archive_string.c revision 318482
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: stable/11/contrib/libarchive/libarchive/archive_string.c 318482 2017-05-18 19:47:43Z mm $"); 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 78struct archive_string_conv { 79 struct archive_string_conv *next; 80 char *from_charset; 81 char *to_charset; 82 unsigned from_cp; 83 unsigned to_cp; 84 /* Set 1 if from_charset and to_charset are the same. */ 85 int same; 86 int flag; 87#define SCONV_TO_CHARSET 1 /* MBS is being converted to specified 88 * charset. */ 89#define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from 90 * specified charset. */ 91#define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */ 92#define SCONV_WIN_CP (1<<3) /* Use Windows API for converting 93 * MBS. */ 94#define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive 95 * 2.x in the wrong assumption. */ 96#define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C. 97 * Before UTF-8 characters are actually 98 * processed. */ 99#define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D. 100 * Before UTF-8 characters are actually 101 * processed. 102 * Currently this only for MAC OS X. */ 103#define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */ 104#define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */ 105#define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */ 106#define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */ 107#define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */ 108#define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */ 109#define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE) 110#define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE) 111 112#if HAVE_ICONV 113 iconv_t cd; 114 iconv_t cd_w;/* Use at archive_mstring on 115 * Windows. */ 116#endif 117 /* A temporary buffer for normalization. */ 118 struct archive_string utftmp; 119 int (*converter[2])(struct archive_string *, const void *, size_t, 120 struct archive_string_conv *); 121 int nconverter; 122}; 123 124#define CP_C_LOCALE 0 /* "C" locale only for this file. */ 125#define CP_UTF16LE 1200 126#define CP_UTF16BE 1201 127 128#define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF) 129#define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF) 130#define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF) 131#define UNICODE_MAX 0x10FFFF 132#define UNICODE_R_CHAR 0xFFFD /* Replacement character. */ 133/* Set U+FFFD(Replacement character) in UTF-8. */ 134static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd}; 135 136static struct archive_string_conv *find_sconv_object(struct archive *, 137 const char *, const char *); 138static void add_sconv_object(struct archive *, struct archive_string_conv *); 139static struct archive_string_conv *create_sconv_object(const char *, 140 const char *, unsigned, int); 141static void free_sconv_object(struct archive_string_conv *); 142static struct archive_string_conv *get_sconv_object(struct archive *, 143 const char *, const char *, int); 144static unsigned make_codepage_from_charset(const char *); 145static unsigned get_current_codepage(void); 146static unsigned get_current_oemcp(void); 147static size_t mbsnbytes(const void *, size_t); 148static size_t utf16nbytes(const void *, size_t); 149#if defined(_WIN32) && !defined(__CYGWIN__) 150static int archive_wstring_append_from_mbs_in_codepage( 151 struct archive_wstring *, const char *, size_t, 152 struct archive_string_conv *); 153static int archive_string_append_from_wcs_in_codepage(struct archive_string *, 154 const wchar_t *, size_t, struct archive_string_conv *); 155static int is_big_endian(void); 156static int strncat_in_codepage(struct archive_string *, const void *, 157 size_t, struct archive_string_conv *); 158static int win_strncat_from_utf16be(struct archive_string *, const void *, 159 size_t, struct archive_string_conv *); 160static int win_strncat_from_utf16le(struct archive_string *, const void *, 161 size_t, struct archive_string_conv *); 162static int win_strncat_to_utf16be(struct archive_string *, const void *, 163 size_t, struct archive_string_conv *); 164static int win_strncat_to_utf16le(struct archive_string *, const void *, 165 size_t, struct archive_string_conv *); 166#endif 167static int best_effort_strncat_from_utf16be(struct archive_string *, 168 const void *, size_t, struct archive_string_conv *); 169static int best_effort_strncat_from_utf16le(struct archive_string *, 170 const void *, size_t, struct archive_string_conv *); 171static int best_effort_strncat_to_utf16be(struct archive_string *, 172 const void *, size_t, struct archive_string_conv *); 173static int best_effort_strncat_to_utf16le(struct archive_string *, 174 const void *, size_t, struct archive_string_conv *); 175#if defined(HAVE_ICONV) 176static int iconv_strncat_in_locale(struct archive_string *, const void *, 177 size_t, struct archive_string_conv *); 178#endif 179static int best_effort_strncat_in_locale(struct archive_string *, 180 const void *, size_t, struct archive_string_conv *); 181static int _utf8_to_unicode(uint32_t *, const char *, size_t); 182static int utf8_to_unicode(uint32_t *, const char *, size_t); 183static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t); 184static int cesu8_to_unicode(uint32_t *, const char *, size_t); 185static size_t unicode_to_utf8(char *, size_t, uint32_t); 186static int utf16_to_unicode(uint32_t *, const char *, size_t, int); 187static size_t unicode_to_utf16be(char *, size_t, uint32_t); 188static size_t unicode_to_utf16le(char *, size_t, uint32_t); 189static int strncat_from_utf8_libarchive2(struct archive_string *, 190 const void *, size_t, struct archive_string_conv *); 191static int strncat_from_utf8_to_utf8(struct archive_string *, const void *, 192 size_t, struct archive_string_conv *); 193static int archive_string_normalize_C(struct archive_string *, const void *, 194 size_t, struct archive_string_conv *); 195static int archive_string_normalize_D(struct archive_string *, const void *, 196 size_t, struct archive_string_conv *); 197static int archive_string_append_unicode(struct archive_string *, 198 const void *, size_t, struct archive_string_conv *); 199 200static struct archive_string * 201archive_string_append(struct archive_string *as, const char *p, size_t s) 202{ 203 if (archive_string_ensure(as, as->length + s + 1) == NULL) 204 return (NULL); 205 if (s) 206 memmove(as->s + as->length, p, s); 207 as->length += s; 208 as->s[as->length] = 0; 209 return (as); 210} 211 212static struct archive_wstring * 213archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s) 214{ 215 if (archive_wstring_ensure(as, as->length + s + 1) == NULL) 216 return (NULL); 217 wmemmove(as->s + as->length, p, s); 218 as->length += s; 219 as->s[as->length] = 0; 220 return (as); 221} 222 223struct archive_string * 224archive_array_append(struct archive_string *as, const char *p, size_t s) 225{ 226 return archive_string_append(as, p, s); 227} 228 229void 230archive_string_concat(struct archive_string *dest, struct archive_string *src) 231{ 232 if (archive_string_append(dest, src->s, src->length) == NULL) 233 __archive_errx(1, "Out of memory"); 234} 235 236void 237archive_wstring_concat(struct archive_wstring *dest, 238 struct archive_wstring *src) 239{ 240 if (archive_wstring_append(dest, src->s, src->length) == NULL) 241 __archive_errx(1, "Out of memory"); 242} 243 244void 245archive_string_free(struct archive_string *as) 246{ 247 as->length = 0; 248 as->buffer_length = 0; 249 free(as->s); 250 as->s = NULL; 251} 252 253void 254archive_wstring_free(struct archive_wstring *as) 255{ 256 as->length = 0; 257 as->buffer_length = 0; 258 free(as->s); 259 as->s = NULL; 260} 261 262struct archive_wstring * 263archive_wstring_ensure(struct archive_wstring *as, size_t s) 264{ 265 return (struct archive_wstring *) 266 archive_string_ensure((struct archive_string *)as, 267 s * sizeof(wchar_t)); 268} 269 270/* Returns NULL on any allocation failure. */ 271struct archive_string * 272archive_string_ensure(struct archive_string *as, size_t s) 273{ 274 char *p; 275 size_t new_length; 276 277 /* If buffer is already big enough, don't reallocate. */ 278 if (as->s && (s <= as->buffer_length)) 279 return (as); 280 281 /* 282 * Growing the buffer at least exponentially ensures that 283 * append operations are always linear in the number of 284 * characters appended. Using a smaller growth rate for 285 * larger buffers reduces memory waste somewhat at the cost of 286 * a larger constant factor. 287 */ 288 if (as->buffer_length < 32) 289 /* Start with a minimum 32-character buffer. */ 290 new_length = 32; 291 else if (as->buffer_length < 8192) 292 /* Buffers under 8k are doubled for speed. */ 293 new_length = as->buffer_length + as->buffer_length; 294 else { 295 /* Buffers 8k and over grow by at least 25% each time. */ 296 new_length = as->buffer_length + as->buffer_length / 4; 297 /* Be safe: If size wraps, fail. */ 298 if (new_length < as->buffer_length) { 299 /* On failure, wipe the string and return NULL. */ 300 archive_string_free(as); 301 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 302 return (NULL); 303 } 304 } 305 /* 306 * The computation above is a lower limit to how much we'll 307 * grow the buffer. In any case, we have to grow it enough to 308 * hold the request. 309 */ 310 if (new_length < s) 311 new_length = s; 312 /* Now we can reallocate the buffer. */ 313 p = (char *)realloc(as->s, new_length); 314 if (p == NULL) { 315 /* On failure, wipe the string and return NULL. */ 316 archive_string_free(as); 317 errno = ENOMEM;/* Make sure errno has ENOMEM. */ 318 return (NULL); 319 } 320 321 as->s = p; 322 as->buffer_length = new_length; 323 return (as); 324} 325 326/* 327 * TODO: See if there's a way to avoid scanning 328 * the source string twice. Then test to see 329 * if it actually helps (remember that we're almost 330 * always called with pretty short arguments, so 331 * such an optimization might not help). 332 */ 333struct archive_string * 334archive_strncat(struct archive_string *as, const void *_p, size_t n) 335{ 336 size_t s; 337 const char *p, *pp; 338 339 p = (const char *)_p; 340 341 /* Like strlen(p), except won't examine positions beyond p[n]. */ 342 s = 0; 343 pp = p; 344 while (s < n && *pp) { 345 pp++; 346 s++; 347 } 348 if ((as = archive_string_append(as, p, s)) == NULL) 349 __archive_errx(1, "Out of memory"); 350 return (as); 351} 352 353struct archive_wstring * 354archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n) 355{ 356 size_t s; 357 const wchar_t *pp; 358 359 /* Like strlen(p), except won't examine positions beyond p[n]. */ 360 s = 0; 361 pp = p; 362 while (s < n && *pp) { 363 pp++; 364 s++; 365 } 366 if ((as = archive_wstring_append(as, p, s)) == NULL) 367 __archive_errx(1, "Out of memory"); 368 return (as); 369} 370 371struct archive_string * 372archive_strcat(struct archive_string *as, const void *p) 373{ 374 /* strcat is just strncat without an effective limit. 375 * Assert that we'll never get called with a source 376 * string over 16MB. 377 * TODO: Review all uses of strcat in the source 378 * and try to replace them with strncat(). 379 */ 380 return archive_strncat(as, p, 0x1000000); 381} 382 383struct archive_wstring * 384archive_wstrcat(struct archive_wstring *as, const wchar_t *p) 385{ 386 /* Ditto. */ 387 return archive_wstrncat(as, p, 0x1000000); 388} 389 390struct archive_string * 391archive_strappend_char(struct archive_string *as, char c) 392{ 393 if ((as = archive_string_append(as, &c, 1)) == NULL) 394 __archive_errx(1, "Out of memory"); 395 return (as); 396} 397 398struct archive_wstring * 399archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c) 400{ 401 if ((as = archive_wstring_append(as, &c, 1)) == NULL) 402 __archive_errx(1, "Out of memory"); 403 return (as); 404} 405 406/* 407 * Get the "current character set" name to use with iconv. 408 * On FreeBSD, the empty character set name "" chooses 409 * the correct character encoding for the current locale, 410 * so this isn't necessary. 411 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly; 412 * on that system, we have to explicitly call nl_langinfo() 413 * to get the right name. Not sure about other platforms. 414 * 415 * NOTE: GNU libiconv does not recognize the character-set name 416 * which some platform nl_langinfo(CODESET) returns, so we should 417 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv. 418 */ 419static const char * 420default_iconv_charset(const char *charset) { 421 if (charset != NULL && charset[0] != '\0') 422 return charset; 423#if HAVE_LOCALE_CHARSET && !defined(__APPLE__) 424 /* locale_charset() is broken on Mac OS */ 425 return locale_charset(); 426#elif HAVE_NL_LANGINFO 427 return nl_langinfo(CODESET); 428#else 429 return ""; 430#endif 431} 432 433#if defined(_WIN32) && !defined(__CYGWIN__) 434 435/* 436 * Convert MBS to WCS. 437 * Note: returns -1 if conversion fails. 438 */ 439int 440archive_wstring_append_from_mbs(struct archive_wstring *dest, 441 const char *p, size_t len) 442{ 443 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL); 444} 445 446static int 447archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest, 448 const char *s, size_t length, struct archive_string_conv *sc) 449{ 450 int count, ret = 0; 451 UINT from_cp; 452 453 if (sc != NULL) 454 from_cp = sc->from_cp; 455 else 456 from_cp = get_current_codepage(); 457 458 if (from_cp == CP_C_LOCALE) { 459 /* 460 * "C" locale special process. 461 */ 462 wchar_t *ws; 463 const unsigned char *mp; 464 465 if (NULL == archive_wstring_ensure(dest, 466 dest->length + length + 1)) 467 return (-1); 468 469 ws = dest->s + dest->length; 470 mp = (const unsigned char *)s; 471 count = 0; 472 while (count < (int)length && *mp) { 473 *ws++ = (wchar_t)*mp++; 474 count++; 475 } 476 } else if (sc != NULL && 477 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) { 478 /* 479 * Normalize UTF-8 and UTF-16BE and convert it directly 480 * to UTF-16 as wchar_t. 481 */ 482 struct archive_string u16; 483 int saved_flag = sc->flag;/* save current flag. */ 484 485 if (is_big_endian()) 486 sc->flag |= SCONV_TO_UTF16BE; 487 else 488 sc->flag |= SCONV_TO_UTF16LE; 489 490 if (sc->flag & SCONV_FROM_UTF16) { 491 /* 492 * UTF-16BE/LE NFD ===> UTF-16 NFC 493 * UTF-16BE/LE NFC ===> UTF-16 NFD 494 */ 495 count = (int)utf16nbytes(s, length); 496 } else { 497 /* 498 * UTF-8 NFD ===> UTF-16 NFC 499 * UTF-8 NFC ===> UTF-16 NFD 500 */ 501 count = (int)mbsnbytes(s, length); 502 } 503 u16.s = (char *)dest->s; 504 u16.length = dest->length << 1;; 505 u16.buffer_length = dest->buffer_length; 506 if (sc->flag & SCONV_NORMALIZATION_C) 507 ret = archive_string_normalize_C(&u16, s, count, sc); 508 else 509 ret = archive_string_normalize_D(&u16, s, count, sc); 510 dest->s = (wchar_t *)u16.s; 511 dest->length = u16.length >> 1; 512 dest->buffer_length = u16.buffer_length; 513 sc->flag = saved_flag;/* restore the saved flag. */ 514 return (ret); 515 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) { 516 count = (int)utf16nbytes(s, length); 517 count >>= 1; /* to be WCS length */ 518 /* Allocate memory for WCS. */ 519 if (NULL == archive_wstring_ensure(dest, 520 dest->length + count + 1)) 521 return (-1); 522 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count); 523 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) { 524 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 525 int b; 526 for (b = 0; b < count; b++) { 527 uint16_t val = archive_le16dec(u16+b); 528 archive_be16enc(u16+b, val); 529 } 530 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) { 531 uint16_t *u16 = (uint16_t *)(dest->s + dest->length); 532 int b; 533 for (b = 0; b < count; b++) { 534 uint16_t val = archive_be16dec(u16+b); 535 archive_le16enc(u16+b, val); 536 } 537 } 538 } else { 539 DWORD mbflag; 540 size_t buffsize; 541 542 if (sc == NULL) 543 mbflag = 0; 544 else if (sc->flag & SCONV_FROM_CHARSET) { 545 /* Do not trust the length which comes from 546 * an archive file. */ 547 length = mbsnbytes(s, length); 548 mbflag = 0; 549 } else 550 mbflag = MB_PRECOMPOSED; 551 552 buffsize = dest->length + length + 1; 553 do { 554 /* Allocate memory for WCS. */ 555 if (NULL == archive_wstring_ensure(dest, buffsize)) 556 return (-1); 557 /* Convert MBS to WCS. */ 558 count = MultiByteToWideChar(from_cp, 559 mbflag, s, (int)length, dest->s + dest->length, 560 (int)(dest->buffer_length >> 1) -1); 561 if (count == 0 && 562 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 563 /* Expand the WCS buffer. */ 564 buffsize = dest->buffer_length << 1; 565 continue; 566 } 567 if (count == 0 && length != 0) 568 ret = -1; 569 break; 570 } while (1); 571 } 572 dest->length += count; 573 dest->s[dest->length] = L'\0'; 574 return (ret); 575} 576 577#else 578 579/* 580 * Convert MBS to WCS. 581 * Note: returns -1 if conversion fails. 582 */ 583int 584archive_wstring_append_from_mbs(struct archive_wstring *dest, 585 const char *p, size_t len) 586{ 587 size_t r; 588 int ret_val = 0; 589 /* 590 * No single byte will be more than one wide character, 591 * so this length estimate will always be big enough. 592 */ 593 size_t wcs_length = len; 594 size_t mbs_length = len; 595 const char *mbs = p; 596 wchar_t *wcs; 597#if HAVE_MBRTOWC 598 mbstate_t shift_state; 599 600 memset(&shift_state, 0, sizeof(shift_state)); 601#endif 602 if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) 603 return (-1); 604 wcs = dest->s + dest->length; 605 /* 606 * We cannot use mbsrtowcs/mbstowcs here because those may convert 607 * extra MBS when strlen(p) > len and one wide character consists of 608 * multi bytes. 609 */ 610 while (*mbs && mbs_length > 0) { 611 if (wcs_length == 0) { 612 dest->length = wcs - dest->s; 613 dest->s[dest->length] = L'\0'; 614 wcs_length = mbs_length; 615 if (NULL == archive_wstring_ensure(dest, 616 dest->length + wcs_length + 1)) 617 return (-1); 618 wcs = dest->s + dest->length; 619 } 620#if HAVE_MBRTOWC 621 r = mbrtowc(wcs, mbs, wcs_length, &shift_state); 622#else 623 r = mbtowc(wcs, mbs, wcs_length); 624#endif 625 if (r == (size_t)-1 || r == (size_t)-2) { 626 ret_val = -1; 627 if (errno == EILSEQ) { 628 ++mbs; 629 --mbs_length; 630 continue; 631 } else 632 break; 633 } 634 if (r == 0 || r > mbs_length) 635 break; 636 wcs++; 637 wcs_length--; 638 mbs += r; 639 mbs_length -= r; 640 } 641 dest->length = wcs - dest->s; 642 dest->s[dest->length] = L'\0'; 643 return (ret_val); 644} 645 646#endif 647 648#if defined(_WIN32) && !defined(__CYGWIN__) 649 650/* 651 * WCS ==> MBS. 652 * Note: returns -1 if conversion fails. 653 * 654 * Win32 builds use WideCharToMultiByte from the Windows API. 655 * (Maybe Cygwin should too? WideCharToMultiByte will know a 656 * lot more about local character encodings than the wcrtomb() 657 * wrapper is going to know.) 658 */ 659int 660archive_string_append_from_wcs(struct archive_string *as, 661 const wchar_t *w, size_t len) 662{ 663 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL); 664} 665 666static int 667archive_string_append_from_wcs_in_codepage(struct archive_string *as, 668 const wchar_t *ws, size_t len, struct archive_string_conv *sc) 669{ 670 BOOL defchar_used, *dp; 671 int count, ret = 0; 672 UINT to_cp; 673 int wslen = (int)len; 674 675 if (sc != NULL) 676 to_cp = sc->to_cp; 677 else 678 to_cp = get_current_codepage(); 679 680 if (to_cp == CP_C_LOCALE) { 681 /* 682 * "C" locale special process. 683 */ 684 const wchar_t *wp = ws; 685 char *p; 686 687 if (NULL == archive_string_ensure(as, 688 as->length + wslen +1)) 689 return (-1); 690 p = as->s + as->length; 691 count = 0; 692 defchar_used = 0; 693 while (count < wslen && *wp) { 694 if (*wp > 255) { 695 *p++ = '?'; 696 wp++; 697 defchar_used = 1; 698 } else 699 *p++ = (char)*wp++; 700 count++; 701 } 702 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) { 703 uint16_t *u16; 704 705 if (NULL == 706 archive_string_ensure(as, as->length + len * 2 + 2)) 707 return (-1); 708 u16 = (uint16_t *)(as->s + as->length); 709 count = 0; 710 defchar_used = 0; 711 if (sc->flag & SCONV_TO_UTF16BE) { 712 while (count < (int)len && *ws) { 713 archive_be16enc(u16+count, *ws); 714 ws++; 715 count++; 716 } 717 } else { 718 while (count < (int)len && *ws) { 719 archive_le16enc(u16+count, *ws); 720 ws++; 721 count++; 722 } 723 } 724 count <<= 1; /* to be byte size */ 725 } else { 726 /* Make sure the MBS buffer has plenty to set. */ 727 if (NULL == 728 archive_string_ensure(as, as->length + len * 2 + 1)) 729 return (-1); 730 do { 731 defchar_used = 0; 732 if (to_cp == CP_UTF8 || sc == NULL) 733 dp = NULL; 734 else 735 dp = &defchar_used; 736 count = WideCharToMultiByte(to_cp, 0, ws, wslen, 737 as->s + as->length, (int)as->buffer_length-1, NULL, dp); 738 if (count == 0 && 739 GetLastError() == ERROR_INSUFFICIENT_BUFFER) { 740 /* Expand the MBS buffer and retry. */ 741 if (NULL == archive_string_ensure(as, 742 as->buffer_length + len)) 743 return (-1); 744 continue; 745 } 746 if (count == 0) 747 ret = -1; 748 break; 749 } while (1); 750 } 751 as->length += count; 752 as->s[as->length] = '\0'; 753 return (defchar_used?-1:ret); 754} 755 756#elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) 757 758/* 759 * Translates a wide character string into current locale character set 760 * and appends to the archive_string. Note: returns -1 if conversion 761 * fails. 762 */ 763int 764archive_string_append_from_wcs(struct archive_string *as, 765 const wchar_t *w, size_t len) 766{ 767 /* We cannot use the standard wcstombs() here because it 768 * cannot tell us how big the output buffer should be. So 769 * I've built a loop around wcrtomb() or wctomb() that 770 * converts a character at a time and resizes the string as 771 * needed. We prefer wcrtomb() when it's available because 772 * it's thread-safe. */ 773 int n, ret_val = 0; 774 char *p; 775 char *end; 776#if HAVE_WCRTOMB 777 mbstate_t shift_state; 778 779 memset(&shift_state, 0, sizeof(shift_state)); 780#else 781 /* Clear the shift state before starting. */ 782 wctomb(NULL, L'\0'); 783#endif 784 /* 785 * Allocate buffer for MBS. 786 * We need this allocation here since it is possible that 787 * as->s is still NULL. 788 */ 789 if (archive_string_ensure(as, as->length + len + 1) == NULL) 790 return (-1); 791 792 p = as->s + as->length; 793 end = as->s + as->buffer_length - MB_CUR_MAX -1; 794 while (*w != L'\0' && len > 0) { 795 if (p >= end) { 796 as->length = p - as->s; 797 as->s[as->length] = '\0'; 798 /* Re-allocate buffer for MBS. */ 799 if (archive_string_ensure(as, 800 as->length + len * 2 + 1) == NULL) 801 return (-1); 802 p = as->s + as->length; 803 end = as->s + as->buffer_length - MB_CUR_MAX -1; 804 } 805#if HAVE_WCRTOMB 806 n = wcrtomb(p, *w++, &shift_state); 807#else 808 n = wctomb(p, *w++); 809#endif 810 if (n == -1) { 811 if (errno == EILSEQ) { 812 /* Skip an illegal wide char. */ 813 *p++ = '?'; 814 ret_val = -1; 815 } else { 816 ret_val = -1; 817 break; 818 } 819 } else 820 p += n; 821 len--; 822 } 823 as->length = p - as->s; 824 as->s[as->length] = '\0'; 825 return (ret_val); 826} 827 828#else /* HAVE_WCTOMB || HAVE_WCRTOMB */ 829 830/* 831 * TODO: Test if __STDC_ISO_10646__ is defined. 832 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion 833 * one character at a time. If a non-Windows platform doesn't have 834 * either of these, fall back to the built-in UTF8 conversion. 835 */ 836int 837archive_string_append_from_wcs(struct archive_string *as, 838 const wchar_t *w, size_t len) 839{ 840 (void)as;/* UNUSED */ 841 (void)w;/* UNUSED */ 842 (void)len;/* UNUSED */ 843 errno = ENOSYS; 844 return (-1); 845} 846 847#endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ 848 849/* 850 * Find a string conversion object by a pair of 'from' charset name 851 * and 'to' charset name from an archive object. 852 * Return NULL if not found. 853 */ 854static struct archive_string_conv * 855find_sconv_object(struct archive *a, const char *fc, const char *tc) 856{ 857 struct archive_string_conv *sc; 858 859 if (a == NULL) 860 return (NULL); 861 862 for (sc = a->sconv; sc != NULL; sc = sc->next) { 863 if (strcmp(sc->from_charset, fc) == 0 && 864 strcmp(sc->to_charset, tc) == 0) 865 break; 866 } 867 return (sc); 868} 869 870/* 871 * Register a string object to an archive object. 872 */ 873static void 874add_sconv_object(struct archive *a, struct archive_string_conv *sc) 875{ 876 struct archive_string_conv **psc; 877 878 /* Add a new sconv to sconv list. */ 879 psc = &(a->sconv); 880 while (*psc != NULL) 881 psc = &((*psc)->next); 882 *psc = sc; 883} 884 885static void 886add_converter(struct archive_string_conv *sc, int (*converter) 887 (struct archive_string *, const void *, size_t, 888 struct archive_string_conv *)) 889{ 890 if (sc == NULL || sc->nconverter >= 2) 891 __archive_errx(1, "Programing error"); 892 sc->converter[sc->nconverter++] = converter; 893} 894 895static void 896setup_converter(struct archive_string_conv *sc) 897{ 898 899 /* Reset. */ 900 sc->nconverter = 0; 901 902 /* 903 * Perform special sequence for the incorrect UTF-8 filenames 904 * made by libarchive2.x. 905 */ 906 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { 907 add_converter(sc, strncat_from_utf8_libarchive2); 908 return; 909 } 910 911 /* 912 * Convert a string to UTF-16BE/LE. 913 */ 914 if (sc->flag & SCONV_TO_UTF16) { 915 /* 916 * If the current locale is UTF-8, we can translate 917 * a UTF-8 string into a UTF-16BE string. 918 */ 919 if (sc->flag & SCONV_FROM_UTF8) { 920 add_converter(sc, archive_string_append_unicode); 921 return; 922 } 923 924#if defined(_WIN32) && !defined(__CYGWIN__) 925 if (sc->flag & SCONV_WIN_CP) { 926 if (sc->flag & SCONV_TO_UTF16BE) 927 add_converter(sc, win_strncat_to_utf16be); 928 else 929 add_converter(sc, win_strncat_to_utf16le); 930 return; 931 } 932#endif 933 934#if defined(HAVE_ICONV) 935 if (sc->cd != (iconv_t)-1) { 936 add_converter(sc, iconv_strncat_in_locale); 937 return; 938 } 939#endif 940 941 if (sc->flag & SCONV_BEST_EFFORT) { 942 if (sc->flag & SCONV_TO_UTF16BE) 943 add_converter(sc, 944 best_effort_strncat_to_utf16be); 945 else 946 add_converter(sc, 947 best_effort_strncat_to_utf16le); 948 } else 949 /* Make sure we have no converter. */ 950 sc->nconverter = 0; 951 return; 952 } 953 954 /* 955 * Convert a string from UTF-16BE/LE. 956 */ 957 if (sc->flag & SCONV_FROM_UTF16) { 958 /* 959 * At least we should normalize a UTF-16BE string. 960 */ 961 if (sc->flag & SCONV_NORMALIZATION_D) 962 add_converter(sc,archive_string_normalize_D); 963 else if (sc->flag & SCONV_NORMALIZATION_C) 964 add_converter(sc, archive_string_normalize_C); 965 966 if (sc->flag & SCONV_TO_UTF8) { 967 /* 968 * If the current locale is UTF-8, we can translate 969 * a UTF-16BE/LE string into a UTF-8 string directly. 970 */ 971 if (!(sc->flag & 972 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 973 add_converter(sc, 974 archive_string_append_unicode); 975 return; 976 } 977 978#if defined(_WIN32) && !defined(__CYGWIN__) 979 if (sc->flag & SCONV_WIN_CP) { 980 if (sc->flag & SCONV_FROM_UTF16BE) 981 add_converter(sc, win_strncat_from_utf16be); 982 else 983 add_converter(sc, win_strncat_from_utf16le); 984 return; 985 } 986#endif 987 988#if defined(HAVE_ICONV) 989 if (sc->cd != (iconv_t)-1) { 990 add_converter(sc, iconv_strncat_in_locale); 991 return; 992 } 993#endif 994 995 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 996 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) 997 add_converter(sc, best_effort_strncat_from_utf16be); 998 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 999 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) 1000 add_converter(sc, best_effort_strncat_from_utf16le); 1001 else 1002 /* Make sure we have no converter. */ 1003 sc->nconverter = 0; 1004 return; 1005 } 1006 1007 if (sc->flag & SCONV_FROM_UTF8) { 1008 /* 1009 * At least we should normalize a UTF-8 string. 1010 */ 1011 if (sc->flag & SCONV_NORMALIZATION_D) 1012 add_converter(sc,archive_string_normalize_D); 1013 else if (sc->flag & SCONV_NORMALIZATION_C) 1014 add_converter(sc, archive_string_normalize_C); 1015 1016 /* 1017 * Copy UTF-8 string with a check of CESU-8. 1018 * Apparently, iconv does not check surrogate pairs in UTF-8 1019 * when both from-charset and to-charset are UTF-8, and then 1020 * we use our UTF-8 copy code. 1021 */ 1022 if (sc->flag & SCONV_TO_UTF8) { 1023 /* 1024 * If the current locale is UTF-8, we can translate 1025 * a UTF-16BE string into a UTF-8 string directly. 1026 */ 1027 if (!(sc->flag & 1028 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) 1029 add_converter(sc, strncat_from_utf8_to_utf8); 1030 return; 1031 } 1032 } 1033 1034#if defined(_WIN32) && !defined(__CYGWIN__) 1035 /* 1036 * On Windows we can use Windows API for a string conversion. 1037 */ 1038 if (sc->flag & SCONV_WIN_CP) { 1039 add_converter(sc, strncat_in_codepage); 1040 return; 1041 } 1042#endif 1043 1044#if HAVE_ICONV 1045 if (sc->cd != (iconv_t)-1) { 1046 add_converter(sc, iconv_strncat_in_locale); 1047 /* 1048 * iconv generally does not support UTF-8-MAC and so 1049 * we have to the output of iconv from NFC to NFD if 1050 * need. 1051 */ 1052 if ((sc->flag & SCONV_FROM_CHARSET) && 1053 (sc->flag & SCONV_TO_UTF8)) { 1054 if (sc->flag & SCONV_NORMALIZATION_D) 1055 add_converter(sc, archive_string_normalize_D); 1056 } 1057 return; 1058 } 1059#endif 1060 1061 /* 1062 * Try conversion in the best effort or no conversion. 1063 */ 1064 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) 1065 add_converter(sc, best_effort_strncat_in_locale); 1066 else 1067 /* Make sure we have no converter. */ 1068 sc->nconverter = 0; 1069} 1070 1071/* 1072 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE 1073 * and CP932 which are referenced in create_sconv_object(). 1074 */ 1075static const char * 1076canonical_charset_name(const char *charset) 1077{ 1078 char cs[16]; 1079 char *p; 1080 const char *s; 1081 1082 if (charset == NULL || charset[0] == '\0' 1083 || strlen(charset) > 15) 1084 return (charset); 1085 1086 /* Copy name to uppercase. */ 1087 p = cs; 1088 s = charset; 1089 while (*s) { 1090 char c = *s++; 1091 if (c >= 'a' && c <= 'z') 1092 c -= 'a' - 'A'; 1093 *p++ = c; 1094 } 1095 *p++ = '\0'; 1096 1097 if (strcmp(cs, "UTF-8") == 0 || 1098 strcmp(cs, "UTF8") == 0) 1099 return ("UTF-8"); 1100 if (strcmp(cs, "UTF-16BE") == 0 || 1101 strcmp(cs, "UTF16BE") == 0) 1102 return ("UTF-16BE"); 1103 if (strcmp(cs, "UTF-16LE") == 0 || 1104 strcmp(cs, "UTF16LE") == 0) 1105 return ("UTF-16LE"); 1106 if (strcmp(cs, "CP932") == 0) 1107 return ("CP932"); 1108 return (charset); 1109} 1110 1111/* 1112 * Create a string conversion object. 1113 */ 1114static struct archive_string_conv * 1115create_sconv_object(const char *fc, const char *tc, 1116 unsigned current_codepage, int flag) 1117{ 1118 struct archive_string_conv *sc; 1119 1120 sc = calloc(1, sizeof(*sc)); 1121 if (sc == NULL) 1122 return (NULL); 1123 sc->next = NULL; 1124 sc->from_charset = strdup(fc); 1125 if (sc->from_charset == NULL) { 1126 free(sc); 1127 return (NULL); 1128 } 1129 sc->to_charset = strdup(tc); 1130 if (sc->to_charset == NULL) { 1131 free(sc->from_charset); 1132 free(sc); 1133 return (NULL); 1134 } 1135 archive_string_init(&sc->utftmp); 1136 1137 if (flag & SCONV_TO_CHARSET) { 1138 /* 1139 * Convert characters from the current locale charset to 1140 * a specified charset. 1141 */ 1142 sc->from_cp = current_codepage; 1143 sc->to_cp = make_codepage_from_charset(tc); 1144#if defined(_WIN32) && !defined(__CYGWIN__) 1145 if (IsValidCodePage(sc->to_cp)) 1146 flag |= SCONV_WIN_CP; 1147#endif 1148 } else if (flag & SCONV_FROM_CHARSET) { 1149 /* 1150 * Convert characters from a specified charset to 1151 * the current locale charset. 1152 */ 1153 sc->to_cp = current_codepage; 1154 sc->from_cp = make_codepage_from_charset(fc); 1155#if defined(_WIN32) && !defined(__CYGWIN__) 1156 if (IsValidCodePage(sc->from_cp)) 1157 flag |= SCONV_WIN_CP; 1158#endif 1159 } 1160 1161 /* 1162 * Check if "from charset" and "to charset" are the same. 1163 */ 1164 if (strcmp(fc, tc) == 0 || 1165 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) 1166 sc->same = 1; 1167 else 1168 sc->same = 0; 1169 1170 /* 1171 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. 1172 */ 1173 if (strcmp(tc, "UTF-8") == 0) 1174 flag |= SCONV_TO_UTF8; 1175 else if (strcmp(tc, "UTF-16BE") == 0) 1176 flag |= SCONV_TO_UTF16BE; 1177 else if (strcmp(tc, "UTF-16LE") == 0) 1178 flag |= SCONV_TO_UTF16LE; 1179 if (strcmp(fc, "UTF-8") == 0) 1180 flag |= SCONV_FROM_UTF8; 1181 else if (strcmp(fc, "UTF-16BE") == 0) 1182 flag |= SCONV_FROM_UTF16BE; 1183 else if (strcmp(fc, "UTF-16LE") == 0) 1184 flag |= SCONV_FROM_UTF16LE; 1185#if defined(_WIN32) && !defined(__CYGWIN__) 1186 if (sc->to_cp == CP_UTF8) 1187 flag |= SCONV_TO_UTF8; 1188 else if (sc->to_cp == CP_UTF16BE) 1189 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; 1190 else if (sc->to_cp == CP_UTF16LE) 1191 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; 1192 if (sc->from_cp == CP_UTF8) 1193 flag |= SCONV_FROM_UTF8; 1194 else if (sc->from_cp == CP_UTF16BE) 1195 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; 1196 else if (sc->from_cp == CP_UTF16LE) 1197 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; 1198#endif 1199 1200 /* 1201 * Set a flag for Unicode NFD. Usually iconv cannot correctly 1202 * handle it. So we have to translate NFD characters to NFC ones 1203 * ourselves before iconv handles. Another reason is to prevent 1204 * that the same sight of two filenames, one is NFC and other 1205 * is NFD, would be in its directory. 1206 * On Mac OS X, although its filesystem layer automatically 1207 * convert filenames to NFD, it would be useful for filename 1208 * comparing to find out the same filenames that we normalize 1209 * that to be NFD ourselves. 1210 */ 1211 if ((flag & SCONV_FROM_CHARSET) && 1212 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { 1213#if defined(__APPLE__) 1214 if (flag & SCONV_TO_UTF8) 1215 flag |= SCONV_NORMALIZATION_D; 1216 else 1217#endif 1218 flag |= SCONV_NORMALIZATION_C; 1219 } 1220#if defined(__APPLE__) 1221 /* 1222 * In case writing an archive file, make sure that a filename 1223 * going to be passed to iconv is a Unicode NFC string since 1224 * a filename in HFS Plus filesystem is a Unicode NFD one and 1225 * iconv cannot handle it with "UTF-8" charset. It is simpler 1226 * than a use of "UTF-8-MAC" charset. 1227 */ 1228 if ((flag & SCONV_TO_CHARSET) && 1229 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1230 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1231 flag |= SCONV_NORMALIZATION_C; 1232 /* 1233 * In case reading an archive file. make sure that a filename 1234 * will be passed to users is a Unicode NFD string in order to 1235 * correctly compare the filename with other one which comes 1236 * from HFS Plus filesystem. 1237 */ 1238 if ((flag & SCONV_FROM_CHARSET) && 1239 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1240 (flag & SCONV_TO_UTF8)) 1241 flag |= SCONV_NORMALIZATION_D; 1242#endif 1243 1244#if defined(HAVE_ICONV) 1245 sc->cd_w = (iconv_t)-1; 1246 /* 1247 * Create an iconv object. 1248 */ 1249 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && 1250 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || 1251 (flag & SCONV_WIN_CP)) { 1252 /* This case we won't use iconv. */ 1253 sc->cd = (iconv_t)-1; 1254 } else { 1255 sc->cd = iconv_open(tc, fc); 1256 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { 1257 /* 1258 * Unfortunately, all of iconv implements do support 1259 * "CP932" character-set, so we should use "SJIS" 1260 * instead if iconv_open failed. 1261 */ 1262 if (strcmp(tc, "CP932") == 0) 1263 sc->cd = iconv_open("SJIS", fc); 1264 else if (strcmp(fc, "CP932") == 0) 1265 sc->cd = iconv_open(tc, "SJIS"); 1266 } 1267#if defined(_WIN32) && !defined(__CYGWIN__) 1268 /* 1269 * archive_mstring on Windows directly convert multi-bytes 1270 * into archive_wstring in order not to depend on locale 1271 * so that you can do a I18N programming. This will be 1272 * used only in archive_mstring_copy_mbs_len_l so far. 1273 */ 1274 if (flag & SCONV_FROM_CHARSET) { 1275 sc->cd_w = iconv_open("UTF-8", fc); 1276 if (sc->cd_w == (iconv_t)-1 && 1277 (sc->flag & SCONV_BEST_EFFORT)) { 1278 if (strcmp(fc, "CP932") == 0) 1279 sc->cd_w = iconv_open("UTF-8", "SJIS"); 1280 } 1281 } 1282#endif /* _WIN32 && !__CYGWIN__ */ 1283 } 1284#endif /* HAVE_ICONV */ 1285 1286 sc->flag = flag; 1287 1288 /* 1289 * Set up converters. 1290 */ 1291 setup_converter(sc); 1292 1293 return (sc); 1294} 1295 1296/* 1297 * Free a string conversion object. 1298 */ 1299static void 1300free_sconv_object(struct archive_string_conv *sc) 1301{ 1302 free(sc->from_charset); 1303 free(sc->to_charset); 1304 archive_string_free(&sc->utftmp); 1305#if HAVE_ICONV 1306 if (sc->cd != (iconv_t)-1) 1307 iconv_close(sc->cd); 1308 if (sc->cd_w != (iconv_t)-1) 1309 iconv_close(sc->cd_w); 1310#endif 1311 free(sc); 1312} 1313 1314#if defined(_WIN32) && !defined(__CYGWIN__) 1315static unsigned 1316my_atoi(const char *p) 1317{ 1318 unsigned cp; 1319 1320 cp = 0; 1321 while (*p) { 1322 if (*p >= '0' && *p <= '9') 1323 cp = cp * 10 + (*p - '0'); 1324 else 1325 return (-1); 1326 p++; 1327 } 1328 return (cp); 1329} 1330 1331/* 1332 * Translate Charset name (as used by iconv) into CodePage (as used by Windows) 1333 * Return -1 if failed. 1334 * 1335 * Note: This translation code may be insufficient. 1336 */ 1337static struct charset { 1338 const char *name; 1339 unsigned cp; 1340} charsets[] = { 1341 /* MUST BE SORTED! */ 1342 {"ASCII", 1252}, 1343 {"ASMO-708", 708}, 1344 {"BIG5", 950}, 1345 {"CHINESE", 936}, 1346 {"CP367", 1252}, 1347 {"CP819", 1252}, 1348 {"CP1025", 21025}, 1349 {"DOS-720", 720}, 1350 {"DOS-862", 862}, 1351 {"EUC-CN", 51936}, 1352 {"EUC-JP", 51932}, 1353 {"EUC-KR", 949}, 1354 {"EUCCN", 51936}, 1355 {"EUCJP", 51932}, 1356 {"EUCKR", 949}, 1357 {"GB18030", 54936}, 1358 {"GB2312", 936}, 1359 {"HEBREW", 1255}, 1360 {"HZ-GB-2312", 52936}, 1361 {"IBM273", 20273}, 1362 {"IBM277", 20277}, 1363 {"IBM278", 20278}, 1364 {"IBM280", 20280}, 1365 {"IBM284", 20284}, 1366 {"IBM285", 20285}, 1367 {"IBM290", 20290}, 1368 {"IBM297", 20297}, 1369 {"IBM367", 1252}, 1370 {"IBM420", 20420}, 1371 {"IBM423", 20423}, 1372 {"IBM424", 20424}, 1373 {"IBM819", 1252}, 1374 {"IBM871", 20871}, 1375 {"IBM880", 20880}, 1376 {"IBM905", 20905}, 1377 {"IBM924", 20924}, 1378 {"ISO-8859-1", 28591}, 1379 {"ISO-8859-13", 28603}, 1380 {"ISO-8859-15", 28605}, 1381 {"ISO-8859-2", 28592}, 1382 {"ISO-8859-3", 28593}, 1383 {"ISO-8859-4", 28594}, 1384 {"ISO-8859-5", 28595}, 1385 {"ISO-8859-6", 28596}, 1386 {"ISO-8859-7", 28597}, 1387 {"ISO-8859-8", 28598}, 1388 {"ISO-8859-9", 28599}, 1389 {"ISO8859-1", 28591}, 1390 {"ISO8859-13", 28603}, 1391 {"ISO8859-15", 28605}, 1392 {"ISO8859-2", 28592}, 1393 {"ISO8859-3", 28593}, 1394 {"ISO8859-4", 28594}, 1395 {"ISO8859-5", 28595}, 1396 {"ISO8859-6", 28596}, 1397 {"ISO8859-7", 28597}, 1398 {"ISO8859-8", 28598}, 1399 {"ISO8859-9", 28599}, 1400 {"JOHAB", 1361}, 1401 {"KOI8-R", 20866}, 1402 {"KOI8-U", 21866}, 1403 {"KS_C_5601-1987", 949}, 1404 {"LATIN1", 1252}, 1405 {"LATIN2", 28592}, 1406 {"MACINTOSH", 10000}, 1407 {"SHIFT-JIS", 932}, 1408 {"SHIFT_JIS", 932}, 1409 {"SJIS", 932}, 1410 {"US", 1252}, 1411 {"US-ASCII", 1252}, 1412 {"UTF-16", 1200}, 1413 {"UTF-16BE", 1201}, 1414 {"UTF-16LE", 1200}, 1415 {"UTF-8", CP_UTF8}, 1416 {"X-EUROPA", 29001}, 1417 {"X-MAC-ARABIC", 10004}, 1418 {"X-MAC-CE", 10029}, 1419 {"X-MAC-CHINESEIMP", 10008}, 1420 {"X-MAC-CHINESETRAD", 10002}, 1421 {"X-MAC-CROATIAN", 10082}, 1422 {"X-MAC-CYRILLIC", 10007}, 1423 {"X-MAC-GREEK", 10006}, 1424 {"X-MAC-HEBREW", 10005}, 1425 {"X-MAC-ICELANDIC", 10079}, 1426 {"X-MAC-JAPANESE", 10001}, 1427 {"X-MAC-KOREAN", 10003}, 1428 {"X-MAC-ROMANIAN", 10010}, 1429 {"X-MAC-THAI", 10021}, 1430 {"X-MAC-TURKISH", 10081}, 1431 {"X-MAC-UKRAINIAN", 10017}, 1432}; 1433static unsigned 1434make_codepage_from_charset(const char *charset) 1435{ 1436 char cs[16]; 1437 char *p; 1438 unsigned cp; 1439 int a, b; 1440 1441 if (charset == NULL || strlen(charset) > 15) 1442 return -1; 1443 1444 /* Copy name to uppercase. */ 1445 p = cs; 1446 while (*charset) { 1447 char c = *charset++; 1448 if (c >= 'a' && c <= 'z') 1449 c -= 'a' - 'A'; 1450 *p++ = c; 1451 } 1452 *p++ = '\0'; 1453 cp = -1; 1454 1455 /* Look it up in the table first, so that we can easily 1456 * override CP367, which we map to 1252 instead of 367. */ 1457 a = 0; 1458 b = sizeof(charsets)/sizeof(charsets[0]); 1459 while (b > a) { 1460 int c = (b + a) / 2; 1461 int r = strcmp(charsets[c].name, cs); 1462 if (r < 0) 1463 a = c + 1; 1464 else if (r > 0) 1465 b = c; 1466 else 1467 return charsets[c].cp; 1468 } 1469 1470 /* If it's not in the table, try to parse it. */ 1471 switch (*cs) { 1472 case 'C': 1473 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { 1474 cp = my_atoi(cs + 2); 1475 } else if (strcmp(cs, "CP_ACP") == 0) 1476 cp = get_current_codepage(); 1477 else if (strcmp(cs, "CP_OEMCP") == 0) 1478 cp = get_current_oemcp(); 1479 break; 1480 case 'I': 1481 if (cs[1] == 'B' && cs[2] == 'M' && 1482 cs[3] >= '0' && cs[3] <= '9') { 1483 cp = my_atoi(cs + 3); 1484 } 1485 break; 1486 case 'W': 1487 if (strncmp(cs, "WINDOWS-", 8) == 0) { 1488 cp = my_atoi(cs + 8); 1489 if (cp != 874 && (cp < 1250 || cp > 1258)) 1490 cp = -1;/* This may invalid code. */ 1491 } 1492 break; 1493 } 1494 return (cp); 1495} 1496 1497/* 1498 * Return ANSI Code Page of current locale set by setlocale(). 1499 */ 1500static unsigned 1501get_current_codepage(void) 1502{ 1503 char *locale, *p; 1504 unsigned cp; 1505 1506 locale = setlocale(LC_CTYPE, NULL); 1507 if (locale == NULL) 1508 return (GetACP()); 1509 if (locale[0] == 'C' && locale[1] == '\0') 1510 return (CP_C_LOCALE); 1511 p = strrchr(locale, '.'); 1512 if (p == NULL) 1513 return (GetACP()); 1514 cp = my_atoi(p+1); 1515 if (cp <= 0) 1516 return (GetACP()); 1517 return (cp); 1518} 1519 1520/* 1521 * Translation table between Locale Name and ACP/OEMCP. 1522 */ 1523static struct { 1524 unsigned acp; 1525 unsigned ocp; 1526 const char *locale; 1527} acp_ocp_map[] = { 1528 { 950, 950, "Chinese_Taiwan" }, 1529 { 936, 936, "Chinese_People's Republic of China" }, 1530 { 950, 950, "Chinese_Taiwan" }, 1531 { 1250, 852, "Czech_Czech Republic" }, 1532 { 1252, 850, "Danish_Denmark" }, 1533 { 1252, 850, "Dutch_Netherlands" }, 1534 { 1252, 850, "Dutch_Belgium" }, 1535 { 1252, 437, "English_United States" }, 1536 { 1252, 850, "English_Australia" }, 1537 { 1252, 850, "English_Canada" }, 1538 { 1252, 850, "English_New Zealand" }, 1539 { 1252, 850, "English_United Kingdom" }, 1540 { 1252, 437, "English_United States" }, 1541 { 1252, 850, "Finnish_Finland" }, 1542 { 1252, 850, "French_France" }, 1543 { 1252, 850, "French_Belgium" }, 1544 { 1252, 850, "French_Canada" }, 1545 { 1252, 850, "French_Switzerland" }, 1546 { 1252, 850, "German_Germany" }, 1547 { 1252, 850, "German_Austria" }, 1548 { 1252, 850, "German_Switzerland" }, 1549 { 1253, 737, "Greek_Greece" }, 1550 { 1250, 852, "Hungarian_Hungary" }, 1551 { 1252, 850, "Icelandic_Iceland" }, 1552 { 1252, 850, "Italian_Italy" }, 1553 { 1252, 850, "Italian_Switzerland" }, 1554 { 932, 932, "Japanese_Japan" }, 1555 { 949, 949, "Korean_Korea" }, 1556 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1557 { 1252, 850, "Norwegian (BokmOl)_Norway" }, 1558 { 1252, 850, "Norwegian-Nynorsk_Norway" }, 1559 { 1250, 852, "Polish_Poland" }, 1560 { 1252, 850, "Portuguese_Portugal" }, 1561 { 1252, 850, "Portuguese_Brazil" }, 1562 { 1251, 866, "Russian_Russia" }, 1563 { 1250, 852, "Slovak_Slovakia" }, 1564 { 1252, 850, "Spanish_Spain" }, 1565 { 1252, 850, "Spanish_Mexico" }, 1566 { 1252, 850, "Spanish_Spain" }, 1567 { 1252, 850, "Swedish_Sweden" }, 1568 { 1254, 857, "Turkish_Turkey" }, 1569 { 0, 0, NULL} 1570}; 1571 1572/* 1573 * Return OEM Code Page of current locale set by setlocale(). 1574 */ 1575static unsigned 1576get_current_oemcp(void) 1577{ 1578 int i; 1579 char *locale, *p; 1580 size_t len; 1581 1582 locale = setlocale(LC_CTYPE, NULL); 1583 if (locale == NULL) 1584 return (GetOEMCP()); 1585 if (locale[0] == 'C' && locale[1] == '\0') 1586 return (CP_C_LOCALE); 1587 1588 p = strrchr(locale, '.'); 1589 if (p == NULL) 1590 return (GetOEMCP()); 1591 len = p - locale; 1592 for (i = 0; acp_ocp_map[i].acp; i++) { 1593 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) 1594 return (acp_ocp_map[i].ocp); 1595 } 1596 return (GetOEMCP()); 1597} 1598#else 1599 1600/* 1601 * POSIX platform does not use CodePage. 1602 */ 1603 1604static unsigned 1605get_current_codepage(void) 1606{ 1607 return (-1);/* Unknown */ 1608} 1609static unsigned 1610make_codepage_from_charset(const char *charset) 1611{ 1612 (void)charset; /* UNUSED */ 1613 return (-1);/* Unknown */ 1614} 1615static unsigned 1616get_current_oemcp(void) 1617{ 1618 return (-1);/* Unknown */ 1619} 1620 1621#endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 1622 1623/* 1624 * Return a string conversion object. 1625 */ 1626static struct archive_string_conv * 1627get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) 1628{ 1629 struct archive_string_conv *sc; 1630 unsigned current_codepage; 1631 1632 /* Check if we have made the sconv object. */ 1633 sc = find_sconv_object(a, fc, tc); 1634 if (sc != NULL) 1635 return (sc); 1636 1637 if (a == NULL) 1638 current_codepage = get_current_codepage(); 1639 else 1640 current_codepage = a->current_codepage; 1641 1642 sc = create_sconv_object(canonical_charset_name(fc), 1643 canonical_charset_name(tc), current_codepage, flag); 1644 if (sc == NULL) { 1645 if (a != NULL) 1646 archive_set_error(a, ENOMEM, 1647 "Could not allocate memory for " 1648 "a string conversion object"); 1649 return (NULL); 1650 } 1651 1652 /* 1653 * If there is no converter for current string conversion object, 1654 * we cannot handle this conversion. 1655 */ 1656 if (sc->nconverter == 0) { 1657 if (a != NULL) { 1658#if HAVE_ICONV 1659 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1660 "iconv_open failed : Cannot handle ``%s''", 1661 (flag & SCONV_TO_CHARSET)?tc:fc); 1662#else 1663 archive_set_error(a, ARCHIVE_ERRNO_MISC, 1664 "A character-set conversion not fully supported " 1665 "on this platform"); 1666#endif 1667 } 1668 /* Failed; free a sconv object. */ 1669 free_sconv_object(sc); 1670 return (NULL); 1671 } 1672 1673 /* 1674 * Success! 1675 */ 1676 if (a != NULL) 1677 add_sconv_object(a, sc); 1678 return (sc); 1679} 1680 1681static const char * 1682get_current_charset(struct archive *a) 1683{ 1684 const char *cur_charset; 1685 1686 if (a == NULL) 1687 cur_charset = default_iconv_charset(""); 1688 else { 1689 cur_charset = default_iconv_charset(a->current_code); 1690 if (a->current_code == NULL) { 1691 a->current_code = strdup(cur_charset); 1692 a->current_codepage = get_current_codepage(); 1693 a->current_oemcp = get_current_oemcp(); 1694 } 1695 } 1696 return (cur_charset); 1697} 1698 1699/* 1700 * Make and Return a string conversion object. 1701 * Return NULL if the platform does not support the specified conversion 1702 * and best_effort is 0. 1703 * If best_effort is set, A string conversion object must be returned 1704 * unless memory allocation for the object fails, but the conversion 1705 * might fail when non-ASCII code is found. 1706 */ 1707struct archive_string_conv * 1708archive_string_conversion_to_charset(struct archive *a, const char *charset, 1709 int best_effort) 1710{ 1711 int flag = SCONV_TO_CHARSET; 1712 1713 if (best_effort) 1714 flag |= SCONV_BEST_EFFORT; 1715 return (get_sconv_object(a, get_current_charset(a), charset, flag)); 1716} 1717 1718struct archive_string_conv * 1719archive_string_conversion_from_charset(struct archive *a, const char *charset, 1720 int best_effort) 1721{ 1722 int flag = SCONV_FROM_CHARSET; 1723 1724 if (best_effort) 1725 flag |= SCONV_BEST_EFFORT; 1726 return (get_sconv_object(a, charset, get_current_charset(a), flag)); 1727} 1728 1729/* 1730 * archive_string_default_conversion_*_archive() are provided for Windows 1731 * platform because other archiver application use CP_OEMCP for 1732 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames 1733 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP 1734 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). 1735 * So we should make a string conversion between CP_ACP and CP_OEMCP 1736 * for compatibility. 1737 */ 1738#if defined(_WIN32) && !defined(__CYGWIN__) 1739struct archive_string_conv * 1740archive_string_default_conversion_for_read(struct archive *a) 1741{ 1742 const char *cur_charset = get_current_charset(a); 1743 char oemcp[16]; 1744 1745 /* NOTE: a check of cur_charset is unneeded but we need 1746 * that get_current_charset() has been surely called at 1747 * this time whatever C compiler optimized. */ 1748 if (cur_charset != NULL && 1749 (a->current_codepage == CP_C_LOCALE || 1750 a->current_codepage == a->current_oemcp)) 1751 return (NULL);/* no conversion. */ 1752 1753 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1754 /* Make sure a null termination must be set. */ 1755 oemcp[sizeof(oemcp)-1] = '\0'; 1756 return (get_sconv_object(a, oemcp, cur_charset, 1757 SCONV_FROM_CHARSET)); 1758} 1759 1760struct archive_string_conv * 1761archive_string_default_conversion_for_write(struct archive *a) 1762{ 1763 const char *cur_charset = get_current_charset(a); 1764 char oemcp[16]; 1765 1766 /* NOTE: a check of cur_charset is unneeded but we need 1767 * that get_current_charset() has been surely called at 1768 * this time whatever C compiler optimized. */ 1769 if (cur_charset != NULL && 1770 (a->current_codepage == CP_C_LOCALE || 1771 a->current_codepage == a->current_oemcp)) 1772 return (NULL);/* no conversion. */ 1773 1774 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); 1775 /* Make sure a null termination must be set. */ 1776 oemcp[sizeof(oemcp)-1] = '\0'; 1777 return (get_sconv_object(a, cur_charset, oemcp, 1778 SCONV_TO_CHARSET)); 1779} 1780#else 1781struct archive_string_conv * 1782archive_string_default_conversion_for_read(struct archive *a) 1783{ 1784 (void)a; /* UNUSED */ 1785 return (NULL); 1786} 1787 1788struct archive_string_conv * 1789archive_string_default_conversion_for_write(struct archive *a) 1790{ 1791 (void)a; /* UNUSED */ 1792 return (NULL); 1793} 1794#endif 1795 1796/* 1797 * Dispose of all character conversion objects in the archive object. 1798 */ 1799void 1800archive_string_conversion_free(struct archive *a) 1801{ 1802 struct archive_string_conv *sc; 1803 struct archive_string_conv *sc_next; 1804 1805 for (sc = a->sconv; sc != NULL; sc = sc_next) { 1806 sc_next = sc->next; 1807 free_sconv_object(sc); 1808 } 1809 a->sconv = NULL; 1810 free(a->current_code); 1811 a->current_code = NULL; 1812} 1813 1814/* 1815 * Return a conversion charset name. 1816 */ 1817const char * 1818archive_string_conversion_charset_name(struct archive_string_conv *sc) 1819{ 1820 if (sc->flag & SCONV_TO_CHARSET) 1821 return (sc->to_charset); 1822 else 1823 return (sc->from_charset); 1824} 1825 1826/* 1827 * Change the behavior of a string conversion. 1828 */ 1829void 1830archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) 1831{ 1832 switch (opt) { 1833 /* 1834 * A filename in UTF-8 was made with libarchive 2.x in a wrong 1835 * assumption that wchar_t was Unicode. 1836 * This option enables simulating the assumption in order to read 1837 * that filename correctly. 1838 */ 1839 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: 1840#if (defined(_WIN32) && !defined(__CYGWIN__)) \ 1841 || defined(__STDC_ISO_10646__) || defined(__APPLE__) 1842 /* 1843 * Nothing to do for it since wchar_t on these platforms 1844 * is really Unicode. 1845 */ 1846 (void)sc; /* UNUSED */ 1847#else 1848 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { 1849 sc->flag |= SCONV_UTF8_LIBARCHIVE_2; 1850 /* Set up string converters. */ 1851 setup_converter(sc); 1852 } 1853#endif 1854 break; 1855 case SCONV_SET_OPT_NORMALIZATION_C: 1856 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) { 1857 sc->flag |= SCONV_NORMALIZATION_C; 1858 sc->flag &= ~SCONV_NORMALIZATION_D; 1859 /* Set up string converters. */ 1860 setup_converter(sc); 1861 } 1862 break; 1863 case SCONV_SET_OPT_NORMALIZATION_D: 1864#if defined(HAVE_ICONV) 1865 /* 1866 * If iconv will take the string, do not change the 1867 * setting of the normalization. 1868 */ 1869 if (!(sc->flag & SCONV_WIN_CP) && 1870 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) && 1871 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8))) 1872 break; 1873#endif 1874 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) { 1875 sc->flag |= SCONV_NORMALIZATION_D; 1876 sc->flag &= ~SCONV_NORMALIZATION_C; 1877 /* Set up string converters. */ 1878 setup_converter(sc); 1879 } 1880 break; 1881 default: 1882 break; 1883 } 1884} 1885 1886/* 1887 * 1888 * Copy one archive_string to another in locale conversion. 1889 * 1890 * archive_strncat_l(); 1891 * archive_strncpy_l(); 1892 * 1893 */ 1894 1895static size_t 1896mbsnbytes(const void *_p, size_t n) 1897{ 1898 size_t s; 1899 const char *p, *pp; 1900 1901 if (_p == NULL) 1902 return (0); 1903 p = (const char *)_p; 1904 1905 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1906 s = 0; 1907 pp = p; 1908 while (s < n && *pp) { 1909 pp++; 1910 s++; 1911 } 1912 return (s); 1913} 1914 1915static size_t 1916utf16nbytes(const void *_p, size_t n) 1917{ 1918 size_t s; 1919 const char *p, *pp; 1920 1921 if (_p == NULL) 1922 return (0); 1923 p = (const char *)_p; 1924 1925 /* Like strlen(p), except won't examine positions beyond p[n]. */ 1926 s = 0; 1927 pp = p; 1928 n >>= 1; 1929 while (s < n && (pp[0] || pp[1])) { 1930 pp += 2; 1931 s++; 1932 } 1933 return (s<<1); 1934} 1935 1936int 1937archive_strncpy_l(struct archive_string *as, const void *_p, size_t n, 1938 struct archive_string_conv *sc) 1939{ 1940 as->length = 0; 1941 return (archive_strncat_l(as, _p, n, sc)); 1942} 1943 1944int 1945archive_strncat_l(struct archive_string *as, const void *_p, size_t n, 1946 struct archive_string_conv *sc) 1947{ 1948 const void *s; 1949 size_t length = 0; 1950 int i, r = 0, r2; 1951 1952 if (_p != NULL && n > 0) { 1953 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) 1954 length = utf16nbytes(_p, n); 1955 else 1956 length = mbsnbytes(_p, n); 1957 } 1958 1959 /* We must allocate memory even if there is no data for conversion 1960 * or copy. This simulates archive_string_append behavior. */ 1961 if (length == 0) { 1962 int tn = 1; 1963 if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) 1964 tn = 2; 1965 if (archive_string_ensure(as, as->length + tn) == NULL) 1966 return (-1); 1967 as->s[as->length] = 0; 1968 if (tn == 2) 1969 as->s[as->length+1] = 0; 1970 return (0); 1971 } 1972 1973 /* 1974 * If sc is NULL, we just make a copy. 1975 */ 1976 if (sc == NULL) { 1977 if (archive_string_append(as, _p, length) == NULL) 1978 return (-1);/* No memory */ 1979 return (0); 1980 } 1981 1982 s = _p; 1983 i = 0; 1984 if (sc->nconverter > 1) { 1985 sc->utftmp.length = 0; 1986 r2 = sc->converter[0](&(sc->utftmp), s, length, sc); 1987 if (r2 != 0 && errno == ENOMEM) 1988 return (r2); 1989 if (r > r2) 1990 r = r2; 1991 s = sc->utftmp.s; 1992 length = sc->utftmp.length; 1993 ++i; 1994 } 1995 r2 = sc->converter[i](as, s, length, sc); 1996 if (r > r2) 1997 r = r2; 1998 return (r); 1999} 2000 2001#if HAVE_ICONV 2002 2003/* 2004 * Return -1 if conversion fails. 2005 */ 2006static int 2007iconv_strncat_in_locale(struct archive_string *as, const void *_p, 2008 size_t length, struct archive_string_conv *sc) 2009{ 2010 ICONV_CONST char *itp; 2011 size_t remaining; 2012 iconv_t cd; 2013 char *outp; 2014 size_t avail, bs; 2015 int return_value = 0; /* success */ 2016 int to_size, from_size; 2017 2018 if (sc->flag & SCONV_TO_UTF16) 2019 to_size = 2; 2020 else 2021 to_size = 1; 2022 if (sc->flag & SCONV_FROM_UTF16) 2023 from_size = 2; 2024 else 2025 from_size = 1; 2026 2027 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) 2028 return (-1); 2029 2030 cd = sc->cd; 2031 itp = (char *)(uintptr_t)_p; 2032 remaining = length; 2033 outp = as->s + as->length; 2034 avail = as->buffer_length - as->length - to_size; 2035 while (remaining >= (size_t)from_size) { 2036 size_t result = iconv(cd, &itp, &remaining, &outp, &avail); 2037 2038 if (result != (size_t)-1) 2039 break; /* Conversion completed. */ 2040 2041 if (errno == EILSEQ || errno == EINVAL) { 2042 /* 2043 * If an output charset is UTF-8 or UTF-16BE/LE, 2044 * unknown character should be U+FFFD 2045 * (replacement character). 2046 */ 2047 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { 2048 size_t rbytes; 2049 if (sc->flag & SCONV_TO_UTF8) 2050 rbytes = sizeof(utf8_replacement_char); 2051 else 2052 rbytes = 2; 2053 2054 if (avail < rbytes) { 2055 as->length = outp - as->s; 2056 bs = as->buffer_length + 2057 (remaining * to_size) + rbytes; 2058 if (NULL == 2059 archive_string_ensure(as, bs)) 2060 return (-1); 2061 outp = as->s + as->length; 2062 avail = as->buffer_length 2063 - as->length - to_size; 2064 } 2065 if (sc->flag & SCONV_TO_UTF8) 2066 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char)); 2067 else if (sc->flag & SCONV_TO_UTF16BE) 2068 archive_be16enc(outp, UNICODE_R_CHAR); 2069 else 2070 archive_le16enc(outp, UNICODE_R_CHAR); 2071 outp += rbytes; 2072 avail -= rbytes; 2073 } else { 2074 /* Skip the illegal input bytes. */ 2075 *outp++ = '?'; 2076 avail--; 2077 } 2078 itp += from_size; 2079 remaining -= from_size; 2080 return_value = -1; /* failure */ 2081 } else { 2082 /* E2BIG no output buffer, 2083 * Increase an output buffer. */ 2084 as->length = outp - as->s; 2085 bs = as->buffer_length + remaining * 2; 2086 if (NULL == archive_string_ensure(as, bs)) 2087 return (-1); 2088 outp = as->s + as->length; 2089 avail = as->buffer_length - as->length - to_size; 2090 } 2091 } 2092 as->length = outp - as->s; 2093 as->s[as->length] = 0; 2094 if (to_size == 2) 2095 as->s[as->length+1] = 0; 2096 return (return_value); 2097} 2098 2099#endif /* HAVE_ICONV */ 2100 2101 2102#if defined(_WIN32) && !defined(__CYGWIN__) 2103 2104/* 2105 * Translate a string from a some CodePage to an another CodePage by 2106 * Windows APIs, and copy the result. Return -1 if conversion fails. 2107 */ 2108static int 2109strncat_in_codepage(struct archive_string *as, 2110 const void *_p, size_t length, struct archive_string_conv *sc) 2111{ 2112 const char *s = (const char *)_p; 2113 struct archive_wstring aws; 2114 size_t l; 2115 int r, saved_flag; 2116 2117 archive_string_init(&aws); 2118 saved_flag = sc->flag; 2119 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); 2120 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); 2121 sc->flag = saved_flag; 2122 if (r != 0) { 2123 archive_wstring_free(&aws); 2124 if (errno != ENOMEM) 2125 archive_string_append(as, s, length); 2126 return (-1); 2127 } 2128 2129 l = as->length; 2130 r = archive_string_append_from_wcs_in_codepage( 2131 as, aws.s, aws.length, sc); 2132 if (r != 0 && errno != ENOMEM && l == as->length) 2133 archive_string_append(as, s, length); 2134 archive_wstring_free(&aws); 2135 return (r); 2136} 2137 2138/* 2139 * Test whether MBS ==> WCS is okay. 2140 */ 2141static int 2142invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2143{ 2144 const char *p = (const char *)_p; 2145 unsigned codepage; 2146 DWORD mbflag = MB_ERR_INVALID_CHARS; 2147 2148 if (sc->flag & SCONV_FROM_CHARSET) 2149 codepage = sc->to_cp; 2150 else 2151 codepage = sc->from_cp; 2152 2153 if (codepage == CP_C_LOCALE) 2154 return (0); 2155 if (codepage != CP_UTF8) 2156 mbflag |= MB_PRECOMPOSED; 2157 2158 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0) 2159 return (-1); /* Invalid */ 2160 return (0); /* Okay */ 2161} 2162 2163#else 2164 2165/* 2166 * Test whether MBS ==> WCS is okay. 2167 */ 2168static int 2169invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) 2170{ 2171 const char *p = (const char *)_p; 2172 size_t r; 2173 2174#if HAVE_MBRTOWC 2175 mbstate_t shift_state; 2176 2177 memset(&shift_state, 0, sizeof(shift_state)); 2178#else 2179 /* Clear the shift state before starting. */ 2180 mbtowc(NULL, NULL, 0); 2181#endif 2182 while (n) { 2183 wchar_t wc; 2184 2185#if HAVE_MBRTOWC 2186 r = mbrtowc(&wc, p, n, &shift_state); 2187#else 2188 r = mbtowc(&wc, p, n); 2189#endif 2190 if (r == (size_t)-1 || r == (size_t)-2) 2191 return (-1);/* Invalid. */ 2192 if (r == 0) 2193 break; 2194 p += r; 2195 n -= r; 2196 } 2197 (void)sc; /* UNUSED */ 2198 return (0); /* All Okey. */ 2199} 2200 2201#endif /* defined(_WIN32) && !defined(__CYGWIN__) */ 2202 2203/* 2204 * Basically returns -1 because we cannot make a conversion of charset 2205 * without iconv but in some cases this would return 0. 2206 * Returns 0 if all copied characters are ASCII. 2207 * Returns 0 if both from-locale and to-locale are the same and those 2208 * can be WCS with no error. 2209 */ 2210static int 2211best_effort_strncat_in_locale(struct archive_string *as, const void *_p, 2212 size_t length, struct archive_string_conv *sc) 2213{ 2214 size_t remaining; 2215 const uint8_t *itp; 2216 int return_value = 0; /* success */ 2217 2218 /* 2219 * If both from-locale and to-locale is the same, this makes a copy. 2220 * And then this checks all copied MBS can be WCS if so returns 0. 2221 */ 2222 if (sc->same) { 2223 if (archive_string_append(as, _p, length) == NULL) 2224 return (-1);/* No memory */ 2225 return (invalid_mbs(_p, length, sc)); 2226 } 2227 2228 /* 2229 * If a character is ASCII, this just copies it. If not, this 2230 * assigns '?' character instead but in UTF-8 locale this assigns 2231 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, 2232 * a Replacement Character in Unicode. 2233 */ 2234 2235 remaining = length; 2236 itp = (const uint8_t *)_p; 2237 while (*itp && remaining > 0) { 2238 if (*itp > 127) { 2239 // Non-ASCII: Substitute with suitable replacement 2240 if (sc->flag & SCONV_TO_UTF8) { 2241 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) { 2242 __archive_errx(1, "Out of memory"); 2243 } 2244 } else { 2245 archive_strappend_char(as, '?'); 2246 } 2247 return_value = -1; 2248 } else { 2249 archive_strappend_char(as, *itp); 2250 } 2251 ++itp; 2252 } 2253 return (return_value); 2254} 2255 2256 2257/* 2258 * Unicode conversion functions. 2259 * - UTF-8 <===> UTF-8 in removing surrogate pairs. 2260 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. 2261 * - UTF-8 made by libarchive 2.x ===> UTF-8. 2262 * - UTF-16BE <===> UTF-8. 2263 * 2264 */ 2265 2266/* 2267 * Utility to convert a single UTF-8 sequence. 2268 * 2269 * Usually return used bytes, return used byte in negative value when 2270 * a unicode character is replaced with U+FFFD. 2271 * See also http://unicode.org/review/pr-121.html Public Review Issue #121 2272 * Recommended Practice for Replacement Characters. 2273 */ 2274static int 2275_utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2276{ 2277 static const char utf8_count[256] = { 2278 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 2279 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 2280 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 2281 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 2282 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 2283 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 2284 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 2285 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 2286 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 2287 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 2288 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 2289 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 2290 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2291 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 2292 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 2293 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ 2294 }; 2295 int ch, i; 2296 int cnt; 2297 uint32_t wc; 2298 2299 /* Sanity check. */ 2300 if (n == 0) 2301 return (0); 2302 /* 2303 * Decode 1-4 bytes depending on the value of the first byte. 2304 */ 2305 ch = (unsigned char)*s; 2306 if (ch == 0) 2307 return (0); /* Standard: return 0 for end-of-string. */ 2308 cnt = utf8_count[ch]; 2309 2310 /* Invalid sequence or there are not plenty bytes. */ 2311 if ((int)n < cnt) { 2312 cnt = (int)n; 2313 for (i = 1; i < cnt; i++) { 2314 if ((s[i] & 0xc0) != 0x80) { 2315 cnt = i; 2316 break; 2317 } 2318 } 2319 goto invalid_sequence; 2320 } 2321 2322 /* Make a Unicode code point from a single UTF-8 sequence. */ 2323 switch (cnt) { 2324 case 1: /* 1 byte sequence. */ 2325 *pwc = ch & 0x7f; 2326 return (cnt); 2327 case 2: /* 2 bytes sequence. */ 2328 if ((s[1] & 0xc0) != 0x80) { 2329 cnt = 1; 2330 goto invalid_sequence; 2331 } 2332 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); 2333 return (cnt); 2334 case 3: /* 3 bytes sequence. */ 2335 if ((s[1] & 0xc0) != 0x80) { 2336 cnt = 1; 2337 goto invalid_sequence; 2338 } 2339 if ((s[2] & 0xc0) != 0x80) { 2340 cnt = 2; 2341 goto invalid_sequence; 2342 } 2343 wc = ((ch & 0x0f) << 12) 2344 | ((s[1] & 0x3f) << 6) 2345 | (s[2] & 0x3f); 2346 if (wc < 0x800) 2347 goto invalid_sequence;/* Overlong sequence. */ 2348 break; 2349 case 4: /* 4 bytes sequence. */ 2350 if ((s[1] & 0xc0) != 0x80) { 2351 cnt = 1; 2352 goto invalid_sequence; 2353 } 2354 if ((s[2] & 0xc0) != 0x80) { 2355 cnt = 2; 2356 goto invalid_sequence; 2357 } 2358 if ((s[3] & 0xc0) != 0x80) { 2359 cnt = 3; 2360 goto invalid_sequence; 2361 } 2362 wc = ((ch & 0x07) << 18) 2363 | ((s[1] & 0x3f) << 12) 2364 | ((s[2] & 0x3f) << 6) 2365 | (s[3] & 0x3f); 2366 if (wc < 0x10000) 2367 goto invalid_sequence;/* Overlong sequence. */ 2368 break; 2369 default: /* Others are all invalid sequence. */ 2370 if (ch == 0xc0 || ch == 0xc1) 2371 cnt = 2; 2372 else if (ch >= 0xf5 && ch <= 0xf7) 2373 cnt = 4; 2374 else if (ch >= 0xf8 && ch <= 0xfb) 2375 cnt = 5; 2376 else if (ch == 0xfc || ch == 0xfd) 2377 cnt = 6; 2378 else 2379 cnt = 1; 2380 if ((int)n < cnt) 2381 cnt = (int)n; 2382 for (i = 1; i < cnt; i++) { 2383 if ((s[i] & 0xc0) != 0x80) { 2384 cnt = i; 2385 break; 2386 } 2387 } 2388 goto invalid_sequence; 2389 } 2390 2391 /* The code point larger than 0x10FFFF is not legal 2392 * Unicode values. */ 2393 if (wc > UNICODE_MAX) 2394 goto invalid_sequence; 2395 /* Correctly gets a Unicode, returns used bytes. */ 2396 *pwc = wc; 2397 return (cnt); 2398invalid_sequence: 2399 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2400 return (cnt * -1); 2401} 2402 2403static int 2404utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2405{ 2406 int cnt; 2407 2408 cnt = _utf8_to_unicode(pwc, s, n); 2409 /* Any of Surrogate pair is not legal Unicode values. */ 2410 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) 2411 return (-3); 2412 return (cnt); 2413} 2414 2415static inline uint32_t 2416combine_surrogate_pair(uint32_t uc, uint32_t uc2) 2417{ 2418 uc -= 0xD800; 2419 uc *= 0x400; 2420 uc += uc2 - 0xDC00; 2421 uc += 0x10000; 2422 return (uc); 2423} 2424 2425/* 2426 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in 2427 * removing surrogate pairs. 2428 * 2429 * CESU-8: The Compatibility Encoding Scheme for UTF-16. 2430 * 2431 * Usually return used bytes, return used byte in negative value when 2432 * a unicode character is replaced with U+FFFD. 2433 */ 2434static int 2435cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) 2436{ 2437 uint32_t wc = 0; 2438 int cnt; 2439 2440 cnt = _utf8_to_unicode(&wc, s, n); 2441 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { 2442 uint32_t wc2 = 0; 2443 if (n - 3 < 3) { 2444 /* Invalid byte sequence. */ 2445 goto invalid_sequence; 2446 } 2447 cnt = _utf8_to_unicode(&wc2, s+3, n-3); 2448 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { 2449 /* Invalid byte sequence. */ 2450 goto invalid_sequence; 2451 } 2452 wc = combine_surrogate_pair(wc, wc2); 2453 cnt = 6; 2454 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { 2455 /* Invalid byte sequence. */ 2456 goto invalid_sequence; 2457 } 2458 *pwc = wc; 2459 return (cnt); 2460invalid_sequence: 2461 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ 2462 if (cnt > 0) 2463 cnt *= -1; 2464 return (cnt); 2465} 2466 2467/* 2468 * Convert a Unicode code point to a single UTF-8 sequence. 2469 * 2470 * NOTE:This function does not check if the Unicode is legal or not. 2471 * Please you definitely check it before calling this. 2472 */ 2473static size_t 2474unicode_to_utf8(char *p, size_t remaining, uint32_t uc) 2475{ 2476 char *_p = p; 2477 2478 /* Invalid Unicode char maps to Replacement character */ 2479 if (uc > UNICODE_MAX) 2480 uc = UNICODE_R_CHAR; 2481 /* Translate code point to UTF8 */ 2482 if (uc <= 0x7f) { 2483 if (remaining == 0) 2484 return (0); 2485 *p++ = (char)uc; 2486 } else if (uc <= 0x7ff) { 2487 if (remaining < 2) 2488 return (0); 2489 *p++ = 0xc0 | ((uc >> 6) & 0x1f); 2490 *p++ = 0x80 | (uc & 0x3f); 2491 } else if (uc <= 0xffff) { 2492 if (remaining < 3) 2493 return (0); 2494 *p++ = 0xe0 | ((uc >> 12) & 0x0f); 2495 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2496 *p++ = 0x80 | (uc & 0x3f); 2497 } else { 2498 if (remaining < 4) 2499 return (0); 2500 *p++ = 0xf0 | ((uc >> 18) & 0x07); 2501 *p++ = 0x80 | ((uc >> 12) & 0x3f); 2502 *p++ = 0x80 | ((uc >> 6) & 0x3f); 2503 *p++ = 0x80 | (uc & 0x3f); 2504 } 2505 return (p - _p); 2506} 2507 2508static int 2509utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) 2510{ 2511 return (utf16_to_unicode(pwc, s, n, 1)); 2512} 2513 2514static int 2515utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) 2516{ 2517 return (utf16_to_unicode(pwc, s, n, 0)); 2518} 2519 2520static int 2521utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) 2522{ 2523 const char *utf16 = s; 2524 unsigned uc; 2525 2526 if (n == 0) 2527 return (0); 2528 if (n == 1) { 2529 /* set the Replacement Character instead. */ 2530 *pwc = UNICODE_R_CHAR; 2531 return (-1); 2532 } 2533 2534 if (be) 2535 uc = archive_be16dec(utf16); 2536 else 2537 uc = archive_le16dec(utf16); 2538 utf16 += 2; 2539 2540 /* If this is a surrogate pair, assemble the full code point.*/ 2541 if (IS_HIGH_SURROGATE_LA(uc)) { 2542 unsigned uc2; 2543 2544 if (n >= 4) { 2545 if (be) 2546 uc2 = archive_be16dec(utf16); 2547 else 2548 uc2 = archive_le16dec(utf16); 2549 } else 2550 uc2 = 0; 2551 if (IS_LOW_SURROGATE_LA(uc2)) { 2552 uc = combine_surrogate_pair(uc, uc2); 2553 utf16 += 2; 2554 } else { 2555 /* Undescribed code point should be U+FFFD 2556 * (replacement character). */ 2557 *pwc = UNICODE_R_CHAR; 2558 return (-2); 2559 } 2560 } 2561 2562 /* 2563 * Surrogate pair values(0xd800 through 0xdfff) are only 2564 * used by UTF-16, so, after above calculation, the code 2565 * must not be surrogate values, and Unicode has no codes 2566 * larger than 0x10ffff. Thus, those are not legal Unicode 2567 * values. 2568 */ 2569 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { 2570 /* Undescribed code point should be U+FFFD 2571 * (replacement character). */ 2572 *pwc = UNICODE_R_CHAR; 2573 return (((int)(utf16 - s)) * -1); 2574 } 2575 *pwc = uc; 2576 return ((int)(utf16 - s)); 2577} 2578 2579static size_t 2580unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) 2581{ 2582 char *utf16 = p; 2583 2584 if (uc > 0xffff) { 2585 /* We have a code point that won't fit into a 2586 * wchar_t; convert it to a surrogate pair. */ 2587 if (remaining < 4) 2588 return (0); 2589 uc -= 0x10000; 2590 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2591 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2592 return (4); 2593 } else { 2594 if (remaining < 2) 2595 return (0); 2596 archive_be16enc(utf16, uc); 2597 return (2); 2598 } 2599} 2600 2601static size_t 2602unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) 2603{ 2604 char *utf16 = p; 2605 2606 if (uc > 0xffff) { 2607 /* We have a code point that won't fit into a 2608 * wchar_t; convert it to a surrogate pair. */ 2609 if (remaining < 4) 2610 return (0); 2611 uc -= 0x10000; 2612 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); 2613 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); 2614 return (4); 2615 } else { 2616 if (remaining < 2) 2617 return (0); 2618 archive_le16enc(utf16, uc); 2619 return (2); 2620 } 2621} 2622 2623/* 2624 * Copy UTF-8 string in checking surrogate pair. 2625 * If any surrogate pair are found, it would be canonicalized. 2626 */ 2627static int 2628strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, 2629 size_t len, struct archive_string_conv *sc) 2630{ 2631 const char *s; 2632 char *p, *endp; 2633 int n, ret = 0; 2634 2635 (void)sc; /* UNUSED */ 2636 2637 if (archive_string_ensure(as, as->length + len + 1) == NULL) 2638 return (-1); 2639 2640 s = (const char *)_p; 2641 p = as->s + as->length; 2642 endp = as->s + as->buffer_length -1; 2643 do { 2644 uint32_t uc; 2645 const char *ss = s; 2646 size_t w; 2647 2648 /* 2649 * Forward byte sequence until a conversion of that is needed. 2650 */ 2651 while ((n = utf8_to_unicode(&uc, s, len)) > 0) { 2652 s += n; 2653 len -= n; 2654 } 2655 if (ss < s) { 2656 if (p + (s - ss) > endp) { 2657 as->length = p - as->s; 2658 if (archive_string_ensure(as, 2659 as->buffer_length + len + 1) == NULL) 2660 return (-1); 2661 p = as->s + as->length; 2662 endp = as->s + as->buffer_length -1; 2663 } 2664 2665 memcpy(p, ss, s - ss); 2666 p += s - ss; 2667 } 2668 2669 /* 2670 * If n is negative, current byte sequence needs a replacement. 2671 */ 2672 if (n < 0) { 2673 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { 2674 /* Current byte sequence may be CESU-8. */ 2675 n = cesu8_to_unicode(&uc, s, len); 2676 } 2677 if (n < 0) { 2678 ret = -1; 2679 n *= -1;/* Use a replaced unicode character. */ 2680 } 2681 2682 /* Rebuild UTF-8 byte sequence. */ 2683 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { 2684 as->length = p - as->s; 2685 if (archive_string_ensure(as, 2686 as->buffer_length + len + 1) == NULL) 2687 return (-1); 2688 p = as->s + as->length; 2689 endp = as->s + as->buffer_length -1; 2690 } 2691 p += w; 2692 s += n; 2693 len -= n; 2694 } 2695 } while (n > 0); 2696 as->length = p - as->s; 2697 as->s[as->length] = '\0'; 2698 return (ret); 2699} 2700 2701static int 2702archive_string_append_unicode(struct archive_string *as, const void *_p, 2703 size_t len, struct archive_string_conv *sc) 2704{ 2705 const char *s; 2706 char *p, *endp; 2707 uint32_t uc; 2708 size_t w; 2709 int n, ret = 0, ts, tm; 2710 int (*parse)(uint32_t *, const char *, size_t); 2711 size_t (*unparse)(char *, size_t, uint32_t); 2712 2713 if (sc->flag & SCONV_TO_UTF16BE) { 2714 unparse = unicode_to_utf16be; 2715 ts = 2; 2716 } else if (sc->flag & SCONV_TO_UTF16LE) { 2717 unparse = unicode_to_utf16le; 2718 ts = 2; 2719 } else if (sc->flag & SCONV_TO_UTF8) { 2720 unparse = unicode_to_utf8; 2721 ts = 1; 2722 } else { 2723 /* 2724 * This case is going to be converted to another 2725 * character-set through iconv. 2726 */ 2727 if (sc->flag & SCONV_FROM_UTF16BE) { 2728 unparse = unicode_to_utf16be; 2729 ts = 2; 2730 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2731 unparse = unicode_to_utf16le; 2732 ts = 2; 2733 } else { 2734 unparse = unicode_to_utf8; 2735 ts = 1; 2736 } 2737 } 2738 2739 if (sc->flag & SCONV_FROM_UTF16BE) { 2740 parse = utf16be_to_unicode; 2741 tm = 1; 2742 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2743 parse = utf16le_to_unicode; 2744 tm = 1; 2745 } else { 2746 parse = cesu8_to_unicode; 2747 tm = ts; 2748 } 2749 2750 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2751 return (-1); 2752 2753 s = (const char *)_p; 2754 p = as->s + as->length; 2755 endp = as->s + as->buffer_length - ts; 2756 while ((n = parse(&uc, s, len)) != 0) { 2757 if (n < 0) { 2758 /* Use a replaced unicode character. */ 2759 n *= -1; 2760 ret = -1; 2761 } 2762 s += n; 2763 len -= n; 2764 while ((w = unparse(p, endp - p, uc)) == 0) { 2765 /* There is not enough output buffer so 2766 * we have to expand it. */ 2767 as->length = p - as->s; 2768 if (archive_string_ensure(as, 2769 as->buffer_length + len * tm + ts) == NULL) 2770 return (-1); 2771 p = as->s + as->length; 2772 endp = as->s + as->buffer_length - ts; 2773 } 2774 p += w; 2775 } 2776 as->length = p - as->s; 2777 as->s[as->length] = '\0'; 2778 if (ts == 2) 2779 as->s[as->length+1] = '\0'; 2780 return (ret); 2781} 2782 2783/* 2784 * Following Constants for Hangul compositions this information comes from 2785 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ 2786 */ 2787#define HC_SBASE 0xAC00 2788#define HC_LBASE 0x1100 2789#define HC_VBASE 0x1161 2790#define HC_TBASE 0x11A7 2791#define HC_LCOUNT 19 2792#define HC_VCOUNT 21 2793#define HC_TCOUNT 28 2794#define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) 2795#define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) 2796 2797static uint32_t 2798get_nfc(uint32_t uc, uint32_t uc2) 2799{ 2800 int t, b; 2801 2802 t = 0; 2803 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; 2804 while (b >= t) { 2805 int m = (t + b) / 2; 2806 if (u_composition_table[m].cp1 < uc) 2807 t = m + 1; 2808 else if (u_composition_table[m].cp1 > uc) 2809 b = m - 1; 2810 else if (u_composition_table[m].cp2 < uc2) 2811 t = m + 1; 2812 else if (u_composition_table[m].cp2 > uc2) 2813 b = m - 1; 2814 else 2815 return (u_composition_table[m].nfc); 2816 } 2817 return (0); 2818} 2819 2820#define FDC_MAX 10 /* The maximum number of Following Decomposable 2821 * Characters. */ 2822 2823/* 2824 * Update first code point. 2825 */ 2826#define UPDATE_UC(new_uc) do { \ 2827 uc = new_uc; \ 2828 ucptr = NULL; \ 2829} while (0) 2830 2831/* 2832 * Replace first code point with second code point. 2833 */ 2834#define REPLACE_UC_WITH_UC2() do { \ 2835 uc = uc2; \ 2836 ucptr = uc2ptr; \ 2837 n = n2; \ 2838} while (0) 2839 2840#define EXPAND_BUFFER() do { \ 2841 as->length = p - as->s; \ 2842 if (archive_string_ensure(as, \ 2843 as->buffer_length + len * tm + ts) == NULL)\ 2844 return (-1); \ 2845 p = as->s + as->length; \ 2846 endp = as->s + as->buffer_length - ts; \ 2847} while (0) 2848 2849#define UNPARSE(p, endp, uc) do { \ 2850 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ 2851 EXPAND_BUFFER(); \ 2852 } \ 2853 p += w; \ 2854} while (0) 2855 2856/* 2857 * Write first code point. 2858 * If the code point has not be changed from its original code, 2859 * this just copies it from its original buffer pointer. 2860 * If not, this converts it to UTF-8 byte sequence and copies it. 2861 */ 2862#define WRITE_UC() do { \ 2863 if (ucptr) { \ 2864 if (p + n > endp) \ 2865 EXPAND_BUFFER(); \ 2866 switch (n) { \ 2867 case 4: \ 2868 *p++ = *ucptr++; \ 2869 /* FALL THROUGH */ \ 2870 case 3: \ 2871 *p++ = *ucptr++; \ 2872 /* FALL THROUGH */ \ 2873 case 2: \ 2874 *p++ = *ucptr++; \ 2875 /* FALL THROUGH */ \ 2876 case 1: \ 2877 *p++ = *ucptr; \ 2878 break; \ 2879 } \ 2880 ucptr = NULL; \ 2881 } else { \ 2882 UNPARSE(p, endp, uc); \ 2883 } \ 2884} while (0) 2885 2886/* 2887 * Collect following decomposable code points. 2888 */ 2889#define COLLECT_CPS(start) do { \ 2890 int _i; \ 2891 for (_i = start; _i < FDC_MAX ; _i++) { \ 2892 nx = parse(&ucx[_i], s, len); \ 2893 if (nx <= 0) \ 2894 break; \ 2895 cx = CCC(ucx[_i]); \ 2896 if (cl >= cx && cl != 228 && cx != 228)\ 2897 break; \ 2898 s += nx; \ 2899 len -= nx; \ 2900 cl = cx; \ 2901 ccx[_i] = cx; \ 2902 } \ 2903 if (_i >= FDC_MAX) { \ 2904 ret = -1; \ 2905 ucx_size = FDC_MAX; \ 2906 } else \ 2907 ucx_size = _i; \ 2908} while (0) 2909 2910/* 2911 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. 2912 * 2913 * TODO: Convert composition exclusions, which are never converted 2914 * from NFC,NFD,NFKC and NFKD, to Form C. 2915 */ 2916static int 2917archive_string_normalize_C(struct archive_string *as, const void *_p, 2918 size_t len, struct archive_string_conv *sc) 2919{ 2920 const char *s = (const char *)_p; 2921 char *p, *endp; 2922 uint32_t uc, uc2; 2923 size_t w; 2924 int always_replace, n, n2, ret = 0, spair, ts, tm; 2925 int (*parse)(uint32_t *, const char *, size_t); 2926 size_t (*unparse)(char *, size_t, uint32_t); 2927 2928 always_replace = 1; 2929 ts = 1;/* text size. */ 2930 if (sc->flag & SCONV_TO_UTF16BE) { 2931 unparse = unicode_to_utf16be; 2932 ts = 2; 2933 if (sc->flag & SCONV_FROM_UTF16BE) 2934 always_replace = 0; 2935 } else if (sc->flag & SCONV_TO_UTF16LE) { 2936 unparse = unicode_to_utf16le; 2937 ts = 2; 2938 if (sc->flag & SCONV_FROM_UTF16LE) 2939 always_replace = 0; 2940 } else if (sc->flag & SCONV_TO_UTF8) { 2941 unparse = unicode_to_utf8; 2942 if (sc->flag & SCONV_FROM_UTF8) 2943 always_replace = 0; 2944 } else { 2945 /* 2946 * This case is going to be converted to another 2947 * character-set through iconv. 2948 */ 2949 always_replace = 0; 2950 if (sc->flag & SCONV_FROM_UTF16BE) { 2951 unparse = unicode_to_utf16be; 2952 ts = 2; 2953 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2954 unparse = unicode_to_utf16le; 2955 ts = 2; 2956 } else { 2957 unparse = unicode_to_utf8; 2958 } 2959 } 2960 2961 if (sc->flag & SCONV_FROM_UTF16BE) { 2962 parse = utf16be_to_unicode; 2963 tm = 1; 2964 spair = 4;/* surrogate pair size in UTF-16. */ 2965 } else if (sc->flag & SCONV_FROM_UTF16LE) { 2966 parse = utf16le_to_unicode; 2967 tm = 1; 2968 spair = 4;/* surrogate pair size in UTF-16. */ 2969 } else { 2970 parse = cesu8_to_unicode; 2971 tm = ts; 2972 spair = 6;/* surrogate pair size in UTF-8. */ 2973 } 2974 2975 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 2976 return (-1); 2977 2978 p = as->s + as->length; 2979 endp = as->s + as->buffer_length - ts; 2980 while ((n = parse(&uc, s, len)) != 0) { 2981 const char *ucptr, *uc2ptr; 2982 2983 if (n < 0) { 2984 /* Use a replaced unicode character. */ 2985 UNPARSE(p, endp, uc); 2986 s += n*-1; 2987 len -= n*-1; 2988 ret = -1; 2989 continue; 2990 } else if (n == spair || always_replace) 2991 /* uc is converted from a surrogate pair. 2992 * this should be treated as a changed code. */ 2993 ucptr = NULL; 2994 else 2995 ucptr = s; 2996 s += n; 2997 len -= n; 2998 2999 /* Read second code point. */ 3000 while ((n2 = parse(&uc2, s, len)) > 0) { 3001 uint32_t ucx[FDC_MAX]; 3002 int ccx[FDC_MAX]; 3003 int cl, cx, i, nx, ucx_size; 3004 int LIndex,SIndex; 3005 uint32_t nfc; 3006 3007 if (n2 == spair || always_replace) 3008 /* uc2 is converted from a surrogate pair. 3009 * this should be treated as a changed code. */ 3010 uc2ptr = NULL; 3011 else 3012 uc2ptr = s; 3013 s += n2; 3014 len -= n2; 3015 3016 /* 3017 * If current second code point is out of decomposable 3018 * code points, finding compositions is unneeded. 3019 */ 3020 if (!IS_DECOMPOSABLE_BLOCK(uc2)) { 3021 WRITE_UC(); 3022 REPLACE_UC_WITH_UC2(); 3023 continue; 3024 } 3025 3026 /* 3027 * Try to combine current code points. 3028 */ 3029 /* 3030 * We have to combine Hangul characters according to 3031 * http://uniicode.org/reports/tr15/#Hangul 3032 */ 3033 if (0 <= (LIndex = uc - HC_LBASE) && 3034 LIndex < HC_LCOUNT) { 3035 /* 3036 * Hangul Composition. 3037 * 1. Two current code points are L and V. 3038 */ 3039 int VIndex = uc2 - HC_VBASE; 3040 if (0 <= VIndex && VIndex < HC_VCOUNT) { 3041 /* Make syllable of form LV. */ 3042 UPDATE_UC(HC_SBASE + 3043 (LIndex * HC_VCOUNT + VIndex) * 3044 HC_TCOUNT); 3045 } else { 3046 WRITE_UC(); 3047 REPLACE_UC_WITH_UC2(); 3048 } 3049 continue; 3050 } else if (0 <= (SIndex = uc - HC_SBASE) && 3051 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { 3052 /* 3053 * Hangul Composition. 3054 * 2. Two current code points are LV and T. 3055 */ 3056 int TIndex = uc2 - HC_TBASE; 3057 if (0 < TIndex && TIndex < HC_TCOUNT) { 3058 /* Make syllable of form LVT. */ 3059 UPDATE_UC(uc + TIndex); 3060 } else { 3061 WRITE_UC(); 3062 REPLACE_UC_WITH_UC2(); 3063 } 3064 continue; 3065 } else if ((nfc = get_nfc(uc, uc2)) != 0) { 3066 /* A composition to current code points 3067 * is found. */ 3068 UPDATE_UC(nfc); 3069 continue; 3070 } else if ((cl = CCC(uc2)) == 0) { 3071 /* Clearly 'uc2' the second code point is not 3072 * a decomposable code. */ 3073 WRITE_UC(); 3074 REPLACE_UC_WITH_UC2(); 3075 continue; 3076 } 3077 3078 /* 3079 * Collect following decomposable code points. 3080 */ 3081 cx = 0; 3082 ucx[0] = uc2; 3083 ccx[0] = cl; 3084 COLLECT_CPS(1); 3085 3086 /* 3087 * Find a composed code in the collected code points. 3088 */ 3089 i = 1; 3090 while (i < ucx_size) { 3091 int j; 3092 3093 if ((nfc = get_nfc(uc, ucx[i])) == 0) { 3094 i++; 3095 continue; 3096 } 3097 3098 /* 3099 * nfc is composed of uc and ucx[i]. 3100 */ 3101 UPDATE_UC(nfc); 3102 3103 /* 3104 * Remove ucx[i] by shifting 3105 * following code points. 3106 */ 3107 for (j = i; j+1 < ucx_size; j++) { 3108 ucx[j] = ucx[j+1]; 3109 ccx[j] = ccx[j+1]; 3110 } 3111 ucx_size --; 3112 3113 /* 3114 * Collect following code points blocked 3115 * by ucx[i] the removed code point. 3116 */ 3117 if (ucx_size > 0 && i == ucx_size && 3118 nx > 0 && cx == cl) { 3119 cl = ccx[ucx_size-1]; 3120 COLLECT_CPS(ucx_size); 3121 } 3122 /* 3123 * Restart finding a composed code with 3124 * the updated uc from the top of the 3125 * collected code points. 3126 */ 3127 i = 0; 3128 } 3129 3130 /* 3131 * Apparently the current code points are not 3132 * decomposed characters or already composed. 3133 */ 3134 WRITE_UC(); 3135 for (i = 0; i < ucx_size; i++) 3136 UNPARSE(p, endp, ucx[i]); 3137 3138 /* 3139 * Flush out remaining canonical combining characters. 3140 */ 3141 if (nx > 0 && cx == cl && len > 0) { 3142 while ((nx = parse(&ucx[0], s, len)) 3143 > 0) { 3144 cx = CCC(ucx[0]); 3145 if (cl > cx) 3146 break; 3147 s += nx; 3148 len -= nx; 3149 cl = cx; 3150 UNPARSE(p, endp, ucx[0]); 3151 } 3152 } 3153 break; 3154 } 3155 if (n2 < 0) { 3156 WRITE_UC(); 3157 /* Use a replaced unicode character. */ 3158 UNPARSE(p, endp, uc2); 3159 s += n2*-1; 3160 len -= n2*-1; 3161 ret = -1; 3162 continue; 3163 } else if (n2 == 0) { 3164 WRITE_UC(); 3165 break; 3166 } 3167 } 3168 as->length = p - as->s; 3169 as->s[as->length] = '\0'; 3170 if (ts == 2) 3171 as->s[as->length+1] = '\0'; 3172 return (ret); 3173} 3174 3175static int 3176get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc) 3177{ 3178 int t, b; 3179 3180 /* 3181 * These are not converted to NFD on Mac OS. 3182 */ 3183 if ((uc >= 0x2000 && uc <= 0x2FFF) || 3184 (uc >= 0xF900 && uc <= 0xFAFF) || 3185 (uc >= 0x2F800 && uc <= 0x2FAFF)) 3186 return (0); 3187 /* 3188 * Those code points are not converted to NFD on Mac OS. 3189 * I do not know the reason because it is undocumented. 3190 * NFC NFD 3191 * 1109A ==> 11099 110BA 3192 * 1109C ==> 1109B 110BA 3193 * 110AB ==> 110A5 110BA 3194 */ 3195 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB) 3196 return (0); 3197 3198 t = 0; 3199 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1; 3200 while (b >= t) { 3201 int m = (t + b) / 2; 3202 if (u_decomposition_table[m].nfc < uc) 3203 t = m + 1; 3204 else if (u_decomposition_table[m].nfc > uc) 3205 b = m - 1; 3206 else { 3207 *cp1 = u_decomposition_table[m].cp1; 3208 *cp2 = u_decomposition_table[m].cp2; 3209 return (1); 3210 } 3211 } 3212 return (0); 3213} 3214 3215#define REPLACE_UC_WITH(cp) do { \ 3216 uc = cp; \ 3217 ucptr = NULL; \ 3218} while (0) 3219 3220/* 3221 * Normalize UTF-8 characters to Form D and copy the result. 3222 */ 3223static int 3224archive_string_normalize_D(struct archive_string *as, const void *_p, 3225 size_t len, struct archive_string_conv *sc) 3226{ 3227 const char *s = (const char *)_p; 3228 char *p, *endp; 3229 uint32_t uc, uc2; 3230 size_t w; 3231 int always_replace, n, n2, ret = 0, spair, ts, tm; 3232 int (*parse)(uint32_t *, const char *, size_t); 3233 size_t (*unparse)(char *, size_t, uint32_t); 3234 3235 always_replace = 1; 3236 ts = 1;/* text size. */ 3237 if (sc->flag & SCONV_TO_UTF16BE) { 3238 unparse = unicode_to_utf16be; 3239 ts = 2; 3240 if (sc->flag & SCONV_FROM_UTF16BE) 3241 always_replace = 0; 3242 } else if (sc->flag & SCONV_TO_UTF16LE) { 3243 unparse = unicode_to_utf16le; 3244 ts = 2; 3245 if (sc->flag & SCONV_FROM_UTF16LE) 3246 always_replace = 0; 3247 } else if (sc->flag & SCONV_TO_UTF8) { 3248 unparse = unicode_to_utf8; 3249 if (sc->flag & SCONV_FROM_UTF8) 3250 always_replace = 0; 3251 } else { 3252 /* 3253 * This case is going to be converted to another 3254 * character-set through iconv. 3255 */ 3256 always_replace = 0; 3257 if (sc->flag & SCONV_FROM_UTF16BE) { 3258 unparse = unicode_to_utf16be; 3259 ts = 2; 3260 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3261 unparse = unicode_to_utf16le; 3262 ts = 2; 3263 } else { 3264 unparse = unicode_to_utf8; 3265 } 3266 } 3267 3268 if (sc->flag & SCONV_FROM_UTF16BE) { 3269 parse = utf16be_to_unicode; 3270 tm = 1; 3271 spair = 4;/* surrogate pair size in UTF-16. */ 3272 } else if (sc->flag & SCONV_FROM_UTF16LE) { 3273 parse = utf16le_to_unicode; 3274 tm = 1; 3275 spair = 4;/* surrogate pair size in UTF-16. */ 3276 } else { 3277 parse = cesu8_to_unicode; 3278 tm = ts; 3279 spair = 6;/* surrogate pair size in UTF-8. */ 3280 } 3281 3282 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) 3283 return (-1); 3284 3285 p = as->s + as->length; 3286 endp = as->s + as->buffer_length - ts; 3287 while ((n = parse(&uc, s, len)) != 0) { 3288 const char *ucptr; 3289 uint32_t cp1, cp2; 3290 int SIndex; 3291 struct { 3292 uint32_t uc; 3293 int ccc; 3294 } fdc[FDC_MAX]; 3295 int fdi, fdj; 3296 int ccc; 3297 3298check_first_code: 3299 if (n < 0) { 3300 /* Use a replaced unicode character. */ 3301 UNPARSE(p, endp, uc); 3302 s += n*-1; 3303 len -= n*-1; 3304 ret = -1; 3305 continue; 3306 } else if (n == spair || always_replace) 3307 /* uc is converted from a surrogate pair. 3308 * this should be treated as a changed code. */ 3309 ucptr = NULL; 3310 else 3311 ucptr = s; 3312 s += n; 3313 len -= n; 3314 3315 /* Hangul Decomposition. */ 3316 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) { 3317 int L = HC_LBASE + SIndex / HC_NCOUNT; 3318 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT; 3319 int T = HC_TBASE + SIndex % HC_TCOUNT; 3320 3321 REPLACE_UC_WITH(L); 3322 WRITE_UC(); 3323 REPLACE_UC_WITH(V); 3324 WRITE_UC(); 3325 if (T != HC_TBASE) { 3326 REPLACE_UC_WITH(T); 3327 WRITE_UC(); 3328 } 3329 continue; 3330 } 3331 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) { 3332 WRITE_UC(); 3333 continue; 3334 } 3335 3336 fdi = 0; 3337 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) { 3338 int k; 3339 3340 for (k = fdi; k > 0; k--) 3341 fdc[k] = fdc[k-1]; 3342 fdc[0].ccc = CCC(cp2); 3343 fdc[0].uc = cp2; 3344 fdi++; 3345 REPLACE_UC_WITH(cp1); 3346 } 3347 3348 /* Read following code points. */ 3349 while ((n2 = parse(&uc2, s, len)) > 0 && 3350 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) { 3351 int j, k; 3352 3353 s += n2; 3354 len -= n2; 3355 for (j = 0; j < fdi; j++) { 3356 if (fdc[j].ccc > ccc) 3357 break; 3358 } 3359 if (j < fdi) { 3360 for (k = fdi; k > j; k--) 3361 fdc[k] = fdc[k-1]; 3362 fdc[j].ccc = ccc; 3363 fdc[j].uc = uc2; 3364 } else { 3365 fdc[fdi].ccc = ccc; 3366 fdc[fdi].uc = uc2; 3367 } 3368 fdi++; 3369 } 3370 3371 WRITE_UC(); 3372 for (fdj = 0; fdj < fdi; fdj++) { 3373 REPLACE_UC_WITH(fdc[fdj].uc); 3374 WRITE_UC(); 3375 } 3376 3377 if (n2 == 0) 3378 break; 3379 REPLACE_UC_WITH(uc2); 3380 n = n2; 3381 goto check_first_code; 3382 } 3383 as->length = p - as->s; 3384 as->s[as->length] = '\0'; 3385 if (ts == 2) 3386 as->s[as->length+1] = '\0'; 3387 return (ret); 3388} 3389 3390/* 3391 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption 3392 * that WCS is Unicode. It is true for several platforms but some are false. 3393 * And then people who did not use UTF-8 locale on the non Unicode WCS 3394 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those 3395 * now cannot get right filename from libarchive 3.x and later since we 3396 * fixed the wrong assumption and it is incompatible to older its versions. 3397 * So we provide special option, "compat-2x.x", for resolving it. 3398 * That option enable the string conversion of libarchive 2.x. 3399 * 3400 * Translates the wrong UTF-8 string made by libarchive 2.x into current 3401 * locale character set and appends to the archive_string. 3402 * Note: returns -1 if conversion fails. 3403 */ 3404static int 3405strncat_from_utf8_libarchive2(struct archive_string *as, 3406 const void *_p, size_t len, struct archive_string_conv *sc) 3407{ 3408 const char *s; 3409 int n; 3410 char *p; 3411 char *end; 3412 uint32_t unicode; 3413#if HAVE_WCRTOMB 3414 mbstate_t shift_state; 3415 3416 memset(&shift_state, 0, sizeof(shift_state)); 3417#else 3418 /* Clear the shift state before starting. */ 3419 wctomb(NULL, L'\0'); 3420#endif 3421 (void)sc; /* UNUSED */ 3422 /* 3423 * Allocate buffer for MBS. 3424 * We need this allocation here since it is possible that 3425 * as->s is still NULL. 3426 */ 3427 if (archive_string_ensure(as, as->length + len + 1) == NULL) 3428 return (-1); 3429 3430 s = (const char *)_p; 3431 p = as->s + as->length; 3432 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3433 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { 3434 wchar_t wc; 3435 3436 if (p >= end) { 3437 as->length = p - as->s; 3438 /* Re-allocate buffer for MBS. */ 3439 if (archive_string_ensure(as, 3440 as->length + len * 2 + 1) == NULL) 3441 return (-1); 3442 p = as->s + as->length; 3443 end = as->s + as->buffer_length - MB_CUR_MAX -1; 3444 } 3445 3446 /* 3447 * As libarchive 2.x, translates the UTF-8 characters into 3448 * wide-characters in the assumption that WCS is Unicode. 3449 */ 3450 if (n < 0) { 3451 n *= -1; 3452 wc = L'?'; 3453 } else 3454 wc = (wchar_t)unicode; 3455 3456 s += n; 3457 len -= n; 3458 /* 3459 * Translates the wide-character into the current locale MBS. 3460 */ 3461#if HAVE_WCRTOMB 3462 n = (int)wcrtomb(p, wc, &shift_state); 3463#else 3464 n = (int)wctomb(p, wc); 3465#endif 3466 if (n == -1) 3467 return (-1); 3468 p += n; 3469 } 3470 as->length = p - as->s; 3471 as->s[as->length] = '\0'; 3472 return (0); 3473} 3474 3475 3476/* 3477 * Conversion functions between current locale dependent MBS and UTF-16BE. 3478 * strncat_from_utf16be() : UTF-16BE --> MBS 3479 * strncat_to_utf16be() : MBS --> UTF16BE 3480 */ 3481 3482#if defined(_WIN32) && !defined(__CYGWIN__) 3483 3484/* 3485 * Convert a UTF-16BE/LE string to current locale and copy the result. 3486 * Return -1 if conversion fails. 3487 */ 3488static int 3489win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, 3490 struct archive_string_conv *sc, int be) 3491{ 3492 struct archive_string tmp; 3493 const char *u16; 3494 int ll; 3495 BOOL defchar; 3496 char *mbs; 3497 size_t mbs_size, b; 3498 int ret = 0; 3499 3500 bytes &= ~1; 3501 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3502 return (-1); 3503 3504 mbs = as->s + as->length; 3505 mbs_size = as->buffer_length - as->length -1; 3506 3507 if (sc->to_cp == CP_C_LOCALE) { 3508 /* 3509 * "C" locale special process. 3510 */ 3511 u16 = _p; 3512 ll = 0; 3513 for (b = 0; b < bytes; b += 2) { 3514 uint16_t val; 3515 if (be) 3516 val = archive_be16dec(u16+b); 3517 else 3518 val = archive_le16dec(u16+b); 3519 if (val > 255) { 3520 *mbs++ = '?'; 3521 ret = -1; 3522 } else 3523 *mbs++ = (char)(val&0xff); 3524 ll++; 3525 } 3526 as->length += ll; 3527 as->s[as->length] = '\0'; 3528 return (ret); 3529 } 3530 3531 archive_string_init(&tmp); 3532 if (be) { 3533 if (is_big_endian()) { 3534 u16 = _p; 3535 } else { 3536 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3537 return (-1); 3538 memcpy(tmp.s, _p, bytes); 3539 for (b = 0; b < bytes; b += 2) { 3540 uint16_t val = archive_be16dec(tmp.s+b); 3541 archive_le16enc(tmp.s+b, val); 3542 } 3543 u16 = tmp.s; 3544 } 3545 } else { 3546 if (!is_big_endian()) { 3547 u16 = _p; 3548 } else { 3549 if (archive_string_ensure(&tmp, bytes+2) == NULL) 3550 return (-1); 3551 memcpy(tmp.s, _p, bytes); 3552 for (b = 0; b < bytes; b += 2) { 3553 uint16_t val = archive_le16dec(tmp.s+b); 3554 archive_be16enc(tmp.s+b, val); 3555 } 3556 u16 = tmp.s; 3557 } 3558 } 3559 3560 do { 3561 defchar = 0; 3562 ll = WideCharToMultiByte(sc->to_cp, 0, 3563 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size, 3564 NULL, &defchar); 3565 /* Exit loop if we succeeded */ 3566 if (ll != 0 || 3567 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3568 break; 3569 } 3570 /* Else expand buffer and loop to try again. */ 3571 ll = WideCharToMultiByte(sc->to_cp, 0, 3572 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL); 3573 if (archive_string_ensure(as, ll +1) == NULL) 3574 return (-1); 3575 mbs = as->s + as->length; 3576 mbs_size = as->buffer_length - as->length -1; 3577 } while (1); 3578 archive_string_free(&tmp); 3579 as->length += ll; 3580 as->s[as->length] = '\0'; 3581 if (ll == 0 || defchar) 3582 ret = -1; 3583 return (ret); 3584} 3585 3586static int 3587win_strncat_from_utf16be(struct archive_string *as, const void *_p, 3588 size_t bytes, struct archive_string_conv *sc) 3589{ 3590 return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); 3591} 3592 3593static int 3594win_strncat_from_utf16le(struct archive_string *as, const void *_p, 3595 size_t bytes, struct archive_string_conv *sc) 3596{ 3597 return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); 3598} 3599 3600static int 3601is_big_endian(void) 3602{ 3603 uint16_t d = 1; 3604 3605 return (archive_be16dec(&d) == 1); 3606} 3607 3608/* 3609 * Convert a current locale string to UTF-16BE/LE and copy the result. 3610 * Return -1 if conversion fails. 3611 */ 3612static int 3613win_strncat_to_utf16(struct archive_string *as16, const void *_p, 3614 size_t length, struct archive_string_conv *sc, int bigendian) 3615{ 3616 const char *s = (const char *)_p; 3617 char *u16; 3618 size_t count, avail; 3619 3620 if (archive_string_ensure(as16, 3621 as16->length + (length + 1) * 2) == NULL) 3622 return (-1); 3623 3624 u16 = as16->s + as16->length; 3625 avail = as16->buffer_length - 2; 3626 if (sc->from_cp == CP_C_LOCALE) { 3627 /* 3628 * "C" locale special process. 3629 */ 3630 count = 0; 3631 while (count < length && *s) { 3632 if (bigendian) 3633 archive_be16enc(u16, *s); 3634 else 3635 archive_le16enc(u16, *s); 3636 u16 += 2; 3637 s++; 3638 count++; 3639 } 3640 as16->length += count << 1; 3641 as16->s[as16->length] = 0; 3642 as16->s[as16->length+1] = 0; 3643 return (0); 3644 } 3645 do { 3646 count = MultiByteToWideChar(sc->from_cp, 3647 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1); 3648 /* Exit loop if we succeeded */ 3649 if (count != 0 || 3650 GetLastError() != ERROR_INSUFFICIENT_BUFFER) { 3651 break; 3652 } 3653 /* Expand buffer and try again */ 3654 count = MultiByteToWideChar(sc->from_cp, 3655 MB_PRECOMPOSED, s, (int)length, NULL, 0); 3656 if (archive_string_ensure(as16, (count +1) * 2) 3657 == NULL) 3658 return (-1); 3659 u16 = as16->s + as16->length; 3660 avail = as16->buffer_length - 2; 3661 } while (1); 3662 as16->length += count * 2; 3663 as16->s[as16->length] = 0; 3664 as16->s[as16->length+1] = 0; 3665 if (count == 0) 3666 return (-1); 3667 3668 if (is_big_endian()) { 3669 if (!bigendian) { 3670 while (count > 0) { 3671 uint16_t v = archive_be16dec(u16); 3672 archive_le16enc(u16, v); 3673 u16 += 2; 3674 count--; 3675 } 3676 } 3677 } else { 3678 if (bigendian) { 3679 while (count > 0) { 3680 uint16_t v = archive_le16dec(u16); 3681 archive_be16enc(u16, v); 3682 u16 += 2; 3683 count--; 3684 } 3685 } 3686 } 3687 return (0); 3688} 3689 3690static int 3691win_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3692 size_t length, struct archive_string_conv *sc) 3693{ 3694 return (win_strncat_to_utf16(as16, _p, length, sc, 1)); 3695} 3696 3697static int 3698win_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3699 size_t length, struct archive_string_conv *sc) 3700{ 3701 return (win_strncat_to_utf16(as16, _p, length, sc, 0)); 3702} 3703 3704#endif /* _WIN32 && !__CYGWIN__ */ 3705 3706/* 3707 * Do the best effort for conversions. 3708 * We cannot handle UTF-16BE character-set without such iconv, 3709 * but there is a chance if a string consists just ASCII code or 3710 * a current locale is UTF-8. 3711 */ 3712 3713/* 3714 * Convert a UTF-16BE string to current locale and copy the result. 3715 * Return -1 if conversion fails. 3716 */ 3717static int 3718best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, 3719 size_t bytes, struct archive_string_conv *sc, int be) 3720{ 3721 const char *utf16 = (const char *)_p; 3722 char *mbs; 3723 uint32_t uc; 3724 int n, ret; 3725 3726 (void)sc; /* UNUSED */ 3727 /* 3728 * Other case, we should do the best effort. 3729 * If all character are ASCII(<0x7f), we can convert it. 3730 * if not , we set a alternative character and return -1. 3731 */ 3732 ret = 0; 3733 if (archive_string_ensure(as, as->length + bytes +1) == NULL) 3734 return (-1); 3735 mbs = as->s + as->length; 3736 3737 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { 3738 if (n < 0) { 3739 n *= -1; 3740 ret = -1; 3741 } 3742 bytes -= n; 3743 utf16 += n; 3744 3745 if (uc > 127) { 3746 /* We cannot handle it. */ 3747 *mbs++ = '?'; 3748 ret = -1; 3749 } else 3750 *mbs++ = (char)uc; 3751 } 3752 as->length = mbs - as->s; 3753 as->s[as->length] = '\0'; 3754 return (ret); 3755} 3756 3757static int 3758best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, 3759 size_t bytes, struct archive_string_conv *sc) 3760{ 3761 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); 3762} 3763 3764static int 3765best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, 3766 size_t bytes, struct archive_string_conv *sc) 3767{ 3768 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); 3769} 3770 3771/* 3772 * Convert a current locale string to UTF-16BE/LE and copy the result. 3773 * Return -1 if conversion fails. 3774 */ 3775static int 3776best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, 3777 size_t length, struct archive_string_conv *sc, int bigendian) 3778{ 3779 const char *s = (const char *)_p; 3780 char *utf16; 3781 size_t remaining; 3782 int ret; 3783 3784 (void)sc; /* UNUSED */ 3785 /* 3786 * Other case, we should do the best effort. 3787 * If all character are ASCII(<0x7f), we can convert it. 3788 * if not , we set a alternative character and return -1. 3789 */ 3790 ret = 0; 3791 remaining = length; 3792 3793 if (archive_string_ensure(as16, 3794 as16->length + (length + 1) * 2) == NULL) 3795 return (-1); 3796 3797 utf16 = as16->s + as16->length; 3798 while (remaining--) { 3799 unsigned c = *s++; 3800 if (c > 127) { 3801 /* We cannot handle it. */ 3802 c = UNICODE_R_CHAR; 3803 ret = -1; 3804 } 3805 if (bigendian) 3806 archive_be16enc(utf16, c); 3807 else 3808 archive_le16enc(utf16, c); 3809 utf16 += 2; 3810 } 3811 as16->length = utf16 - as16->s; 3812 as16->s[as16->length] = 0; 3813 as16->s[as16->length+1] = 0; 3814 return (ret); 3815} 3816 3817static int 3818best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, 3819 size_t length, struct archive_string_conv *sc) 3820{ 3821 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); 3822} 3823 3824static int 3825best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, 3826 size_t length, struct archive_string_conv *sc) 3827{ 3828 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); 3829} 3830 3831 3832/* 3833 * Multistring operations. 3834 */ 3835 3836void 3837archive_mstring_clean(struct archive_mstring *aes) 3838{ 3839 archive_wstring_free(&(aes->aes_wcs)); 3840 archive_string_free(&(aes->aes_mbs)); 3841 archive_string_free(&(aes->aes_utf8)); 3842 archive_string_free(&(aes->aes_mbs_in_locale)); 3843 aes->aes_set = 0; 3844} 3845 3846void 3847archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) 3848{ 3849 dest->aes_set = src->aes_set; 3850 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); 3851 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); 3852 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); 3853} 3854 3855int 3856archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, 3857 const char **p) 3858{ 3859 struct archive_string_conv *sc; 3860 int r; 3861 3862 /* If we already have a UTF8 form, return that immediately. */ 3863 if (aes->aes_set & AES_SET_UTF8) { 3864 *p = aes->aes_utf8.s; 3865 return (0); 3866 } 3867 3868 *p = NULL; 3869 if (aes->aes_set & AES_SET_MBS) { 3870 sc = archive_string_conversion_to_charset(a, "UTF-8", 1); 3871 if (sc == NULL) 3872 return (-1);/* Couldn't allocate memory for sc. */ 3873 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s, 3874 aes->aes_mbs.length, sc); 3875 if (a == NULL) 3876 free_sconv_object(sc); 3877 if (r == 0) { 3878 aes->aes_set |= AES_SET_UTF8; 3879 *p = aes->aes_utf8.s; 3880 return (0);/* success. */ 3881 } else 3882 return (-1);/* failure. */ 3883 } 3884 return (0);/* success. */ 3885} 3886 3887int 3888archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, 3889 const char **p) 3890{ 3891 int r, ret = 0; 3892 3893 (void)a; /* UNUSED */ 3894 /* If we already have an MBS form, return that immediately. */ 3895 if (aes->aes_set & AES_SET_MBS) { 3896 *p = aes->aes_mbs.s; 3897 return (ret); 3898 } 3899 3900 *p = NULL; 3901 /* If there's a WCS form, try converting with the native locale. */ 3902 if (aes->aes_set & AES_SET_WCS) { 3903 archive_string_empty(&(aes->aes_mbs)); 3904 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3905 aes->aes_wcs.s, aes->aes_wcs.length); 3906 *p = aes->aes_mbs.s; 3907 if (r == 0) { 3908 aes->aes_set |= AES_SET_MBS; 3909 return (ret); 3910 } else 3911 ret = -1; 3912 } 3913 3914 /* 3915 * Only a UTF-8 form cannot avail because its conversion already 3916 * failed at archive_mstring_update_utf8(). 3917 */ 3918 return (ret); 3919} 3920 3921int 3922archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, 3923 const wchar_t **wp) 3924{ 3925 int r, ret = 0; 3926 3927 (void)a;/* UNUSED */ 3928 /* Return WCS form if we already have it. */ 3929 if (aes->aes_set & AES_SET_WCS) { 3930 *wp = aes->aes_wcs.s; 3931 return (ret); 3932 } 3933 3934 *wp = NULL; 3935 /* Try converting MBS to WCS using native locale. */ 3936 if (aes->aes_set & AES_SET_MBS) { 3937 archive_wstring_empty(&(aes->aes_wcs)); 3938 r = archive_wstring_append_from_mbs(&(aes->aes_wcs), 3939 aes->aes_mbs.s, aes->aes_mbs.length); 3940 if (r == 0) { 3941 aes->aes_set |= AES_SET_WCS; 3942 *wp = aes->aes_wcs.s; 3943 } else 3944 ret = -1;/* failure. */ 3945 } 3946 return (ret); 3947} 3948 3949int 3950archive_mstring_get_mbs_l(struct archive_mstring *aes, 3951 const char **p, size_t *length, struct archive_string_conv *sc) 3952{ 3953 int r, ret = 0; 3954 3955#if defined(_WIN32) && !defined(__CYGWIN__) 3956 /* 3957 * Internationalization programming on Windows must use Wide 3958 * characters because Windows platform cannot make locale UTF-8. 3959 */ 3960 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { 3961 archive_string_empty(&(aes->aes_mbs_in_locale)); 3962 r = archive_string_append_from_wcs_in_codepage( 3963 &(aes->aes_mbs_in_locale), aes->aes_wcs.s, 3964 aes->aes_wcs.length, sc); 3965 if (r == 0) { 3966 *p = aes->aes_mbs_in_locale.s; 3967 if (length != NULL) 3968 *length = aes->aes_mbs_in_locale.length; 3969 return (0); 3970 } else if (errno == ENOMEM) 3971 return (-1); 3972 else 3973 ret = -1; 3974 } 3975#endif 3976 3977 /* If there is not an MBS form but is a WCS form, try converting 3978 * with the native locale to be used for translating it to specified 3979 * character-set. */ 3980 if ((aes->aes_set & AES_SET_MBS) == 0 && 3981 (aes->aes_set & AES_SET_WCS) != 0) { 3982 archive_string_empty(&(aes->aes_mbs)); 3983 r = archive_string_append_from_wcs(&(aes->aes_mbs), 3984 aes->aes_wcs.s, aes->aes_wcs.length); 3985 if (r == 0) 3986 aes->aes_set |= AES_SET_MBS; 3987 else if (errno == ENOMEM) 3988 return (-1); 3989 else 3990 ret = -1; 3991 } 3992 /* If we already have an MBS form, use it to be translated to 3993 * specified character-set. */ 3994 if (aes->aes_set & AES_SET_MBS) { 3995 if (sc == NULL) { 3996 /* Conversion is unneeded. */ 3997 *p = aes->aes_mbs.s; 3998 if (length != NULL) 3999 *length = aes->aes_mbs.length; 4000 return (0); 4001 } 4002 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale), 4003 aes->aes_mbs.s, aes->aes_mbs.length, sc); 4004 *p = aes->aes_mbs_in_locale.s; 4005 if (length != NULL) 4006 *length = aes->aes_mbs_in_locale.length; 4007 } else { 4008 *p = NULL; 4009 if (length != NULL) 4010 *length = 0; 4011 } 4012 return (ret); 4013} 4014 4015int 4016archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) 4017{ 4018 if (mbs == NULL) { 4019 aes->aes_set = 0; 4020 return (0); 4021 } 4022 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); 4023} 4024 4025int 4026archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, 4027 size_t len) 4028{ 4029 if (mbs == NULL) { 4030 aes->aes_set = 0; 4031 return (0); 4032 } 4033 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4034 archive_strncpy(&(aes->aes_mbs), mbs, len); 4035 archive_string_empty(&(aes->aes_utf8)); 4036 archive_wstring_empty(&(aes->aes_wcs)); 4037 return (0); 4038} 4039 4040int 4041archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) 4042{ 4043 return archive_mstring_copy_wcs_len(aes, wcs, 4044 wcs == NULL ? 0 : wcslen(wcs)); 4045} 4046 4047int 4048archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8) 4049{ 4050 if (utf8 == NULL) { 4051 aes->aes_set = 0; 4052 } 4053 aes->aes_set = AES_SET_UTF8; 4054 archive_string_empty(&(aes->aes_mbs)); 4055 archive_string_empty(&(aes->aes_wcs)); 4056 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8)); 4057 return (int)strlen(utf8); 4058} 4059 4060int 4061archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, 4062 size_t len) 4063{ 4064 if (wcs == NULL) { 4065 aes->aes_set = 0; 4066 } 4067 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ 4068 archive_string_empty(&(aes->aes_mbs)); 4069 archive_string_empty(&(aes->aes_utf8)); 4070 archive_wstrncpy(&(aes->aes_wcs), wcs, len); 4071 return (0); 4072} 4073 4074int 4075archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, 4076 const char *mbs, size_t len, struct archive_string_conv *sc) 4077{ 4078 int r; 4079 4080 if (mbs == NULL) { 4081 aes->aes_set = 0; 4082 return (0); 4083 } 4084 archive_string_empty(&(aes->aes_mbs)); 4085 archive_wstring_empty(&(aes->aes_wcs)); 4086 archive_string_empty(&(aes->aes_utf8)); 4087#if defined(_WIN32) && !defined(__CYGWIN__) 4088 /* 4089 * Internationalization programming on Windows must use Wide 4090 * characters because Windows platform cannot make locale UTF-8. 4091 */ 4092 if (sc == NULL) { 4093 if (archive_string_append(&(aes->aes_mbs), 4094 mbs, mbsnbytes(mbs, len)) == NULL) { 4095 aes->aes_set = 0; 4096 r = -1; 4097 } else { 4098 aes->aes_set = AES_SET_MBS; 4099 r = 0; 4100 } 4101#if defined(HAVE_ICONV) 4102 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { 4103 /* 4104 * This case happens only when MultiByteToWideChar() cannot 4105 * handle sc->from_cp, and we have to iconv in order to 4106 * translate character-set to wchar_t,UTF-16. 4107 */ 4108 iconv_t cd = sc->cd; 4109 unsigned from_cp; 4110 int flag; 4111 4112 /* 4113 * Translate multi-bytes from some character-set to UTF-8. 4114 */ 4115 sc->cd = sc->cd_w; 4116 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc); 4117 sc->cd = cd; 4118 if (r != 0) { 4119 aes->aes_set = 0; 4120 return (r); 4121 } 4122 aes->aes_set = AES_SET_UTF8; 4123 4124 /* 4125 * Append the UTF-8 string into wstring. 4126 */ 4127 flag = sc->flag; 4128 sc->flag &= ~(SCONV_NORMALIZATION_C 4129 | SCONV_TO_UTF16| SCONV_FROM_UTF16); 4130 from_cp = sc->from_cp; 4131 sc->from_cp = CP_UTF8; 4132 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), 4133 aes->aes_utf8.s, aes->aes_utf8.length, sc); 4134 sc->flag = flag; 4135 sc->from_cp = from_cp; 4136 if (r == 0) 4137 aes->aes_set |= AES_SET_WCS; 4138#endif 4139 } else { 4140 r = archive_wstring_append_from_mbs_in_codepage( 4141 &(aes->aes_wcs), mbs, len, sc); 4142 if (r == 0) 4143 aes->aes_set = AES_SET_WCS; 4144 else 4145 aes->aes_set = 0; 4146 } 4147#else 4148 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc); 4149 if (r == 0) 4150 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ 4151 else 4152 aes->aes_set = 0; 4153#endif 4154 return (r); 4155} 4156 4157/* 4158 * The 'update' form tries to proactively update all forms of 4159 * this string (WCS and MBS) and returns an error if any of 4160 * them fail. This is used by the 'pax' handler, for instance, 4161 * to detect and report character-conversion failures early while 4162 * still allowing clients to get potentially useful values from 4163 * the more tolerant lazy conversions. (get_mbs and get_wcs will 4164 * strive to give the user something useful, so you can get hopefully 4165 * usable values even if some of the character conversions are failing.) 4166 */ 4167int 4168archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, 4169 const char *utf8) 4170{ 4171 struct archive_string_conv *sc; 4172 int r; 4173 4174 if (utf8 == NULL) { 4175 aes->aes_set = 0; 4176 return (0); /* Succeeded in clearing everything. */ 4177 } 4178 4179 /* Save the UTF8 string. */ 4180 archive_strcpy(&(aes->aes_utf8), utf8); 4181 4182 /* Empty the mbs and wcs strings. */ 4183 archive_string_empty(&(aes->aes_mbs)); 4184 archive_wstring_empty(&(aes->aes_wcs)); 4185 4186 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ 4187 4188 /* Try converting UTF-8 to MBS, return false on failure. */ 4189 sc = archive_string_conversion_from_charset(a, "UTF-8", 1); 4190 if (sc == NULL) 4191 return (-1);/* Couldn't allocate memory for sc. */ 4192 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc); 4193 if (a == NULL) 4194 free_sconv_object(sc); 4195 if (r != 0) 4196 return (-1); 4197 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ 4198 4199 /* Try converting MBS to WCS, return false on failure. */ 4200 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, 4201 aes->aes_mbs.length)) 4202 return (-1); 4203 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; 4204 4205 /* All conversions succeeded. */ 4206 return (0); 4207} 4208