xref: /freebsd/contrib/libarchive/libarchive/archive_string.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
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 
29 /*
30  * Basic resizable string support, to simplify manipulating arbitrary-sized
31  * strings while minimizing heap activity.
32  *
33  * In particular, the buffer used by a string object is only grown, it
34  * never shrinks, so you can clear and reuse the same string object
35  * without incurring additional memory allocations.
36  */
37 
38 #ifdef HAVE_ERRNO_H
39 #include <errno.h>
40 #endif
41 #ifdef HAVE_ICONV_H
42 #include <iconv.h>
43 #endif
44 #ifdef HAVE_LANGINFO_H
45 #include <langinfo.h>
46 #endif
47 #ifdef HAVE_LOCALCHARSET_H
48 #include <localcharset.h>
49 #endif
50 #ifdef HAVE_STDLIB_H
51 #include <stdlib.h>
52 #endif
53 #ifdef HAVE_STRING_H
54 #include <string.h>
55 #endif
56 #ifdef HAVE_WCHAR_H
57 #include <wchar.h>
58 #endif
59 #if defined(_WIN32) && !defined(__CYGWIN__)
60 #include <windows.h>
61 #include <locale.h>
62 #endif
63 
64 #include "archive_endian.h"
65 #include "archive_private.h"
66 #include "archive_string.h"
67 #include "archive_string_composition.h"
68 
69 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
70 #define wmemcpy(a,b,i)  (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
71 #endif
72 
73 #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove)
74 #define wmemmove(a,b,i)  (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t))
75 #endif
76 
77 #undef max
78 #define max(a, b)       ((a)>(b)?(a):(b))
79 
80 struct archive_string_conv {
81 	struct archive_string_conv	*next;
82 	char				*from_charset;
83 	char				*to_charset;
84 	unsigned			 from_cp;
85 	unsigned			 to_cp;
86 	/* Set 1 if from_charset and to_charset are the same. */
87 	int				 same;
88 	int				 flag;
89 #define SCONV_TO_CHARSET	1	/* MBS is being converted to specified
90 					 * charset. */
91 #define SCONV_FROM_CHARSET	(1<<1)	/* MBS is being converted from
92 					 * specified charset. */
93 #define SCONV_BEST_EFFORT 	(1<<2)	/* Copy at least ASCII code. */
94 #define SCONV_WIN_CP	 	(1<<3)	/* Use Windows API for converting
95 					 * MBS. */
96 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4)	/* Incorrect UTF-8 made by libarchive
97 					 * 2.x in the wrong assumption. */
98 #define SCONV_NORMALIZATION_C	(1<<6)	/* Need normalization to be Form C.
99 					 * Before UTF-8 characters are actually
100 					 * processed. */
101 #define SCONV_NORMALIZATION_D	(1<<7)	/* Need normalization to be Form D.
102 					 * Before UTF-8 characters are actually
103 					 * processed.
104 					 * Currently this only for MAC OS X. */
105 #define SCONV_TO_UTF8		(1<<8)	/* "to charset" side is UTF-8. */
106 #define SCONV_FROM_UTF8		(1<<9)	/* "from charset" side is UTF-8. */
107 #define SCONV_TO_UTF16BE 	(1<<10)	/* "to charset" side is UTF-16BE. */
108 #define SCONV_FROM_UTF16BE 	(1<<11)	/* "from charset" side is UTF-16BE. */
109 #define SCONV_TO_UTF16LE 	(1<<12)	/* "to charset" side is UTF-16LE. */
110 #define SCONV_FROM_UTF16LE 	(1<<13)	/* "from charset" side is UTF-16LE. */
111 #define SCONV_TO_UTF16		(SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
112 #define SCONV_FROM_UTF16	(SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
113 
114 #if HAVE_ICONV
115 	iconv_t				 cd;
116 	iconv_t				 cd_w;/* Use at archive_mstring on
117 				 	       * Windows. */
118 #endif
119 	/* A temporary buffer for normalization. */
120 	struct archive_string		 utftmp;
121 	int (*converter[2])(struct archive_string *, const void *, size_t,
122 	    struct archive_string_conv *);
123 	int				 nconverter;
124 };
125 
126 #define CP_C_LOCALE	0	/* "C" locale only for this file. */
127 #define CP_UTF16LE	1200
128 #define CP_UTF16BE	1201
129 
130 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
131 #define IS_LOW_SURROGATE_LA(uc)	 ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
132 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
133 #define UNICODE_MAX		0x10FFFF
134 #define UNICODE_R_CHAR		0xFFFD	/* Replacement character. */
135 /* Set U+FFFD(Replacement character) in UTF-8. */
136 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd};
137 
138 static struct archive_string_conv *find_sconv_object(struct archive *,
139 	const char *, const char *);
140 static void add_sconv_object(struct archive *, struct archive_string_conv *);
141 static struct archive_string_conv *create_sconv_object(const char *,
142 	const char *, unsigned, int);
143 static void free_sconv_object(struct archive_string_conv *);
144 static struct archive_string_conv *get_sconv_object(struct archive *,
145 	const char *, const char *, int);
146 static unsigned make_codepage_from_charset(const char *);
147 static unsigned get_current_codepage(void);
148 static unsigned get_current_oemcp(void);
149 static size_t mbsnbytes(const void *, size_t);
150 static size_t utf16nbytes(const void *, size_t);
151 #if defined(_WIN32) && !defined(__CYGWIN__)
152 static int archive_wstring_append_from_mbs_in_codepage(
153     struct archive_wstring *, const char *, size_t,
154     struct archive_string_conv *);
155 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
156     const wchar_t *, size_t, struct archive_string_conv *);
157 static int is_big_endian(void);
158 static int strncat_in_codepage(struct archive_string *, const void *,
159     size_t, struct archive_string_conv *);
160 static int win_strncat_from_utf16be(struct archive_string *, const void *,
161     size_t, struct archive_string_conv *);
162 static int win_strncat_from_utf16le(struct archive_string *, const void *,
163     size_t, struct archive_string_conv *);
164 static int win_strncat_to_utf16be(struct archive_string *, const void *,
165     size_t, struct archive_string_conv *);
166 static int win_strncat_to_utf16le(struct archive_string *, const void *,
167     size_t, struct archive_string_conv *);
168 #endif
169 static int best_effort_strncat_from_utf16be(struct archive_string *,
170     const void *, size_t, struct archive_string_conv *);
171 static int best_effort_strncat_from_utf16le(struct archive_string *,
172     const void *, size_t, struct archive_string_conv *);
173 static int best_effort_strncat_to_utf16be(struct archive_string *,
174     const void *, size_t, struct archive_string_conv *);
175 static int best_effort_strncat_to_utf16le(struct archive_string *,
176     const void *, size_t, struct archive_string_conv *);
177 #if defined(HAVE_ICONV)
178 static int iconv_strncat_in_locale(struct archive_string *, const void *,
179     size_t, struct archive_string_conv *);
180 #endif
181 static int best_effort_strncat_in_locale(struct archive_string *,
182     const void *, size_t, struct archive_string_conv *);
183 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
184 static int utf8_to_unicode(uint32_t *, const char *, size_t);
185 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
186 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
187 static size_t unicode_to_utf8(char *, size_t, uint32_t);
188 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
189 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
190 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
191 static int strncat_from_utf8_libarchive2(struct archive_string *,
192     const void *, size_t, struct archive_string_conv *);
193 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
194     size_t, struct archive_string_conv *);
195 static int archive_string_normalize_C(struct archive_string *, const void *,
196     size_t, struct archive_string_conv *);
197 static int archive_string_normalize_D(struct archive_string *, const void *,
198     size_t, struct archive_string_conv *);
199 static int archive_string_append_unicode(struct archive_string *,
200     const void *, size_t, struct archive_string_conv *);
201 
202 static struct archive_string *
203 archive_string_append(struct archive_string *as, const char *p, size_t s)
204 {
205 	if (archive_string_ensure(as, as->length + s + 1) == NULL)
206 		return (NULL);
207 	if (s)
208 		memmove(as->s + as->length, p, s);
209 	as->length += s;
210 	as->s[as->length] = 0;
211 	return (as);
212 }
213 
214 static struct archive_wstring *
215 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
216 {
217 	if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
218 		return (NULL);
219 	if (s)
220 		wmemmove(as->s + as->length, p, s);
221 	as->length += s;
222 	as->s[as->length] = 0;
223 	return (as);
224 }
225 
226 struct archive_string *
227 archive_array_append(struct archive_string *as, const char *p, size_t s)
228 {
229 	return archive_string_append(as, p, s);
230 }
231 
232 void
233 archive_string_concat(struct archive_string *dest, struct archive_string *src)
234 {
235 	if (archive_string_append(dest, src->s, src->length) == NULL)
236 		__archive_errx(1, "Out of memory");
237 }
238 
239 void
240 archive_wstring_concat(struct archive_wstring *dest,
241     struct archive_wstring *src)
242 {
243 	if (archive_wstring_append(dest, src->s, src->length) == NULL)
244 		__archive_errx(1, "Out of memory");
245 }
246 
247 void
248 archive_string_free(struct archive_string *as)
249 {
250 	as->length = 0;
251 	as->buffer_length = 0;
252 	free(as->s);
253 	as->s = NULL;
254 }
255 
256 void
257 archive_wstring_free(struct archive_wstring *as)
258 {
259 	as->length = 0;
260 	as->buffer_length = 0;
261 	free(as->s);
262 	as->s = NULL;
263 }
264 
265 struct archive_wstring *
266 archive_wstring_ensure(struct archive_wstring *as, size_t s)
267 {
268 	return (struct archive_wstring *)
269 		archive_string_ensure((struct archive_string *)as,
270 					s * sizeof(wchar_t));
271 }
272 
273 /* Returns NULL on any allocation failure. */
274 struct archive_string *
275 archive_string_ensure(struct archive_string *as, size_t s)
276 {
277 	char *p;
278 	size_t new_length;
279 
280 	/* If buffer is already big enough, don't reallocate. */
281 	if (as->s && (s <= as->buffer_length))
282 		return (as);
283 
284 	/*
285 	 * Growing the buffer at least exponentially ensures that
286 	 * append operations are always linear in the number of
287 	 * characters appended.  Using a smaller growth rate for
288 	 * larger buffers reduces memory waste somewhat at the cost of
289 	 * a larger constant factor.
290 	 */
291 	if (as->buffer_length < 32)
292 		/* Start with a minimum 32-character buffer. */
293 		new_length = 32;
294 	else if (as->buffer_length < 8192)
295 		/* Buffers under 8k are doubled for speed. */
296 		new_length = as->buffer_length + as->buffer_length;
297 	else {
298 		/* Buffers 8k and over grow by at least 25% each time. */
299 		new_length = as->buffer_length + as->buffer_length / 4;
300 		/* Be safe: If size wraps, fail. */
301 		if (new_length < as->buffer_length) {
302 			/* On failure, wipe the string and return NULL. */
303 			archive_string_free(as);
304 			errno = ENOMEM;/* Make sure errno has ENOMEM. */
305 			return (NULL);
306 		}
307 	}
308 	/*
309 	 * The computation above is a lower limit to how much we'll
310 	 * grow the buffer.  In any case, we have to grow it enough to
311 	 * hold the request.
312 	 */
313 	if (new_length < s)
314 		new_length = s;
315 	/* Now we can reallocate the buffer. */
316 	p = realloc(as->s, new_length);
317 	if (p == NULL) {
318 		/* On failure, wipe the string and return NULL. */
319 		archive_string_free(as);
320 		errno = ENOMEM;/* Make sure errno has ENOMEM. */
321 		return (NULL);
322 	}
323 
324 	as->s = p;
325 	as->buffer_length = new_length;
326 	return (as);
327 }
328 
329 /*
330  * TODO: See if there's a way to avoid scanning
331  * the source string twice.  Then test to see
332  * if it actually helps (remember that we're almost
333  * always called with pretty short arguments, so
334  * such an optimization might not help).
335  */
336 struct archive_string *
337 archive_strncat(struct archive_string *as, const void *_p, size_t n)
338 {
339 	size_t s;
340 	const char *p, *pp;
341 
342 	p = (const char *)_p;
343 
344 	/* Like strlen(p), except won't examine positions beyond p[n]. */
345 	s = 0;
346 	pp = p;
347 	while (s < n && *pp) {
348 		pp++;
349 		s++;
350 	}
351 	if ((as = archive_string_append(as, p, s)) == NULL)
352 		__archive_errx(1, "Out of memory");
353 	return (as);
354 }
355 
356 struct archive_wstring *
357 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
358 {
359 	size_t s;
360 	const wchar_t *pp;
361 
362 	/* Like strlen(p), except won't examine positions beyond p[n]. */
363 	s = 0;
364 	pp = p;
365 	while (s < n && *pp) {
366 		pp++;
367 		s++;
368 	}
369 	if ((as = archive_wstring_append(as, p, s)) == NULL)
370 		__archive_errx(1, "Out of memory");
371 	return (as);
372 }
373 
374 struct archive_string *
375 archive_strcat(struct archive_string *as, const void *p)
376 {
377 	/* strcat is just strncat without an effective limit.
378 	 * Assert that we'll never get called with a source
379 	 * string over 16MB.
380 	 * TODO: Review all uses of strcat in the source
381 	 * and try to replace them with strncat().
382 	 */
383 	return archive_strncat(as, p, 0x1000000);
384 }
385 
386 struct archive_wstring *
387 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
388 {
389 	/* Ditto. */
390 	return archive_wstrncat(as, p, 0x1000000);
391 }
392 
393 struct archive_string *
394 archive_strappend_char(struct archive_string *as, char c)
395 {
396 	if ((as = archive_string_append(as, &c, 1)) == NULL)
397 		__archive_errx(1, "Out of memory");
398 	return (as);
399 }
400 
401 struct archive_wstring *
402 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
403 {
404 	if ((as = archive_wstring_append(as, &c, 1)) == NULL)
405 		__archive_errx(1, "Out of memory");
406 	return (as);
407 }
408 
409 /*
410  * Get the "current character set" name to use with iconv.
411  * On FreeBSD, the empty character set name "" chooses
412  * the correct character encoding for the current locale,
413  * so this isn't necessary.
414  * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
415  * on that system, we have to explicitly call nl_langinfo()
416  * to get the right name.  Not sure about other platforms.
417  *
418  * NOTE: GNU libiconv does not recognize the character-set name
419  * which some platform nl_langinfo(CODESET) returns, so we should
420  * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
421  */
422 static const char *
423 default_iconv_charset(const char *charset) {
424 	if (charset != NULL && charset[0] != '\0')
425 		return charset;
426 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
427 	/* locale_charset() is broken on Mac OS */
428 	return locale_charset();
429 #elif HAVE_NL_LANGINFO
430 	return nl_langinfo(CODESET);
431 #else
432 	return "";
433 #endif
434 }
435 
436 #if defined(_WIN32) && !defined(__CYGWIN__)
437 
438 /*
439  * Convert MBS to WCS.
440  * Note: returns -1 if conversion fails.
441  */
442 int
443 archive_wstring_append_from_mbs(struct archive_wstring *dest,
444     const char *p, size_t len)
445 {
446 	return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
447 }
448 
449 static int
450 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
451     const char *s, size_t length, struct archive_string_conv *sc)
452 {
453 	int count, ret = 0;
454 	UINT from_cp;
455 
456 	if (sc != NULL)
457 		from_cp = sc->from_cp;
458 	else
459 		from_cp = get_current_codepage();
460 
461 	if (from_cp == CP_C_LOCALE) {
462 		/*
463 		 * "C" locale special processing.
464 		 */
465 		wchar_t *ws;
466 		const unsigned char *mp;
467 
468 		if (NULL == archive_wstring_ensure(dest,
469 		    dest->length + length + 1))
470 			return (-1);
471 
472 		ws = dest->s + dest->length;
473 		mp = (const unsigned char *)s;
474 		count = 0;
475 		while (count < (int)length && *mp) {
476 			*ws++ = (wchar_t)*mp++;
477 			count++;
478 		}
479 	} else if (sc != NULL &&
480 	    (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
481 		/*
482 		 * Normalize UTF-8 and UTF-16BE and convert it directly
483 		 * to UTF-16 as wchar_t.
484 		 */
485 		struct archive_string u16;
486 		int saved_flag = sc->flag;/* save current flag. */
487 
488 		if (is_big_endian())
489 			sc->flag |= SCONV_TO_UTF16BE;
490 		else
491 			sc->flag |= SCONV_TO_UTF16LE;
492 
493 		if (sc->flag & SCONV_FROM_UTF16) {
494 			/*
495 			 *  UTF-16BE/LE NFD ===> UTF-16 NFC
496 			 *  UTF-16BE/LE NFC ===> UTF-16 NFD
497 			 */
498 			count = (int)utf16nbytes(s, length);
499 		} else {
500 			/*
501 			 *  UTF-8 NFD ===> UTF-16 NFC
502 			 *  UTF-8 NFC ===> UTF-16 NFD
503 			 */
504 			count = (int)mbsnbytes(s, length);
505 		}
506 		u16.s = (char *)dest->s;
507 		u16.length = dest->length << 1;;
508 		u16.buffer_length = dest->buffer_length;
509 		if (sc->flag & SCONV_NORMALIZATION_C)
510 			ret = archive_string_normalize_C(&u16, s, count, sc);
511 		else
512 			ret = archive_string_normalize_D(&u16, s, count, sc);
513 		dest->s = (wchar_t *)u16.s;
514 		dest->length = u16.length >> 1;
515 		dest->buffer_length = u16.buffer_length;
516 		sc->flag = saved_flag;/* restore the saved flag. */
517 		return (ret);
518 	} else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
519 		count = (int)utf16nbytes(s, length);
520 		count >>= 1; /* to be WCS length */
521 		/* Allocate memory for WCS. */
522 		if (NULL == archive_wstring_ensure(dest,
523 		    dest->length + count + 1))
524 			return (-1);
525 		wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
526 		if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
527 			uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
528 			int b;
529 			for (b = 0; b < count; b++) {
530 				uint16_t val = archive_le16dec(u16+b);
531 				archive_be16enc(u16+b, val);
532 			}
533 		} else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
534 			uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
535 			int b;
536 			for (b = 0; b < count; b++) {
537 				uint16_t val = archive_be16dec(u16+b);
538 				archive_le16enc(u16+b, val);
539 			}
540 		}
541 	} else {
542 		DWORD mbflag;
543 		size_t buffsize;
544 
545 		if (sc == NULL)
546 			mbflag = 0;
547 		else if (sc->flag & SCONV_FROM_CHARSET) {
548 			/* Do not trust the length which comes from
549 			 * an archive file. */
550 			length = mbsnbytes(s, length);
551 			mbflag = 0;
552 		} else
553 			mbflag = MB_PRECOMPOSED;
554 
555 		mbflag |= MB_ERR_INVALID_CHARS;
556 
557 		buffsize = dest->length + length + 1;
558 		do {
559 			/* Allocate memory for WCS. */
560 			if (NULL == archive_wstring_ensure(dest, buffsize))
561 				return (-1);
562 			/* Convert MBS to WCS. */
563 			count = MultiByteToWideChar(from_cp,
564 			    mbflag, s, (int)length, dest->s + dest->length,
565 			    (int)(dest->buffer_length >> 1) -1);
566 			if (count == 0 &&
567 			    GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
568 				/* Expand the WCS buffer. */
569 				buffsize = dest->buffer_length << 1;
570 				continue;
571 			}
572 			if (count == 0 && length != 0)
573 				ret = -1;
574 			break;
575 		} while (1);
576 	}
577 	dest->length += count;
578 	dest->s[dest->length] = L'\0';
579 	return (ret);
580 }
581 
582 #else
583 
584 /*
585  * Convert MBS to WCS.
586  * Note: returns -1 if conversion fails.
587  */
588 int
589 archive_wstring_append_from_mbs(struct archive_wstring *dest,
590     const char *p, size_t len)
591 {
592 	size_t r;
593 	int ret_val = 0;
594 	/*
595 	 * No single byte will be more than one wide character,
596 	 * so this length estimate will always be big enough.
597 	 */
598 	// size_t wcs_length = len;
599 	size_t mbs_length = len;
600 	const char *mbs = p;
601 	wchar_t *wcs;
602 #if HAVE_MBRTOWC
603 	mbstate_t shift_state;
604 
605 	memset(&shift_state, 0, sizeof(shift_state));
606 #endif
607 	/*
608 	 * As we decided to have wcs_length == mbs_length == len
609 	 * we can use len here instead of wcs_length
610 	 */
611 	if (NULL == archive_wstring_ensure(dest, dest->length + len + 1))
612 		return (-1);
613 	wcs = dest->s + dest->length;
614 	/*
615 	 * We cannot use mbsrtowcs/mbstowcs here because those may convert
616 	 * extra MBS when strlen(p) > len and one wide character consists of
617 	 * multi bytes.
618 	 */
619 	while (*mbs && mbs_length > 0) {
620 		/*
621 		 * The buffer we allocated is always big enough.
622 		 * Keep this code path in a comment if we decide to choose
623 		 * smaller wcs_length in the future
624 		 */
625 /*
626 		if (wcs_length == 0) {
627 			dest->length = wcs - dest->s;
628 			dest->s[dest->length] = L'\0';
629 			wcs_length = mbs_length;
630 			if (NULL == archive_wstring_ensure(dest,
631 			    dest->length + wcs_length + 1))
632 				return (-1);
633 			wcs = dest->s + dest->length;
634 		}
635 */
636 #if HAVE_MBRTOWC
637 		r = mbrtowc(wcs, mbs, mbs_length, &shift_state);
638 #else
639 		r = mbtowc(wcs, mbs, mbs_length);
640 #endif
641 		if (r == (size_t)-1 || r == (size_t)-2) {
642 			ret_val = -1;
643 			break;
644 		}
645 		if (r == 0 || r > mbs_length)
646 			break;
647 		wcs++;
648 		// wcs_length--;
649 		mbs += r;
650 		mbs_length -= r;
651 	}
652 	dest->length = wcs - dest->s;
653 	dest->s[dest->length] = L'\0';
654 	return (ret_val);
655 }
656 
657 #endif
658 
659 #if defined(_WIN32) && !defined(__CYGWIN__)
660 
661 /*
662  * WCS ==> MBS.
663  * Note: returns -1 if conversion fails.
664  *
665  * Win32 builds use WideCharToMultiByte from the Windows API.
666  * (Maybe Cygwin should too?  WideCharToMultiByte will know a
667  * lot more about local character encodings than the wcrtomb()
668  * wrapper is going to know.)
669  */
670 int
671 archive_string_append_from_wcs(struct archive_string *as,
672     const wchar_t *w, size_t len)
673 {
674 	return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
675 }
676 
677 static int
678 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
679     const wchar_t *ws, size_t len, struct archive_string_conv *sc)
680 {
681 	BOOL defchar_used, *dp;
682 	int count, ret = 0;
683 	UINT to_cp;
684 	int wslen = (int)len;
685 
686 	if (sc != NULL)
687 		to_cp = sc->to_cp;
688 	else
689 		to_cp = get_current_codepage();
690 
691 	if (to_cp == CP_C_LOCALE) {
692 		/*
693 		 * "C" locale special processing.
694 		 */
695 		const wchar_t *wp = ws;
696 		char *p;
697 
698 		if (NULL == archive_string_ensure(as,
699 		    as->length + wslen +1))
700 			return (-1);
701 		p = as->s + as->length;
702 		count = 0;
703 		defchar_used = 0;
704 		while (count < wslen && *wp) {
705 			if (*wp > 255) {
706 				*p++ = '?';
707 				wp++;
708 				defchar_used = 1;
709 			} else
710 				*p++ = (char)*wp++;
711 			count++;
712 		}
713 	} else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
714 		uint16_t *u16;
715 
716 		if (NULL ==
717 		    archive_string_ensure(as, as->length + len * 2 + 2))
718 			return (-1);
719 		u16 = (uint16_t *)(as->s + as->length);
720 		count = 0;
721 		defchar_used = 0;
722 		if (sc->flag & SCONV_TO_UTF16BE) {
723 			while (count < (int)len && *ws) {
724 				archive_be16enc(u16+count, *ws);
725 				ws++;
726 				count++;
727 			}
728 		} else {
729 			while (count < (int)len && *ws) {
730 				archive_le16enc(u16+count, *ws);
731 				ws++;
732 				count++;
733 			}
734 		}
735 		count <<= 1; /* to be byte size */
736 	} else {
737 		/* Make sure the MBS buffer has plenty to set. */
738 		if (NULL ==
739 		    archive_string_ensure(as, as->length + len * 2 + 1))
740 			return (-1);
741 		do {
742 			defchar_used = 0;
743 			if (to_cp == CP_UTF8 || sc == NULL)
744 				dp = NULL;
745 			else
746 				dp = &defchar_used;
747 			count = WideCharToMultiByte(to_cp, 0, ws, wslen,
748 			    as->s + as->length,
749 			    (int)as->buffer_length - (int)as->length - 1, NULL, dp);
750 			if (count == 0 &&
751 			    GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
752 				/* Expand the MBS buffer and retry. */
753 				if (NULL == archive_string_ensure(as,
754 					as->buffer_length + len))
755 					return (-1);
756 				continue;
757 			}
758 			if (count == 0)
759 				ret = -1;
760 			break;
761 		} while (1);
762 	}
763 	as->length += count;
764 	as->s[as->length] = '\0';
765 	return (defchar_used?-1:ret);
766 }
767 
768 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
769 
770 /*
771  * Translates a wide character string into current locale character set
772  * and appends to the archive_string.  Note: returns -1 if conversion
773  * fails.
774  */
775 int
776 archive_string_append_from_wcs(struct archive_string *as,
777     const wchar_t *w, size_t len)
778 {
779 	/* We cannot use the standard wcstombs() here because it
780 	 * cannot tell us how big the output buffer should be.  So
781 	 * I've built a loop around wcrtomb() or wctomb() that
782 	 * converts a character at a time and resizes the string as
783 	 * needed.  We prefer wcrtomb() when it's available because
784 	 * it's thread-safe. */
785 	int n, ret_val = 0;
786 	char *p;
787 	char *end;
788 #if HAVE_WCRTOMB
789 	mbstate_t shift_state;
790 
791 	memset(&shift_state, 0, sizeof(shift_state));
792 #else
793 	/* Clear the shift state before starting. */
794 	wctomb(NULL, L'\0');
795 #endif
796 	/*
797 	 * Allocate buffer for MBS.
798 	 * We need this allocation here since it is possible that
799 	 * as->s is still NULL.
800 	 */
801 	if (archive_string_ensure(as, as->length + len + 1) == NULL)
802 		return (-1);
803 
804 	p = as->s + as->length;
805 	end = as->s + as->buffer_length - MB_CUR_MAX -1;
806 	while (*w != L'\0' && len > 0) {
807 		if (p >= end) {
808 			as->length = p - as->s;
809 			as->s[as->length] = '\0';
810 			/* Re-allocate buffer for MBS. */
811 			if (archive_string_ensure(as,
812 			    as->length + max(len * 2,
813 			    (size_t)MB_CUR_MAX) + 1) == NULL)
814 				return (-1);
815 			p = as->s + as->length;
816 			end = as->s + as->buffer_length - MB_CUR_MAX -1;
817 		}
818 #if HAVE_WCRTOMB
819 		n = wcrtomb(p, *w++, &shift_state);
820 #else
821 		n = wctomb(p, *w++);
822 #endif
823 		if (n == -1) {
824 			if (errno == EILSEQ) {
825 				/* Skip an illegal wide char. */
826 				*p++ = '?';
827 				ret_val = -1;
828 			} else {
829 				ret_val = -1;
830 				break;
831 			}
832 		} else
833 			p += n;
834 		len--;
835 	}
836 	as->length = p - as->s;
837 	as->s[as->length] = '\0';
838 	return (ret_val);
839 }
840 
841 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
842 
843 /*
844  * TODO: Test if __STDC_ISO_10646__ is defined.
845  * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
846  * one character at a time.  If a non-Windows platform doesn't have
847  * either of these, fall back to the built-in UTF8 conversion.
848  */
849 int
850 archive_string_append_from_wcs(struct archive_string *as,
851     const wchar_t *w, size_t len)
852 {
853 	(void)as;/* UNUSED */
854 	(void)w;/* UNUSED */
855 	(void)len;/* UNUSED */
856 	errno = ENOSYS;
857 	return (-1);
858 }
859 
860 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
861 
862 /*
863  * Find a string conversion object by a pair of 'from' charset name
864  * and 'to' charset name from an archive object.
865  * Return NULL if not found.
866  */
867 static struct archive_string_conv *
868 find_sconv_object(struct archive *a, const char *fc, const char *tc)
869 {
870 	struct archive_string_conv *sc;
871 
872 	if (a == NULL)
873 		return (NULL);
874 
875 	for (sc = a->sconv; sc != NULL; sc = sc->next) {
876 		if (strcmp(sc->from_charset, fc) == 0 &&
877 		    strcmp(sc->to_charset, tc) == 0)
878 			break;
879 	}
880 	return (sc);
881 }
882 
883 /*
884  * Register a string object to an archive object.
885  */
886 static void
887 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
888 {
889 	struct archive_string_conv **psc;
890 
891 	/* Add a new sconv to sconv list. */
892 	psc = &(a->sconv);
893 	while (*psc != NULL)
894 		psc = &((*psc)->next);
895 	*psc = sc;
896 }
897 
898 static void
899 add_converter(struct archive_string_conv *sc, int (*converter)
900     (struct archive_string *, const void *, size_t,
901      struct archive_string_conv *))
902 {
903 	if (sc == NULL || sc->nconverter >= 2)
904 		__archive_errx(1, "Programming error");
905 	sc->converter[sc->nconverter++] = converter;
906 }
907 
908 static void
909 setup_converter(struct archive_string_conv *sc)
910 {
911 
912 	/* Reset. */
913 	sc->nconverter = 0;
914 
915 	/*
916 	 * Perform special sequence for the incorrect UTF-8 filenames
917 	 * made by libarchive2.x.
918 	 */
919 	if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
920 		add_converter(sc, strncat_from_utf8_libarchive2);
921 		return;
922 	}
923 
924 	/*
925 	 * Convert a string to UTF-16BE/LE.
926 	 */
927 	if (sc->flag & SCONV_TO_UTF16) {
928 		/*
929 		 * If the current locale is UTF-8, we can translate
930 		 * a UTF-8 string into a UTF-16BE string.
931 		 */
932 		if (sc->flag & SCONV_FROM_UTF8) {
933 			add_converter(sc, archive_string_append_unicode);
934 			return;
935 		}
936 
937 #if defined(_WIN32) && !defined(__CYGWIN__)
938 		if (sc->flag & SCONV_WIN_CP) {
939 			if (sc->flag & SCONV_TO_UTF16BE)
940 				add_converter(sc, win_strncat_to_utf16be);
941 			else
942 				add_converter(sc, win_strncat_to_utf16le);
943 			return;
944 		}
945 #endif
946 
947 #if defined(HAVE_ICONV)
948 		if (sc->cd != (iconv_t)-1) {
949 			add_converter(sc, iconv_strncat_in_locale);
950 			return;
951 		}
952 #endif
953 
954 		if (sc->flag & SCONV_BEST_EFFORT) {
955 			if (sc->flag & SCONV_TO_UTF16BE)
956 				add_converter(sc,
957 					best_effort_strncat_to_utf16be);
958 			else
959 				add_converter(sc,
960 					best_effort_strncat_to_utf16le);
961 		} else
962 			/* Make sure we have no converter. */
963 			sc->nconverter = 0;
964 		return;
965 	}
966 
967 	/*
968 	 * Convert a string from UTF-16BE/LE.
969 	 */
970 	if (sc->flag & SCONV_FROM_UTF16) {
971 		/*
972 		 * At least we should normalize a UTF-16BE string.
973 		 */
974 		if (sc->flag & SCONV_NORMALIZATION_D)
975 			add_converter(sc,archive_string_normalize_D);
976 		else if (sc->flag & SCONV_NORMALIZATION_C)
977 			add_converter(sc, archive_string_normalize_C);
978 
979 		if (sc->flag & SCONV_TO_UTF8) {
980 			/*
981 			 * If the current locale is UTF-8, we can translate
982 			 * a UTF-16BE/LE string into a UTF-8 string directly.
983 			 */
984 			if (!(sc->flag &
985 			    (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
986 				add_converter(sc,
987 				    archive_string_append_unicode);
988 			return;
989 		}
990 
991 #if defined(_WIN32) && !defined(__CYGWIN__)
992 		if (sc->flag & SCONV_WIN_CP) {
993 			if (sc->flag & SCONV_FROM_UTF16BE)
994 				add_converter(sc, win_strncat_from_utf16be);
995 			else
996 				add_converter(sc, win_strncat_from_utf16le);
997 			return;
998 		}
999 #endif
1000 
1001 #if defined(HAVE_ICONV)
1002 		if (sc->cd != (iconv_t)-1) {
1003 			add_converter(sc, iconv_strncat_in_locale);
1004 			return;
1005 		}
1006 #endif
1007 
1008 		if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1009 		    == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1010 			add_converter(sc, best_effort_strncat_from_utf16be);
1011 		else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1012 		    == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1013 			add_converter(sc, best_effort_strncat_from_utf16le);
1014 		else
1015 			/* Make sure we have no converter. */
1016 			sc->nconverter = 0;
1017 		return;
1018 	}
1019 
1020 	if (sc->flag & SCONV_FROM_UTF8) {
1021 		/*
1022 		 * At least we should normalize a UTF-8 string.
1023 		 */
1024 		if (sc->flag & SCONV_NORMALIZATION_D)
1025 			add_converter(sc,archive_string_normalize_D);
1026 		else if (sc->flag & SCONV_NORMALIZATION_C)
1027 			add_converter(sc, archive_string_normalize_C);
1028 
1029 		/*
1030 		 * Copy UTF-8 string with a check of CESU-8.
1031 		 * Apparently, iconv does not check surrogate pairs in UTF-8
1032 		 * when both from-charset and to-charset are UTF-8, and then
1033 		 * we use our UTF-8 copy code.
1034 		 */
1035 		if (sc->flag & SCONV_TO_UTF8) {
1036 			/*
1037 			 * If the current locale is UTF-8, we can translate
1038 			 * a UTF-16BE string into a UTF-8 string directly.
1039 			 */
1040 			if (!(sc->flag &
1041 			    (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1042 				add_converter(sc, strncat_from_utf8_to_utf8);
1043 			return;
1044 		}
1045 	}
1046 
1047 #if defined(_WIN32) && !defined(__CYGWIN__)
1048 	/*
1049 	 * On Windows we can use Windows API for a string conversion.
1050 	 */
1051 	if (sc->flag & SCONV_WIN_CP) {
1052 		add_converter(sc, strncat_in_codepage);
1053 		return;
1054 	}
1055 #endif
1056 
1057 #if HAVE_ICONV
1058 	if (sc->cd != (iconv_t)-1) {
1059 		add_converter(sc, iconv_strncat_in_locale);
1060 		/*
1061 		 * iconv generally does not support UTF-8-MAC and so
1062 		 * we have to the output of iconv from NFC to NFD if
1063 		 * need.
1064 		 */
1065 		if ((sc->flag & SCONV_FROM_CHARSET) &&
1066 		    (sc->flag & SCONV_TO_UTF8)) {
1067 			if (sc->flag & SCONV_NORMALIZATION_D)
1068 				add_converter(sc, archive_string_normalize_D);
1069 		}
1070 		return;
1071 	}
1072 #endif
1073 
1074 	/*
1075 	 * Try conversion in the best effort or no conversion.
1076 	 */
1077 	if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1078 		add_converter(sc, best_effort_strncat_in_locale);
1079 	else
1080 		/* Make sure we have no converter. */
1081 		sc->nconverter = 0;
1082 }
1083 
1084 /*
1085  * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1086  * and CP932 which are referenced in create_sconv_object().
1087  */
1088 static const char *
1089 canonical_charset_name(const char *charset)
1090 {
1091 	char cs[16];
1092 	char *p;
1093 	const char *s;
1094 
1095 	if (charset == NULL || charset[0] == '\0'
1096 	    || strlen(charset) > 15)
1097 		return (charset);
1098 
1099 	/* Copy name to uppercase. */
1100 	p = cs;
1101 	s = charset;
1102 	while (*s) {
1103 		char c = *s++;
1104 		if (c >= 'a' && c <= 'z')
1105 			c -= 'a' - 'A';
1106 		*p++ = c;
1107 	}
1108 	*p++ = '\0';
1109 
1110 	if (strcmp(cs, "UTF-8") == 0 ||
1111 	    strcmp(cs, "UTF8") == 0)
1112 		return ("UTF-8");
1113 	if (strcmp(cs, "UTF-16BE") == 0 ||
1114 	    strcmp(cs, "UTF16BE") == 0)
1115 		return ("UTF-16BE");
1116 	if (strcmp(cs, "UTF-16LE") == 0 ||
1117 	    strcmp(cs, "UTF16LE") == 0)
1118 		return ("UTF-16LE");
1119 	if (strcmp(cs, "CP932") == 0)
1120 		return ("CP932");
1121 	return (charset);
1122 }
1123 
1124 /*
1125  * Create a string conversion object.
1126  */
1127 static struct archive_string_conv *
1128 create_sconv_object(const char *fc, const char *tc,
1129     unsigned current_codepage, int flag)
1130 {
1131 	struct archive_string_conv *sc;
1132 
1133 	sc = calloc(1, sizeof(*sc));
1134 	if (sc == NULL)
1135 		return (NULL);
1136 	sc->next = NULL;
1137 	sc->from_charset = strdup(fc);
1138 	if (sc->from_charset == NULL) {
1139 		free(sc);
1140 		return (NULL);
1141 	}
1142 	sc->to_charset = strdup(tc);
1143 	if (sc->to_charset == NULL) {
1144 		free(sc->from_charset);
1145 		free(sc);
1146 		return (NULL);
1147 	}
1148 	archive_string_init(&sc->utftmp);
1149 
1150 	if (flag & SCONV_TO_CHARSET) {
1151 		/*
1152 		 * Convert characters from the current locale charset to
1153 		 * a specified charset.
1154 		 */
1155 		sc->from_cp = current_codepage;
1156 		sc->to_cp = make_codepage_from_charset(tc);
1157 #if defined(_WIN32) && !defined(__CYGWIN__)
1158 		if (IsValidCodePage(sc->to_cp))
1159 			flag |= SCONV_WIN_CP;
1160 #endif
1161 	} else if (flag & SCONV_FROM_CHARSET) {
1162 		/*
1163 		 * Convert characters from a specified charset to
1164 		 * the current locale charset.
1165 		 */
1166 		sc->to_cp = current_codepage;
1167 		sc->from_cp = make_codepage_from_charset(fc);
1168 #if defined(_WIN32) && !defined(__CYGWIN__)
1169 		if (IsValidCodePage(sc->from_cp))
1170 			flag |= SCONV_WIN_CP;
1171 #endif
1172 	}
1173 
1174 	/*
1175 	 * Check if "from charset" and "to charset" are the same.
1176 	 */
1177 	if (strcmp(fc, tc) == 0 ||
1178 	    (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1179 		sc->same = 1;
1180 	else
1181 		sc->same = 0;
1182 
1183 	/*
1184 	 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1185 	 */
1186 	if (strcmp(tc, "UTF-8") == 0)
1187 		flag |= SCONV_TO_UTF8;
1188 	else if (strcmp(tc, "UTF-16BE") == 0)
1189 		flag |= SCONV_TO_UTF16BE;
1190 	else if (strcmp(tc, "UTF-16LE") == 0)
1191 		flag |= SCONV_TO_UTF16LE;
1192 	if (strcmp(fc, "UTF-8") == 0)
1193 		flag |= SCONV_FROM_UTF8;
1194 	else if (strcmp(fc, "UTF-16BE") == 0)
1195 		flag |= SCONV_FROM_UTF16BE;
1196 	else if (strcmp(fc, "UTF-16LE") == 0)
1197 		flag |= SCONV_FROM_UTF16LE;
1198 #if defined(_WIN32) && !defined(__CYGWIN__)
1199 	if (sc->to_cp == CP_UTF8)
1200 		flag |= SCONV_TO_UTF8;
1201 	else if (sc->to_cp == CP_UTF16BE)
1202 		flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1203 	else if (sc->to_cp == CP_UTF16LE)
1204 		flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1205 	if (sc->from_cp == CP_UTF8)
1206 		flag |= SCONV_FROM_UTF8;
1207 	else if (sc->from_cp == CP_UTF16BE)
1208 		flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1209 	else if (sc->from_cp == CP_UTF16LE)
1210 		flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1211 #endif
1212 
1213 	/*
1214 	 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1215 	 * handle it. So we have to translate NFD characters to NFC ones
1216 	 * ourselves before iconv handles. Another reason is to prevent
1217 	 * that the same sight of two filenames, one is NFC and other
1218 	 * is NFD, would be in its directory.
1219 	 * On Mac OS X, although its filesystem layer automatically
1220 	 * convert filenames to NFD, it would be useful for filename
1221 	 * comparing to find out the same filenames that we normalize
1222 	 * that to be NFD ourselves.
1223 	 */
1224 	if ((flag & SCONV_FROM_CHARSET) &&
1225 	    (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1226 #if defined(__APPLE__)
1227 		if (flag & SCONV_TO_UTF8)
1228 			flag |= SCONV_NORMALIZATION_D;
1229 		else
1230 #endif
1231 			flag |= SCONV_NORMALIZATION_C;
1232 	}
1233 #if defined(__APPLE__)
1234 	/*
1235 	 * In case writing an archive file, make sure that a filename
1236 	 * going to be passed to iconv is a Unicode NFC string since
1237 	 * a filename in HFS Plus filesystem is a Unicode NFD one and
1238 	 * iconv cannot handle it with "UTF-8" charset. It is simpler
1239 	 * than a use of "UTF-8-MAC" charset.
1240 	 */
1241 	if ((flag & SCONV_TO_CHARSET) &&
1242 	    (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1243 	    !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1244 		flag |= SCONV_NORMALIZATION_C;
1245 	/*
1246 	 * In case reading an archive file. make sure that a filename
1247 	 * will be passed to users is a Unicode NFD string in order to
1248 	 * correctly compare the filename with other one which comes
1249 	 * from HFS Plus filesystem.
1250 	 */
1251 	if ((flag & SCONV_FROM_CHARSET) &&
1252 	   !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1253 	    (flag & SCONV_TO_UTF8))
1254 		flag |= SCONV_NORMALIZATION_D;
1255 #endif
1256 
1257 #if defined(HAVE_ICONV)
1258 	sc->cd_w = (iconv_t)-1;
1259 	/*
1260 	 * Create an iconv object.
1261 	 */
1262 	if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1263 	    (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1264 	    (flag & SCONV_WIN_CP)) {
1265 		/* This case we won't use iconv. */
1266 		sc->cd = (iconv_t)-1;
1267 	} else {
1268 		sc->cd = iconv_open(tc, fc);
1269 		if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1270 			/*
1271 			 * Unfortunately, all of iconv implements do support
1272 			 * "CP932" character-set, so we should use "SJIS"
1273 			 * instead if iconv_open failed.
1274 			 */
1275 			if (strcmp(tc, "CP932") == 0)
1276 				sc->cd = iconv_open("SJIS", fc);
1277 			else if (strcmp(fc, "CP932") == 0)
1278 				sc->cd = iconv_open(tc, "SJIS");
1279 		}
1280 #if defined(_WIN32) && !defined(__CYGWIN__)
1281 		/*
1282 		 * archive_mstring on Windows directly convert multi-bytes
1283 		 * into archive_wstring in order not to depend on locale
1284 		 * so that you can do a I18N programming. This will be
1285 		 * used only in archive_mstring_copy_mbs_len_l so far.
1286 		 */
1287 		if (flag & SCONV_FROM_CHARSET) {
1288 			sc->cd_w = iconv_open("UTF-8", fc);
1289 			if (sc->cd_w == (iconv_t)-1 &&
1290 			    (sc->flag & SCONV_BEST_EFFORT)) {
1291 				if (strcmp(fc, "CP932") == 0)
1292 					sc->cd_w = iconv_open("UTF-8", "SJIS");
1293 			}
1294 		}
1295 #endif /* _WIN32 && !__CYGWIN__ */
1296 	}
1297 #endif	/* HAVE_ICONV */
1298 
1299 	sc->flag = flag;
1300 
1301 	/*
1302 	 * Set up converters.
1303 	 */
1304 	setup_converter(sc);
1305 
1306 	return (sc);
1307 }
1308 
1309 /*
1310  * Free a string conversion object.
1311  */
1312 static void
1313 free_sconv_object(struct archive_string_conv *sc)
1314 {
1315 	free(sc->from_charset);
1316 	free(sc->to_charset);
1317 	archive_string_free(&sc->utftmp);
1318 #if HAVE_ICONV
1319 	if (sc->cd != (iconv_t)-1)
1320 		iconv_close(sc->cd);
1321 	if (sc->cd_w != (iconv_t)-1)
1322 		iconv_close(sc->cd_w);
1323 #endif
1324 	free(sc);
1325 }
1326 
1327 #if defined(_WIN32) && !defined(__CYGWIN__)
1328 # if defined(WINAPI_FAMILY_PARTITION) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
1329 #  define GetOEMCP() CP_OEMCP
1330 # endif
1331 
1332 static unsigned
1333 my_atoi(const char *p)
1334 {
1335 	unsigned cp;
1336 
1337 	cp = 0;
1338 	while (*p) {
1339 		if (*p >= '0' && *p <= '9')
1340 			cp = cp * 10 + (*p - '0');
1341 		else
1342 			return (-1);
1343 		p++;
1344 	}
1345 	return (cp);
1346 }
1347 
1348 /*
1349  * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1350  * Return -1 if failed.
1351  *
1352  * Note: This translation code may be insufficient.
1353  */
1354 static struct charset {
1355 	const char *name;
1356 	unsigned cp;
1357 } charsets[] = {
1358 	/* MUST BE SORTED! */
1359 	{"ASCII", 1252},
1360 	{"ASMO-708", 708},
1361 	{"BIG5", 950},
1362 	{"CHINESE", 936},
1363 	{"CP367", 1252},
1364 	{"CP819", 1252},
1365 	{"CP1025", 21025},
1366 	{"DOS-720", 720},
1367 	{"DOS-862", 862},
1368 	{"EUC-CN", 51936},
1369 	{"EUC-JP", 51932},
1370 	{"EUC-KR", 949},
1371 	{"EUCCN", 51936},
1372 	{"EUCJP", 51932},
1373 	{"EUCKR", 949},
1374 	{"GB18030", 54936},
1375 	{"GB2312", 936},
1376 	{"HEBREW", 1255},
1377 	{"HZ-GB-2312", 52936},
1378 	{"IBM273", 20273},
1379 	{"IBM277", 20277},
1380 	{"IBM278", 20278},
1381 	{"IBM280", 20280},
1382 	{"IBM284", 20284},
1383 	{"IBM285", 20285},
1384 	{"IBM290", 20290},
1385 	{"IBM297", 20297},
1386 	{"IBM367", 1252},
1387 	{"IBM420", 20420},
1388 	{"IBM423", 20423},
1389 	{"IBM424", 20424},
1390 	{"IBM819", 1252},
1391 	{"IBM871", 20871},
1392 	{"IBM880", 20880},
1393 	{"IBM905", 20905},
1394 	{"IBM924", 20924},
1395 	{"ISO-8859-1", 28591},
1396 	{"ISO-8859-13", 28603},
1397 	{"ISO-8859-15", 28605},
1398 	{"ISO-8859-2", 28592},
1399 	{"ISO-8859-3", 28593},
1400 	{"ISO-8859-4", 28594},
1401 	{"ISO-8859-5", 28595},
1402 	{"ISO-8859-6", 28596},
1403 	{"ISO-8859-7", 28597},
1404 	{"ISO-8859-8", 28598},
1405 	{"ISO-8859-9", 28599},
1406 	{"ISO8859-1", 28591},
1407 	{"ISO8859-13", 28603},
1408 	{"ISO8859-15", 28605},
1409 	{"ISO8859-2", 28592},
1410 	{"ISO8859-3", 28593},
1411 	{"ISO8859-4", 28594},
1412 	{"ISO8859-5", 28595},
1413 	{"ISO8859-6", 28596},
1414 	{"ISO8859-7", 28597},
1415 	{"ISO8859-8", 28598},
1416 	{"ISO8859-9", 28599},
1417 	{"JOHAB", 1361},
1418 	{"KOI8-R", 20866},
1419 	{"KOI8-U", 21866},
1420 	{"KS_C_5601-1987", 949},
1421 	{"LATIN1", 1252},
1422 	{"LATIN2", 28592},
1423 	{"MACINTOSH", 10000},
1424 	{"SHIFT-JIS", 932},
1425 	{"SHIFT_JIS", 932},
1426 	{"SJIS", 932},
1427 	{"US", 1252},
1428 	{"US-ASCII", 1252},
1429 	{"UTF-16", 1200},
1430 	{"UTF-16BE", 1201},
1431 	{"UTF-16LE", 1200},
1432 	{"UTF-8", CP_UTF8},
1433 	{"X-EUROPA", 29001},
1434 	{"X-MAC-ARABIC", 10004},
1435 	{"X-MAC-CE", 10029},
1436 	{"X-MAC-CHINESEIMP", 10008},
1437 	{"X-MAC-CHINESETRAD", 10002},
1438 	{"X-MAC-CROATIAN", 10082},
1439 	{"X-MAC-CYRILLIC", 10007},
1440 	{"X-MAC-GREEK", 10006},
1441 	{"X-MAC-HEBREW", 10005},
1442 	{"X-MAC-ICELANDIC", 10079},
1443 	{"X-MAC-JAPANESE", 10001},
1444 	{"X-MAC-KOREAN", 10003},
1445 	{"X-MAC-ROMANIAN", 10010},
1446 	{"X-MAC-THAI", 10021},
1447 	{"X-MAC-TURKISH", 10081},
1448 	{"X-MAC-UKRAINIAN", 10017},
1449 };
1450 static unsigned
1451 make_codepage_from_charset(const char *charset)
1452 {
1453 	char cs[16];
1454 	char *p;
1455 	unsigned cp;
1456 	int a, b;
1457 
1458 	if (charset == NULL || strlen(charset) > 15)
1459 		return -1;
1460 
1461 	/* Copy name to uppercase. */
1462 	p = cs;
1463 	while (*charset) {
1464 		char c = *charset++;
1465 		if (c >= 'a' && c <= 'z')
1466 			c -= 'a' - 'A';
1467 		*p++ = c;
1468 	}
1469 	*p++ = '\0';
1470 	cp = -1;
1471 
1472 	/* Look it up in the table first, so that we can easily
1473 	 * override CP367, which we map to 1252 instead of 367. */
1474 	a = 0;
1475 	b = sizeof(charsets)/sizeof(charsets[0]);
1476 	while (b > a) {
1477 		int c = (b + a) / 2;
1478 		int r = strcmp(charsets[c].name, cs);
1479 		if (r < 0)
1480 			a = c + 1;
1481 		else if (r > 0)
1482 			b = c;
1483 		else
1484 			return charsets[c].cp;
1485 	}
1486 
1487 	/* If it's not in the table, try to parse it. */
1488 	switch (*cs) {
1489 	case 'C':
1490 		if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1491 			cp = my_atoi(cs + 2);
1492 		} else if (strcmp(cs, "CP_ACP") == 0)
1493 			cp = get_current_codepage();
1494 		else if (strcmp(cs, "CP_OEMCP") == 0)
1495 			cp = get_current_oemcp();
1496 		break;
1497 	case 'I':
1498 		if (cs[1] == 'B' && cs[2] == 'M' &&
1499 		    cs[3] >= '0' && cs[3] <= '9') {
1500 			cp = my_atoi(cs + 3);
1501 		}
1502 		break;
1503 	case 'W':
1504 		if (strncmp(cs, "WINDOWS-", 8) == 0) {
1505 			cp = my_atoi(cs + 8);
1506 			if (cp != 874 && (cp < 1250 || cp > 1258))
1507 				cp = -1;/* This may invalid code. */
1508 		}
1509 		break;
1510 	}
1511 	return (cp);
1512 }
1513 
1514 /*
1515  * Return ANSI Code Page of current locale set by setlocale().
1516  */
1517 static unsigned
1518 get_current_codepage(void)
1519 {
1520 	char *locale, *p;
1521 	unsigned cp;
1522 
1523 	locale = setlocale(LC_CTYPE, NULL);
1524 	if (locale == NULL)
1525 		return (GetACP());
1526 	if (locale[0] == 'C' && locale[1] == '\0')
1527 		return (CP_C_LOCALE);
1528 	p = strrchr(locale, '.');
1529 	if (p == NULL)
1530 		return (GetACP());
1531 	if ((strcmp(p+1, "utf8") == 0) || (strcmp(p+1, "UTF-8") == 0))
1532 		return CP_UTF8;
1533 	cp = my_atoi(p+1);
1534 	if ((int)cp <= 0)
1535 		return (GetACP());
1536 	return (cp);
1537 }
1538 
1539 /*
1540  * Translation table between Locale Name and ACP/OEMCP.
1541  */
1542 static struct {
1543 	unsigned acp;
1544 	unsigned ocp;
1545 	const char *locale;
1546 } acp_ocp_map[] = {
1547 	{  950,  950, "Chinese_Taiwan" },
1548 	{  936,  936, "Chinese_People's Republic of China" },
1549 	{  950,  950, "Chinese_Taiwan" },
1550 	{ 1250,  852, "Czech_Czech Republic" },
1551 	{ 1252,  850, "Danish_Denmark" },
1552 	{ 1252,  850, "Dutch_Netherlands" },
1553 	{ 1252,  850, "Dutch_Belgium" },
1554 	{ 1252,  437, "English_United States" },
1555 	{ 1252,  850, "English_Australia" },
1556 	{ 1252,  850, "English_Canada" },
1557 	{ 1252,  850, "English_New Zealand" },
1558 	{ 1252,  850, "English_United Kingdom" },
1559 	{ 1252,  437, "English_United States" },
1560 	{ 1252,  850, "Finnish_Finland" },
1561 	{ 1252,  850, "French_France" },
1562 	{ 1252,  850, "French_Belgium" },
1563 	{ 1252,  850, "French_Canada" },
1564 	{ 1252,  850, "French_Switzerland" },
1565 	{ 1252,  850, "German_Germany" },
1566 	{ 1252,  850, "German_Austria" },
1567 	{ 1252,  850, "German_Switzerland" },
1568 	{ 1253,  737, "Greek_Greece" },
1569 	{ 1250,  852, "Hungarian_Hungary" },
1570 	{ 1252,  850, "Icelandic_Iceland" },
1571 	{ 1252,  850, "Italian_Italy" },
1572 	{ 1252,  850, "Italian_Switzerland" },
1573 	{  932,  932, "Japanese_Japan" },
1574 	{  949,  949, "Korean_Korea" },
1575 	{ 1252,  850, "Norwegian (BokmOl)_Norway" },
1576 	{ 1252,  850, "Norwegian (BokmOl)_Norway" },
1577 	{ 1252,  850, "Norwegian-Nynorsk_Norway" },
1578 	{ 1250,  852, "Polish_Poland" },
1579 	{ 1252,  850, "Portuguese_Portugal" },
1580 	{ 1252,  850, "Portuguese_Brazil" },
1581 	{ 1251,  866, "Russian_Russia" },
1582 	{ 1250,  852, "Slovak_Slovakia" },
1583 	{ 1252,  850, "Spanish_Spain" },
1584 	{ 1252,  850, "Spanish_Mexico" },
1585 	{ 1252,  850, "Spanish_Spain" },
1586 	{ 1252,  850, "Swedish_Sweden" },
1587 	{ 1254,  857, "Turkish_Turkey" },
1588 	{ 0, 0, NULL}
1589 };
1590 
1591 /*
1592  * Return OEM Code Page of current locale set by setlocale().
1593  */
1594 static unsigned
1595 get_current_oemcp(void)
1596 {
1597 	int i;
1598 	char *locale, *p;
1599 	size_t len;
1600 
1601 	locale = setlocale(LC_CTYPE, NULL);
1602 	if (locale == NULL)
1603 		return (GetOEMCP());
1604 	if (locale[0] == 'C' && locale[1] == '\0')
1605 		return (CP_C_LOCALE);
1606 
1607 	p = strrchr(locale, '.');
1608 	if (p == NULL)
1609 		return (GetOEMCP());
1610 	len = p - locale;
1611 	for (i = 0; acp_ocp_map[i].acp; i++) {
1612 		if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1613 			return (acp_ocp_map[i].ocp);
1614 	}
1615 	return (GetOEMCP());
1616 }
1617 #else
1618 
1619 /*
1620  * POSIX platform does not use CodePage.
1621  */
1622 
1623 static unsigned
1624 get_current_codepage(void)
1625 {
1626 	return (-1);/* Unknown */
1627 }
1628 static unsigned
1629 make_codepage_from_charset(const char *charset)
1630 {
1631 	(void)charset; /* UNUSED */
1632 	return (-1);/* Unknown */
1633 }
1634 static unsigned
1635 get_current_oemcp(void)
1636 {
1637 	return (-1);/* Unknown */
1638 }
1639 
1640 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1641 
1642 /*
1643  * Return a string conversion object.
1644  */
1645 static struct archive_string_conv *
1646 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1647 {
1648 	struct archive_string_conv *sc;
1649 	unsigned current_codepage;
1650 
1651 	/* Check if we have made the sconv object. */
1652 	sc = find_sconv_object(a, fc, tc);
1653 	if (sc != NULL)
1654 		return (sc);
1655 
1656 	if (a == NULL)
1657 		current_codepage = get_current_codepage();
1658 	else
1659 		current_codepage = a->current_codepage;
1660 
1661 	sc = create_sconv_object(canonical_charset_name(fc),
1662 	    canonical_charset_name(tc), current_codepage, flag);
1663 	if (sc == NULL) {
1664 		if (a != NULL)
1665 			archive_set_error(a, ENOMEM,
1666 			    "Could not allocate memory for "
1667 			    "a string conversion object");
1668 		return (NULL);
1669 	}
1670 
1671 	/*
1672 	 * If there is no converter for current string conversion object,
1673 	 * we cannot handle this conversion.
1674 	 */
1675 	if (sc->nconverter == 0) {
1676 		if (a != NULL) {
1677 #if HAVE_ICONV
1678 			archive_set_error(a, ARCHIVE_ERRNO_MISC,
1679 			    "iconv_open failed : Cannot handle ``%s''",
1680 			    (flag & SCONV_TO_CHARSET)?tc:fc);
1681 #else
1682 			archive_set_error(a, ARCHIVE_ERRNO_MISC,
1683 			    "A character-set conversion not fully supported "
1684 			    "on this platform");
1685 #endif
1686 		}
1687 		/* Failed; free a sconv object. */
1688 		free_sconv_object(sc);
1689 		return (NULL);
1690 	}
1691 
1692 	/*
1693 	 * Success!
1694 	 */
1695 	if (a != NULL)
1696 		add_sconv_object(a, sc);
1697 	return (sc);
1698 }
1699 
1700 static const char *
1701 get_current_charset(struct archive *a)
1702 {
1703 	const char *cur_charset;
1704 
1705 	if (a == NULL)
1706 		cur_charset = default_iconv_charset("");
1707 	else {
1708 		cur_charset = default_iconv_charset(a->current_code);
1709 		if (a->current_code == NULL) {
1710 			a->current_code = strdup(cur_charset);
1711 			a->current_codepage = get_current_codepage();
1712 			a->current_oemcp = get_current_oemcp();
1713 		}
1714 	}
1715 	return (cur_charset);
1716 }
1717 
1718 /*
1719  * Make and Return a string conversion object.
1720  * Return NULL if the platform does not support the specified conversion
1721  * and best_effort is 0.
1722  * If best_effort is set, A string conversion object must be returned
1723  * unless memory allocation for the object fails, but the conversion
1724  * might fail when non-ASCII code is found.
1725  */
1726 struct archive_string_conv *
1727 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1728     int best_effort)
1729 {
1730 	int flag = SCONV_TO_CHARSET;
1731 
1732 	if (best_effort)
1733 		flag |= SCONV_BEST_EFFORT;
1734 	return (get_sconv_object(a, get_current_charset(a), charset, flag));
1735 }
1736 
1737 struct archive_string_conv *
1738 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1739     int best_effort)
1740 {
1741 	int flag = SCONV_FROM_CHARSET;
1742 
1743 	if (best_effort)
1744 		flag |= SCONV_BEST_EFFORT;
1745 	return (get_sconv_object(a, charset, get_current_charset(a), flag));
1746 }
1747 
1748 /*
1749  * archive_string_default_conversion_*_archive() are provided for Windows
1750  * platform because other archiver application use CP_OEMCP for
1751  * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1752  * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1753  * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1754  * So we should make a string conversion between CP_ACP and CP_OEMCP
1755  * for compatibility.
1756  */
1757 #if defined(_WIN32) && !defined(__CYGWIN__)
1758 struct archive_string_conv *
1759 archive_string_default_conversion_for_read(struct archive *a)
1760 {
1761 	const char *cur_charset = get_current_charset(a);
1762 	char oemcp[16];
1763 
1764 	/* NOTE: a check of cur_charset is unneeded but we need
1765 	 * that get_current_charset() has been surely called at
1766 	 * this time whatever C compiler optimized. */
1767 	if (cur_charset != NULL &&
1768 	    (a->current_codepage == CP_C_LOCALE ||
1769 	     a->current_codepage == a->current_oemcp))
1770 		return (NULL);/* no conversion. */
1771 
1772 	_snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1773 	/* Make sure a null termination must be set. */
1774 	oemcp[sizeof(oemcp)-1] = '\0';
1775 	return (get_sconv_object(a, oemcp, cur_charset,
1776 	    SCONV_FROM_CHARSET));
1777 }
1778 
1779 struct archive_string_conv *
1780 archive_string_default_conversion_for_write(struct archive *a)
1781 {
1782 	const char *cur_charset = get_current_charset(a);
1783 	char oemcp[16];
1784 
1785 	/* NOTE: a check of cur_charset is unneeded but we need
1786 	 * that get_current_charset() has been surely called at
1787 	 * this time whatever C compiler optimized. */
1788 	if (cur_charset != NULL &&
1789 	    (a->current_codepage == CP_C_LOCALE ||
1790 	     a->current_codepage == a->current_oemcp))
1791 		return (NULL);/* no conversion. */
1792 
1793 	_snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1794 	/* Make sure a null termination must be set. */
1795 	oemcp[sizeof(oemcp)-1] = '\0';
1796 	return (get_sconv_object(a, cur_charset, oemcp,
1797 	    SCONV_TO_CHARSET));
1798 }
1799 #else
1800 struct archive_string_conv *
1801 archive_string_default_conversion_for_read(struct archive *a)
1802 {
1803 	(void)a; /* UNUSED */
1804 	return (NULL);
1805 }
1806 
1807 struct archive_string_conv *
1808 archive_string_default_conversion_for_write(struct archive *a)
1809 {
1810 	(void)a; /* UNUSED */
1811 	return (NULL);
1812 }
1813 #endif
1814 
1815 /*
1816  * Dispose of all character conversion objects in the archive object.
1817  */
1818 void
1819 archive_string_conversion_free(struct archive *a)
1820 {
1821 	struct archive_string_conv *sc;
1822 	struct archive_string_conv *sc_next;
1823 
1824 	for (sc = a->sconv; sc != NULL; sc = sc_next) {
1825 		sc_next = sc->next;
1826 		free_sconv_object(sc);
1827 	}
1828 	a->sconv = NULL;
1829 	free(a->current_code);
1830 	a->current_code = NULL;
1831 }
1832 
1833 /*
1834  * Return a conversion charset name.
1835  */
1836 const char *
1837 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1838 {
1839 	if (sc->flag & SCONV_TO_CHARSET)
1840 		return (sc->to_charset);
1841 	else
1842 		return (sc->from_charset);
1843 }
1844 
1845 /*
1846  * Change the behavior of a string conversion.
1847  */
1848 void
1849 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1850 {
1851 	switch (opt) {
1852 	/*
1853 	 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1854 	 * assumption that wchar_t was Unicode.
1855 	 * This option enables simulating the assumption in order to read
1856 	 * that filename correctly.
1857 	 */
1858 	case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1859 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1860 	 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1861 		/*
1862 		 * Nothing to do for it since wchar_t on these platforms
1863 		 * is really Unicode.
1864 		 */
1865 		(void)sc; /* UNUSED */
1866 #else
1867 		if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1868 			sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1869 			/* Set up string converters. */
1870 			setup_converter(sc);
1871 		}
1872 #endif
1873 		break;
1874 	case SCONV_SET_OPT_NORMALIZATION_C:
1875 		if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1876 			sc->flag |= SCONV_NORMALIZATION_C;
1877 			sc->flag &= ~SCONV_NORMALIZATION_D;
1878 			/* Set up string converters. */
1879 			setup_converter(sc);
1880 		}
1881 		break;
1882 	case SCONV_SET_OPT_NORMALIZATION_D:
1883 #if defined(HAVE_ICONV)
1884 		/*
1885 		 * If iconv will take the string, do not change the
1886 		 * setting of the normalization.
1887 		 */
1888 		if (!(sc->flag & SCONV_WIN_CP) &&
1889 		     (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1890 		    !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1891 			break;
1892 #endif
1893 		if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1894 			sc->flag |= SCONV_NORMALIZATION_D;
1895 			sc->flag &= ~SCONV_NORMALIZATION_C;
1896 			/* Set up string converters. */
1897 			setup_converter(sc);
1898 		}
1899 		break;
1900 	default:
1901 		break;
1902 	}
1903 }
1904 
1905 /*
1906  *
1907  * Copy one archive_string to another in locale conversion.
1908  *
1909  *	archive_strncat_l();
1910  *	archive_strncpy_l();
1911  *
1912  */
1913 
1914 static size_t
1915 mbsnbytes(const void *_p, size_t n)
1916 {
1917 	size_t s;
1918 	const char *p, *pp;
1919 
1920 	if (_p == NULL)
1921 		return (0);
1922 	p = (const char *)_p;
1923 
1924 	/* Like strlen(p), except won't examine positions beyond p[n]. */
1925 	s = 0;
1926 	pp = p;
1927 	while (s < n && *pp) {
1928 		pp++;
1929 		s++;
1930 	}
1931 	return (s);
1932 }
1933 
1934 static size_t
1935 utf16nbytes(const void *_p, size_t n)
1936 {
1937 	size_t s;
1938 	const char *p, *pp;
1939 
1940 	if (_p == NULL)
1941 		return (0);
1942 	p = (const char *)_p;
1943 
1944 	/* Like strlen(p), except won't examine positions beyond p[n]. */
1945 	s = 0;
1946 	pp = p;
1947 	n >>= 1;
1948 	while (s < n && (pp[0] || pp[1])) {
1949 		pp += 2;
1950 		s++;
1951 	}
1952 	return (s<<1);
1953 }
1954 
1955 int
1956 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1957     struct archive_string_conv *sc)
1958 {
1959 	as->length = 0;
1960 	return (archive_strncat_l(as, _p, n, sc));
1961 }
1962 
1963 int
1964 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1965     struct archive_string_conv *sc)
1966 {
1967 	const void *s;
1968 	size_t length = 0;
1969 	int i, r = 0, r2;
1970 
1971 	if (_p != NULL && n > 0) {
1972 		if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
1973 			length = utf16nbytes(_p, n);
1974 		else
1975 			length = mbsnbytes(_p, n);
1976 	}
1977 
1978 	/* We must allocate memory even if there is no data for conversion
1979 	 * or copy. This simulates archive_string_append behavior. */
1980 	if (length == 0) {
1981 		int tn = 1;
1982 		if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1983 			tn = 2;
1984 		if (archive_string_ensure(as, as->length + tn) == NULL)
1985 			return (-1);
1986 		as->s[as->length] = 0;
1987 		if (tn == 2)
1988 			as->s[as->length+1] = 0;
1989 		return (0);
1990 	}
1991 
1992 	/*
1993 	 * If sc is NULL, we just make a copy.
1994 	 */
1995 	if (sc == NULL) {
1996 		if (archive_string_append(as, _p, length) == NULL)
1997 			return (-1);/* No memory */
1998 		return (0);
1999 	}
2000 
2001 	s = _p;
2002 	i = 0;
2003 	if (sc->nconverter > 1) {
2004 		sc->utftmp.length = 0;
2005 		r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2006 		if (r2 != 0 && errno == ENOMEM)
2007 			return (r2);
2008 		if (r > r2)
2009 			r = r2;
2010 		s = sc->utftmp.s;
2011 		length = sc->utftmp.length;
2012 		++i;
2013 	}
2014 	r2 = sc->converter[i](as, s, length, sc);
2015 	if (r > r2)
2016 		r = r2;
2017 	return (r);
2018 }
2019 
2020 #if HAVE_ICONV
2021 
2022 /*
2023  * Return -1 if conversion fails.
2024  */
2025 static int
2026 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2027     size_t length, struct archive_string_conv *sc)
2028 {
2029 	ICONV_CONST char *itp;
2030 	size_t remaining;
2031 	iconv_t cd;
2032 	char *outp;
2033 	size_t avail, bs;
2034 	int return_value = 0; /* success */
2035 	int to_size, from_size;
2036 
2037 	if (sc->flag & SCONV_TO_UTF16)
2038 		to_size = 2;
2039 	else
2040 		to_size = 1;
2041 	if (sc->flag & SCONV_FROM_UTF16)
2042 		from_size = 2;
2043 	else
2044 		from_size = 1;
2045 
2046 	if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2047 		return (-1);
2048 
2049 	cd = sc->cd;
2050 	itp = (char *)(uintptr_t)_p;
2051 	remaining = length;
2052 	outp = as->s + as->length;
2053 	avail = as->buffer_length - as->length - to_size;
2054 	while (remaining >= (size_t)from_size) {
2055 		size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2056 
2057 		if (result != (size_t)-1)
2058 			break; /* Conversion completed. */
2059 
2060 		if (errno == EILSEQ || errno == EINVAL) {
2061 			/*
2062 		 	 * If an output charset is UTF-8 or UTF-16BE/LE,
2063 			 * unknown character should be U+FFFD
2064 			 * (replacement character).
2065 			 */
2066 			if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2067 				size_t rbytes;
2068 				if (sc->flag & SCONV_TO_UTF8)
2069 					rbytes = sizeof(utf8_replacement_char);
2070 				else
2071 					rbytes = 2;
2072 
2073 				if (avail < rbytes) {
2074 					as->length = outp - as->s;
2075 					bs = as->buffer_length +
2076 					    (remaining * to_size) + rbytes;
2077 					if (NULL ==
2078 					    archive_string_ensure(as, bs))
2079 						return (-1);
2080 					outp = as->s + as->length;
2081 					avail = as->buffer_length
2082 					    - as->length - to_size;
2083 				}
2084 				if (sc->flag & SCONV_TO_UTF8)
2085 					memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2086 				else if (sc->flag & SCONV_TO_UTF16BE)
2087 					archive_be16enc(outp, UNICODE_R_CHAR);
2088 				else
2089 					archive_le16enc(outp, UNICODE_R_CHAR);
2090 				outp += rbytes;
2091 				avail -= rbytes;
2092 			} else {
2093 				/* Skip the illegal input bytes. */
2094 				*outp++ = '?';
2095 				avail--;
2096 			}
2097 			itp += from_size;
2098 			remaining -= from_size;
2099 			return_value = -1; /* failure */
2100 		} else {
2101 			/* E2BIG no output buffer,
2102 			 * Increase an output buffer.  */
2103 			as->length = outp - as->s;
2104 			bs = as->buffer_length + remaining * 2;
2105 			if (NULL == archive_string_ensure(as, bs))
2106 				return (-1);
2107 			outp = as->s + as->length;
2108 			avail = as->buffer_length - as->length - to_size;
2109 		}
2110 	}
2111 	as->length = outp - as->s;
2112 	as->s[as->length] = 0;
2113 	if (to_size == 2)
2114 		as->s[as->length+1] = 0;
2115 	return (return_value);
2116 }
2117 
2118 #endif /* HAVE_ICONV */
2119 
2120 
2121 #if defined(_WIN32) && !defined(__CYGWIN__)
2122 
2123 /*
2124  * Translate a string from a some CodePage to an another CodePage by
2125  * Windows APIs, and copy the result. Return -1 if conversion fails.
2126  */
2127 static int
2128 strncat_in_codepage(struct archive_string *as,
2129     const void *_p, size_t length, struct archive_string_conv *sc)
2130 {
2131 	const char *s = (const char *)_p;
2132 	struct archive_wstring aws;
2133 	size_t l;
2134 	int r, saved_flag;
2135 
2136 	archive_string_init(&aws);
2137 	saved_flag = sc->flag;
2138 	sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2139 	r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2140 	sc->flag = saved_flag;
2141 	if (r != 0) {
2142 		archive_wstring_free(&aws);
2143 		if (errno != ENOMEM)
2144 			archive_string_append(as, s, length);
2145 		return (-1);
2146 	}
2147 
2148 	l = as->length;
2149 	r = archive_string_append_from_wcs_in_codepage(
2150 	    as, aws.s, aws.length, sc);
2151 	if (r != 0 && errno != ENOMEM && l == as->length)
2152 		archive_string_append(as, s, length);
2153 	archive_wstring_free(&aws);
2154 	return (r);
2155 }
2156 
2157 /*
2158  * Test whether MBS ==> WCS is okay.
2159  */
2160 static int
2161 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2162 {
2163 	const char *p = (const char *)_p;
2164 	unsigned codepage;
2165 	DWORD mbflag = MB_ERR_INVALID_CHARS;
2166 
2167 	if (sc->flag & SCONV_FROM_CHARSET)
2168 		codepage = sc->to_cp;
2169 	else
2170 		codepage = sc->from_cp;
2171 
2172 	if (codepage == CP_C_LOCALE)
2173 		return (0);
2174 	if (codepage != CP_UTF8)
2175 		mbflag |= MB_PRECOMPOSED;
2176 
2177 	if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2178 		return (-1); /* Invalid */
2179 	return (0); /* Okay */
2180 }
2181 
2182 #else
2183 
2184 /*
2185  * Test whether MBS ==> WCS is okay.
2186  */
2187 static int
2188 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2189 {
2190 	const char *p = (const char *)_p;
2191 	size_t r;
2192 
2193 #if HAVE_MBRTOWC
2194 	mbstate_t shift_state;
2195 
2196 	memset(&shift_state, 0, sizeof(shift_state));
2197 #else
2198 	/* Clear the shift state before starting. */
2199 	mbtowc(NULL, NULL, 0);
2200 #endif
2201 	while (n) {
2202 		wchar_t wc;
2203 
2204 #if HAVE_MBRTOWC
2205 		r = mbrtowc(&wc, p, n, &shift_state);
2206 #else
2207 		r = mbtowc(&wc, p, n);
2208 #endif
2209 		if (r == (size_t)-1 || r == (size_t)-2)
2210 			return (-1);/* Invalid. */
2211 		if (r == 0)
2212 			break;
2213 		p += r;
2214 		n -= r;
2215 	}
2216 	(void)sc; /* UNUSED */
2217 	return (0); /* All Okey. */
2218 }
2219 
2220 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2221 
2222 /*
2223  * Basically returns -1 because we cannot make a conversion of charset
2224  * without iconv but in some cases this would return 0.
2225  * Returns 0 if all copied characters are ASCII.
2226  * Returns 0 if both from-locale and to-locale are the same and those
2227  * can be WCS with no error.
2228  */
2229 static int
2230 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2231     size_t length, struct archive_string_conv *sc)
2232 {
2233 	size_t remaining;
2234 	const uint8_t *itp;
2235 	int return_value = 0; /* success */
2236 
2237 	/*
2238 	 * If both from-locale and to-locale is the same, this makes a copy.
2239 	 * And then this checks all copied MBS can be WCS if so returns 0.
2240 	 */
2241 	if (sc->same) {
2242 		if (archive_string_append(as, _p, length) == NULL)
2243 			return (-1);/* No memory */
2244 		return (invalid_mbs(_p, length, sc));
2245 	}
2246 
2247 	/*
2248 	 * If a character is ASCII, this just copies it. If not, this
2249 	 * assigns '?' character instead but in UTF-8 locale this assigns
2250 	 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2251 	 * a Replacement Character in Unicode.
2252 	 */
2253 
2254 	remaining = length;
2255 	itp = (const uint8_t *)_p;
2256 	while (*itp && remaining > 0) {
2257 		if (*itp > 127) {
2258 			// Non-ASCII: Substitute with suitable replacement
2259 			if (sc->flag & SCONV_TO_UTF8) {
2260 				if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2261 					__archive_errx(1, "Out of memory");
2262 				}
2263 			} else {
2264 				archive_strappend_char(as, '?');
2265 			}
2266 			return_value = -1;
2267 		} else {
2268 			archive_strappend_char(as, *itp);
2269 		}
2270 		++itp;
2271 	}
2272 	return (return_value);
2273 }
2274 
2275 
2276 /*
2277  * Unicode conversion functions.
2278  *   - UTF-8 <===> UTF-8 in removing surrogate pairs.
2279  *   - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2280  *   - UTF-8 made by libarchive 2.x ===> UTF-8.
2281  *   - UTF-16BE <===> UTF-8.
2282  *
2283  */
2284 
2285 /*
2286  * Utility to convert a single UTF-8 sequence.
2287  *
2288  * Usually return used bytes, return used byte in negative value when
2289  * a unicode character is replaced with U+FFFD.
2290  * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2291  * Recommended Practice for Replacement Characters.
2292  */
2293 static int
2294 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2295 {
2296 	static const char utf8_count[256] = {
2297 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2298 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2299 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2300 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2301 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2302 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2303 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2304 		 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2305 		 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2306 		 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2307 		 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2308 		 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2309 		 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2310 		 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2311 		 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2312 		 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2313 	};
2314 	int ch, i;
2315 	int cnt;
2316 	uint32_t wc;
2317 
2318 	/* Sanity check. */
2319 	if (n == 0)
2320 		return (0);
2321 	/*
2322 	 * Decode 1-4 bytes depending on the value of the first byte.
2323 	 */
2324 	ch = (unsigned char)*s;
2325 	if (ch == 0)
2326 		return (0); /* Standard:  return 0 for end-of-string. */
2327 	cnt = utf8_count[ch];
2328 
2329 	/* Invalid sequence or there are not plenty bytes. */
2330 	if ((int)n < cnt) {
2331 		cnt = (int)n;
2332 		for (i = 1; i < cnt; i++) {
2333 			if ((s[i] & 0xc0) != 0x80) {
2334 				cnt = i;
2335 				break;
2336 			}
2337 		}
2338 		goto invalid_sequence;
2339 	}
2340 
2341 	/* Make a Unicode code point from a single UTF-8 sequence. */
2342 	switch (cnt) {
2343 	case 1:	/* 1 byte sequence. */
2344 		*pwc = ch & 0x7f;
2345 		return (cnt);
2346 	case 2:	/* 2 bytes sequence. */
2347 		if ((s[1] & 0xc0) != 0x80) {
2348 			cnt = 1;
2349 			goto invalid_sequence;
2350 		}
2351 		*pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2352 		return (cnt);
2353 	case 3:	/* 3 bytes sequence. */
2354 		if ((s[1] & 0xc0) != 0x80) {
2355 			cnt = 1;
2356 			goto invalid_sequence;
2357 		}
2358 		if ((s[2] & 0xc0) != 0x80) {
2359 			cnt = 2;
2360 			goto invalid_sequence;
2361 		}
2362 		wc = ((ch & 0x0f) << 12)
2363 		    | ((s[1] & 0x3f) << 6)
2364 		    | (s[2] & 0x3f);
2365 		if (wc < 0x800)
2366 			goto invalid_sequence;/* Overlong sequence. */
2367 		break;
2368 	case 4:	/* 4 bytes sequence. */
2369 		if ((s[1] & 0xc0) != 0x80) {
2370 			cnt = 1;
2371 			goto invalid_sequence;
2372 		}
2373 		if ((s[2] & 0xc0) != 0x80) {
2374 			cnt = 2;
2375 			goto invalid_sequence;
2376 		}
2377 		if ((s[3] & 0xc0) != 0x80) {
2378 			cnt = 3;
2379 			goto invalid_sequence;
2380 		}
2381 		wc = ((ch & 0x07) << 18)
2382 		    | ((s[1] & 0x3f) << 12)
2383 		    | ((s[2] & 0x3f) << 6)
2384 		    | (s[3] & 0x3f);
2385 		if (wc < 0x10000)
2386 			goto invalid_sequence;/* Overlong sequence. */
2387 		break;
2388 	default: /* Others are all invalid sequence. */
2389 		if (ch == 0xc0 || ch == 0xc1)
2390 			cnt = 2;
2391 		else if (ch >= 0xf5 && ch <= 0xf7)
2392 			cnt = 4;
2393 		else if (ch >= 0xf8 && ch <= 0xfb)
2394 			cnt = 5;
2395 		else if (ch == 0xfc || ch == 0xfd)
2396 			cnt = 6;
2397 		else
2398 			cnt = 1;
2399 		if ((int)n < cnt)
2400 			cnt = (int)n;
2401 		for (i = 1; i < cnt; i++) {
2402 			if ((s[i] & 0xc0) != 0x80) {
2403 				cnt = i;
2404 				break;
2405 			}
2406 		}
2407 		goto invalid_sequence;
2408 	}
2409 
2410 	/* The code point larger than 0x10FFFF is not legal
2411 	 * Unicode values. */
2412 	if (wc > UNICODE_MAX)
2413 		goto invalid_sequence;
2414 	/* Correctly gets a Unicode, returns used bytes. */
2415 	*pwc = wc;
2416 	return (cnt);
2417 invalid_sequence:
2418 	*pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2419 	return (cnt * -1);
2420 }
2421 
2422 static int
2423 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2424 {
2425 	int cnt;
2426 
2427 	cnt = _utf8_to_unicode(pwc, s, n);
2428 	/* Any of Surrogate pair is not legal Unicode values. */
2429 	if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2430 		return (-3);
2431 	return (cnt);
2432 }
2433 
2434 static inline uint32_t
2435 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2436 {
2437 	uc -= 0xD800;
2438 	uc *= 0x400;
2439 	uc += uc2 - 0xDC00;
2440 	uc += 0x10000;
2441 	return (uc);
2442 }
2443 
2444 /*
2445  * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2446  * removing surrogate pairs.
2447  *
2448  * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2449  *
2450  * Usually return used bytes, return used byte in negative value when
2451  * a unicode character is replaced with U+FFFD.
2452  */
2453 static int
2454 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2455 {
2456 	uint32_t wc = 0;
2457 	int cnt;
2458 
2459 	cnt = _utf8_to_unicode(&wc, s, n);
2460 	if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2461 		uint32_t wc2 = 0;
2462 		if (n - 3 < 3) {
2463 			/* Invalid byte sequence. */
2464 			goto invalid_sequence;
2465 		}
2466 		cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2467 		if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2468 			/* Invalid byte sequence. */
2469 			goto invalid_sequence;
2470 		}
2471 		wc = combine_surrogate_pair(wc, wc2);
2472 		cnt = 6;
2473 	} else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2474 		/* Invalid byte sequence. */
2475 		goto invalid_sequence;
2476 	}
2477 	*pwc = wc;
2478 	return (cnt);
2479 invalid_sequence:
2480 	*pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2481 	if (cnt > 0)
2482 		cnt *= -1;
2483 	return (cnt);
2484 }
2485 
2486 /*
2487  * Convert a Unicode code point to a single UTF-8 sequence.
2488  *
2489  * NOTE:This function does not check if the Unicode is legal or not.
2490  * Please you definitely check it before calling this.
2491  */
2492 static size_t
2493 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2494 {
2495 	char *_p = p;
2496 
2497 	/* Invalid Unicode char maps to Replacement character */
2498 	if (uc > UNICODE_MAX)
2499 		uc = UNICODE_R_CHAR;
2500 	/* Translate code point to UTF8 */
2501 	if (uc <= 0x7f) {
2502 		if (remaining == 0)
2503 			return (0);
2504 		*p++ = (char)uc;
2505 	} else if (uc <= 0x7ff) {
2506 		if (remaining < 2)
2507 			return (0);
2508 		*p++ = 0xc0 | ((uc >> 6) & 0x1f);
2509 		*p++ = 0x80 | (uc & 0x3f);
2510 	} else if (uc <= 0xffff) {
2511 		if (remaining < 3)
2512 			return (0);
2513 		*p++ = 0xe0 | ((uc >> 12) & 0x0f);
2514 		*p++ = 0x80 | ((uc >> 6) & 0x3f);
2515 		*p++ = 0x80 | (uc & 0x3f);
2516 	} else {
2517 		if (remaining < 4)
2518 			return (0);
2519 		*p++ = 0xf0 | ((uc >> 18) & 0x07);
2520 		*p++ = 0x80 | ((uc >> 12) & 0x3f);
2521 		*p++ = 0x80 | ((uc >> 6) & 0x3f);
2522 		*p++ = 0x80 | (uc & 0x3f);
2523 	}
2524 	return (p - _p);
2525 }
2526 
2527 static int
2528 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2529 {
2530 	return (utf16_to_unicode(pwc, s, n, 1));
2531 }
2532 
2533 static int
2534 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2535 {
2536 	return (utf16_to_unicode(pwc, s, n, 0));
2537 }
2538 
2539 static int
2540 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2541 {
2542 	const char *utf16 = s;
2543 	unsigned uc;
2544 
2545 	if (n == 0)
2546 		return (0);
2547 	if (n == 1) {
2548 		/* set the Replacement Character instead. */
2549 		*pwc = UNICODE_R_CHAR;
2550 		return (-1);
2551 	}
2552 
2553 	if (be)
2554 		uc = archive_be16dec(utf16);
2555 	else
2556 		uc = archive_le16dec(utf16);
2557 	utf16 += 2;
2558 
2559 	/* If this is a surrogate pair, assemble the full code point.*/
2560 	if (IS_HIGH_SURROGATE_LA(uc)) {
2561 		unsigned uc2;
2562 
2563 		if (n >= 4) {
2564 			if (be)
2565 				uc2 = archive_be16dec(utf16);
2566 			else
2567 				uc2 = archive_le16dec(utf16);
2568 		} else
2569 			uc2 = 0;
2570 		if (IS_LOW_SURROGATE_LA(uc2)) {
2571 			uc = combine_surrogate_pair(uc, uc2);
2572 			utf16 += 2;
2573 		} else {
2574 	 		/* Undescribed code point should be U+FFFD
2575 		 	* (replacement character). */
2576 			*pwc = UNICODE_R_CHAR;
2577 			return (-2);
2578 		}
2579 	}
2580 
2581 	/*
2582 	 * Surrogate pair values(0xd800 through 0xdfff) are only
2583 	 * used by UTF-16, so, after above calculation, the code
2584 	 * must not be surrogate values, and Unicode has no codes
2585 	 * larger than 0x10ffff. Thus, those are not legal Unicode
2586 	 * values.
2587 	 */
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);
2593 	}
2594 	*pwc = uc;
2595 	return ((int)(utf16 - s));
2596 }
2597 
2598 static size_t
2599 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2600 {
2601 	char *utf16 = p;
2602 
2603 	if (uc > 0xffff) {
2604 		/* We have a code point that won't fit into a
2605 		 * wchar_t; convert it to a surrogate pair. */
2606 		if (remaining < 4)
2607 			return (0);
2608 		uc -= 0x10000;
2609 		archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2610 		archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2611 		return (4);
2612 	} else {
2613 		if (remaining < 2)
2614 			return (0);
2615 		archive_be16enc(utf16, uc);
2616 		return (2);
2617 	}
2618 }
2619 
2620 static size_t
2621 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2622 {
2623 	char *utf16 = p;
2624 
2625 	if (uc > 0xffff) {
2626 		/* We have a code point that won't fit into a
2627 		 * wchar_t; convert it to a surrogate pair. */
2628 		if (remaining < 4)
2629 			return (0);
2630 		uc -= 0x10000;
2631 		archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2632 		archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2633 		return (4);
2634 	} else {
2635 		if (remaining < 2)
2636 			return (0);
2637 		archive_le16enc(utf16, uc);
2638 		return (2);
2639 	}
2640 }
2641 
2642 /*
2643  * Append new UTF-8 string to existing UTF-8 string.
2644  * Existing string is assumed to already be in proper form;
2645  * the new string will have invalid sequences replaced and
2646  * surrogate pairs canonicalized.
2647  */
2648 static int
2649 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2650     size_t len, struct archive_string_conv *sc)
2651 {
2652 	int ret = 0;
2653 	const char *src = _src;
2654 	(void)sc; /* UNUSED */
2655 
2656 	/* Pre-extend the destination */
2657 	if (archive_string_ensure(as, as->length + len + 1) == NULL)
2658 		return (-1);
2659 
2660 	/* Invariant: src points to the first UTF8 byte that hasn't
2661 	 * been copied to the destination `as`. */
2662 	for (;;) {
2663 		int n;
2664 		uint32_t uc;
2665 		const char *e = src;
2666 
2667 		/* Skip UTF-8 sequences until we reach end-of-string or
2668 		 * a code point that needs conversion. */
2669 		while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2670 			e += n;
2671 			len -= n;
2672 		}
2673 		/* Copy the part that doesn't need conversion */
2674 		if (e > src) {
2675 			if (archive_string_append(as, src, e - src) == NULL)
2676 				return (-1);
2677 			src = e;
2678 		}
2679 
2680 		if (n == 0) {
2681 			/* We reached end-of-string */
2682 			return (ret);
2683 		} else {
2684 			/* Next code point needs conversion */
2685 			char t[4];
2686 			size_t w;
2687 
2688 			/* Try decoding a surrogate pair */
2689 			if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2690 				n = cesu8_to_unicode(&uc, src, len);
2691 			}
2692 			/* Not a (valid) surrogate, so use a replacement char */
2693 			if (n < 0) {
2694 				ret = -1; /* Return -1 if we used any replacement */
2695 				n *= -1;
2696 			}
2697 			/* Consume converted code point */
2698 			src += n;
2699 			len -= n;
2700 			/* Convert and append new UTF-8 sequence. */
2701 			w = unicode_to_utf8(t, sizeof(t), uc);
2702 			if (archive_string_append(as, t, w) == NULL)
2703 				return (-1);
2704 		}
2705 	}
2706 }
2707 
2708 static int
2709 archive_string_append_unicode(struct archive_string *as, const void *_p,
2710     size_t len, struct archive_string_conv *sc)
2711 {
2712 	const char *s;
2713 	char *p, *endp;
2714 	uint32_t uc;
2715 	size_t w;
2716 	int n, ret = 0, ts, tm;
2717 	int (*parse)(uint32_t *, const char *, size_t);
2718 	size_t (*unparse)(char *, size_t, uint32_t);
2719 
2720 	if (sc->flag & SCONV_TO_UTF16BE) {
2721 		unparse = unicode_to_utf16be;
2722 		ts = 2;
2723 	} else if (sc->flag & SCONV_TO_UTF16LE) {
2724 		unparse = unicode_to_utf16le;
2725 		ts = 2;
2726 	} else if (sc->flag & SCONV_TO_UTF8) {
2727 		unparse = unicode_to_utf8;
2728 		ts = 1;
2729 	} else {
2730 		/*
2731 		 * This case is going to be converted to another
2732 		 * character-set through iconv.
2733 		 */
2734 		if (sc->flag & SCONV_FROM_UTF16BE) {
2735 			unparse = unicode_to_utf16be;
2736 			ts = 2;
2737 		} else if (sc->flag & SCONV_FROM_UTF16LE) {
2738 			unparse = unicode_to_utf16le;
2739 			ts = 2;
2740 		} else {
2741 			unparse = unicode_to_utf8;
2742 			ts = 1;
2743 		}
2744 	}
2745 
2746 	if (sc->flag & SCONV_FROM_UTF16BE) {
2747 		parse = utf16be_to_unicode;
2748 		tm = 1;
2749 	} else if (sc->flag & SCONV_FROM_UTF16LE) {
2750 		parse = utf16le_to_unicode;
2751 		tm = 1;
2752 	} else {
2753 		parse = cesu8_to_unicode;
2754 		tm = ts;
2755 	}
2756 
2757 	if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2758 		return (-1);
2759 
2760 	s = (const char *)_p;
2761 	p = as->s + as->length;
2762 	endp = as->s + as->buffer_length - ts;
2763 	while ((n = parse(&uc, s, len)) != 0) {
2764 		if (n < 0) {
2765 			/* Use a replaced unicode character. */
2766 			n *= -1;
2767 			ret = -1;
2768 		}
2769 		s += n;
2770 		len -= n;
2771 		while ((w = unparse(p, endp - p, uc)) == 0) {
2772 			/* There is not enough output buffer so
2773 			 * we have to expand it. */
2774 			as->length = p - as->s;
2775 			if (archive_string_ensure(as,
2776 			    as->buffer_length + len * tm + ts) == NULL)
2777 				return (-1);
2778 			p = as->s + as->length;
2779 			endp = as->s + as->buffer_length - ts;
2780 		}
2781 		p += w;
2782 	}
2783 	as->length = p - as->s;
2784 	as->s[as->length] = '\0';
2785 	if (ts == 2)
2786 		as->s[as->length+1] = '\0';
2787 	return (ret);
2788 }
2789 
2790 /*
2791  * Following Constants for Hangul compositions this information comes from
2792  * Unicode Standard Annex #15  http://unicode.org/reports/tr15/
2793  */
2794 #define HC_SBASE	0xAC00
2795 #define HC_LBASE	0x1100
2796 #define HC_VBASE	0x1161
2797 #define HC_TBASE	0x11A7
2798 #define HC_LCOUNT	19
2799 #define HC_VCOUNT	21
2800 #define HC_TCOUNT	28
2801 #define HC_NCOUNT	(HC_VCOUNT * HC_TCOUNT)
2802 #define HC_SCOUNT	(HC_LCOUNT * HC_NCOUNT)
2803 
2804 static uint32_t
2805 get_nfc(uint32_t uc, uint32_t uc2)
2806 {
2807 	int t, b;
2808 
2809 	t = 0;
2810 	b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2811 	while (b >= t) {
2812 		int m = (t + b) / 2;
2813 		if (u_composition_table[m].cp1 < uc)
2814 			t = m + 1;
2815 		else if (u_composition_table[m].cp1 > uc)
2816 			b = m - 1;
2817 		else if (u_composition_table[m].cp2 < uc2)
2818 			t = m + 1;
2819 		else if (u_composition_table[m].cp2 > uc2)
2820 			b = m - 1;
2821 		else
2822 			return (u_composition_table[m].nfc);
2823 	}
2824 	return (0);
2825 }
2826 
2827 #define FDC_MAX 10	/* The maximum number of Following Decomposable
2828 			 * Characters. */
2829 
2830 /*
2831  * Update first code point.
2832  */
2833 #define UPDATE_UC(new_uc)	do {		\
2834 	uc = new_uc;				\
2835 	ucptr = NULL;				\
2836 } while (0)
2837 
2838 /*
2839  * Replace first code point with second code point.
2840  */
2841 #define REPLACE_UC_WITH_UC2() do {		\
2842 	uc = uc2;				\
2843 	ucptr = uc2ptr;				\
2844 	n = n2;					\
2845 } while (0)
2846 
2847 #define EXPAND_BUFFER() do {			\
2848 	as->length = p - as->s;			\
2849 	if (archive_string_ensure(as,		\
2850 	    as->buffer_length + len * tm + ts) == NULL)\
2851 		return (-1);			\
2852 	p = as->s + as->length;			\
2853 	endp = as->s + as->buffer_length - ts;	\
2854 } while (0)
2855 
2856 #define UNPARSE(p, endp, uc)	do {		\
2857 	while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2858 		EXPAND_BUFFER();		\
2859 	}					\
2860 	p += w;					\
2861 } while (0)
2862 
2863 /*
2864  * Write first code point.
2865  * If the code point has not be changed from its original code,
2866  * this just copies it from its original buffer pointer.
2867  * If not, this converts it to UTF-8 byte sequence and copies it.
2868  */
2869 #define WRITE_UC()	do {			\
2870 	if (ucptr) {				\
2871 		if (p + n > endp)		\
2872 			EXPAND_BUFFER();	\
2873 		switch (n) {			\
2874 		case 4:				\
2875 			*p++ = *ucptr++;	\
2876 			/* FALL THROUGH */	\
2877 		case 3:				\
2878 			*p++ = *ucptr++;	\
2879 			/* FALL THROUGH */	\
2880 		case 2:				\
2881 			*p++ = *ucptr++;	\
2882 			/* FALL THROUGH */	\
2883 		case 1:				\
2884 			*p++ = *ucptr;		\
2885 			break;			\
2886 		}				\
2887 		ucptr = NULL;			\
2888 	} else {				\
2889 		UNPARSE(p, endp, uc);		\
2890 	}					\
2891 } while (0)
2892 
2893 /*
2894  * Collect following decomposable code points.
2895  */
2896 #define COLLECT_CPS(start)	do {		\
2897 	int _i;					\
2898 	for (_i = start; _i < FDC_MAX ; _i++) {	\
2899 		nx = parse(&ucx[_i], s, len);	\
2900 		if (nx <= 0)			\
2901 			break;			\
2902 		cx = CCC(ucx[_i]);		\
2903 		if (cl >= cx && cl != 228 && cx != 228)\
2904 			break;			\
2905 		s += nx;			\
2906 		len -= nx;			\
2907 		cl = cx;			\
2908 		ccx[_i] = cx;			\
2909 	}					\
2910 	if (_i >= FDC_MAX) {			\
2911 		ret = -1;			\
2912 		ucx_size = FDC_MAX;		\
2913 	} else					\
2914 		ucx_size = _i;			\
2915 } while (0)
2916 
2917 /*
2918  * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2919  *
2920  * TODO: Convert composition exclusions, which are never converted
2921  * from NFC,NFD,NFKC and NFKD, to Form C.
2922  */
2923 static int
2924 archive_string_normalize_C(struct archive_string *as, const void *_p,
2925     size_t len, struct archive_string_conv *sc)
2926 {
2927 	const char *s = (const char *)_p;
2928 	char *p, *endp;
2929 	uint32_t uc, uc2;
2930 	size_t w;
2931 	int always_replace, n, n2, ret = 0, spair, ts, tm;
2932 	int (*parse)(uint32_t *, const char *, size_t);
2933 	size_t (*unparse)(char *, size_t, uint32_t);
2934 
2935 	always_replace = 1;
2936 	ts = 1;/* text size. */
2937 	if (sc->flag & SCONV_TO_UTF16BE) {
2938 		unparse = unicode_to_utf16be;
2939 		ts = 2;
2940 		if (sc->flag & SCONV_FROM_UTF16BE)
2941 			always_replace = 0;
2942 	} else if (sc->flag & SCONV_TO_UTF16LE) {
2943 		unparse = unicode_to_utf16le;
2944 		ts = 2;
2945 		if (sc->flag & SCONV_FROM_UTF16LE)
2946 			always_replace = 0;
2947 	} else if (sc->flag & SCONV_TO_UTF8) {
2948 		unparse = unicode_to_utf8;
2949 		if (sc->flag & SCONV_FROM_UTF8)
2950 			always_replace = 0;
2951 	} else {
2952 		/*
2953 		 * This case is going to be converted to another
2954 		 * character-set through iconv.
2955 		 */
2956 		always_replace = 0;
2957 		if (sc->flag & SCONV_FROM_UTF16BE) {
2958 			unparse = unicode_to_utf16be;
2959 			ts = 2;
2960 		} else if (sc->flag & SCONV_FROM_UTF16LE) {
2961 			unparse = unicode_to_utf16le;
2962 			ts = 2;
2963 		} else {
2964 			unparse = unicode_to_utf8;
2965 		}
2966 	}
2967 
2968 	if (sc->flag & SCONV_FROM_UTF16BE) {
2969 		parse = utf16be_to_unicode;
2970 		tm = 1;
2971 		spair = 4;/* surrogate pair size in UTF-16. */
2972 	} else if (sc->flag & SCONV_FROM_UTF16LE) {
2973 		parse = utf16le_to_unicode;
2974 		tm = 1;
2975 		spair = 4;/* surrogate pair size in UTF-16. */
2976 	} else {
2977 		parse = cesu8_to_unicode;
2978 		tm = ts;
2979 		spair = 6;/* surrogate pair size in UTF-8. */
2980 	}
2981 
2982 	if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2983 		return (-1);
2984 
2985 	p = as->s + as->length;
2986 	endp = as->s + as->buffer_length - ts;
2987 	while ((n = parse(&uc, s, len)) != 0) {
2988 		const char *ucptr, *uc2ptr;
2989 
2990 		if (n < 0) {
2991 			/* Use a replaced unicode character. */
2992 			UNPARSE(p, endp, uc);
2993 			s += n*-1;
2994 			len -= n*-1;
2995 			ret = -1;
2996 			continue;
2997 		} else if (n == spair || always_replace)
2998 			/* uc is converted from a surrogate pair.
2999 			 * this should be treated as a changed code. */
3000 			ucptr = NULL;
3001 		else
3002 			ucptr = s;
3003 		s += n;
3004 		len -= n;
3005 
3006 		/* Read second code point. */
3007 		while ((n2 = parse(&uc2, s, len)) > 0) {
3008 			uint32_t ucx[FDC_MAX];
3009 			int ccx[FDC_MAX];
3010 			int cl, cx, i, nx, ucx_size;
3011 			int LIndex,SIndex;
3012 			uint32_t nfc;
3013 
3014 			if (n2 == spair || always_replace)
3015 				/* uc2 is converted from a surrogate pair.
3016 			 	 * this should be treated as a changed code. */
3017 				uc2ptr = NULL;
3018 			else
3019 				uc2ptr = s;
3020 			s += n2;
3021 			len -= n2;
3022 
3023 			/*
3024 			 * If current second code point is out of decomposable
3025 			 * code points, finding compositions is unneeded.
3026 			 */
3027 			if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3028 				WRITE_UC();
3029 				REPLACE_UC_WITH_UC2();
3030 				continue;
3031 			}
3032 
3033 			/*
3034 			 * Try to combine current code points.
3035 			 */
3036 			/*
3037 			 * We have to combine Hangul characters according to
3038 			 * http://uniicode.org/reports/tr15/#Hangul
3039 			 */
3040 			if (0 <= (LIndex = uc - HC_LBASE) &&
3041 			    LIndex < HC_LCOUNT) {
3042 				/*
3043 				 * Hangul Composition.
3044 				 * 1. Two current code points are L and V.
3045 				 */
3046 				int VIndex = uc2 - HC_VBASE;
3047 				if (0 <= VIndex && VIndex < HC_VCOUNT) {
3048 					/* Make syllable of form LV. */
3049 					UPDATE_UC(HC_SBASE +
3050 					    (LIndex * HC_VCOUNT + VIndex) *
3051 					     HC_TCOUNT);
3052 				} else {
3053 					WRITE_UC();
3054 					REPLACE_UC_WITH_UC2();
3055 				}
3056 				continue;
3057 			} else if (0 <= (SIndex = uc - HC_SBASE) &&
3058 			    SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3059 				/*
3060 				 * Hangul Composition.
3061 				 * 2. Two current code points are LV and T.
3062 				 */
3063 				int TIndex = uc2 - HC_TBASE;
3064 				if (0 < TIndex && TIndex < HC_TCOUNT) {
3065 					/* Make syllable of form LVT. */
3066 					UPDATE_UC(uc + TIndex);
3067 				} else {
3068 					WRITE_UC();
3069 					REPLACE_UC_WITH_UC2();
3070 				}
3071 				continue;
3072 			} else if ((nfc = get_nfc(uc, uc2)) != 0) {
3073 				/* A composition to current code points
3074 				 * is found. */
3075 				UPDATE_UC(nfc);
3076 				continue;
3077 			} else if ((cl = CCC(uc2)) == 0) {
3078 				/* Clearly 'uc2' the second code point is not
3079 				 * a decomposable code. */
3080 				WRITE_UC();
3081 				REPLACE_UC_WITH_UC2();
3082 				continue;
3083 			}
3084 
3085 			/*
3086 			 * Collect following decomposable code points.
3087 			 */
3088 			cx = 0;
3089 			ucx[0] = uc2;
3090 			ccx[0] = cl;
3091 			COLLECT_CPS(1);
3092 
3093 			/*
3094 			 * Find a composed code in the collected code points.
3095 			 */
3096 			i = 1;
3097 			while (i < ucx_size) {
3098 				int j;
3099 
3100 				if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3101 					i++;
3102 					continue;
3103 				}
3104 
3105 				/*
3106 				 * nfc is composed of uc and ucx[i].
3107 				 */
3108 				UPDATE_UC(nfc);
3109 
3110 				/*
3111 				 * Remove ucx[i] by shifting
3112 				 * following code points.
3113 				 */
3114 				for (j = i; j+1 < ucx_size; j++) {
3115 					ucx[j] = ucx[j+1];
3116 					ccx[j] = ccx[j+1];
3117 				}
3118 				ucx_size --;
3119 
3120 				/*
3121 				 * Collect following code points blocked
3122 				 * by ucx[i] the removed code point.
3123 				 */
3124 				if (ucx_size > 0 && i == ucx_size &&
3125 				    nx > 0 && cx == cl) {
3126 					cl =  ccx[ucx_size-1];
3127 					COLLECT_CPS(ucx_size);
3128 				}
3129 				/*
3130 				 * Restart finding a composed code with
3131 				 * the updated uc from the top of the
3132 				 * collected code points.
3133 				 */
3134 				i = 0;
3135 			}
3136 
3137 			/*
3138 			 * Apparently the current code points are not
3139 			 * decomposed characters or already composed.
3140 			 */
3141 			WRITE_UC();
3142 			for (i = 0; i < ucx_size; i++)
3143 				UNPARSE(p, endp, ucx[i]);
3144 
3145 			/*
3146 			 * Flush out remaining canonical combining characters.
3147 			 */
3148 			if (nx > 0 && cx == cl && len > 0) {
3149 				while ((nx = parse(&ucx[0], s, len))
3150 				    > 0) {
3151 					cx = CCC(ucx[0]);
3152 					if (cl > cx)
3153 						break;
3154 					s += nx;
3155 					len -= nx;
3156 					cl = cx;
3157 					UNPARSE(p, endp, ucx[0]);
3158 				}
3159 			}
3160 			break;
3161 		}
3162 		if (n2 < 0) {
3163 			WRITE_UC();
3164 			/* Use a replaced unicode character. */
3165 			UNPARSE(p, endp, uc2);
3166 			s += n2*-1;
3167 			len -= n2*-1;
3168 			ret = -1;
3169 			continue;
3170 		} else if (n2 == 0) {
3171 			WRITE_UC();
3172 			break;
3173 		}
3174 	}
3175 	as->length = p - as->s;
3176 	as->s[as->length] = '\0';
3177 	if (ts == 2)
3178 		as->s[as->length+1] = '\0';
3179 	return (ret);
3180 }
3181 
3182 static int
3183 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3184 {
3185 	int t, b;
3186 
3187 	/*
3188 	 * These are not converted to NFD on Mac OS.
3189 	 */
3190 	if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3191 	    (uc >= 0xF900 && uc <= 0xFAFF) ||
3192 	    (uc >= 0x2F800 && uc <= 0x2FAFF))
3193 		return (0);
3194 	/*
3195 	 * Those code points are not converted to NFD on Mac OS.
3196 	 * I do not know the reason because it is undocumented.
3197 	 *   NFC        NFD
3198 	 *   1109A  ==> 11099 110BA
3199 	 *   1109C  ==> 1109B 110BA
3200 	 *   110AB  ==> 110A5 110BA
3201 	 */
3202 	if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3203 		return (0);
3204 
3205 	t = 0;
3206 	b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3207 	while (b >= t) {
3208 		int m = (t + b) / 2;
3209 		if (u_decomposition_table[m].nfc < uc)
3210 			t = m + 1;
3211 		else if (u_decomposition_table[m].nfc > uc)
3212 			b = m - 1;
3213 		else {
3214 			*cp1 = u_decomposition_table[m].cp1;
3215 			*cp2 = u_decomposition_table[m].cp2;
3216 			return (1);
3217 		}
3218 	}
3219 	return (0);
3220 }
3221 
3222 #define REPLACE_UC_WITH(cp) do {		\
3223 	uc = cp;				\
3224 	ucptr = NULL;				\
3225 } while (0)
3226 
3227 /*
3228  * Normalize UTF-8 characters to Form D and copy the result.
3229  */
3230 static int
3231 archive_string_normalize_D(struct archive_string *as, const void *_p,
3232     size_t len, struct archive_string_conv *sc)
3233 {
3234 	const char *s = (const char *)_p;
3235 	char *p, *endp;
3236 	uint32_t uc, uc2;
3237 	size_t w;
3238 	int always_replace, n, n2, ret = 0, spair, ts, tm;
3239 	int (*parse)(uint32_t *, const char *, size_t);
3240 	size_t (*unparse)(char *, size_t, uint32_t);
3241 
3242 	always_replace = 1;
3243 	ts = 1;/* text size. */
3244 	if (sc->flag & SCONV_TO_UTF16BE) {
3245 		unparse = unicode_to_utf16be;
3246 		ts = 2;
3247 		if (sc->flag & SCONV_FROM_UTF16BE)
3248 			always_replace = 0;
3249 	} else if (sc->flag & SCONV_TO_UTF16LE) {
3250 		unparse = unicode_to_utf16le;
3251 		ts = 2;
3252 		if (sc->flag & SCONV_FROM_UTF16LE)
3253 			always_replace = 0;
3254 	} else if (sc->flag & SCONV_TO_UTF8) {
3255 		unparse = unicode_to_utf8;
3256 		if (sc->flag & SCONV_FROM_UTF8)
3257 			always_replace = 0;
3258 	} else {
3259 		/*
3260 		 * This case is going to be converted to another
3261 		 * character-set through iconv.
3262 		 */
3263 		always_replace = 0;
3264 		if (sc->flag & SCONV_FROM_UTF16BE) {
3265 			unparse = unicode_to_utf16be;
3266 			ts = 2;
3267 		} else if (sc->flag & SCONV_FROM_UTF16LE) {
3268 			unparse = unicode_to_utf16le;
3269 			ts = 2;
3270 		} else {
3271 			unparse = unicode_to_utf8;
3272 		}
3273 	}
3274 
3275 	if (sc->flag & SCONV_FROM_UTF16BE) {
3276 		parse = utf16be_to_unicode;
3277 		tm = 1;
3278 		spair = 4;/* surrogate pair size in UTF-16. */
3279 	} else if (sc->flag & SCONV_FROM_UTF16LE) {
3280 		parse = utf16le_to_unicode;
3281 		tm = 1;
3282 		spair = 4;/* surrogate pair size in UTF-16. */
3283 	} else {
3284 		parse = cesu8_to_unicode;
3285 		tm = ts;
3286 		spair = 6;/* surrogate pair size in UTF-8. */
3287 	}
3288 
3289 	if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3290 		return (-1);
3291 
3292 	p = as->s + as->length;
3293 	endp = as->s + as->buffer_length - ts;
3294 	while ((n = parse(&uc, s, len)) != 0) {
3295 		const char *ucptr;
3296 		uint32_t cp1, cp2;
3297 		int SIndex;
3298 		struct {
3299 			uint32_t uc;
3300 			int ccc;
3301 		} fdc[FDC_MAX];
3302 		int fdi, fdj;
3303 		int ccc;
3304 
3305 check_first_code:
3306 		if (n < 0) {
3307 			/* Use a replaced unicode character. */
3308 			UNPARSE(p, endp, uc);
3309 			s += n*-1;
3310 			len -= n*-1;
3311 			ret = -1;
3312 			continue;
3313 		} else if (n == spair || always_replace)
3314 			/* uc is converted from a surrogate pair.
3315 			 * this should be treated as a changed code. */
3316 			ucptr = NULL;
3317 		else
3318 			ucptr = s;
3319 		s += n;
3320 		len -= n;
3321 
3322 		/* Hangul Decomposition. */
3323 		if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3324 			int L = HC_LBASE + SIndex / HC_NCOUNT;
3325 			int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3326 			int T = HC_TBASE + SIndex % HC_TCOUNT;
3327 
3328 			REPLACE_UC_WITH(L);
3329 			WRITE_UC();
3330 			REPLACE_UC_WITH(V);
3331 			WRITE_UC();
3332 			if (T != HC_TBASE) {
3333 				REPLACE_UC_WITH(T);
3334 				WRITE_UC();
3335 			}
3336 			continue;
3337 		}
3338 		if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3339 			WRITE_UC();
3340 			continue;
3341 		}
3342 
3343 		fdi = 0;
3344 		while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3345 			int k;
3346 
3347 			for (k = fdi; k > 0; k--)
3348 				fdc[k] = fdc[k-1];
3349 			fdc[0].ccc = CCC(cp2);
3350 			fdc[0].uc = cp2;
3351 			fdi++;
3352 			REPLACE_UC_WITH(cp1);
3353 		}
3354 
3355 		/* Read following code points. */
3356 		while ((n2 = parse(&uc2, s, len)) > 0 &&
3357 		    (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3358 			int j, k;
3359 
3360 			s += n2;
3361 			len -= n2;
3362 			for (j = 0; j < fdi; j++) {
3363 				if (fdc[j].ccc > ccc)
3364 					break;
3365 			}
3366 			if (j < fdi) {
3367 				for (k = fdi; k > j; k--)
3368 					fdc[k] = fdc[k-1];
3369 				fdc[j].ccc = ccc;
3370 				fdc[j].uc = uc2;
3371 			} else {
3372 				fdc[fdi].ccc = ccc;
3373 				fdc[fdi].uc = uc2;
3374 			}
3375 			fdi++;
3376 		}
3377 
3378 		WRITE_UC();
3379 		for (fdj = 0; fdj < fdi; fdj++) {
3380 			REPLACE_UC_WITH(fdc[fdj].uc);
3381 			WRITE_UC();
3382 		}
3383 
3384 		if (n2 == 0)
3385 			break;
3386 		REPLACE_UC_WITH(uc2);
3387 		n = n2;
3388 		goto check_first_code;
3389 	}
3390 	as->length = p - as->s;
3391 	as->s[as->length] = '\0';
3392 	if (ts == 2)
3393 		as->s[as->length+1] = '\0';
3394 	return (ret);
3395 }
3396 
3397 /*
3398  * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3399  * that WCS is Unicode. It is true for several platforms but some are false.
3400  * And then people who did not use UTF-8 locale on the non Unicode WCS
3401  * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3402  * now cannot get right filename from libarchive 3.x and later since we
3403  * fixed the wrong assumption and it is incompatible to older its versions.
3404  * So we provide special option, "compat-2x.x", for resolving it.
3405  * That option enable the string conversion of libarchive 2.x.
3406  *
3407  * Translates the wrong UTF-8 string made by libarchive 2.x into current
3408  * locale character set and appends to the archive_string.
3409  * Note: returns -1 if conversion fails.
3410  */
3411 static int
3412 strncat_from_utf8_libarchive2(struct archive_string *as,
3413     const void *_p, size_t len, struct archive_string_conv *sc)
3414 {
3415 	const char *s;
3416 	int n;
3417 	char *p;
3418 	char *end;
3419 	uint32_t unicode;
3420 #if HAVE_WCRTOMB
3421 	mbstate_t shift_state;
3422 
3423 	memset(&shift_state, 0, sizeof(shift_state));
3424 #else
3425 	/* Clear the shift state before starting. */
3426 	wctomb(NULL, L'\0');
3427 #endif
3428 	(void)sc; /* UNUSED */
3429 	/*
3430 	 * Allocate buffer for MBS.
3431 	 * We need this allocation here since it is possible that
3432 	 * as->s is still NULL.
3433 	 */
3434 	if (archive_string_ensure(as, as->length + len + 1) == NULL)
3435 		return (-1);
3436 
3437 	s = (const char *)_p;
3438 	p = as->s + as->length;
3439 	end = as->s + as->buffer_length - MB_CUR_MAX -1;
3440 	while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3441 		wchar_t wc;
3442 
3443 		if (p >= end) {
3444 			as->length = p - as->s;
3445 			/* Re-allocate buffer for MBS. */
3446 			if (archive_string_ensure(as,
3447 			    as->length + max(len * 2,
3448 			    (size_t)MB_CUR_MAX) + 1) == NULL)
3449 				return (-1);
3450 			p = as->s + as->length;
3451 			end = as->s + as->buffer_length - MB_CUR_MAX -1;
3452 		}
3453 
3454 		/*
3455 		 * As libarchive 2.x, translates the UTF-8 characters into
3456 		 * wide-characters in the assumption that WCS is Unicode.
3457 		 */
3458 		if (n < 0) {
3459 			n *= -1;
3460 			wc = L'?';
3461 		} else
3462 			wc = (wchar_t)unicode;
3463 
3464 		s += n;
3465 		len -= n;
3466 		/*
3467 		 * Translates the wide-character into the current locale MBS.
3468 		 */
3469 #if HAVE_WCRTOMB
3470 		n = (int)wcrtomb(p, wc, &shift_state);
3471 #else
3472 		n = (int)wctomb(p, wc);
3473 #endif
3474 		if (n == -1)
3475 			return (-1);
3476 		p += n;
3477 	}
3478 	as->length = p - as->s;
3479 	as->s[as->length] = '\0';
3480 	return (0);
3481 }
3482 
3483 
3484 /*
3485  * Conversion functions between current locale dependent MBS and UTF-16BE.
3486  *   strncat_from_utf16be() : UTF-16BE --> MBS
3487  *   strncat_to_utf16be()   : MBS --> UTF16BE
3488  */
3489 
3490 #if defined(_WIN32) && !defined(__CYGWIN__)
3491 
3492 /*
3493  * Convert a UTF-16BE/LE string to current locale and copy the result.
3494  * Return -1 if conversion fails.
3495  */
3496 static int
3497 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3498     struct archive_string_conv *sc, int be)
3499 {
3500 	struct archive_string tmp;
3501 	const char *u16;
3502 	int ll;
3503 	BOOL defchar;
3504 	char *mbs;
3505 	size_t mbs_size, b;
3506 	int ret = 0;
3507 
3508 	bytes &= ~1;
3509 	if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3510 		return (-1);
3511 
3512 	mbs = as->s + as->length;
3513 	mbs_size = as->buffer_length - as->length -1;
3514 
3515 	if (sc->to_cp == CP_C_LOCALE) {
3516 		/*
3517 		 * "C" locale special process.
3518 		 */
3519 		u16 = _p;
3520 		ll = 0;
3521 		for (b = 0; b < bytes; b += 2) {
3522 			uint16_t val;
3523 			if (be)
3524 				val = archive_be16dec(u16+b);
3525 			else
3526 				val = archive_le16dec(u16+b);
3527 			if (val > 255) {
3528 				*mbs++ = '?';
3529 				ret = -1;
3530 			} else
3531 				*mbs++ = (char)(val&0xff);
3532 			ll++;
3533 		}
3534 		as->length += ll;
3535 		as->s[as->length] = '\0';
3536 		return (ret);
3537 	}
3538 
3539 	archive_string_init(&tmp);
3540 	if (be) {
3541 		if (is_big_endian()) {
3542 			u16 = _p;
3543 		} else {
3544 			if (archive_string_ensure(&tmp, bytes+2) == NULL)
3545 				return (-1);
3546 			memcpy(tmp.s, _p, bytes);
3547 			for (b = 0; b < bytes; b += 2) {
3548 				uint16_t val = archive_be16dec(tmp.s+b);
3549 				archive_le16enc(tmp.s+b, val);
3550 			}
3551 			u16 = tmp.s;
3552 		}
3553 	} else {
3554 		if (!is_big_endian()) {
3555 			u16 = _p;
3556 		} else {
3557 			if (archive_string_ensure(&tmp, bytes+2) == NULL)
3558 				return (-1);
3559 			memcpy(tmp.s, _p, bytes);
3560 			for (b = 0; b < bytes; b += 2) {
3561 				uint16_t val = archive_le16dec(tmp.s+b);
3562 				archive_be16enc(tmp.s+b, val);
3563 			}
3564 			u16 = tmp.s;
3565 		}
3566 	}
3567 
3568 	do {
3569 		defchar = 0;
3570 		ll = WideCharToMultiByte(sc->to_cp, 0,
3571 		    (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3572 			NULL, &defchar);
3573 		/* Exit loop if we succeeded */
3574 		if (ll != 0 ||
3575 		    GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3576 			break;
3577 		}
3578 		/* Else expand buffer and loop to try again. */
3579 		ll = WideCharToMultiByte(sc->to_cp, 0,
3580 		    (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3581 		if (archive_string_ensure(as, ll +1) == NULL)
3582 			return (-1);
3583 		mbs = as->s + as->length;
3584 		mbs_size = as->buffer_length - as->length -1;
3585 	} while (1);
3586 	archive_string_free(&tmp);
3587 	as->length += ll;
3588 	as->s[as->length] = '\0';
3589 	if (ll == 0 || defchar)
3590 		ret = -1;
3591 	return (ret);
3592 }
3593 
3594 static int
3595 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3596     size_t bytes, struct archive_string_conv *sc)
3597 {
3598 	return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3599 }
3600 
3601 static int
3602 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3603     size_t bytes, struct archive_string_conv *sc)
3604 {
3605 	return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3606 }
3607 
3608 static int
3609 is_big_endian(void)
3610 {
3611 	uint16_t d = 1;
3612 
3613 	return (archive_be16dec(&d) == 1);
3614 }
3615 
3616 /*
3617  * Convert a current locale string to UTF-16BE/LE and copy the result.
3618  * Return -1 if conversion fails.
3619  */
3620 static int
3621 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3622     size_t length, struct archive_string_conv *sc, int bigendian)
3623 {
3624 	const char *s = (const char *)_p;
3625 	char *u16;
3626 	size_t count, avail;
3627 
3628 	if (archive_string_ensure(as16,
3629 	    as16->length + (length + 1) * 2) == NULL)
3630 		return (-1);
3631 
3632 	u16 = as16->s + as16->length;
3633 	avail = as16->buffer_length - 2;
3634 	if (sc->from_cp == CP_C_LOCALE) {
3635 		/*
3636 		 * "C" locale special process.
3637 		 */
3638 		count = 0;
3639 		while (count < length && *s) {
3640 			if (bigendian)
3641 				archive_be16enc(u16, *s);
3642 			else
3643 				archive_le16enc(u16, *s);
3644 			u16 += 2;
3645 			s++;
3646 			count++;
3647 		}
3648 		as16->length += count << 1;
3649 		as16->s[as16->length] = 0;
3650 		as16->s[as16->length+1] = 0;
3651 		return (0);
3652 	}
3653 	do {
3654 		count = MultiByteToWideChar(sc->from_cp,
3655 		    MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3656 		/* Exit loop if we succeeded */
3657 		if (count != 0 ||
3658 		    GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3659 			break;
3660 		}
3661 		/* Expand buffer and try again */
3662 		count = MultiByteToWideChar(sc->from_cp,
3663 		    MB_PRECOMPOSED, s, (int)length, NULL, 0);
3664 		if (archive_string_ensure(as16, (count +1) * 2)
3665 		    == NULL)
3666 			return (-1);
3667 		u16 = as16->s + as16->length;
3668 		avail = as16->buffer_length - 2;
3669 	} while (1);
3670 	as16->length += count * 2;
3671 	as16->s[as16->length] = 0;
3672 	as16->s[as16->length+1] = 0;
3673 	if (count == 0)
3674 		return (-1);
3675 
3676 	if (is_big_endian()) {
3677 		if (!bigendian) {
3678 			while (count > 0) {
3679 				uint16_t v = archive_be16dec(u16);
3680 				archive_le16enc(u16, v);
3681 				u16 += 2;
3682 				count--;
3683 			}
3684 		}
3685 	} else {
3686 		if (bigendian) {
3687 			while (count > 0) {
3688 				uint16_t v = archive_le16dec(u16);
3689 				archive_be16enc(u16, v);
3690 				u16 += 2;
3691 				count--;
3692 			}
3693 		}
3694 	}
3695 	return (0);
3696 }
3697 
3698 static int
3699 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3700     size_t length, struct archive_string_conv *sc)
3701 {
3702 	return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3703 }
3704 
3705 static int
3706 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3707     size_t length, struct archive_string_conv *sc)
3708 {
3709 	return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3710 }
3711 
3712 #endif /* _WIN32 && !__CYGWIN__ */
3713 
3714 /*
3715  * Do the best effort for conversions.
3716  * We cannot handle UTF-16BE character-set without such iconv,
3717  * but there is a chance if a string consists just ASCII code or
3718  * a current locale is UTF-8.
3719  */
3720 
3721 /*
3722  * Convert a UTF-16BE string to current locale and copy the result.
3723  * Return -1 if conversion fails.
3724  */
3725 static int
3726 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3727     size_t bytes, struct archive_string_conv *sc, int be)
3728 {
3729 	const char *utf16 = (const char *)_p;
3730 	char *mbs;
3731 	uint32_t uc;
3732 	int n, ret;
3733 
3734 	(void)sc; /* UNUSED */
3735 	/*
3736 	 * Other case, we should do the best effort.
3737 	 * If all character are ASCII(<0x7f), we can convert it.
3738 	 * if not , we set a alternative character and return -1.
3739 	 */
3740 	ret = 0;
3741 	if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3742 		return (-1);
3743 	mbs = as->s + as->length;
3744 
3745 	while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3746 		if (n < 0) {
3747 			n *= -1;
3748 			ret =  -1;
3749 		}
3750 		bytes -= n;
3751 		utf16 += n;
3752 
3753 		if (uc > 127) {
3754 			/* We cannot handle it. */
3755 			*mbs++ = '?';
3756 			ret =  -1;
3757 		} else
3758 			*mbs++ = (char)uc;
3759 	}
3760 	as->length = mbs - as->s;
3761 	as->s[as->length] = '\0';
3762 	return (ret);
3763 }
3764 
3765 static int
3766 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3767     size_t bytes, struct archive_string_conv *sc)
3768 {
3769 	return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3770 }
3771 
3772 static int
3773 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3774     size_t bytes, struct archive_string_conv *sc)
3775 {
3776 	return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3777 }
3778 
3779 /*
3780  * Convert a current locale string to UTF-16BE/LE and copy the result.
3781  * Return -1 if conversion fails.
3782  */
3783 static int
3784 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3785     size_t length, struct archive_string_conv *sc, int bigendian)
3786 {
3787 	const char *s = (const char *)_p;
3788 	char *utf16;
3789 	size_t remaining;
3790 	int ret;
3791 
3792 	(void)sc; /* UNUSED */
3793 	/*
3794 	 * Other case, we should do the best effort.
3795 	 * If all character are ASCII(<0x7f), we can convert it.
3796 	 * if not , we set a alternative character and return -1.
3797 	 */
3798 	ret = 0;
3799 	remaining = length;
3800 
3801 	if (archive_string_ensure(as16,
3802 	    as16->length + (length + 1) * 2) == NULL)
3803 		return (-1);
3804 
3805 	utf16 = as16->s + as16->length;
3806 	while (remaining--) {
3807 		unsigned c = *s++;
3808 		if (c > 127) {
3809 			/* We cannot handle it. */
3810 			c = UNICODE_R_CHAR;
3811 			ret = -1;
3812 		}
3813 		if (bigendian)
3814 			archive_be16enc(utf16, c);
3815 		else
3816 			archive_le16enc(utf16, c);
3817 		utf16 += 2;
3818 	}
3819 	as16->length = utf16 - as16->s;
3820 	as16->s[as16->length] = 0;
3821 	as16->s[as16->length+1] = 0;
3822 	return (ret);
3823 }
3824 
3825 static int
3826 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3827     size_t length, struct archive_string_conv *sc)
3828 {
3829 	return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3830 }
3831 
3832 static int
3833 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3834     size_t length, struct archive_string_conv *sc)
3835 {
3836 	return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3837 }
3838 
3839 
3840 /*
3841  * Multistring operations.
3842  */
3843 
3844 void
3845 archive_mstring_clean(struct archive_mstring *aes)
3846 {
3847 	archive_wstring_free(&(aes->aes_wcs));
3848 	archive_string_free(&(aes->aes_mbs));
3849 	archive_string_free(&(aes->aes_utf8));
3850 	archive_string_free(&(aes->aes_mbs_in_locale));
3851 	aes->aes_set = 0;
3852 }
3853 
3854 void
3855 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3856 {
3857 	dest->aes_set = src->aes_set;
3858 	archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3859 	archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3860 	archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3861 }
3862 
3863 int
3864 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3865   const char **p)
3866 {
3867 	struct archive_string_conv *sc;
3868 	int r;
3869 
3870 	/* If we already have a UTF8 form, return that immediately. */
3871 	if (aes->aes_set & AES_SET_UTF8) {
3872 		*p = aes->aes_utf8.s;
3873 		return (0);
3874 	}
3875 
3876 	*p = NULL;
3877 #if defined(_WIN32) && !defined(__CYGWIN__)
3878 	/*
3879 	 * On Windows, first try converting from WCS because (1) there's no
3880 	 * guarantee that the conversion to MBS will succeed, e.g. when using
3881 	 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3882 	 * convert back to WCS again before finally converting to UTF-8
3883 	 */
3884 	if ((aes->aes_set & AES_SET_WCS) != 0) {
3885 		sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3886 		if (sc == NULL)
3887 			return (-1);/* Couldn't allocate memory for sc. */
3888 		archive_string_empty(&(aes->aes_utf8));
3889 		r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3890 			aes->aes_wcs.s, aes->aes_wcs.length, sc);
3891 		if (a == NULL)
3892 			free_sconv_object(sc);
3893 		if (r == 0) {
3894 			aes->aes_set |= AES_SET_UTF8;
3895 			*p = aes->aes_utf8.s;
3896 			return (0);/* success. */
3897 		} else
3898 			return (-1);/* failure. */
3899 	}
3900 #endif
3901 	/* Try converting WCS to MBS first if MBS does not exist yet. */
3902 	if ((aes->aes_set & AES_SET_MBS) == 0) {
3903 		const char *pm; /* unused */
3904 		archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3905 	}
3906 	if (aes->aes_set & AES_SET_MBS) {
3907 		sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3908 		if (sc == NULL)
3909 			return (-1);/* Couldn't allocate memory for sc. */
3910 		r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3911 		    aes->aes_mbs.length, sc);
3912 		if (a == NULL)
3913 			free_sconv_object(sc);
3914 		if (r == 0) {
3915 			aes->aes_set |= AES_SET_UTF8;
3916 			*p = aes->aes_utf8.s;
3917 			return (0);/* success. */
3918 		} else
3919 			return (-1);/* failure. */
3920 	}
3921 	return (0);/* success. */
3922 }
3923 
3924 int
3925 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3926     const char **p)
3927 {
3928 	struct archive_string_conv *sc;
3929 	int r, ret = 0;
3930 
3931 	/* If we already have an MBS form, return that immediately. */
3932 	if (aes->aes_set & AES_SET_MBS) {
3933 		*p = aes->aes_mbs.s;
3934 		return (ret);
3935 	}
3936 
3937 	*p = NULL;
3938 	/* If there's a WCS form, try converting with the native locale. */
3939 	if (aes->aes_set & AES_SET_WCS) {
3940 		archive_string_empty(&(aes->aes_mbs));
3941 		r = archive_string_append_from_wcs(&(aes->aes_mbs),
3942 		    aes->aes_wcs.s, aes->aes_wcs.length);
3943 		*p = aes->aes_mbs.s;
3944 		if (r == 0) {
3945 			aes->aes_set |= AES_SET_MBS;
3946 			return (ret);
3947 		} else
3948 			ret = -1;
3949 	}
3950 
3951 	/* If there's a UTF-8 form, try converting with the native locale. */
3952 	if (aes->aes_set & AES_SET_UTF8) {
3953 		archive_string_empty(&(aes->aes_mbs));
3954 		sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
3955 		if (sc == NULL)
3956 			return (-1);/* Couldn't allocate memory for sc. */
3957 		r = archive_strncpy_l(&(aes->aes_mbs),
3958 			aes->aes_utf8.s, aes->aes_utf8.length, sc);
3959 		if (a == NULL)
3960 			free_sconv_object(sc);
3961 		*p = aes->aes_mbs.s;
3962 		if (r == 0) {
3963 			aes->aes_set |= AES_SET_MBS;
3964 			ret = 0;/* success; overwrite previous error. */
3965 		} else
3966 			ret = -1;/* failure. */
3967 	}
3968 	return (ret);
3969 }
3970 
3971 int
3972 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3973     const wchar_t **wp)
3974 {
3975 	int r, ret = 0;
3976 
3977 	(void)a;/* UNUSED */
3978 	/* Return WCS form if we already have it. */
3979 	if (aes->aes_set & AES_SET_WCS) {
3980 		*wp = aes->aes_wcs.s;
3981 		return (ret);
3982 	}
3983 
3984 	*wp = NULL;
3985 #if defined(_WIN32) && !defined(__CYGWIN__)
3986 	/*
3987 	 * On Windows, prefer converting from UTF-8 directly to WCS because:
3988 	 * (1) there's no guarantee that the string can be represented in MBS (e.g.
3989 	 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
3990 	 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
3991 	 * away that intermediate result
3992 	 */
3993 	if (aes->aes_set & AES_SET_UTF8) {
3994 		struct archive_string_conv *sc;
3995 
3996 		sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
3997 		if (sc != NULL) {
3998 			archive_wstring_empty((&aes->aes_wcs));
3999 			r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4000 			    aes->aes_utf8.s, aes->aes_utf8.length, sc);
4001 			if (a == NULL)
4002 				free_sconv_object(sc);
4003 			if (r == 0) {
4004 				aes->aes_set |= AES_SET_WCS;
4005 				*wp = aes->aes_wcs.s;
4006 				return (0);
4007 			}
4008 		}
4009 	}
4010 #endif
4011 	/* Try converting UTF8 to MBS first if MBS does not exist yet. */
4012 	if ((aes->aes_set & AES_SET_MBS) == 0) {
4013 		const char *p; /* unused */
4014 		archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4015 	}
4016 	/* Try converting MBS to WCS using native locale. */
4017 	if (aes->aes_set & AES_SET_MBS) {
4018 		archive_wstring_empty(&(aes->aes_wcs));
4019 		r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4020 		    aes->aes_mbs.s, aes->aes_mbs.length);
4021 		if (r == 0) {
4022 			aes->aes_set |= AES_SET_WCS;
4023 			*wp = aes->aes_wcs.s;
4024 		} else
4025 			ret = -1;/* failure. */
4026 	}
4027 	return (ret);
4028 }
4029 
4030 int
4031 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4032     const char **p, size_t *length, struct archive_string_conv *sc)
4033 {
4034 	int ret = 0;
4035 #if defined(_WIN32) && !defined(__CYGWIN__)
4036 	int r;
4037 
4038 	/*
4039 	 * Internationalization programming on Windows must use Wide
4040 	 * characters because Windows platform cannot make locale UTF-8.
4041 	 */
4042 	if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4043 		archive_string_empty(&(aes->aes_mbs_in_locale));
4044 		r = archive_string_append_from_wcs_in_codepage(
4045 		    &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4046 		    aes->aes_wcs.length, sc);
4047 		if (r == 0) {
4048 			*p = aes->aes_mbs_in_locale.s;
4049 			if (length != NULL)
4050 				*length = aes->aes_mbs_in_locale.length;
4051 			return (0);
4052 		} else if (errno == ENOMEM)
4053 			return (-1);
4054 		else
4055 			ret = -1;
4056 	}
4057 #endif
4058 
4059 	/* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4060 	 * with the native locale to be used for translating it to specified
4061 	 * character-set. */
4062 	if ((aes->aes_set & AES_SET_MBS) == 0) {
4063 		const char *pm; /* unused */
4064 		archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
4065 	}
4066 	/* If we already have an MBS form, use it to be translated to
4067 	 * specified character-set. */
4068 	if (aes->aes_set & AES_SET_MBS) {
4069 		if (sc == NULL) {
4070 			/* Conversion is unneeded. */
4071 			*p = aes->aes_mbs.s;
4072 			if (length != NULL)
4073 				*length = aes->aes_mbs.length;
4074 			return (0);
4075 		}
4076 		ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4077 		    aes->aes_mbs.s, aes->aes_mbs.length, sc);
4078 		*p = aes->aes_mbs_in_locale.s;
4079 		if (length != NULL)
4080 			*length = aes->aes_mbs_in_locale.length;
4081 	} else {
4082 		*p = NULL;
4083 		if (length != NULL)
4084 			*length = 0;
4085 	}
4086 	return (ret);
4087 }
4088 
4089 int
4090 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4091 {
4092 	if (mbs == NULL) {
4093 		aes->aes_set = 0;
4094 		return (0);
4095 	}
4096 	return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4097 }
4098 
4099 int
4100 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4101     size_t len)
4102 {
4103 	if (mbs == NULL) {
4104 		aes->aes_set = 0;
4105 		return (0);
4106 	}
4107 	aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4108 	archive_strncpy(&(aes->aes_mbs), mbs, len);
4109 	archive_string_empty(&(aes->aes_utf8));
4110 	archive_wstring_empty(&(aes->aes_wcs));
4111 	return (0);
4112 }
4113 
4114 int
4115 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4116 {
4117 	return archive_mstring_copy_wcs_len(aes, wcs,
4118 				wcs == NULL ? 0 : wcslen(wcs));
4119 }
4120 
4121 int
4122 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4123 {
4124   if (utf8 == NULL) {
4125     aes->aes_set = 0;
4126     return (0);
4127   }
4128   aes->aes_set = AES_SET_UTF8;
4129   archive_string_empty(&(aes->aes_mbs));
4130   archive_string_empty(&(aes->aes_wcs));
4131   archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4132   return (int)strlen(utf8);
4133 }
4134 
4135 int
4136 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4137     size_t len)
4138 {
4139 	if (wcs == NULL) {
4140 		aes->aes_set = 0;
4141 		return (0);
4142 	}
4143 	aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4144 	archive_string_empty(&(aes->aes_mbs));
4145 	archive_string_empty(&(aes->aes_utf8));
4146 	archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4147 	return (0);
4148 }
4149 
4150 int
4151 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4152     const char *mbs, size_t len, struct archive_string_conv *sc)
4153 {
4154 	int r;
4155 
4156 	if (mbs == NULL) {
4157 		aes->aes_set = 0;
4158 		return (0);
4159 	}
4160 	archive_string_empty(&(aes->aes_mbs));
4161 	archive_wstring_empty(&(aes->aes_wcs));
4162 	archive_string_empty(&(aes->aes_utf8));
4163 #if defined(_WIN32) && !defined(__CYGWIN__)
4164 	/*
4165 	 * Internationalization programming on Windows must use Wide
4166 	 * characters because Windows platform cannot make locale UTF-8.
4167 	 */
4168 	if (sc == NULL) {
4169 		if (archive_string_append(&(aes->aes_mbs),
4170 			mbs, mbsnbytes(mbs, len)) == NULL) {
4171 			aes->aes_set = 0;
4172 			r = -1;
4173 		} else {
4174 			aes->aes_set = AES_SET_MBS;
4175 			r = 0;
4176 		}
4177 #if defined(HAVE_ICONV)
4178 	} else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4179 		/*
4180 		 * This case happens only when MultiByteToWideChar() cannot
4181 		 * handle sc->from_cp, and we have to iconv in order to
4182 		 * translate character-set to wchar_t,UTF-16.
4183 		 */
4184 		iconv_t cd = sc->cd;
4185 		unsigned from_cp;
4186 		int flag;
4187 
4188 		/*
4189 		 * Translate multi-bytes from some character-set to UTF-8.
4190 		 */
4191 		sc->cd = sc->cd_w;
4192 		r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4193 		sc->cd = cd;
4194 		if (r != 0) {
4195 			aes->aes_set = 0;
4196 			return (r);
4197 		}
4198 		aes->aes_set = AES_SET_UTF8;
4199 
4200 		/*
4201 		 * Append the UTF-8 string into wstring.
4202 		 */
4203 		flag = sc->flag;
4204 		sc->flag &= ~(SCONV_NORMALIZATION_C
4205 				| SCONV_TO_UTF16| SCONV_FROM_UTF16);
4206 		from_cp = sc->from_cp;
4207 		sc->from_cp = CP_UTF8;
4208 		r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4209 			aes->aes_utf8.s, aes->aes_utf8.length, sc);
4210 		sc->flag = flag;
4211 		sc->from_cp = from_cp;
4212 		if (r == 0)
4213 			aes->aes_set |= AES_SET_WCS;
4214 #endif
4215 	} else {
4216 		r = archive_wstring_append_from_mbs_in_codepage(
4217 		    &(aes->aes_wcs), mbs, len, sc);
4218 		if (r == 0)
4219 			aes->aes_set = AES_SET_WCS;
4220 		else
4221 			aes->aes_set = 0;
4222 	}
4223 #else
4224 	r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4225 	if (r == 0)
4226 		aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4227 	else
4228 		aes->aes_set = 0;
4229 #endif
4230 	return (r);
4231 }
4232 
4233 /*
4234  * The 'update' form tries to proactively update all forms of
4235  * this string (WCS and MBS) and returns an error if any of
4236  * them fail.  This is used by the 'pax' handler, for instance,
4237  * to detect and report character-conversion failures early while
4238  * still allowing clients to get potentially useful values from
4239  * the more tolerant lazy conversions.  (get_mbs and get_wcs will
4240  * strive to give the user something useful, so you can get hopefully
4241  * usable values even if some of the character conversions are failing.)
4242  */
4243 int
4244 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4245     const char *utf8)
4246 {
4247 	struct archive_string_conv *sc;
4248 	int r;
4249 
4250 	if (utf8 == NULL) {
4251 		aes->aes_set = 0;
4252 		return (0); /* Succeeded in clearing everything. */
4253 	}
4254 
4255 	/* Save the UTF8 string. */
4256 	archive_strcpy(&(aes->aes_utf8), utf8);
4257 
4258 	/* Empty the mbs and wcs strings. */
4259 	archive_string_empty(&(aes->aes_mbs));
4260 	archive_wstring_empty(&(aes->aes_wcs));
4261 
4262 	aes->aes_set = AES_SET_UTF8;	/* Only UTF8 is set now. */
4263 
4264 	sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4265 	if (sc == NULL)
4266 		return (-1);/* Couldn't allocate memory for sc. */
4267 
4268 #if defined(_WIN32) && !defined(__CYGWIN__)
4269 	/* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4270 	 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4271 	 * intermediate result, and (2) WCS will still be set even if we fail to
4272 	 * convert to MBS (e.g. with ACP that can't represent the characters) */
4273 	r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4274 		aes->aes_utf8.s, aes->aes_utf8.length, sc);
4275 
4276 	if (a == NULL)
4277 		free_sconv_object(sc);
4278 	if (r != 0)
4279 		return (-1); /* This will guarantee we can't convert to MBS */
4280 	aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4281 
4282 	/* Try converting WCS to MBS, return false on failure. */
4283 	if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4284 	    aes->aes_wcs.length))
4285 		return (-1);
4286 #else
4287 	/* Try converting UTF-8 to MBS, return false on failure. */
4288 	r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4289 
4290 	if (a == NULL)
4291 		free_sconv_object(sc);
4292 	if (r != 0)
4293 		return (-1);
4294 	aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4295 
4296 	/* Try converting MBS to WCS, return false on failure. */
4297 	if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4298 	    aes->aes_mbs.length))
4299 		return (-1);
4300 #endif
4301 
4302 	/* All conversions succeeded. */
4303 	aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4304 
4305 	return (0);
4306 }
4307