2 * Copyright (c) 2003-2011 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
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.
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.
27 #include "archive_platform.h"
28 __FBSDID("$FreeBSD$");
31 * Basic resizable string support, to simplify manipulating arbitrary-sized
32 * strings while minimizing heap activity.
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.
45 #ifdef HAVE_LANGINFO_H
48 #ifdef HAVE_LOCALCHARSET_H
49 #include <localcharset.h>
60 #if defined(_WIN32) && !defined(__CYGWIN__)
65 #include "archive_endian.h"
66 #include "archive_private.h"
67 #include "archive_string.h"
68 #include "archive_string_composition.h"
70 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
71 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
74 struct archive_string_conv {
75 struct archive_string_conv *next;
80 /* Set 1 if from_charset and to_charset are the same. */
83 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified
85 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from
86 * specified charset. */
87 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */
88 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting
90 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive
91 * 2.x in the wrong assumption. */
92 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C.
93 * Before UTF-8 characters are actually
95 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D.
96 * Before UTF-8 characters are actually
98 * Currently this only for MAC OS X. */
99 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */
100 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */
101 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */
102 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */
103 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */
104 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */
105 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
106 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
110 iconv_t cd_w;/* Use at archive_mstring on
113 /* A temporary buffer for normalization. */
114 struct archive_string utftmp;
115 int (*converter[2])(struct archive_string *, const void *, size_t,
116 struct archive_string_conv *);
120 #define CP_C_LOCALE 0 /* "C" locale only for this file. */
121 #define CP_UTF16LE 1200
122 #define CP_UTF16BE 1201
124 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
125 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
126 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
127 #define UNICODE_MAX 0x10FFFF
128 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */
129 /* Set U+FFFD(Replacement character) in UTF-8. */
130 #define UTF8_SET_R_CHAR(outp) do { \
135 #define UTF8_R_CHAR_SIZE 3
137 static struct archive_string_conv *find_sconv_object(struct archive *,
138 const char *, const char *);
139 static void add_sconv_object(struct archive *, struct archive_string_conv *);
140 static struct archive_string_conv *create_sconv_object(const char *,
141 const char *, unsigned, int);
142 static void free_sconv_object(struct archive_string_conv *);
143 static struct archive_string_conv *get_sconv_object(struct archive *,
144 const char *, const char *, int);
145 static unsigned make_codepage_from_charset(const char *);
146 static unsigned get_current_codepage(void);
147 static unsigned get_current_oemcp(void);
148 static size_t mbsnbytes(const void *, size_t);
149 static size_t utf16nbytes(const void *, size_t);
150 #if defined(_WIN32) && !defined(__CYGWIN__)
151 static int archive_wstring_append_from_mbs_in_codepage(
152 struct archive_wstring *, const char *, size_t,
153 struct archive_string_conv *);
154 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
155 const wchar_t *, size_t, struct archive_string_conv *);
156 static int is_big_endian(void);
157 static int strncat_in_codepage(struct archive_string *, const void *,
158 size_t, struct archive_string_conv *);
159 static int win_strncat_from_utf16be(struct archive_string *, const void *,
160 size_t, struct archive_string_conv *);
161 static int win_strncat_from_utf16le(struct archive_string *, const void *,
162 size_t, struct archive_string_conv *);
163 static int win_strncat_to_utf16be(struct archive_string *, const void *,
164 size_t, struct archive_string_conv *);
165 static int win_strncat_to_utf16le(struct archive_string *, const void *,
166 size_t, struct archive_string_conv *);
168 static int best_effort_strncat_from_utf16be(struct archive_string *,
169 const void *, size_t, struct archive_string_conv *);
170 static int best_effort_strncat_from_utf16le(struct archive_string *,
171 const void *, size_t, struct archive_string_conv *);
172 static int best_effort_strncat_to_utf16be(struct archive_string *,
173 const void *, size_t, struct archive_string_conv *);
174 static int best_effort_strncat_to_utf16le(struct archive_string *,
175 const void *, size_t, struct archive_string_conv *);
176 #if defined(HAVE_ICONV)
177 static int iconv_strncat_in_locale(struct archive_string *, const void *,
178 size_t, struct archive_string_conv *);
180 static int best_effort_strncat_in_locale(struct archive_string *,
181 const void *, size_t, struct archive_string_conv *);
182 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
183 static int utf8_to_unicode(uint32_t *, const char *, size_t);
184 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
185 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
186 static size_t unicode_to_utf8(char *, size_t, uint32_t);
187 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
188 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
189 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
190 static int strncat_from_utf8_libarchive2(struct archive_string *,
191 const void *, size_t, struct archive_string_conv *);
192 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
193 size_t, struct archive_string_conv *);
194 static int archive_string_normalize_C(struct archive_string *, const void *,
195 size_t, struct archive_string_conv *);
196 static int archive_string_normalize_D(struct archive_string *, const void *,
197 size_t, struct archive_string_conv *);
198 static int archive_string_append_unicode(struct archive_string *,
199 const void *, size_t, struct archive_string_conv *);
201 static struct archive_string *
202 archive_string_append(struct archive_string *as, const char *p, size_t s)
204 if (archive_string_ensure(as, as->length + s + 1) == NULL)
206 memcpy(as->s + as->length, p, s);
208 as->s[as->length] = 0;
212 static struct archive_wstring *
213 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
215 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
217 wmemcpy(as->s + as->length, p, s);
219 as->s[as->length] = 0;
224 archive_string_concat(struct archive_string *dest, struct archive_string *src)
226 if (archive_string_append(dest, src->s, src->length) == NULL)
227 __archive_errx(1, "Out of memory");
231 archive_wstring_concat(struct archive_wstring *dest,
232 struct archive_wstring *src)
234 if (archive_wstring_append(dest, src->s, src->length) == NULL)
235 __archive_errx(1, "Out of memory");
239 archive_string_free(struct archive_string *as)
242 as->buffer_length = 0;
248 archive_wstring_free(struct archive_wstring *as)
251 as->buffer_length = 0;
256 struct archive_wstring *
257 archive_wstring_ensure(struct archive_wstring *as, size_t s)
259 return (struct archive_wstring *)
260 archive_string_ensure((struct archive_string *)as,
261 s * sizeof(wchar_t));
264 /* Returns NULL on any allocation failure. */
265 struct archive_string *
266 archive_string_ensure(struct archive_string *as, size_t s)
271 /* If buffer is already big enough, don't reallocate. */
272 if (as->s && (s <= as->buffer_length))
276 * Growing the buffer at least exponentially ensures that
277 * append operations are always linear in the number of
278 * characters appended. Using a smaller growth rate for
279 * larger buffers reduces memory waste somewhat at the cost of
280 * a larger constant factor.
282 if (as->buffer_length < 32)
283 /* Start with a minimum 32-character buffer. */
285 else if (as->buffer_length < 8192)
286 /* Buffers under 8k are doubled for speed. */
287 new_length = as->buffer_length + as->buffer_length;
289 /* Buffers 8k and over grow by at least 25% each time. */
290 new_length = as->buffer_length + as->buffer_length / 4;
291 /* Be safe: If size wraps, fail. */
292 if (new_length < as->buffer_length) {
293 /* On failure, wipe the string and return NULL. */
294 archive_string_free(as);
295 errno = ENOMEM;/* Make sure errno has ENOMEM. */
300 * The computation above is a lower limit to how much we'll
301 * grow the buffer. In any case, we have to grow it enough to
306 /* Now we can reallocate the buffer. */
307 p = (char *)realloc(as->s, new_length);
309 /* On failure, wipe the string and return NULL. */
310 archive_string_free(as);
311 errno = ENOMEM;/* Make sure errno has ENOMEM. */
316 as->buffer_length = new_length;
321 * TODO: See if there's a way to avoid scanning
322 * the source string twice. Then test to see
323 * if it actually helps (remember that we're almost
324 * always called with pretty short arguments, so
325 * such an optimization might not help).
327 struct archive_string *
328 archive_strncat(struct archive_string *as, const void *_p, size_t n)
333 p = (const char *)_p;
335 /* Like strlen(p), except won't examine positions beyond p[n]. */
338 while (s < n && *pp) {
342 if ((as = archive_string_append(as, p, s)) == NULL)
343 __archive_errx(1, "Out of memory");
347 struct archive_wstring *
348 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
353 /* Like strlen(p), except won't examine positions beyond p[n]. */
356 while (s < n && *pp) {
360 if ((as = archive_wstring_append(as, p, s)) == NULL)
361 __archive_errx(1, "Out of memory");
365 struct archive_string *
366 archive_strcat(struct archive_string *as, const void *p)
368 /* strcat is just strncat without an effective limit.
369 * Assert that we'll never get called with a source
371 * TODO: Review all uses of strcat in the source
372 * and try to replace them with strncat().
374 return archive_strncat(as, p, 0x1000000);
377 struct archive_wstring *
378 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
381 return archive_wstrncat(as, p, 0x1000000);
384 struct archive_string *
385 archive_strappend_char(struct archive_string *as, char c)
387 if ((as = archive_string_append(as, &c, 1)) == NULL)
388 __archive_errx(1, "Out of memory");
392 struct archive_wstring *
393 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
395 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
396 __archive_errx(1, "Out of memory");
401 * Get the "current character set" name to use with iconv.
402 * On FreeBSD, the empty character set name "" chooses
403 * the correct character encoding for the current locale,
404 * so this isn't necessary.
405 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
406 * on that system, we have to explicitly call nl_langinfo()
407 * to get the right name. Not sure about other platforms.
409 * NOTE: GNU libiconv does not recognize the character-set name
410 * which some platform nl_langinfo(CODESET) returns, so we should
411 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
414 default_iconv_charset(const char *charset) {
415 if (charset != NULL && charset[0] != '\0')
417 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
418 /* locale_charset() is broken on Mac OS */
419 return locale_charset();
420 #elif HAVE_NL_LANGINFO
421 return nl_langinfo(CODESET);
427 #if defined(_WIN32) && !defined(__CYGWIN__)
430 * Convert MBS to WCS.
431 * Note: returns -1 if conversion fails.
434 archive_wstring_append_from_mbs(struct archive_wstring *dest,
435 const char *p, size_t len)
437 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
441 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
442 const char *s, size_t length, struct archive_string_conv *sc)
448 from_cp = sc->from_cp;
450 from_cp = get_current_codepage();
452 if (from_cp == CP_C_LOCALE) {
454 * "C" locale special process.
457 const unsigned char *mp;
459 if (NULL == archive_wstring_ensure(dest,
460 dest->length + length + 1))
463 ws = dest->s + dest->length;
464 mp = (const unsigned char *)s;
466 while (count < (int)length && *mp) {
467 *ws++ = (wchar_t)*mp++;
470 } else if (sc != NULL &&
471 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
473 * Normalize UTF-8 and UTF-16BE and convert it directly
474 * to UTF-16 as wchar_t.
476 struct archive_string u16;
477 int saved_flag = sc->flag;/* save current flag. */
480 sc->flag |= SCONV_TO_UTF16BE;
482 sc->flag |= SCONV_TO_UTF16LE;
484 if (sc->flag & SCONV_FROM_UTF16) {
486 * UTF-16BE/LE NFD ===> UTF-16 NFC
487 * UTF-16BE/LE NFC ===> UTF-16 NFD
489 count = (int)utf16nbytes(s, length);
492 * UTF-8 NFD ===> UTF-16 NFC
493 * UTF-8 NFC ===> UTF-16 NFD
495 count = (int)mbsnbytes(s, length);
497 u16.s = (char *)dest->s;
498 u16.length = dest->length << 1;;
499 u16.buffer_length = dest->buffer_length;
500 if (sc->flag & SCONV_NORMALIZATION_C)
501 ret = archive_string_normalize_C(&u16, s, count, sc);
503 ret = archive_string_normalize_D(&u16, s, count, sc);
504 dest->s = (wchar_t *)u16.s;
505 dest->length = u16.length >> 1;
506 dest->buffer_length = u16.buffer_length;
507 sc->flag = saved_flag;/* restore the saved flag. */
509 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
510 count = (int)utf16nbytes(s, length);
511 count >>= 1; /* to be WCS length */
512 /* Allocate memory for WCS. */
513 if (NULL == archive_wstring_ensure(dest,
514 dest->length + count + 1))
516 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
517 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
518 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
520 for (b = 0; b < count; b++) {
521 uint16_t val = archive_le16dec(u16+b);
522 archive_be16enc(u16+b, val);
524 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
525 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
527 for (b = 0; b < count; b++) {
528 uint16_t val = archive_be16dec(u16+b);
529 archive_le16enc(u16+b, val);
538 else if (sc->flag & SCONV_FROM_CHARSET) {
539 /* Do not trust the length which comes from
540 * an archive file. */
541 length = mbsnbytes(s, length);
544 mbflag = MB_PRECOMPOSED;
546 buffsize = dest->length + length + 1;
548 /* Allocate memory for WCS. */
549 if (NULL == archive_wstring_ensure(dest, buffsize))
551 /* Convert MBS to WCS. */
552 count = MultiByteToWideChar(from_cp,
553 mbflag, s, (int)length, dest->s + dest->length,
554 (int)(dest->buffer_length >> 1) -1);
556 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
557 /* Expand the WCS buffer. */
558 buffsize = dest->buffer_length << 1;
561 if (count == 0 && length != 0)
565 dest->length += count;
566 dest->s[dest->length] = L'\0';
573 * Convert MBS to WCS.
574 * Note: returns -1 if conversion fails.
577 archive_wstring_append_from_mbs(struct archive_wstring *dest,
578 const char *p, size_t len)
583 * No single byte will be more than one wide character,
584 * so this length estimate will always be big enough.
586 size_t wcs_length = len;
587 size_t mbs_length = len;
591 mbstate_t shift_state;
593 memset(&shift_state, 0, sizeof(shift_state));
595 if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1))
597 wcs = dest->s + dest->length;
599 * We cannot use mbsrtowcs/mbstowcs here because those may convert
600 * extra MBS when strlen(p) > len and one wide character consis of
603 while (*mbs && mbs_length > 0) {
604 if (wcs_length == 0) {
605 dest->length = wcs - dest->s;
606 dest->s[dest->length] = L'\0';
607 wcs_length = mbs_length;
608 if (NULL == archive_wstring_ensure(dest,
609 dest->length + wcs_length + 1))
611 wcs = dest->s + dest->length;
614 r = mbrtowc(wcs, mbs, wcs_length, &shift_state);
616 r = mbtowc(wcs, mbs, wcs_length);
618 if (r == (size_t)-1 || r == (size_t)-2) {
620 if (errno == EILSEQ) {
627 if (r == 0 || r > mbs_length)
634 dest->length = wcs - dest->s;
635 dest->s[dest->length] = L'\0';
641 #if defined(_WIN32) && !defined(__CYGWIN__)
645 * Note: returns -1 if conversion fails.
647 * Win32 builds use WideCharToMultiByte from the Windows API.
648 * (Maybe Cygwin should too? WideCharToMultiByte will know a
649 * lot more about local character encodings than the wcrtomb()
650 * wrapper is going to know.)
653 archive_string_append_from_wcs(struct archive_string *as,
654 const wchar_t *w, size_t len)
656 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
660 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
661 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
663 BOOL defchar_used, *dp;
666 int wslen = (int)len;
671 to_cp = get_current_codepage();
673 if (to_cp == CP_C_LOCALE) {
675 * "C" locale special process.
677 const wchar_t *wp = ws;
680 if (NULL == archive_string_ensure(as,
681 as->length + wslen +1))
683 p = as->s + as->length;
686 while (count < wslen && *wp) {
695 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
699 archive_string_ensure(as, as->length + len * 2 + 2))
701 u16 = (uint16_t *)(as->s + as->length);
704 if (sc->flag & SCONV_TO_UTF16BE) {
705 while (count < (int)len && *ws) {
706 archive_be16enc(u16+count, *ws);
711 while (count < (int)len && *ws) {
712 archive_le16enc(u16+count, *ws);
717 count <<= 1; /* to be byte size */
719 /* Make sure the MBS buffer has plenty to set. */
721 archive_string_ensure(as, as->length + len * 2 + 1))
725 if (to_cp == CP_UTF8 || sc == NULL)
729 count = WideCharToMultiByte(to_cp, 0, ws, wslen,
730 as->s + as->length, (int)as->buffer_length-1, NULL, dp);
732 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
733 /* Expand the MBS buffer and retry. */
734 if (NULL == archive_string_ensure(as,
735 as->buffer_length + len))
744 as->s[as->length] = '\0';
745 return (defchar_used?-1:ret);
748 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
751 * Translates a wide character string into current locale character set
752 * and appends to the archive_string. Note: returns -1 if conversion
756 archive_string_append_from_wcs(struct archive_string *as,
757 const wchar_t *w, size_t len)
759 /* We cannot use the standard wcstombs() here because it
760 * cannot tell us how big the output buffer should be. So
761 * I've built a loop around wcrtomb() or wctomb() that
762 * converts a character at a time and resizes the string as
763 * needed. We prefer wcrtomb() when it's available because
764 * it's thread-safe. */
769 mbstate_t shift_state;
771 memset(&shift_state, 0, sizeof(shift_state));
773 /* Clear the shift state before starting. */
777 * Allocate buffer for MBS.
778 * We need this allocation here since it is possible that
779 * as->s is still NULL.
781 if (archive_string_ensure(as, as->length + len + 1) == NULL)
784 p = as->s + as->length;
785 end = as->s + as->buffer_length - MB_CUR_MAX -1;
786 while (*w != L'\0' && len > 0) {
788 as->length = p - as->s;
789 as->s[as->length] = '\0';
790 /* Re-allocate buffer for MBS. */
791 if (archive_string_ensure(as,
792 as->length + len * 2 + 1) == NULL)
794 p = as->s + as->length;
795 end = as->s + as->buffer_length - MB_CUR_MAX -1;
798 n = wcrtomb(p, *w++, &shift_state);
803 if (errno == EILSEQ) {
804 /* Skip an illegal wide char. */
815 as->length = p - as->s;
816 as->s[as->length] = '\0';
820 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
823 * TODO: Test if __STDC_ISO_10646__ is defined.
824 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
825 * one character at a time. If a non-Windows platform doesn't have
826 * either of these, fall back to the built-in UTF8 conversion.
829 archive_string_append_from_wcs(struct archive_string *as,
830 const wchar_t *w, size_t len)
832 (void)as;/* UNUSED */
834 (void)len;/* UNUSED */
839 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
842 * Find a string conversion object by a pair of 'from' charset name
843 * and 'to' charset name from an archive object.
844 * Return NULL if not found.
846 static struct archive_string_conv *
847 find_sconv_object(struct archive *a, const char *fc, const char *tc)
849 struct archive_string_conv *sc;
854 for (sc = a->sconv; sc != NULL; sc = sc->next) {
855 if (strcmp(sc->from_charset, fc) == 0 &&
856 strcmp(sc->to_charset, tc) == 0)
863 * Register a string object to an archive object.
866 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
868 struct archive_string_conv **psc;
870 /* Add a new sconv to sconv list. */
873 psc = &((*psc)->next);
878 add_converter(struct archive_string_conv *sc, int (*converter)
879 (struct archive_string *, const void *, size_t,
880 struct archive_string_conv *))
882 if (sc == NULL || sc->nconverter >= 2)
883 __archive_errx(1, "Programing error");
884 sc->converter[sc->nconverter++] = converter;
888 setup_converter(struct archive_string_conv *sc)
895 * Perform special sequence for the incorrect UTF-8 filenames
896 * made by libarchive2.x.
898 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
899 add_converter(sc, strncat_from_utf8_libarchive2);
904 * Convert a string to UTF-16BE/LE.
906 if (sc->flag & SCONV_TO_UTF16) {
908 * If the current locale is UTF-8, we can translate
909 * a UTF-8 string into a UTF-16BE string.
911 if (sc->flag & SCONV_FROM_UTF8) {
912 add_converter(sc, archive_string_append_unicode);
916 #if defined(_WIN32) && !defined(__CYGWIN__)
917 if (sc->flag & SCONV_WIN_CP) {
918 if (sc->flag & SCONV_TO_UTF16BE)
919 add_converter(sc, win_strncat_to_utf16be);
921 add_converter(sc, win_strncat_to_utf16le);
926 #if defined(HAVE_ICONV)
927 if (sc->cd != (iconv_t)-1) {
928 add_converter(sc, iconv_strncat_in_locale);
933 if (sc->flag & SCONV_BEST_EFFORT) {
934 if (sc->flag & SCONV_TO_UTF16BE)
936 best_effort_strncat_to_utf16be);
939 best_effort_strncat_to_utf16le);
941 /* Make sure we have no converter. */
947 * Convert a string from UTF-16BE/LE.
949 if (sc->flag & SCONV_FROM_UTF16) {
951 * At least we should normalize a UTF-16BE string.
953 if (sc->flag & SCONV_NORMALIZATION_D)
954 add_converter(sc,archive_string_normalize_D);
955 else if (sc->flag & SCONV_NORMALIZATION_C)
956 add_converter(sc, archive_string_normalize_C);
958 if (sc->flag & SCONV_TO_UTF8) {
960 * If the current locale is UTF-8, we can translate
961 * a UTF-16BE/LE string into a UTF-8 string directly.
964 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
966 archive_string_append_unicode);
970 #if defined(_WIN32) && !defined(__CYGWIN__)
971 if (sc->flag & SCONV_WIN_CP) {
972 if (sc->flag & SCONV_FROM_UTF16BE)
973 add_converter(sc, win_strncat_from_utf16be);
975 add_converter(sc, win_strncat_from_utf16le);
980 #if defined(HAVE_ICONV)
981 if (sc->cd != (iconv_t)-1) {
982 add_converter(sc, iconv_strncat_in_locale);
987 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
988 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
989 add_converter(sc, best_effort_strncat_from_utf16be);
990 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
991 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
992 add_converter(sc, best_effort_strncat_from_utf16le);
994 /* Make sure we have no converter. */
999 if (sc->flag & SCONV_FROM_UTF8) {
1001 * At least we should normalize a UTF-8 string.
1003 if (sc->flag & SCONV_NORMALIZATION_D)
1004 add_converter(sc,archive_string_normalize_D);
1005 else if (sc->flag & SCONV_NORMALIZATION_C)
1006 add_converter(sc, archive_string_normalize_C);
1009 * Copy UTF-8 string with a check of CESU-8.
1010 * Apparently, iconv does not check surrogate pairs in UTF-8
1011 * when both from-charset and to-charset are UTF-8, and then
1012 * we use our UTF-8 copy code.
1014 if (sc->flag & SCONV_TO_UTF8) {
1016 * If the current locale is UTF-8, we can translate
1017 * a UTF-16BE string into a UTF-8 string directly.
1020 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1021 add_converter(sc, strncat_from_utf8_to_utf8);
1026 #if defined(_WIN32) && !defined(__CYGWIN__)
1028 * On Windows we can use Windows API for a string conversion.
1030 if (sc->flag & SCONV_WIN_CP) {
1031 add_converter(sc, strncat_in_codepage);
1037 if (sc->cd != (iconv_t)-1) {
1038 add_converter(sc, iconv_strncat_in_locale);
1040 * iconv generally does not support UTF-8-MAC and so
1041 * we have to the output of iconv from NFC to NFD if
1044 if ((sc->flag & SCONV_FROM_CHARSET) &&
1045 (sc->flag & SCONV_TO_UTF8)) {
1046 if (sc->flag & SCONV_NORMALIZATION_D)
1047 add_converter(sc, archive_string_normalize_D);
1054 * Try conversion in the best effort or no conversion.
1056 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1057 add_converter(sc, best_effort_strncat_in_locale);
1059 /* Make sure we have no converter. */
1064 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1065 * and CP932 which are referenced in create_sconv_object().
1068 canonical_charset_name(const char *charset)
1074 if (charset == NULL || charset[0] == '\0'
1075 || strlen(charset) > 15)
1078 /* Copy name to uppercase. */
1083 if (c >= 'a' && c <= 'z')
1089 if (strcmp(cs, "UTF-8") == 0 ||
1090 strcmp(cs, "UTF8") == 0)
1092 if (strcmp(cs, "UTF-16BE") == 0 ||
1093 strcmp(cs, "UTF16BE") == 0)
1094 return ("UTF-16BE");
1095 if (strcmp(cs, "UTF-16LE") == 0 ||
1096 strcmp(cs, "UTF16LE") == 0)
1097 return ("UTF-16LE");
1098 if (strcmp(cs, "CP932") == 0)
1104 * Create a string conversion object.
1106 static struct archive_string_conv *
1107 create_sconv_object(const char *fc, const char *tc,
1108 unsigned current_codepage, int flag)
1110 struct archive_string_conv *sc;
1112 sc = calloc(1, sizeof(*sc));
1116 sc->from_charset = strdup(fc);
1117 if (sc->from_charset == NULL) {
1121 sc->to_charset = strdup(tc);
1122 if (sc->to_charset == NULL) {
1123 free(sc->from_charset);
1127 archive_string_init(&sc->utftmp);
1129 if (flag & SCONV_TO_CHARSET) {
1131 * Convert characters from the current locale charset to
1132 * a specified charset.
1134 sc->from_cp = current_codepage;
1135 sc->to_cp = make_codepage_from_charset(tc);
1136 #if defined(_WIN32) && !defined(__CYGWIN__)
1137 if (IsValidCodePage(sc->to_cp))
1138 flag |= SCONV_WIN_CP;
1140 } else if (flag & SCONV_FROM_CHARSET) {
1142 * Convert characters from a specified charset to
1143 * the current locale charset.
1145 sc->to_cp = current_codepage;
1146 sc->from_cp = make_codepage_from_charset(fc);
1147 #if defined(_WIN32) && !defined(__CYGWIN__)
1148 if (IsValidCodePage(sc->from_cp))
1149 flag |= SCONV_WIN_CP;
1154 * Check if "from charset" and "to charset" are the same.
1156 if (strcmp(fc, tc) == 0 ||
1157 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1163 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1165 if (strcmp(tc, "UTF-8") == 0)
1166 flag |= SCONV_TO_UTF8;
1167 else if (strcmp(tc, "UTF-16BE") == 0)
1168 flag |= SCONV_TO_UTF16BE;
1169 else if (strcmp(tc, "UTF-16LE") == 0)
1170 flag |= SCONV_TO_UTF16LE;
1171 if (strcmp(fc, "UTF-8") == 0)
1172 flag |= SCONV_FROM_UTF8;
1173 else if (strcmp(fc, "UTF-16BE") == 0)
1174 flag |= SCONV_FROM_UTF16BE;
1175 else if (strcmp(fc, "UTF-16LE") == 0)
1176 flag |= SCONV_FROM_UTF16LE;
1177 #if defined(_WIN32) && !defined(__CYGWIN__)
1178 if (sc->to_cp == CP_UTF8)
1179 flag |= SCONV_TO_UTF8;
1180 else if (sc->to_cp == CP_UTF16BE)
1181 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1182 else if (sc->to_cp == CP_UTF16LE)
1183 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1184 if (sc->from_cp == CP_UTF8)
1185 flag |= SCONV_FROM_UTF8;
1186 else if (sc->from_cp == CP_UTF16BE)
1187 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1188 else if (sc->from_cp == CP_UTF16LE)
1189 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1193 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1194 * handle it. So we have to translate NFD characters to NFC ones
1195 * ourselves before iconv handles. Another reason is to prevent
1196 * that the same sight of two filenames, one is NFC and other
1197 * is NFD, would be in its directory.
1198 * On Mac OS X, although its filesystem layer automatically
1199 * convert filenames to NFD, it would be useful for filename
1200 * comparing to find out the same filenames that we normalize
1201 * that to be NFD ourselves.
1203 if ((flag & SCONV_FROM_CHARSET) &&
1204 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1205 #if defined(__APPLE__)
1206 if (flag & SCONV_TO_UTF8)
1207 flag |= SCONV_NORMALIZATION_D;
1210 flag |= SCONV_NORMALIZATION_C;
1212 #if defined(__APPLE__)
1214 * In case writing an archive file, make sure that a filename
1215 * going to be passed to iconv is a Unicode NFC string since
1216 * a filename in HFS Plus filesystem is a Unicode NFD one and
1217 * iconv cannot handle it with "UTF-8" charset. It is simpler
1218 * than a use of "UTF-8-MAC" charset.
1220 if ((flag & SCONV_TO_CHARSET) &&
1221 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1222 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1223 flag |= SCONV_NORMALIZATION_C;
1225 * In case reading an archive file. make sure that a filename
1226 * will be passed to users is a Unicode NFD string in order to
1227 * correctly compare the filename with other one which comes
1228 * from HFS Plus filesystem.
1230 if ((flag & SCONV_FROM_CHARSET) &&
1231 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1232 (flag & SCONV_TO_UTF8))
1233 flag |= SCONV_NORMALIZATION_D;
1236 #if defined(HAVE_ICONV)
1237 sc->cd_w = (iconv_t)-1;
1239 * Create an iconv object.
1241 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1242 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1243 (flag & SCONV_WIN_CP)) {
1244 /* This case we won't use iconv. */
1245 sc->cd = (iconv_t)-1;
1247 sc->cd = iconv_open(tc, fc);
1248 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1250 * Unfortunaly, all of iconv implements do support
1251 * "CP932" character-set, so we should use "SJIS"
1252 * instead if iconv_open failed.
1254 if (strcmp(tc, "CP932") == 0)
1255 sc->cd = iconv_open("SJIS", fc);
1256 else if (strcmp(fc, "CP932") == 0)
1257 sc->cd = iconv_open(tc, "SJIS");
1259 #if defined(_WIN32) && !defined(__CYGWIN__)
1261 * archive_mstring on Windows directly convert multi-bytes
1262 * into archive_wstring in order not to depend on locale
1263 * so that you can do a I18N programing. This will be
1264 * used only in archive_mstring_copy_mbs_len_l so far.
1266 if (flag & SCONV_FROM_CHARSET) {
1267 sc->cd_w = iconv_open("UTF-8", fc);
1268 if (sc->cd_w == (iconv_t)-1 &&
1269 (sc->flag & SCONV_BEST_EFFORT)) {
1270 if (strcmp(fc, "CP932") == 0)
1271 sc->cd_w = iconv_open("UTF-8", "SJIS");
1274 #endif /* _WIN32 && !__CYGWIN__ */
1276 #endif /* HAVE_ICONV */
1281 * Set up converters.
1283 setup_converter(sc);
1289 * Free a string conversion object.
1292 free_sconv_object(struct archive_string_conv *sc)
1294 free(sc->from_charset);
1295 free(sc->to_charset);
1296 archive_string_free(&sc->utftmp);
1298 if (sc->cd != (iconv_t)-1)
1299 iconv_close(sc->cd);
1300 if (sc->cd_w != (iconv_t)-1)
1301 iconv_close(sc->cd_w);
1306 #if defined(_WIN32) && !defined(__CYGWIN__)
1308 my_atoi(const char *p)
1314 if (*p >= '0' && *p <= '9')
1315 cp = cp * 10 + (*p - '0');
1324 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1325 * Return -1 if failed.
1327 * Note: This translation code may be insufficient.
1329 static struct charset {
1333 /* MUST BE SORTED! */
1352 {"HZ-GB-2312", 52936},
1370 {"ISO-8859-1", 28591},
1371 {"ISO-8859-13", 28603},
1372 {"ISO-8859-15", 28605},
1373 {"ISO-8859-2", 28592},
1374 {"ISO-8859-3", 28593},
1375 {"ISO-8859-4", 28594},
1376 {"ISO-8859-5", 28595},
1377 {"ISO-8859-6", 28596},
1378 {"ISO-8859-7", 28597},
1379 {"ISO-8859-8", 28598},
1380 {"ISO-8859-9", 28599},
1381 {"ISO8859-1", 28591},
1382 {"ISO8859-13", 28603},
1383 {"ISO8859-15", 28605},
1384 {"ISO8859-2", 28592},
1385 {"ISO8859-3", 28593},
1386 {"ISO8859-4", 28594},
1387 {"ISO8859-5", 28595},
1388 {"ISO8859-6", 28596},
1389 {"ISO8859-7", 28597},
1390 {"ISO8859-8", 28598},
1391 {"ISO8859-9", 28599},
1395 {"KS_C_5601-1987", 949},
1398 {"MACINTOSH", 10000},
1408 {"X-EUROPA", 29001},
1409 {"X-MAC-ARABIC", 10004},
1410 {"X-MAC-CE", 10029},
1411 {"X-MAC-CHINESEIMP", 10008},
1412 {"X-MAC-CHINESETRAD", 10002},
1413 {"X-MAC-CROATIAN", 10082},
1414 {"X-MAC-CYRILLIC", 10007},
1415 {"X-MAC-GREEK", 10006},
1416 {"X-MAC-HEBREW", 10005},
1417 {"X-MAC-ICELANDIC", 10079},
1418 {"X-MAC-JAPANESE", 10001},
1419 {"X-MAC-KOREAN", 10003},
1420 {"X-MAC-ROMANIAN", 10010},
1421 {"X-MAC-THAI", 10021},
1422 {"X-MAC-TURKISH", 10081},
1423 {"X-MAC-UKRAINIAN", 10017},
1426 make_codepage_from_charset(const char *charset)
1433 if (charset == NULL || strlen(charset) > 15)
1436 /* Copy name to uppercase. */
1439 char c = *charset++;
1440 if (c >= 'a' && c <= 'z')
1447 /* Look it up in the table first, so that we can easily
1448 * override CP367, which we map to 1252 instead of 367. */
1450 b = sizeof(charsets)/sizeof(charsets[0]);
1452 int c = (b + a) / 2;
1453 int r = strcmp(charsets[c].name, cs);
1459 return charsets[c].cp;
1462 /* If it's not in the table, try to parse it. */
1465 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1466 cp = my_atoi(cs + 2);
1467 } else if (strcmp(cs, "CP_ACP") == 0)
1468 cp = get_current_codepage();
1469 else if (strcmp(cs, "CP_OEMCP") == 0)
1470 cp = get_current_oemcp();
1473 if (cs[1] == 'B' && cs[2] == 'M' &&
1474 cs[3] >= '0' && cs[3] <= '9') {
1475 cp = my_atoi(cs + 3);
1479 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1480 cp = my_atoi(cs + 8);
1481 if (cp != 874 && (cp < 1250 || cp > 1258))
1482 cp = -1;/* This may invalid code. */
1490 * Return ANSI Code Page of current locale set by setlocale().
1493 get_current_codepage(void)
1498 locale = setlocale(LC_CTYPE, NULL);
1501 if (locale[0] == 'C' && locale[1] == '\0')
1502 return (CP_C_LOCALE);
1503 p = strrchr(locale, '.');
1513 * Translation table between Locale Name and ACP/OEMCP.
1520 { 950, 950, "Chinese_Taiwan" },
1521 { 936, 936, "Chinese_People's Republic of China" },
1522 { 950, 950, "Chinese_Taiwan" },
1523 { 1250, 852, "Czech_Czech Republic" },
1524 { 1252, 850, "Danish_Denmark" },
1525 { 1252, 850, "Dutch_Netherlands" },
1526 { 1252, 850, "Dutch_Belgium" },
1527 { 1252, 437, "English_United States" },
1528 { 1252, 850, "English_Australia" },
1529 { 1252, 850, "English_Canada" },
1530 { 1252, 850, "English_New Zealand" },
1531 { 1252, 850, "English_United Kingdom" },
1532 { 1252, 437, "English_United States" },
1533 { 1252, 850, "Finnish_Finland" },
1534 { 1252, 850, "French_France" },
1535 { 1252, 850, "French_Belgium" },
1536 { 1252, 850, "French_Canada" },
1537 { 1252, 850, "French_Switzerland" },
1538 { 1252, 850, "German_Germany" },
1539 { 1252, 850, "German_Austria" },
1540 { 1252, 850, "German_Switzerland" },
1541 { 1253, 737, "Greek_Greece" },
1542 { 1250, 852, "Hungarian_Hungary" },
1543 { 1252, 850, "Icelandic_Iceland" },
1544 { 1252, 850, "Italian_Italy" },
1545 { 1252, 850, "Italian_Switzerland" },
1546 { 932, 932, "Japanese_Japan" },
1547 { 949, 949, "Korean_Korea" },
1548 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1549 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1550 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1551 { 1250, 852, "Polish_Poland" },
1552 { 1252, 850, "Portuguese_Portugal" },
1553 { 1252, 850, "Portuguese_Brazil" },
1554 { 1251, 866, "Russian_Russia" },
1555 { 1250, 852, "Slovak_Slovakia" },
1556 { 1252, 850, "Spanish_Spain" },
1557 { 1252, 850, "Spanish_Mexico" },
1558 { 1252, 850, "Spanish_Spain" },
1559 { 1252, 850, "Swedish_Sweden" },
1560 { 1254, 857, "Turkish_Turkey" },
1565 * Return OEM Code Page of current locale set by setlocale().
1568 get_current_oemcp(void)
1574 locale = setlocale(LC_CTYPE, NULL);
1576 return (GetOEMCP());
1577 if (locale[0] == 'C' && locale[1] == '\0')
1578 return (CP_C_LOCALE);
1580 p = strrchr(locale, '.');
1582 return (GetOEMCP());
1584 for (i = 0; acp_ocp_map[i].acp; i++) {
1585 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1586 return (acp_ocp_map[i].ocp);
1588 return (GetOEMCP());
1593 * POSIX platform does not use CodePage.
1597 get_current_codepage(void)
1599 return (-1);/* Unknown */
1602 make_codepage_from_charset(const char *charset)
1604 (void)charset; /* UNUSED */
1605 return (-1);/* Unknown */
1608 get_current_oemcp(void)
1610 return (-1);/* Unknown */
1613 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1616 * Return a string conversion object.
1618 static struct archive_string_conv *
1619 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1621 struct archive_string_conv *sc;
1622 unsigned current_codepage;
1624 /* Check if we have made the sconv object. */
1625 sc = find_sconv_object(a, fc, tc);
1630 current_codepage = get_current_codepage();
1632 current_codepage = a->current_codepage;
1634 sc = create_sconv_object(canonical_charset_name(fc),
1635 canonical_charset_name(tc), current_codepage, flag);
1638 archive_set_error(a, ENOMEM,
1639 "Could not allocate memory for "
1640 "a string conversion object");
1645 * If there is no converter for current string conversion object,
1646 * we cannot handle this conversion.
1648 if (sc->nconverter == 0) {
1651 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1652 "iconv_open failed : Cannot handle ``%s''",
1653 (flag & SCONV_TO_CHARSET)?tc:fc);
1655 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1656 "A character-set conversion not fully supported "
1657 "on this platform");
1660 /* Failed; free a sconv object. */
1661 free_sconv_object(sc);
1669 add_sconv_object(a, sc);
1674 get_current_charset(struct archive *a)
1676 const char *cur_charset;
1679 cur_charset = default_iconv_charset("");
1681 cur_charset = default_iconv_charset(a->current_code);
1682 if (a->current_code == NULL) {
1683 a->current_code = strdup(cur_charset);
1684 a->current_codepage = get_current_codepage();
1685 a->current_oemcp = get_current_oemcp();
1688 return (cur_charset);
1692 * Make and Return a string conversion object.
1693 * Return NULL if the platform does not support the specified conversion
1694 * and best_effort is 0.
1695 * If best_effort is set, A string conversion object must be returned
1696 * unless memory allocation for the object fails, but the conversion
1697 * might fail when non-ASCII code is found.
1699 struct archive_string_conv *
1700 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1703 int flag = SCONV_TO_CHARSET;
1706 flag |= SCONV_BEST_EFFORT;
1707 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1710 struct archive_string_conv *
1711 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1714 int flag = SCONV_FROM_CHARSET;
1717 flag |= SCONV_BEST_EFFORT;
1718 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1722 * archive_string_default_conversion_*_archive() are provided for Windows
1723 * platform because other archiver application use CP_OEMCP for
1724 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1725 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1726 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1727 * So we should make a string conversion between CP_ACP and CP_OEMCP
1728 * for compatibillty.
1730 #if defined(_WIN32) && !defined(__CYGWIN__)
1731 struct archive_string_conv *
1732 archive_string_default_conversion_for_read(struct archive *a)
1734 const char *cur_charset = get_current_charset(a);
1737 /* NOTE: a check of cur_charset is unneeded but we need
1738 * that get_current_charset() has been surely called at
1739 * this time whatever C compiler optimized. */
1740 if (cur_charset != NULL &&
1741 (a->current_codepage == CP_C_LOCALE ||
1742 a->current_codepage == a->current_oemcp))
1743 return (NULL);/* no conversion. */
1745 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1746 /* Make sure a null termination must be set. */
1747 oemcp[sizeof(oemcp)-1] = '\0';
1748 return (get_sconv_object(a, oemcp, cur_charset,
1749 SCONV_FROM_CHARSET));
1752 struct archive_string_conv *
1753 archive_string_default_conversion_for_write(struct archive *a)
1755 const char *cur_charset = get_current_charset(a);
1758 /* NOTE: a check of cur_charset is unneeded but we need
1759 * that get_current_charset() has been surely called at
1760 * this time whatever C compiler optimized. */
1761 if (cur_charset != NULL &&
1762 (a->current_codepage == CP_C_LOCALE ||
1763 a->current_codepage == a->current_oemcp))
1764 return (NULL);/* no conversion. */
1766 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1767 /* Make sure a null termination must be set. */
1768 oemcp[sizeof(oemcp)-1] = '\0';
1769 return (get_sconv_object(a, cur_charset, oemcp,
1773 struct archive_string_conv *
1774 archive_string_default_conversion_for_read(struct archive *a)
1776 (void)a; /* UNUSED */
1780 struct archive_string_conv *
1781 archive_string_default_conversion_for_write(struct archive *a)
1783 (void)a; /* UNUSED */
1789 * Dispose of all character conversion objects in the archive object.
1792 archive_string_conversion_free(struct archive *a)
1794 struct archive_string_conv *sc;
1795 struct archive_string_conv *sc_next;
1797 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1799 free_sconv_object(sc);
1802 free(a->current_code);
1803 a->current_code = NULL;
1807 * Return a conversion charset name.
1810 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1812 if (sc->flag & SCONV_TO_CHARSET)
1813 return (sc->to_charset);
1815 return (sc->from_charset);
1819 * Change the behavior of a string conversion.
1822 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1826 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1827 * assumption that wchar_t was Unicode.
1828 * This option enables simulating the assumption in order to read
1829 * that filname correctly.
1831 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1832 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1833 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1835 * Nothing to do for it since wchar_t on these platforms
1836 * is really Unicode.
1838 (void)sc; /* UNUSED */
1840 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1841 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1842 /* Set up string converters. */
1843 setup_converter(sc);
1847 case SCONV_SET_OPT_NORMALIZATION_C:
1848 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1849 sc->flag |= SCONV_NORMALIZATION_C;
1850 sc->flag &= ~SCONV_NORMALIZATION_D;
1851 /* Set up string converters. */
1852 setup_converter(sc);
1855 case SCONV_SET_OPT_NORMALIZATION_D:
1856 #if defined(HAVE_ICONV)
1858 * If iconv will take the string, do not change the
1859 * setting of the normalization.
1861 if (!(sc->flag & SCONV_WIN_CP) &&
1862 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1863 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1866 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1867 sc->flag |= SCONV_NORMALIZATION_D;
1868 sc->flag &= ~SCONV_NORMALIZATION_C;
1869 /* Set up string converters. */
1870 setup_converter(sc);
1880 * Copy one archive_string to another in locale conversion.
1882 * archive_strncat_l();
1883 * archive_strncpy_l();
1888 mbsnbytes(const void *_p, size_t n)
1895 p = (const char *)_p;
1897 /* Like strlen(p), except won't examine positions beyond p[n]. */
1900 while (s < n && *pp) {
1908 utf16nbytes(const void *_p, size_t n)
1915 p = (const char *)_p;
1917 /* Like strlen(p), except won't examine positions beyond p[n]. */
1921 while (s < n && (pp[0] || pp[1])) {
1929 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1930 struct archive_string_conv *sc)
1933 return (archive_strncat_l(as, _p, n, sc));
1937 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1938 struct archive_string_conv *sc)
1944 /* We must allocate memory even if there is no data for conversion
1945 * or copy. This simulates archive_string_append behavior. */
1946 if (_p == NULL || n == 0) {
1948 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1950 if (archive_string_ensure(as, as->length + tn) == NULL)
1952 as->s[as->length] = 0;
1954 as->s[as->length+1] = 0;
1959 * If sc is NULL, we just make a copy.
1962 length = mbsnbytes(_p, n);
1963 if (archive_string_append(as, _p, length) == NULL)
1964 return (-1);/* No memory */
1968 if (sc->flag & SCONV_FROM_UTF16)
1969 length = utf16nbytes(_p, n);
1971 length = mbsnbytes(_p, n);
1974 if (sc->nconverter > 1) {
1975 sc->utftmp.length = 0;
1976 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
1977 if (r2 != 0 && errno == ENOMEM)
1982 length = sc->utftmp.length;
1985 r2 = sc->converter[i](as, s, length, sc);
1994 * Return -1 if conversion failes.
1997 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
1998 size_t length, struct archive_string_conv *sc)
2000 ICONV_CONST char *itp;
2005 int return_value = 0; /* success */
2006 int to_size, from_size;
2008 if (sc->flag & SCONV_TO_UTF16)
2012 if (sc->flag & SCONV_FROM_UTF16)
2017 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2021 itp = (char *)(uintptr_t)_p;
2023 outp = as->s + as->length;
2024 avail = as->buffer_length - as->length - to_size;
2025 while (remaining >= (size_t)from_size) {
2026 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2028 if (result != (size_t)-1)
2029 break; /* Conversion completed. */
2031 if (errno == EILSEQ || errno == EINVAL) {
2033 * If an output charset is UTF-8 or UTF-16BE/LE,
2034 * unknown character should be U+FFFD
2035 * (replacement character).
2037 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2039 if (sc->flag & SCONV_TO_UTF8)
2040 rbytes = UTF8_R_CHAR_SIZE;
2044 if (avail < rbytes) {
2045 as->length = outp - as->s;
2046 bs = as->buffer_length +
2047 (remaining * to_size) + rbytes;
2049 archive_string_ensure(as, bs))
2051 outp = as->s + as->length;
2052 avail = as->buffer_length
2053 - as->length - to_size;
2055 if (sc->flag & SCONV_TO_UTF8)
2056 UTF8_SET_R_CHAR(outp);
2057 else if (sc->flag & SCONV_TO_UTF16BE)
2058 archive_be16enc(outp, UNICODE_R_CHAR);
2060 archive_le16enc(outp, UNICODE_R_CHAR);
2064 /* Skip the illegal input bytes. */
2069 remaining -= from_size;
2070 return_value = -1; /* failure */
2072 /* E2BIG no output buffer,
2073 * Increase an output buffer. */
2074 as->length = outp - as->s;
2075 bs = as->buffer_length + remaining * 2;
2076 if (NULL == archive_string_ensure(as, bs))
2078 outp = as->s + as->length;
2079 avail = as->buffer_length - as->length - to_size;
2082 as->length = outp - as->s;
2083 as->s[as->length] = 0;
2085 as->s[as->length+1] = 0;
2086 return (return_value);
2089 #endif /* HAVE_ICONV */
2092 #if defined(_WIN32) && !defined(__CYGWIN__)
2095 * Translate a string from a some CodePage to an another CodePage by
2096 * Windows APIs, and copy the result. Return -1 if conversion failes.
2099 strncat_in_codepage(struct archive_string *as,
2100 const void *_p, size_t length, struct archive_string_conv *sc)
2102 const char *s = (const char *)_p;
2103 struct archive_wstring aws;
2107 archive_string_init(&aws);
2108 saved_flag = sc->flag;
2109 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2110 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2111 sc->flag = saved_flag;
2113 archive_wstring_free(&aws);
2114 if (errno != ENOMEM)
2115 archive_string_append(as, s, length);
2120 r = archive_string_append_from_wcs_in_codepage(
2121 as, aws.s, aws.length, sc);
2122 if (r != 0 && errno != ENOMEM && l == as->length)
2123 archive_string_append(as, s, length);
2124 archive_wstring_free(&aws);
2129 * Test whether MBS ==> WCS is okay.
2132 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2134 const char *p = (const char *)_p;
2136 DWORD mbflag = MB_ERR_INVALID_CHARS;
2138 if (sc->flag & SCONV_FROM_CHARSET)
2139 codepage = sc->to_cp;
2141 codepage = sc->from_cp;
2143 if (codepage == CP_C_LOCALE)
2145 if (codepage != CP_UTF8)
2146 mbflag |= MB_PRECOMPOSED;
2148 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2149 return (-1); /* Invalid */
2150 return (0); /* Okay */
2156 * Test whether MBS ==> WCS is okay.
2159 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2161 const char *p = (const char *)_p;
2165 mbstate_t shift_state;
2167 memset(&shift_state, 0, sizeof(shift_state));
2169 /* Clear the shift state before starting. */
2170 mbtowc(NULL, NULL, 0);
2176 r = mbrtowc(&wc, p, n, &shift_state);
2178 r = mbtowc(&wc, p, n);
2180 if (r == (size_t)-1 || r == (size_t)-2)
2181 return (-1);/* Invalid. */
2187 (void)sc; /* UNUSED */
2188 return (0); /* All Okey. */
2191 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2194 * Basically returns -1 because we cannot make a conversion of charset
2195 * without iconv but in some cases this would return 0.
2196 * Returns 0 if all copied characters are ASCII.
2197 * Returns 0 if both from-locale and to-locale are the same and those
2198 * can be WCS with no error.
2201 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2202 size_t length, struct archive_string_conv *sc)
2208 int return_value = 0; /* success */
2211 * If both from-locale and to-locale is the same, this makes a copy.
2212 * And then this checks all copied MBS can be WCS if so returns 0.
2215 if (archive_string_append(as, _p, length) == NULL)
2216 return (-1);/* No memory */
2217 return (invalid_mbs(_p, length, sc));
2221 * If a character is ASCII, this just copies it. If not, this
2222 * assigns '?' charater instead but in UTF-8 locale this assigns
2223 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2224 * a Replacement Character in Unicode.
2226 if (archive_string_ensure(as, as->length + length + 1) == NULL)
2230 itp = (const uint8_t *)_p;
2231 otp = as->s + as->length;
2232 avail = as->buffer_length - as->length -1;
2233 while (*itp && remaining > 0) {
2234 if (*itp > 127 && (sc->flag & SCONV_TO_UTF8)) {
2235 if (avail < UTF8_R_CHAR_SIZE) {
2236 as->length = otp - as->s;
2237 if (NULL == archive_string_ensure(as,
2238 as->buffer_length + remaining +
2241 otp = as->s + as->length;
2242 avail = as->buffer_length - as->length -1;
2245 * When coping a string in UTF-8, unknown character
2246 * should be U+FFFD (replacement character).
2248 UTF8_SET_R_CHAR(otp);
2249 otp += UTF8_R_CHAR_SIZE;
2250 avail -= UTF8_R_CHAR_SIZE;
2254 } else if (*itp > 127) {
2260 *otp++ = (char)*itp++;
2264 as->length = otp - as->s;
2265 as->s[as->length] = '\0';
2266 return (return_value);
2271 * Unicode conversion functions.
2272 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2273 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2274 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2275 * - UTF-16BE <===> UTF-8.
2280 * Utility to convert a single UTF-8 sequence.
2282 * Usually return used bytes, return used byte in negative value when
2283 * a unicode character is replaced with U+FFFD.
2284 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2285 * Recommended Practice for Replacement Characters.
2288 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2290 static const char utf8_count[256] = {
2291 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2292 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2293 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2294 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2295 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2296 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2299 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2300 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2301 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2302 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2303 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2304 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2305 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2306 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2316 * Decode 1-4 bytes depending on the value of the first byte.
2318 ch = (unsigned char)*s;
2320 return (0); /* Standard: return 0 for end-of-string. */
2321 cnt = utf8_count[ch];
2323 /* Invalide sequence or there are not plenty bytes. */
2326 for (i = 1; i < cnt; i++) {
2327 if ((s[i] & 0xc0) != 0x80) {
2332 goto invalid_sequence;
2335 /* Make a Unicode code point from a single UTF-8 sequence. */
2337 case 1: /* 1 byte sequence. */
2340 case 2: /* 2 bytes sequence. */
2341 if ((s[1] & 0xc0) != 0x80) {
2343 goto invalid_sequence;
2345 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2347 case 3: /* 3 bytes sequence. */
2348 if ((s[1] & 0xc0) != 0x80) {
2350 goto invalid_sequence;
2352 if ((s[2] & 0xc0) != 0x80) {
2354 goto invalid_sequence;
2356 wc = ((ch & 0x0f) << 12)
2357 | ((s[1] & 0x3f) << 6)
2360 goto invalid_sequence;/* Overlong sequence. */
2362 case 4: /* 4 bytes sequence. */
2363 if ((s[1] & 0xc0) != 0x80) {
2365 goto invalid_sequence;
2367 if ((s[2] & 0xc0) != 0x80) {
2369 goto invalid_sequence;
2371 if ((s[3] & 0xc0) != 0x80) {
2373 goto invalid_sequence;
2375 wc = ((ch & 0x07) << 18)
2376 | ((s[1] & 0x3f) << 12)
2377 | ((s[2] & 0x3f) << 6)
2380 goto invalid_sequence;/* Overlong sequence. */
2382 default: /* Others are all invalid sequence. */
2383 if (ch == 0xc0 || ch == 0xc1)
2385 else if (ch >= 0xf5 && ch <= 0xf7)
2387 else if (ch >= 0xf8 && ch <= 0xfb)
2389 else if (ch == 0xfc || ch == 0xfd)
2395 for (i = 1; i < cnt; i++) {
2396 if ((s[i] & 0xc0) != 0x80) {
2401 goto invalid_sequence;
2404 /* The code point larger than 0x10FFFF is not leagal
2405 * Unicode values. */
2406 if (wc > UNICODE_MAX)
2407 goto invalid_sequence;
2408 /* Correctly gets a Unicode, returns used bytes. */
2412 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2417 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2421 cnt = _utf8_to_unicode(pwc, s, n);
2422 /* Any of Surrogate pair is not leagal Unicode values. */
2423 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2428 static inline uint32_t
2429 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2439 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2440 * removing surrogate pairs.
2442 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2444 * Usually return used bytes, return used byte in negative value when
2445 * a unicode character is replaced with U+FFFD.
2448 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2453 cnt = _utf8_to_unicode(&wc, s, n);
2454 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2457 /* Invalid byte sequence. */
2458 goto invalid_sequence;
2460 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2461 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2462 /* Invalid byte sequence. */
2463 goto invalid_sequence;
2465 wc = combine_surrogate_pair(wc, wc2);
2467 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2468 /* Invalid byte sequence. */
2469 goto invalid_sequence;
2474 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2481 * Convert a Unicode code point to a single UTF-8 sequence.
2483 * NOTE:This function does not check if the Unicode is leagal or not.
2484 * Please you definitely check it before calling this.
2487 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2491 /* Translate code point to UTF8 */
2496 } else if (uc <= 0x7ff) {
2499 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2500 *p++ = 0x80 | (uc & 0x3f);
2501 } else if (uc <= 0xffff) {
2504 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2505 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2506 *p++ = 0x80 | (uc & 0x3f);
2507 } else if (uc <= UNICODE_MAX) {
2510 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2511 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2512 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2513 *p++ = 0x80 | (uc & 0x3f);
2516 * Undescribed code point should be U+FFFD
2517 * (replacement character).
2519 if (remaining < UTF8_R_CHAR_SIZE)
2522 p += UTF8_R_CHAR_SIZE;
2528 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2530 return (utf16_to_unicode(pwc, s, n, 1));
2534 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2536 return (utf16_to_unicode(pwc, s, n, 0));
2540 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2542 const char *utf16 = s;
2548 /* set the Replacement Character instead. */
2549 *pwc = UNICODE_R_CHAR;
2554 uc = archive_be16dec(utf16);
2556 uc = archive_le16dec(utf16);
2559 /* If this is a surrogate pair, assemble the full code point.*/
2560 if (IS_HIGH_SURROGATE_LA(uc)) {
2565 uc2 = archive_be16dec(utf16);
2567 uc2 = archive_le16dec(utf16);
2570 if (IS_LOW_SURROGATE_LA(uc2)) {
2571 uc = combine_surrogate_pair(uc, uc2);
2574 /* Undescribed code point should be U+FFFD
2575 * (replacement character). */
2576 *pwc = UNICODE_R_CHAR;
2582 * Surrogate pair values(0xd800 through 0xdfff) are only
2583 * used by UTF-16, so, after above culculation, the code
2584 * must not be surrogate values, and Unicode has no codes
2585 * larger than 0x10ffff. Thus, those are not leagal Unicode
2588 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2589 /* Undescribed code point should be U+FFFD
2590 * (replacement character). */
2591 *pwc = UNICODE_R_CHAR;
2592 return (((int)(utf16 - s)) * -1);
2595 return ((int)(utf16 - s));
2599 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2604 /* We have a code point that won't fit into a
2605 * wchar_t; convert it to a surrogate pair. */
2609 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2610 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2615 archive_be16enc(utf16, uc);
2621 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2626 /* We have a code point that won't fit into a
2627 * wchar_t; convert it to a surrogate pair. */
2631 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2632 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2637 archive_le16enc(utf16, uc);
2643 * Copy UTF-8 string in checking surrogate pair.
2644 * If any surrogate pair are found, it would be canonicalized.
2647 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p,
2648 size_t len, struct archive_string_conv *sc)
2654 (void)sc; /* UNUSED */
2656 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2659 s = (const char *)_p;
2660 p = as->s + as->length;
2661 endp = as->s + as->buffer_length -1;
2668 * Forward byte sequence until a conversion of that is needed.
2670 while ((n = utf8_to_unicode(&uc, s, len)) > 0) {
2675 if (p + (s - ss) > endp) {
2676 as->length = p - as->s;
2677 if (archive_string_ensure(as,
2678 as->buffer_length + len + 1) == NULL)
2680 p = as->s + as->length;
2681 endp = as->s + as->buffer_length -1;
2684 memcpy(p, ss, s - ss);
2689 * If n is negative, current byte sequence needs a replacement.
2692 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2693 /* Current byte sequence may be CESU-8. */
2694 n = cesu8_to_unicode(&uc, s, len);
2698 n *= -1;/* Use a replaced unicode character. */
2701 /* Rebuild UTF-8 byte sequence. */
2702 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) {
2703 as->length = p - as->s;
2704 if (archive_string_ensure(as,
2705 as->buffer_length + len + 1) == NULL)
2707 p = as->s + as->length;
2708 endp = as->s + as->buffer_length -1;
2715 as->length = p - as->s;
2716 as->s[as->length] = '\0';
2721 archive_string_append_unicode(struct archive_string *as, const void *_p,
2722 size_t len, struct archive_string_conv *sc)
2728 int n, ret = 0, ts, tm;
2729 int (*parse)(uint32_t *, const char *, size_t);
2730 size_t (*unparse)(char *, size_t, uint32_t);
2732 if (sc->flag & SCONV_TO_UTF16BE) {
2733 unparse = unicode_to_utf16be;
2735 } else if (sc->flag & SCONV_TO_UTF16LE) {
2736 unparse = unicode_to_utf16le;
2738 } else if (sc->flag & SCONV_TO_UTF8) {
2739 unparse = unicode_to_utf8;
2743 * This case is going to be converted to another
2744 * character-set through iconv.
2746 if (sc->flag & SCONV_FROM_UTF16BE) {
2747 unparse = unicode_to_utf16be;
2749 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2750 unparse = unicode_to_utf16le;
2753 unparse = unicode_to_utf8;
2758 if (sc->flag & SCONV_FROM_UTF16BE) {
2759 parse = utf16be_to_unicode;
2761 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2762 parse = utf16le_to_unicode;
2765 parse = cesu8_to_unicode;
2769 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2772 s = (const char *)_p;
2773 p = as->s + as->length;
2774 endp = as->s + as->buffer_length - ts;
2775 while ((n = parse(&uc, s, len)) != 0) {
2777 /* Use a replaced unicode character. */
2783 while ((w = unparse(p, endp - p, uc)) == 0) {
2784 /* There is not enough output buffer so
2785 * we have to expand it. */
2786 as->length = p - as->s;
2787 if (archive_string_ensure(as,
2788 as->buffer_length + len * tm + ts) == NULL)
2790 p = as->s + as->length;
2791 endp = as->s + as->buffer_length - ts;
2795 as->length = p - as->s;
2796 as->s[as->length] = '\0';
2798 as->s[as->length+1] = '\0';
2803 * Following Constants for Hangul compositions this information comes from
2804 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2806 #define HC_SBASE 0xAC00
2807 #define HC_LBASE 0x1100
2808 #define HC_VBASE 0x1161
2809 #define HC_TBASE 0x11A7
2810 #define HC_LCOUNT 19
2811 #define HC_VCOUNT 21
2812 #define HC_TCOUNT 28
2813 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2814 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2817 get_nfc(uint32_t uc, uint32_t uc2)
2822 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2824 int m = (t + b) / 2;
2825 if (u_composition_table[m].cp1 < uc)
2827 else if (u_composition_table[m].cp1 > uc)
2829 else if (u_composition_table[m].cp2 < uc2)
2831 else if (u_composition_table[m].cp2 > uc2)
2834 return (u_composition_table[m].nfc);
2839 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2843 * Update first code point.
2845 #define UPDATE_UC(new_uc) do { \
2851 * Replace first code point with second code point.
2853 #define REPLACE_UC_WITH_UC2() do { \
2859 #define EXPAND_BUFFER() do { \
2860 as->length = p - as->s; \
2861 if (archive_string_ensure(as, \
2862 as->buffer_length + len * tm + ts) == NULL)\
2864 p = as->s + as->length; \
2865 endp = as->s + as->buffer_length - ts; \
2868 #define UNPARSE(p, endp, uc) do { \
2869 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2876 * Write first code point.
2877 * If the code point has not be changed from its original code,
2878 * this just copies it from its original buffer pointer.
2879 * If not, this converts it to UTF-8 byte sequence and copies it.
2881 #define WRITE_UC() do { \
2888 /* FALL THROUGH */ \
2891 /* FALL THROUGH */ \
2894 /* FALL THROUGH */ \
2901 UNPARSE(p, endp, uc); \
2906 * Collect following decomposable code points.
2908 #define COLLECT_CPS(start) do { \
2910 for (_i = start; _i < FDC_MAX ; _i++) { \
2911 nx = parse(&ucx[_i], s, len); \
2914 cx = CCC(ucx[_i]); \
2915 if (cl >= cx && cl != 228 && cx != 228)\
2922 if (_i >= FDC_MAX) { \
2924 ucx_size = FDC_MAX; \
2930 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2932 * TODO: Convert composition exclusions,which are never converted
2933 * from NFC,NFD,NFKC and NFKD, to Form C.
2936 archive_string_normalize_C(struct archive_string *as, const void *_p,
2937 size_t len, struct archive_string_conv *sc)
2939 const char *s = (const char *)_p;
2943 int always_replace, n, n2, ret = 0, spair, ts, tm;
2944 int (*parse)(uint32_t *, const char *, size_t);
2945 size_t (*unparse)(char *, size_t, uint32_t);
2948 ts = 1;/* text size. */
2949 if (sc->flag & SCONV_TO_UTF16BE) {
2950 unparse = unicode_to_utf16be;
2952 if (sc->flag & SCONV_FROM_UTF16BE)
2954 } else if (sc->flag & SCONV_TO_UTF16LE) {
2955 unparse = unicode_to_utf16le;
2957 if (sc->flag & SCONV_FROM_UTF16LE)
2959 } else if (sc->flag & SCONV_TO_UTF8) {
2960 unparse = unicode_to_utf8;
2961 if (sc->flag & SCONV_FROM_UTF8)
2965 * This case is going to be converted to another
2966 * character-set through iconv.
2969 if (sc->flag & SCONV_FROM_UTF16BE) {
2970 unparse = unicode_to_utf16be;
2972 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2973 unparse = unicode_to_utf16le;
2976 unparse = unicode_to_utf8;
2980 if (sc->flag & SCONV_FROM_UTF16BE) {
2981 parse = utf16be_to_unicode;
2983 spair = 4;/* surrogate pair size in UTF-16. */
2984 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2985 parse = utf16le_to_unicode;
2987 spair = 4;/* surrogate pair size in UTF-16. */
2989 parse = cesu8_to_unicode;
2991 spair = 6;/* surrogate pair size in UTF-8. */
2994 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2997 p = as->s + as->length;
2998 endp = as->s + as->buffer_length - ts;
2999 while ((n = parse(&uc, s, len)) != 0) {
3000 const char *ucptr, *uc2ptr;
3003 /* Use a replaced unicode character. */
3004 UNPARSE(p, endp, uc);
3009 } else if (n == spair || always_replace)
3010 /* uc is converted from a surrogate pair.
3011 * this should be treated as a changed code. */
3018 /* Read second code point. */
3019 while ((n2 = parse(&uc2, s, len)) > 0) {
3020 uint32_t ucx[FDC_MAX];
3022 int cl, cx, i, nx, ucx_size;
3026 if (n2 == spair || always_replace)
3027 /* uc2 is converted from a surrogate pair.
3028 * this should be treated as a changed code. */
3036 * If current second code point is out of decomposable
3037 * code points, finding compositions is unneeded.
3039 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3041 REPLACE_UC_WITH_UC2();
3046 * Try to combine current code points.
3049 * We have to combine Hangul characters according to
3050 * http://uniicode.org/reports/tr15/#Hangul
3052 if (0 <= (LIndex = uc - HC_LBASE) &&
3053 LIndex < HC_LCOUNT) {
3055 * Hangul Composition.
3056 * 1. Two current code points are L and V.
3058 int VIndex = uc2 - HC_VBASE;
3059 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3060 /* Make syllable of form LV. */
3061 UPDATE_UC(HC_SBASE +
3062 (LIndex * HC_VCOUNT + VIndex) *
3066 REPLACE_UC_WITH_UC2();
3069 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3070 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3072 * Hangul Composition.
3073 * 2. Two current code points are LV and T.
3075 int TIndex = uc2 - HC_TBASE;
3076 if (0 < TIndex && TIndex < HC_TCOUNT) {
3077 /* Make syllable of form LVT. */
3078 UPDATE_UC(uc + TIndex);
3081 REPLACE_UC_WITH_UC2();
3084 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3085 /* A composition to current code points
3089 } else if ((cl = CCC(uc2)) == 0) {
3090 /* Clearly 'uc2' the second code point is not
3091 * a decomposable code. */
3093 REPLACE_UC_WITH_UC2();
3098 * Collect following decomposable code points.
3106 * Find a composed code in the collected code points.
3109 while (i < ucx_size) {
3112 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3118 * nfc is composed of uc and ucx[i].
3123 * Remove ucx[i] by shifting
3124 * following code points.
3126 for (j = i; j+1 < ucx_size; j++) {
3133 * Collect following code points blocked
3134 * by ucx[i] the removed code point.
3136 if (ucx_size > 0 && i == ucx_size &&
3137 nx > 0 && cx == cl) {
3138 cl = ccx[ucx_size-1];
3139 COLLECT_CPS(ucx_size);
3142 * Restart finding a composed code with
3143 * the updated uc from the top of the
3144 * collected code points.
3150 * Apparently the current code points are not
3151 * decomposed characters or already composed.
3154 for (i = 0; i < ucx_size; i++)
3155 UNPARSE(p, endp, ucx[i]);
3158 * Flush out remaining canonical combining characters.
3160 if (nx > 0 && cx == cl && len > 0) {
3161 while ((nx = parse(&ucx[0], s, len))
3169 UNPARSE(p, endp, ucx[0]);
3176 /* Use a replaced unicode character. */
3177 UNPARSE(p, endp, uc2);
3182 } else if (n2 == 0) {
3187 as->length = p - as->s;
3188 as->s[as->length] = '\0';
3190 as->s[as->length+1] = '\0';
3195 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3200 * These are not converted to NFD on Mac OS.
3202 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3203 (uc >= 0xF900 && uc <= 0xFAFF) ||
3204 (uc >= 0x2F800 && uc <= 0x2FAFF))
3207 * Those code points are not converted to NFD on Mac OS.
3208 * I do not know the reason because it is undocumented.
3210 * 1109A ==> 11099 110BA
3211 * 1109C ==> 1109B 110BA
3212 * 110AB ==> 110A5 110BA
3214 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3218 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3220 int m = (t + b) / 2;
3221 if (u_decomposition_table[m].nfc < uc)
3223 else if (u_decomposition_table[m].nfc > uc)
3226 *cp1 = u_decomposition_table[m].cp1;
3227 *cp2 = u_decomposition_table[m].cp2;
3234 #define REPLACE_UC_WITH(cp) do { \
3240 * Normalize UTF-8 characters to Form D and copy the result.
3243 archive_string_normalize_D(struct archive_string *as, const void *_p,
3244 size_t len, struct archive_string_conv *sc)
3246 const char *s = (const char *)_p;
3250 int always_replace, n, n2, ret = 0, spair, ts, tm;
3251 int (*parse)(uint32_t *, const char *, size_t);
3252 size_t (*unparse)(char *, size_t, uint32_t);
3255 ts = 1;/* text size. */
3256 if (sc->flag & SCONV_TO_UTF16BE) {
3257 unparse = unicode_to_utf16be;
3259 if (sc->flag & SCONV_FROM_UTF16BE)
3261 } else if (sc->flag & SCONV_TO_UTF16LE) {
3262 unparse = unicode_to_utf16le;
3264 if (sc->flag & SCONV_FROM_UTF16LE)
3266 } else if (sc->flag & SCONV_TO_UTF8) {
3267 unparse = unicode_to_utf8;
3268 if (sc->flag & SCONV_FROM_UTF8)
3272 * This case is going to be converted to another
3273 * character-set through iconv.
3276 if (sc->flag & SCONV_FROM_UTF16BE) {
3277 unparse = unicode_to_utf16be;
3279 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3280 unparse = unicode_to_utf16le;
3283 unparse = unicode_to_utf8;
3287 if (sc->flag & SCONV_FROM_UTF16BE) {
3288 parse = utf16be_to_unicode;
3290 spair = 4;/* surrogate pair size in UTF-16. */
3291 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3292 parse = utf16le_to_unicode;
3294 spair = 4;/* surrogate pair size in UTF-16. */
3296 parse = cesu8_to_unicode;
3298 spair = 6;/* surrogate pair size in UTF-8. */
3301 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3304 p = as->s + as->length;
3305 endp = as->s + as->buffer_length - ts;
3306 while ((n = parse(&uc, s, len)) != 0) {
3319 /* Use a replaced unicode character. */
3320 UNPARSE(p, endp, uc);
3325 } else if (n == spair || always_replace)
3326 /* uc is converted from a surrogate pair.
3327 * this should be treated as a changed code. */
3334 /* Hangul Decomposition. */
3335 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3336 int L = HC_LBASE + SIndex / HC_NCOUNT;
3337 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3338 int T = HC_TBASE + SIndex % HC_TCOUNT;
3344 if (T != HC_TBASE) {
3350 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3356 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3359 for (k = fdi; k > 0; k--)
3361 fdc[0].ccc = CCC(cp2);
3364 REPLACE_UC_WITH(cp1);
3367 /* Read following code points. */
3368 while ((n2 = parse(&uc2, s, len)) > 0 &&
3369 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3374 for (j = 0; j < fdi; j++) {
3375 if (fdc[j].ccc > ccc)
3379 for (k = fdi; k > j; k--)
3391 for (fdj = 0; fdj < fdi; fdj++) {
3392 REPLACE_UC_WITH(fdc[fdj].uc);
3398 REPLACE_UC_WITH(uc2);
3400 goto check_first_code;
3402 as->length = p - as->s;
3403 as->s[as->length] = '\0';
3405 as->s[as->length+1] = '\0';
3410 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3411 * that WCS is Unicode. It is true for several platforms but some are false.
3412 * And then people who did not use UTF-8 locale on the non Unicode WCS
3413 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3414 * now cannot get right filename from libarchive 3.x and later since we
3415 * fixed the wrong assumption and it is incompatible to older its versions.
3416 * So we provide special option, "compat-2x.x", for resolving it.
3417 * That option enable the string conversion of libarchive 2.x.
3419 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3420 * locale character set and appends to the archive_string.
3421 * Note: returns -1 if conversion fails.
3424 strncat_from_utf8_libarchive2(struct archive_string *as,
3425 const void *_p, size_t len, struct archive_string_conv *sc)
3433 mbstate_t shift_state;
3435 memset(&shift_state, 0, sizeof(shift_state));
3437 /* Clear the shift state before starting. */
3438 wctomb(NULL, L'\0');
3440 (void)sc; /* UNUSED */
3442 * Allocate buffer for MBS.
3443 * We need this allocation here since it is possible that
3444 * as->s is still NULL.
3446 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3449 s = (const char *)_p;
3450 p = as->s + as->length;
3451 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3452 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3456 as->length = p - as->s;
3457 /* Re-allocate buffer for MBS. */
3458 if (archive_string_ensure(as,
3459 as->length + len * 2 + 1) == NULL)
3461 p = as->s + as->length;
3462 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3466 * As libarchie 2.x, translates the UTF-8 characters into
3467 * wide-characters in the assumption that WCS is Unicode.
3473 wc = (wchar_t)unicode;
3478 * Translates the wide-character into the current locale MBS.
3481 n = (int)wcrtomb(p, wc, &shift_state);
3483 n = (int)wctomb(p, wc);
3489 as->length = p - as->s;
3490 as->s[as->length] = '\0';
3496 * Conversion functions between current locale dependent MBS and UTF-16BE.
3497 * strncat_from_utf16be() : UTF-16BE --> MBS
3498 * strncat_to_utf16be() : MBS --> UTF16BE
3501 #if defined(_WIN32) && !defined(__CYGWIN__)
3504 * Convert a UTF-16BE/LE string to current locale and copy the result.
3505 * Return -1 if conversion failes.
3508 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3509 struct archive_string_conv *sc, int be)
3511 struct archive_string tmp;
3520 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3523 mbs = as->s + as->length;
3524 mbs_size = as->buffer_length - as->length -1;
3526 if (sc->to_cp == CP_C_LOCALE) {
3528 * "C" locale special process.
3532 for (b = 0; b < bytes; b += 2) {
3535 val = archive_be16dec(u16+b);
3537 val = archive_le16dec(u16+b);
3542 *mbs++ = (char)(val&0xff);
3546 as->s[as->length] = '\0';
3550 archive_string_init(&tmp);
3552 if (is_big_endian()) {
3555 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3557 memcpy(tmp.s, _p, bytes);
3558 for (b = 0; b < bytes; b += 2) {
3559 uint16_t val = archive_be16dec(tmp.s+b);
3560 archive_le16enc(tmp.s+b, val);
3565 if (!is_big_endian()) {
3568 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3570 memcpy(tmp.s, _p, bytes);
3571 for (b = 0; b < bytes; b += 2) {
3572 uint16_t val = archive_le16dec(tmp.s+b);
3573 archive_be16enc(tmp.s+b, val);
3581 ll = WideCharToMultiByte(sc->to_cp, 0,
3582 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3585 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
3586 /* Need more buffer for MBS. */
3587 ll = WideCharToMultiByte(sc->to_cp, 0,
3588 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3589 if (archive_string_ensure(as, ll +1) == NULL)
3591 mbs = as->s + as->length;
3592 mbs_size = as->buffer_length - as->length -1;
3596 archive_string_free(&tmp);
3598 as->s[as->length] = '\0';
3599 if (ll == 0 || defchar)
3605 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3606 size_t bytes, struct archive_string_conv *sc)
3608 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3612 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3613 size_t bytes, struct archive_string_conv *sc)
3615 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3623 return (archive_be16dec(&d) == 1);
3627 * Convert a current locale string to UTF-16BE/LE and copy the result.
3628 * Return -1 if conversion failes.
3631 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3632 size_t length, struct archive_string_conv *sc, int bigendian)
3634 const char *s = (const char *)_p;
3636 size_t count, avail;
3638 if (archive_string_ensure(as16,
3639 as16->length + (length + 1) * 2) == NULL)
3642 u16 = as16->s + as16->length;
3643 avail = as16->buffer_length - 2;
3644 if (sc->from_cp == CP_C_LOCALE) {
3646 * "C" locale special process.
3649 while (count < length && *s) {
3651 archive_be16enc(u16, *s);
3653 archive_le16enc(u16, *s);
3658 as16->length += count << 1;
3659 as16->s[as16->length] = 0;
3660 as16->s[as16->length+1] = 0;
3664 count = MultiByteToWideChar(sc->from_cp,
3665 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3667 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
3668 /* Need more buffer for UTF-16 string */
3669 count = MultiByteToWideChar(sc->from_cp,
3670 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3671 if (archive_string_ensure(as16, (count +1) * 2)
3674 u16 = as16->s + as16->length;
3675 avail = as16->buffer_length - 2;
3679 as16->length += count * 2;
3680 as16->s[as16->length] = 0;
3681 as16->s[as16->length+1] = 0;
3685 if (is_big_endian()) {
3688 uint16_t v = archive_be16dec(u16);
3689 archive_le16enc(u16, v);
3697 uint16_t v = archive_le16dec(u16);
3698 archive_be16enc(u16, v);
3708 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3709 size_t length, struct archive_string_conv *sc)
3711 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3715 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3716 size_t length, struct archive_string_conv *sc)
3718 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3721 #endif /* _WIN32 && !__CYGWIN__ */
3724 * Do the best effort for conversions.
3725 * We cannot handle UTF-16BE character-set without such iconv,
3726 * but there is a chance if a string consists just ASCII code or
3727 * a current locale is UTF-8.
3731 * Convert a UTF-16BE string to current locale and copy the result.
3732 * Return -1 if conversion failes.
3735 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3736 size_t bytes, struct archive_string_conv *sc, int be)
3738 const char *utf16 = (const char *)_p;
3743 (void)sc; /* UNUSED */
3745 * Other case, we should do the best effort.
3746 * If all character are ASCII(<0x7f), we can convert it.
3747 * if not , we set a alternative character and return -1.
3750 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3752 mbs = as->s + as->length;
3754 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3763 /* We cannot handle it. */
3769 as->length = mbs - as->s;
3770 as->s[as->length] = '\0';
3775 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3776 size_t bytes, struct archive_string_conv *sc)
3778 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3782 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3783 size_t bytes, struct archive_string_conv *sc)
3785 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3789 * Convert a current locale string to UTF-16BE/LE and copy the result.
3790 * Return -1 if conversion failes.
3793 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3794 size_t length, struct archive_string_conv *sc, int bigendian)
3796 const char *s = (const char *)_p;
3801 (void)sc; /* UNUSED */
3803 * Other case, we should do the best effort.
3804 * If all character are ASCII(<0x7f), we can convert it.
3805 * if not , we set a alternative character and return -1.
3810 if (archive_string_ensure(as16,
3811 as16->length + (length + 1) * 2) == NULL)
3814 utf16 = as16->s + as16->length;
3815 while (remaining--) {
3818 /* We cannot handle it. */
3823 archive_be16enc(utf16, c);
3825 archive_le16enc(utf16, c);
3828 as16->length = utf16 - as16->s;
3829 as16->s[as16->length] = 0;
3830 as16->s[as16->length+1] = 0;
3835 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3836 size_t length, struct archive_string_conv *sc)
3838 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3842 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3843 size_t length, struct archive_string_conv *sc)
3845 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3850 * Multistring operations.
3854 archive_mstring_clean(struct archive_mstring *aes)
3856 archive_wstring_free(&(aes->aes_wcs));
3857 archive_string_free(&(aes->aes_mbs));
3858 archive_string_free(&(aes->aes_utf8));
3859 archive_string_free(&(aes->aes_mbs_in_locale));
3864 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3866 dest->aes_set = src->aes_set;
3867 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3868 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3869 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3873 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3876 struct archive_string_conv *sc;
3879 /* If we already have a UTF8 form, return that immediately. */
3880 if (aes->aes_set & AES_SET_UTF8) {
3881 *p = aes->aes_utf8.s;
3886 if (aes->aes_set & AES_SET_MBS) {
3887 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3889 return (-1);/* Couldn't allocate memory for sc. */
3890 r = archive_strncpy_l(&(aes->aes_mbs), aes->aes_mbs.s,
3891 aes->aes_mbs.length, sc);
3893 free_sconv_object(sc);
3895 aes->aes_set |= AES_SET_UTF8;
3896 *p = aes->aes_utf8.s;
3897 return (0);/* success. */
3899 return (-1);/* failure. */
3901 return (0);/* success. */
3905 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3910 (void)a; /* UNUSED */
3911 /* If we already have an MBS form, return that immediately. */
3912 if (aes->aes_set & AES_SET_MBS) {
3913 *p = aes->aes_mbs.s;
3918 /* If there's a WCS form, try converting with the native locale. */
3919 if (aes->aes_set & AES_SET_WCS) {
3920 archive_string_empty(&(aes->aes_mbs));
3921 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3922 aes->aes_wcs.s, aes->aes_wcs.length);
3923 *p = aes->aes_mbs.s;
3925 aes->aes_set |= AES_SET_MBS;
3932 * Only a UTF-8 form cannot avail because its conversion already
3933 * failed at archive_mstring_update_utf8().
3939 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3944 (void)a;/* UNUSED */
3945 /* Return WCS form if we already have it. */
3946 if (aes->aes_set & AES_SET_WCS) {
3947 *wp = aes->aes_wcs.s;
3952 /* Try converting MBS to WCS using native locale. */
3953 if (aes->aes_set & AES_SET_MBS) {
3954 archive_wstring_empty(&(aes->aes_wcs));
3955 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
3956 aes->aes_mbs.s, aes->aes_mbs.length);
3958 aes->aes_set |= AES_SET_WCS;
3959 *wp = aes->aes_wcs.s;
3961 ret = -1;/* failure. */
3967 archive_mstring_get_mbs_l(struct archive_mstring *aes,
3968 const char **p, size_t *length, struct archive_string_conv *sc)
3972 #if defined(_WIN32) && !defined(__CYGWIN__)
3974 * Internationalization programing on Windows must use Wide
3975 * characters because Windows platform cannot make locale UTF-8.
3977 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
3978 archive_string_empty(&(aes->aes_mbs_in_locale));
3979 r = archive_string_append_from_wcs_in_codepage(
3980 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
3981 aes->aes_wcs.length, sc);
3983 *p = aes->aes_mbs_in_locale.s;
3985 *length = aes->aes_mbs_in_locale.length;
3987 } else if (errno == ENOMEM)
3994 /* If there is not an MBS form but is a WCS form, try converting
3995 * with the native locale to be used for translating it to specified
3997 if ((aes->aes_set & AES_SET_MBS) == 0 &&
3998 (aes->aes_set & AES_SET_WCS) != 0) {
3999 archive_string_empty(&(aes->aes_mbs));
4000 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4001 aes->aes_wcs.s, aes->aes_wcs.length);
4003 aes->aes_set |= AES_SET_MBS;
4004 else if (errno == ENOMEM)
4009 /* If we already have an MBS form, use it to be translated to
4010 * specified character-set. */
4011 if (aes->aes_set & AES_SET_MBS) {
4013 /* Conversion is unneeded. */
4014 *p = aes->aes_mbs.s;
4016 *length = aes->aes_mbs.length;
4019 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4020 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4021 *p = aes->aes_mbs_in_locale.s;
4023 *length = aes->aes_mbs_in_locale.length;
4033 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4039 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4043 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4050 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4051 archive_strncpy(&(aes->aes_mbs), mbs, len);
4052 archive_string_empty(&(aes->aes_utf8));
4053 archive_wstring_empty(&(aes->aes_wcs));
4058 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4060 return archive_mstring_copy_wcs_len(aes, wcs,
4061 wcs == NULL ? 0 : wcslen(wcs));
4065 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4071 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4072 archive_string_empty(&(aes->aes_mbs));
4073 archive_string_empty(&(aes->aes_utf8));
4074 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4079 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4080 const char *mbs, size_t len, struct archive_string_conv *sc)
4088 archive_string_empty(&(aes->aes_mbs));
4089 archive_wstring_empty(&(aes->aes_wcs));
4090 archive_string_empty(&(aes->aes_utf8));
4091 #if defined(_WIN32) && !defined(__CYGWIN__)
4093 * Internationalization programing on Windows must use Wide
4094 * characters because Windows platform cannot make locale UTF-8.
4097 if (archive_string_append(&(aes->aes_mbs),
4098 mbs, mbsnbytes(mbs, len)) == NULL) {
4102 aes->aes_set = AES_SET_MBS;
4105 #if defined(HAVE_ICONV)
4106 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4108 * This case happens only when MultiByteToWideChar() cannot
4109 * handle sc->from_cp, and we have to iconv in order to
4110 * translate character-set to wchar_t,UTF-16.
4112 iconv_t cd = sc->cd;
4117 * Translate multi-bytes from some character-set to UTF-8.
4120 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4126 aes->aes_set = AES_SET_UTF8;
4129 * Append the UTF-8 string into wstring.
4132 sc->flag &= ~(SCONV_NORMALIZATION_C
4133 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4134 from_cp = sc->from_cp;
4135 sc->from_cp = CP_UTF8;
4136 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4137 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4139 sc->from_cp = from_cp;
4141 aes->aes_set |= AES_SET_WCS;
4144 r = archive_wstring_append_from_mbs_in_codepage(
4145 &(aes->aes_wcs), mbs, len, sc);
4147 aes->aes_set = AES_SET_WCS;
4152 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4154 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4162 * The 'update' form tries to proactively update all forms of
4163 * this string (WCS and MBS) and returns an error if any of
4164 * them fail. This is used by the 'pax' handler, for instance,
4165 * to detect and report character-conversion failures early while
4166 * still allowing clients to get potentially useful values from
4167 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4168 * strive to give the user something useful, so you can get hopefully
4169 * usable values even if some of the character conversions are failing.)
4172 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4175 struct archive_string_conv *sc;
4180 return (0); /* Succeeded in clearing everything. */
4183 /* Save the UTF8 string. */
4184 archive_strcpy(&(aes->aes_utf8), utf8);
4186 /* Empty the mbs and wcs strings. */
4187 archive_string_empty(&(aes->aes_mbs));
4188 archive_wstring_empty(&(aes->aes_wcs));
4190 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4192 /* Try converting UTF-8 to MBS, return false on failure. */
4193 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4195 return (-1);/* Couldn't allocate memory for sc. */
4196 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4198 free_sconv_object(sc);
4201 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4203 /* Try converting MBS to WCS, return false on failure. */
4204 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4205 aes->aes_mbs.length))
4207 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4209 /* All conversions succeeded. */