Note: returns -1 if conversion fails. 3421 */ 3422 static int 3423 strncat_from_utf8_libarchive2(struct archive_string *as, 3424 const void *_p, size_t len, struct archive_string_conv *sc) 3425 { 3426 const char *s; 3427 int n; 3428 char *p; 3429 char *end; 3430 uint32_t unicode; 3431 #if HAVE_WCRTOMB 3432 mbstate_t shift_state; 3433 3434 memset(&shift_state, 0, sizeof(shift_state)); 3435 #else 3436 /* Clear the shift state before starting. */ 3437 wctomb(NULL, L'\0'); 3438 #endif 3439 (void)sc; /* UNUSED */ 3440 /* 3441 * Allocate buffer for MBS. 3442 * We need this allocation here since it is possible that 3443 * as->s is still NULL. 3444 */ 3445 if (archive_string_ensure(as, as->length + len + 1) == NULL) 3446 return (-1); 3447 3448 s = (const char *)_p; 3449 p = as->s + as->length; 3450 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3451 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { 3452 wchar_t wc; 3453 3454 if (p >= end) { 3455 as->length = p - as->s; 3456 /* Re-allocate buffer for MBS. */ 3457 if (archive_string_ensure(as, 3458 as->length + max(len * 2, 3459 (size_t)MB_CUR_MAX) + 1) == NULL) 3460 return (-1); 3461 p = as->s + as->length; 3462 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3463 } 3464 3465 /* 3466 * As libarchive 2.x, translates the UTF-8 characters into 3467 * wide-characters in the assumption that WCS is Unicode. 3468 */ 3469 if (n < 0) { 3470 n *= -1; 3471 wc = L'?'; 3472 } else 3473 wc = (wchar_t)unicode; 3474 3475 s += n; 3476 len -= n; 3477 /* 3478 * Translates the wide-character into the current locale MBS. 3479 */ 3480 #if HAVE_WCRTOMB 3481 n = (int)wcrtomb(p, wc, &shift_state); 3482 #else 3483 n = (int)wctomb(p, wc); 3484 #endif 3485 if (n == -1) 3486 return (-1); 3487 p += n; 3488 } 3489 as->length = p - as->s; 3490 as->s[as->length] = '\0'; 3491 return (0); 3492 } 3493 3494 3495 /* 3496 * Conversion functions between current locale dependent MBS and UTF-16BE. 3497 * strncat_from_utf16be() : UTF-16BE --> MBS 3498 * strncat_to_utf16be() : MBS --> UTF16BE 3499 */ 3500 3501 #if defined(_WIN32) && !defined(__CYGWIN__) 3502 3503 /* 3504 * Convert a UTF-16BE/LE string to current locale and copy the result. 3505 * Return -1 if conversion fails. 3506 */ 3507 static int 3508 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, 3509 struct archive_string_conv *sc, int be) 3510 { 3511 struct archive_string tmp; 3512 const char *u16; 3513 int ll; 3514 BOOL defchar; 3515 char *mbs; 3516 size_t mbs_size, b; 3517 int ret = 0; 3518 3519 bytes &= ~1; 3520 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3521 return (-1); 3522 3523 mbs = as->s + as->length; 3524 mbs_size = as->buffer_length - as->length -1; 3525 3526 if (sc->to_cp == CP_C_LOCALE) { 3527 /* 3528 * "C" locale special process. 3529 */ 3530 u16 = _p; 3531 ll = 0; 3532 for (b = 0; b < bytes; b += 2) { 3533 uint16_t val; 3534 if (be) 3535 val = archive_be16dec(u16+b); 3536 else 3537 val = archive_le16dec(u16+b); 3538 if (val > 255) { 3539 *mbs++ = '?'; 3540 ret = -1; 3541 } else 3542 *mbs++ = (char)(val&0xff); 3543 ll++; 3544 } 3545 as->length += ll; 3546 as->s[as->length] = '\0'; 3547 return (ret); 3548 } 3549 3550 archive_string_init(&tmp); 3551 if (be) { 3552 if (is_big_endian()) { 3553 u16 = _p; 3554 } else { 3555 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3556 return (-1); 3557 memcpy(tmp.s, _p, bytes); 3558 for (b = 0; b < bytes; b += 2) { 3559 uint16_t val = archive_be16dec(tmp.s+b); 3560 archive_le16enc(tmp.s+b, val); 3561 } 3562 u16 = tmp.s; 3563 } 3564 } else { 3565 if (!is_big_endian()) { 3566 u16 = _p; 3567 } else { 3568 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3569 return (-1); 3570 memcpy(tmp.s, _p, bytes); 3571 for (b = 0; b < bytes; b += 2) { 3572 uint16_t val = archive_le16dec(tmp.s+b); 3573 archive_be16enc(tmp.s+b, val); 3574 } 3575 u16 = tmp.s; 3576 } 3577 } 3578 3579 do { 3580 defchar = 0; 3581 ll = WideCharToMultiByte(sc->to_cp, 0, 3582 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, 3583 NULL, &defchar); 3584 /* Exit loop if we succeeded */ 3585 if (ll != 0 || 3586 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3587 break; 3588 } 3589 /* Else expand buffer and loop to try again. */ 3590 ll = WideCharToMultiByte(sc->to_cp, 0, 3591 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); 3592 if (archive_string_ensure(as, ll +1) == NULL) 3593 return (-1); 3594 mbs = as->s + as->length; 3595 mbs_size = as->buffer_length - as->length -1; 3596 } while (1); 3597 archive_string_free(&tmp); 3598 as->length += ll; 3599 as->s[as->length] = '\0'; 3600 if (ll == 0 || defchar) 3601 ret = -1; 3602 return (ret); 3603 } 3604 3605 static int 3606 win_strncat_from_utf16be(struct archive_string *as, const void *_p, 3607 size_t bytes, struct archive_string_conv *sc) 3608 { 3609 return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); 3610 } 3611 3612 static int 3613 win_strncat_from_utf16le(struct archive_string *as, const void *_p, 3614 size_t bytes, struct archive_string_conv *sc) 3615 { 3616 return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); 3617 } 3618 3619 static int 3620 is_big_endian(void) 3621 { 3622 uint16_t d = 1; 3623 3624 return (archive_be16dec(&d) == 1); 3625 } 3626 3627 /* 3628 * Convert a current locale string to UTF-16BE/LE and copy the result. 3629 * Return -1 if conversion fails. 3630 */ 3631 static int 3632 win_strncat_to_utf16(struct archive_string *as16, const void *_p, 3633 size_t length, struct archive_string_conv *sc, int bigendian) 3634 { 3635 const char *s = (const char *)_p; 3636 char *u16; 3637 size_t count, avail; 3638 3639 if (archive_string_ensure(as16, 3640 as16->length + (length + 1) * 2) == NULL) 3641 return (-1); 3642 3643 u16 = as16->s + as16->length; 3644 avail = as16->buffer_length - 2; 3645 if (sc->from_cp == CP_C_LOCALE) { 3646 /* 3647 * "C" locale special process. 3648 */ 3649 count = 0; 3650 while (count < length && *s) { 3651 if (bigendian) 3652 archive_be16enc(u16, *s); 3653 else 3654 archive_le16enc(u16, *s); 3655 u16 += 2; 3656 s++; 3657 count++; 3658 } 3659 as16->length += count << 1; 3660 as16->s[as16->length] = 0; 3661 as16->s[as16->length+1] = 0; 3662 return (0); 3663 } 3664 do { 3665 count = MultiByteToWideChar(sc->from_cp, 3666 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); 3667 /* Exit loop if we succeeded */ 3668 if (count != 0 || 3669 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3670 break; 3671 } 3672 /* Expand buffer and try again */ 3673 count = MultiByteToWideChar(sc->from_cp, 3674 MB_PRECOMPOSED, s, (int)length, NULL, 0); 3675 if (archive_string_ensure(as16, (count +1) * 2) 3676 == NULL) 3677 return (-1); 3678 u16 = as16->s + as16->length; 3679 avail = as16->buffer_length - 2; 3680 } while (1); 3681 as16->length += count * 2; 3682 as16->s[as16->length] = 0; 3683 as16->s[as16->length+1] = 0; 3684 if (count == 0) 3685 return (-1); 3686 3687 if (is_big_endian()) { 3688 if (!bigendian) { 3689 while (count > 0) { 3690 uint16_t v = archive_be16dec(u16); 3691 archive_le16enc(u16, v); 3692 u16 += 2; 3693 count--; 3694 } 3695 } 3696 } else { 3697 if (bigendian) { 3698 while (count > 0) { 3699 uint16_t v = archive_le16dec(u16); 3700 archive_be16enc(u16, v); 3701 u16 += 2; 3702 count--; 3703 } 3704 } 3705 } 3706 return (0); 3707 } 3708 3709 static int 3710 win_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3711 size_t length, struct archive_string_conv *sc) 3712 { 3713 return (win_strncat_to_utf16(as16, _p, length, sc, 1)); 3714 } 3715 3716 static int 3717 win_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3718 size_t length, struct archive_string_conv *sc) 3719 { 3720 return (win_strncat_to_utf16(as16, _p, length, sc, 0)); 3721 } 3722 3723 #endif /* _WIN32 && !__CYGWIN__ */ 3724 3725 /* 3726 * Do the best effort for conversions. 3727 * We cannot handle UTF-16BE character-set without such iconv, 3728 * but there is a chance if a string consists just ASCII code or 3729 * a current locale is UTF-8. 3730 */ 3731 3732 /* 3733 * Convert a UTF-16BE string to current locale and copy the result. 3734 * Return -1 if conversion fails. 3735 */ 3736 static int 3737 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, 3738 size_t bytes, struct archive_string_conv *sc, int be) 3739 { 3740 const char *utf16 = (const char *)_p; 3741 char *mbs; 3742 uint32_t uc; 3743 int n, ret; 3744 3745 (void)sc; /* UNUSED */ 3746 /* 3747 * Other case, we should do the best effort. 3748 * If all character are ASCII(<0x7f), we can convert it. 3749 * if not , we set a alternative character and return -1. 3750 */ 3751 ret = 0; 3752 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3753 return (-1); 3754 mbs = as->s + as->length; 3755 3756 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { 3757 if (n < 0) { 3758 n *= -1; 3759 ret = -1; 3760 } 3761 bytes -= n; 3762 utf16 += n; 3763 3764 if (uc > 127) { 3765 /* We cannot handle it. */ 3766 *mbs++ = '?'; 3767 ret = -1; 3768 } else 3769 *mbs++ = (char)uc; 3770 } 3771 as->length = mbs - as->s; 3772 as->s[as->length] = '\0'; 3773 return (ret); 3774 } 3775 3776 static int 3777 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, 3778 size_t bytes, struct archive_string_conv *sc) 3779 { 3780 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); 3781 } 3782 3783 static int 3784 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, 3785 size_t bytes, struct archive_string_conv *sc) 3786 { 3787 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); 3788 } 3789 3790 /* 3791 * Convert a current locale string to UTF-16BE/LE and copy the result. 3792 * Return -1 if conversion fails. 3793 */ 3794 static int 3795 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, 3796 size_t length, struct archive_string_conv *sc, int bigendian) 3797 { 3798 const char *s = (const char *)_p; 3799 char *utf16; 3800 size_t remaining; 3801 int ret; 3802 3803 (void)sc; /* UNUSED */ 3804 /* 3805 * Other case, we should do the best effort. 3806 * If all character are ASCII(<0x7f), we can convert it. 3807 * if not , we set a alternative character and return -1. 3808 */ 3809 ret = 0; 3810 remaining = length; 3811 3812 if (archive_string_ensure(as16, 3813 as16->length + (length + 1) * 2) == NULL) 3814 return (-1); 3815 3816 utf16 = as16->s + as16->length; 3817 while (remaining--) { 3818 unsigned c = *s++; 3819 if (c > 127) { 3820 /* We cannot handle it. */ 3821 c = UNICODE_R_CHAR; 3822 ret = -1; 3823 } 3824 if (bigendian) 3825 archive_be16enc(utf16, c); 3826 else 3827 archive_le16enc(utf16, c); 3828 utf16 += 2; 3829 } 3830 as16->length = utf16 - as16->s; 3831 as16->s[as16->length] = 0; 3832 as16->s[as16->length+1] = 0; 3833 return (ret); 3834 } 3835 3836 static int 3837 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3838 size_t length, struct archive_string_conv *sc) 3839 { 3840 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); 3841 } 3842 3843 static int 3844 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3845 size_t length, struct archive_string_conv *sc) 3846 { 3847 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); 3848 } 3849 3850 3851 /* 3852 * Multistring operations. 3853 */ 3854 3855 void 3856 archive_mstring_clean(struct archive_mstring *aes) 3857 { 3858 archive_wstring_free(&(aes->aes_wcs)); 3859 archive_string_free(&(aes->aes_mbs)); 3860 archive_string_free(&(aes->aes_utf8)); 3861 archive_string_free(&(aes->aes_mbs_in_locale)); 3862 aes->aes_set = 0; 3863 } 3864 3865 void 3866 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) 3867 { 3868 dest->aes_set = src->aes_set; 3869 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); 3870 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); 3871 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); 3872 } 3873 3874 int 3875 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, 3876 const char **p) 3877 { 3878 struct archive_string_conv *sc; 3879 int r; 3880 3881 /* If we already have a UTF8 form, return that immediately. */ 3882 if (aes->aes_set & AES_SET_UTF8) { 3883 *p = aes->aes_utf8.s; 3884 return (0); 3885 } 3886 3887 *p = NULL; 3888 /* Try converting WCS to MBS first if MBS does not exist yet. */ 3889 if ((aes->aes_set & AES_SET_MBS) == 0) { 3890 const char *pm; /* unused */ 3891 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */ 3892 } 3893 if (aes->aes_set & AES_SET_MBS) { 3894 sc = archive_string_conversion_to_charset(a, "UTF-8", 1); 3895 if (sc == NULL) 3896 return (-1);/* Couldn't allocate memory for sc. */ 3897 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, 3898 aes->aes_mbs.length, sc); 3899 if (a == NULL) 3900 free_sconv_object(sc); 3901 if (r == 0) { 3902 aes->aes_set |= AES_SET_UTF8; 3903 *p = aes->aes_utf8.s; 3904 return (0);/* success. */ 3905 } else 3906 return (-1);/* failure. */ 3907 } 3908 return (0);/* success. */ 3909 } 3910 3911 int 3912 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, 3913 const char **p) 3914 { 3915 struct archive_string_conv *sc; 3916 int r, ret = 0; 3917 3918 /* If we already have an MBS form, return that immediately. */ 3919 if (aes->aes_set & AES_SET_MBS) { 3920 *p = aes->aes_mbs.s; 3921 return (ret); 3922 } 3923 3924 *p = NULL; 3925 /* If there's a WCS form, try converting with the native locale. */ 3926 if (aes->aes_set & AES_SET_WCS) { 3927 archive_string_empty(&(aes->aes_mbs)); 3928 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3929 aes->aes_wcs.s, aes->aes_wcs.length); 3930 *p = aes->aes_mbs.s; 3931 if (r == 0) { 3932 aes->aes_set |= AES_SET_MBS; 3933 return (ret); 3934 } else 3935 ret = -1; 3936 } 3937 3938 /* If there's a UTF-8 form, try converting with the native locale. */ 3939 if (aes->aes_set & AES_SET_UTF8) { 3940 archive_string_empty(&(aes->aes_mbs)); 3941 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 3942 if (sc == NULL) 3943 return (-1);/* Couldn't allocate memory for sc. */ 3944 r = archive_strncpy_l(&(aes->aes_mbs), 3945 aes->aes_utf8.s, aes->aes_utf8.length, sc); 3946 if (a == NULL) 3947 free_sconv_object(sc); 3948 *p = aes->aes_mbs.s; 3949 if (r == 0) { 3950 aes->aes_set |= AES_SET_MBS; 3951 ret = 0;/* success; overwrite previous error. */ 3952 } else 3953 ret = -1;/* failure. */ 3954 } 3955 return (ret); 3956 } 3957 3958 int 3959 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, 3960 const wchar_t **wp) 3961 { 3962 int r, ret = 0; 3963 3964 (void)a;/* UNUSED */ 3965 /* Return WCS form if we already have it. */ 3966 if (aes->aes_set & AES_SET_WCS) { 3967 *wp = aes->aes_wcs.s; 3968 return (ret); 3969 } 3970 3971 *wp = NULL; 3972 /* Try converting UTF8 to MBS first if MBS does not exist yet. */ 3973 if ((aes->aes_set & AES_SET_MBS) == 0) { 3974 const char *p; /* unused */ 3975 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */ 3976 } 3977 /* Try converting MBS to WCS using native locale. */ 3978 if (aes->aes_set & AES_SET_MBS) { 3979 archive_wstring_empty(&(aes->aes_wcs)); 3980 r = archive_wstring_append_from_mbs(&(aes->aes_wcs), 3981 aes->aes_mbs.s, aes->aes_mbs.length); 3982 if (r == 0) { 3983 aes->aes_set |= AES_SET_WCS; 3984 *wp = aes->aes_wcs.s; 3985 } else 3986 ret = -1;/* failure. */ 3987 } 3988 return (ret); 3989 } 3990 3991 int 3992 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes, 3993 const char **p, size_t *length, struct archive_string_conv *sc) 3994 { 3995 int ret = 0; 3996 #if defined(_WIN32) && !defined(__CYGWIN__) 3997 int r; 3998 3999 /* 4000 * Internationalization programming on Windows must use Wide 4001 * characters because Windows platform cannot make locale UTF-8. 4002 */ 4003 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { 4004 archive_string_empty(&(aes->aes_mbs_in_locale)); 4005 r = archive_string_append_from_wcs_in_codepage( 4006 &(aes->aes_mbs_in_locale), aes->aes_wcs.s, 4007 aes->aes_wcs.length, sc); 4008 if (r == 0) { 4009 *p = aes->aes_mbs_in_locale.s; 4010 if (length != NULL) 4011 *length = aes->aes_mbs_in_locale.length; 4012 return (0); 4013 } else if (errno == ENOMEM) 4014 return (-1); 4015 else 4016 ret = -1; 4017 } 4018 #endif 4019 4020 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting 4021 * with the native locale to be used for translating it to specified 4022 * character-set. */ 4023 if ((aes->aes_set & AES_SET_MBS) == 0) { 4024 const char *pm; /* unused */ 4025 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */ 4026 } 4027 /* If we already have an MBS form, use it to be translated to 4028 * specified character-set. */ 4029 if (aes->aes_set & AES_SET_MBS) { 4030 if (sc == NULL) { 4031 /* Conversion is unneeded. */ 4032 *p = aes->aes_mbs.s; 4033 if (length != NULL) 4034 *length = aes->aes_mbs.length; 4035 return (0); 4036 } 4037 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), 4038 aes->aes_mbs.s, aes->aes_mbs.length, sc); 4039 *p = aes->aes_mbs_in_locale.s; 4040 if (length != NULL) 4041 *length = aes->aes_mbs_in_locale.length; 4042 } else { 4043 *p = NULL; 4044 if (length != NULL) 4045 *length = 0; 4046 } 4047 return (ret); 4048 } 4049 4050 int 4051 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) 4052 { 4053 if (mbs == NULL) { 4054 aes->aes_set = 0; 4055 return (0); 4056 } 4057 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); 4058 } 4059 4060 int 4061 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, 4062 size_t len) 4063 { 4064 if (mbs == NULL) { 4065 aes->aes_set = 0; 4066 return (0); 4067 } 4068 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4069 archive_strncpy(&(aes->aes_mbs), mbs, len); 4070 archive_string_empty(&(aes->aes_utf8)); 4071 archive_wstring_empty(&(aes->aes_wcs)); 4072 return (0); 4073 } 4074 4075 int 4076 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) 4077 { 4078 return archive_mstring_copy_wcs_len(aes, wcs, 4079 wcs == NULL ? 0 : wcslen(wcs)); 4080 } 4081 4082 int 4083 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) 4084 { 4085 if (utf8 == NULL) { 4086 aes->aes_set = 0; 4087 return (0); 4088 } 4089 aes->aes_set = AES_SET_UTF8; 4090 archive_string_empty(&(aes->aes_mbs)); 4091 archive_string_empty(&(aes->aes_wcs)); 4092 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); 4093 return (int)strlen(utf8); 4094 } 4095 4096 int 4097 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, 4098 size_t len) 4099 { 4100 if (wcs == NULL) { 4101 aes->aes_set = 0; 4102 return (0); 4103 } 4104 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ 4105 archive_string_empty(&(aes->aes_mbs)); 4106 archive_string_empty(&(aes->aes_utf8)); 4107 archive_wstrncpy(&(aes->aes_wcs), wcs, len); 4108 return (0); 4109 } 4110 4111 int 4112 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, 4113 const char *mbs, size_t len, struct archive_string_conv *sc) 4114 { 4115 int r; 4116 4117 if (mbs == NULL) { 4118 aes->aes_set = 0; 4119 return (0); 4120 } 4121 archive_string_empty(&(aes->aes_mbs)); 4122 archive_wstring_empty(&(aes->aes_wcs)); 4123 archive_string_empty(&(aes->aes_utf8)); 4124 #if defined(_WIN32) && !defined(__CYGWIN__) 4125 /* 4126 * Internationalization programming on Windows must use Wide 4127 * characters because Windows platform cannot make locale UTF-8. 4128 */ 4129 if (sc == NULL) { 4130 if (archive_string_append(&(aes->aes_mbs), 4131 mbs, mbsnbytes(mbs, len)) == NULL) { 4132 aes->aes_set = 0; 4133 r = -1; 4134 } else { 4135 aes->aes_set = AES_SET_MBS; 4136 r = 0; 4137 } 4138 #if defined(HAVE_ICONV) 4139 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { 4140 /* 4141 * This case happens only when MultiByteToWideChar() cannot 4142 * handle sc->from_cp, and we have to iconv in order to 4143 * translate character-set to wchar_t,UTF-16. 4144 */ 4145 iconv_t cd = sc->cd; 4146 unsigned from_cp; 4147 int flag; 4148 4149 /* 4150 * Translate multi-bytes from some character-set to UTF-8. 4151 */ 4152 sc->cd = sc->cd_w; 4153 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); 4154 sc->cd = cd; 4155 if (r != 0) { 4156 aes->aes_set = 0; 4157 return (r); 4158 } 4159 aes->aes_set = AES_SET_UTF8; 4160 4161 /* 4162 * Append the UTF-8 string into wstring. 4163 */ 4164 flag = sc->flag; 4165 sc->flag &= ~(SCONV_NORMALIZATION_C 4166 | SCONV_TO_UTF16| SCONV_FROM_UTF16); 4167 from_cp = sc->from_cp; 4168 sc->from_cp = CP_UTF8; 4169 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), 4170 aes->aes_utf8.s, aes->aes_utf8.length, sc); 4171 sc->flag = flag; 4172 sc->from_cp = from_cp; 4173 if (r == 0) 4174 aes->aes_set |= AES_SET_WCS; 4175 #endif 4176 } else { 4177 r = archive_wstring_append_from_mbs_in_codepage( 4178 &(aes->aes_wcs), mbs, len, sc); 4179 if (r == 0) 4180 aes->aes_set = AES_SET_WCS; 4181 else 4182 aes->aes_set = 0; 4183 } 4184 #else 4185 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); 4186 if (r == 0) 4187 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4188 else 4189 aes->aes_set = 0; 4190 #endif 4191 return (r); 4192 } 4193 4194 /* 4195 * The 'update' form tries to proactively update all forms of 4196 * this string (WCS and MBS) and returns an error if any of 4197 * them fail. This is used by the 'pax' handler, for instance, 4198 * to detect and report character-conversion failures early while 4199 * still allowing clients to get potentially useful values from 4200 * the more tolerant lazy conversions. (get_mbs and get_wcs will 4201 * strive to give the user something useful, so you can get hopefully 4202 * usable values even if some of the character conversions are failing.) 4203 */ 4204 int 4205 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, 4206 const char *utf8) 4207 { 4208 struct archive_string_conv *sc; 4209 int r; 4210 4211 if (utf8 == NULL) { 4212 aes->aes_set = 0; 4213 return (0); /* Succeeded in clearing everything. */ 4214 } 4215 4216 /* Save the UTF8 string. */ 4217 archive_strcpy(&(aes->aes_utf8), utf8); 4218 4219 /* Empty the mbs and wcs strings. */ 4220 archive_string_empty(&(aes->aes_mbs)); 4221 archive_wstring_empty(&(aes->aes_wcs)); 4222 4223 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ 4224 4225 /* Try converting UTF-8 to MBS, return false on failure. */ 4226 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 4227 if (sc == NULL) 4228 return (-1);/* Couldn't allocate memory for sc. */ 4229 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); 4230 if (a == NULL) 4231 free_sconv_object(sc); 4232 if (r != 0) 4233 return (-1); 4234 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ 4235 4236 /* Try converting MBS to WCS, return false on failure. */ 4237 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, 4238 aes->aes_mbs.length)) 4239 return (-1); 4240 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; 4241 4242 /* All conversions succeeded. */ 4243 return (0); 4244 } 4245