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 strncat_in_codepage(struct archive_string *, const void *,
158 size_t, struct archive_string_conv *);
159 static int win_strncat_from_utf16be(struct archive_string *, const void *,
160 size_t, struct archive_string_conv *);
161 static int win_strncat_from_utf16le(struct archive_string *, const void *,
162 size_t, struct archive_string_conv *);
163 static int win_strncat_to_utf16be(struct archive_string *, const void *,
164 size_t, struct archive_string_conv *);
165 static int win_strncat_to_utf16le(struct archive_string *, const void *,
166 size_t, struct archive_string_conv *);
167 #endif
168 static int best_effort_strncat_from_utf16be(struct archive_string *,
169 const void *, size_t, struct archive_string_conv *);
170 static int best_effort_strncat_from_utf16le(struct archive_string *,
171 const void *, size_t, struct archive_string_conv *);
172 static int best_effort_strncat_to_utf16be(struct archive_string *,
173 const void *, size_t, struct archive_string_conv *);
174 static int best_effort_strncat_to_utf16le(struct archive_string *,
175 const void *, size_t, struct archive_string_conv *);
176 #if defined(HAVE_ICONV)
177 static int iconv_strncat_in_locale(struct archive_string *, const void *,
178 size_t, struct archive_string_conv *);
179 #endif
180 static int best_effort_strncat_in_locale(struct archive_string *,
181 const void *, size_t, struct archive_string_conv *);
182 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
183 static int utf8_to_unicode(uint32_t *, const char *, size_t);
184 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
185 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
186 static size_t unicode_to_utf8(char *, size_t, uint32_t);
187 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
188 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
189 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
190 static int strncat_from_utf8_libarchive2(struct archive_string *,
191 const void *, size_t, struct archive_string_conv *);
192 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
193 size_t, struct archive_string_conv *);
194 static int archive_string_normalize_C(struct archive_string *, const void *,
195 size_t, struct archive_string_conv *);
196 static int archive_string_normalize_D(struct archive_string *, const void *,
197 size_t, struct archive_string_conv *);
198 static int archive_string_append_unicode(struct archive_string *,
199 const void *, size_t, struct archive_string_conv *);
200
201 #if defined __LITTLE_ENDIAN__
202 #define IS_BIG_ENDIAN 0
203 #elif defined __BIG_ENDIAN__
204 #define IS_BIG_ENDIAN 1
205 #elif defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
206 #define IS_BIG_ENDIAN 0
207 #elif defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
208 #define IS_BIG_ENDIAN 1
209 #elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_AMD64) || defined(_M_X64) || defined(_M_ARM64))
210 #define IS_BIG_ENDIAN 0
211 #else
212 // Detect endianness at runtime.
213 static int
is_big_endian(void)214 is_big_endian(void)
215 {
216 uint16_t d = 1;
217
218 return (archive_be16dec(&d) == 1);
219 }
220
221 #define IS_BIG_ENDIAN is_big_endian()
222 #endif
223
224 static struct archive_string *
archive_string_append(struct archive_string * as,const char * p,size_t s)225 archive_string_append(struct archive_string *as, const char *p, size_t s)
226 {
227 if (archive_string_ensure(as, as->length + s + 1) == NULL)
228 return (NULL);
229 if (s)
230 memmove(as->s + as->length, p, s);
231 as->length += s;
232 as->s[as->length] = 0;
233 return (as);
234 }
235
236 static struct archive_wstring *
archive_wstring_append(struct archive_wstring * as,const wchar_t * p,size_t s)237 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
238 {
239 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
240 return (NULL);
241 if (s)
242 wmemmove(as->s + as->length, p, s);
243 as->length += s;
244 as->s[as->length] = 0;
245 return (as);
246 }
247
248 struct archive_string *
archive_array_append(struct archive_string * as,const char * p,size_t s)249 archive_array_append(struct archive_string *as, const char *p, size_t s)
250 {
251 return archive_string_append(as, p, s);
252 }
253
254 void
archive_string_concat(struct archive_string * dest,struct archive_string * src)255 archive_string_concat(struct archive_string *dest, struct archive_string *src)
256 {
257 if (archive_string_append(dest, src->s, src->length) == NULL)
258 __archive_errx(1, "Out of memory");
259 }
260
261 void
archive_wstring_concat(struct archive_wstring * dest,struct archive_wstring * src)262 archive_wstring_concat(struct archive_wstring *dest,
263 struct archive_wstring *src)
264 {
265 if (archive_wstring_append(dest, src->s, src->length) == NULL)
266 __archive_errx(1, "Out of memory");
267 }
268
269 void
archive_string_free(struct archive_string * as)270 archive_string_free(struct archive_string *as)
271 {
272 as->length = 0;
273 as->buffer_length = 0;
274 free(as->s);
275 as->s = NULL;
276 }
277
278 void
archive_wstring_free(struct archive_wstring * as)279 archive_wstring_free(struct archive_wstring *as)
280 {
281 as->length = 0;
282 as->buffer_length = 0;
283 free(as->s);
284 as->s = NULL;
285 }
286
287 struct archive_wstring *
archive_wstring_ensure(struct archive_wstring * as,size_t s)288 archive_wstring_ensure(struct archive_wstring *as, size_t s)
289 {
290 return (struct archive_wstring *)
291 archive_string_ensure((struct archive_string *)as,
292 s * sizeof(wchar_t));
293 }
294
295 /* Returns NULL on any allocation failure. */
296 struct archive_string *
archive_string_ensure(struct archive_string * as,size_t s)297 archive_string_ensure(struct archive_string *as, size_t s)
298 {
299 char *p;
300 size_t new_length;
301
302 /* If buffer is already big enough, don't reallocate. */
303 if (as->s && (s <= as->buffer_length))
304 return (as);
305
306 /*
307 * Growing the buffer at least exponentially ensures that
308 * append operations are always linear in the number of
309 * characters appended. Using a smaller growth rate for
310 * larger buffers reduces memory waste somewhat at the cost of
311 * a larger constant factor.
312 */
313 if (as->buffer_length < 32)
314 /* Start with a minimum 32-character buffer. */
315 new_length = 32;
316 else if (as->buffer_length < 8192)
317 /* Buffers under 8k are doubled for speed. */
318 new_length = as->buffer_length + as->buffer_length;
319 else {
320 /* Buffers 8k and over grow by at least 25% each time. */
321 new_length = as->buffer_length + as->buffer_length / 4;
322 /* Be safe: If size wraps, fail. */
323 if (new_length < as->buffer_length) {
324 /* On failure, wipe the string and return NULL. */
325 archive_string_free(as);
326 errno = ENOMEM;/* Make sure errno has ENOMEM. */
327 return (NULL);
328 }
329 }
330 /*
331 * The computation above is a lower limit to how much we'll
332 * grow the buffer. In any case, we have to grow it enough to
333 * hold the request.
334 */
335 if (new_length < s)
336 new_length = s;
337 /* Now we can reallocate the buffer. */
338 p = realloc(as->s, new_length);
339 if (p == NULL) {
340 /* On failure, wipe the string and return NULL. */
341 archive_string_free(as);
342 errno = ENOMEM;/* Make sure errno has ENOMEM. */
343 return (NULL);
344 }
345
346 as->s = p;
347 as->buffer_length = new_length;
348 return (as);
349 }
350
351 /*
352 * TODO: See if there's a way to avoid scanning
353 * the source string twice. Then test to see
354 * if it actually helps (remember that we're almost
355 * always called with pretty short arguments, so
356 * such an optimization might not help).
357 */
358 struct archive_string *
archive_strncat(struct archive_string * as,const void * _p,size_t n)359 archive_strncat(struct archive_string *as, const void *_p, size_t n)
360 {
361 size_t s;
362 const char *p, *pp;
363
364 p = (const char *)_p;
365
366 /* Like strlen(p), except won't examine positions beyond p[n]. */
367 s = 0;
368 pp = p;
369 while (s < n && *pp) {
370 pp++;
371 s++;
372 }
373 if ((as = archive_string_append(as, p, s)) == NULL)
374 __archive_errx(1, "Out of memory");
375 return (as);
376 }
377
378 struct archive_wstring *
archive_wstrncat(struct archive_wstring * as,const wchar_t * p,size_t n)379 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
380 {
381 size_t s;
382 const wchar_t *pp;
383
384 /* Like strlen(p), except won't examine positions beyond p[n]. */
385 s = 0;
386 pp = p;
387 while (s < n && *pp) {
388 pp++;
389 s++;
390 }
391 if ((as = archive_wstring_append(as, p, s)) == NULL)
392 __archive_errx(1, "Out of memory");
393 return (as);
394 }
395
396 struct archive_string *
archive_strcat(struct archive_string * as,const void * p)397 archive_strcat(struct archive_string *as, const void *p)
398 {
399 /* strcat is just strncat without an effective limit.
400 * Assert that we'll never get called with a source
401 * string over 16MB.
402 * TODO: Review all uses of strcat in the source
403 * and try to replace them with strncat().
404 */
405 return archive_strncat(as, p, 0x1000000);
406 }
407
408 struct archive_wstring *
archive_wstrcat(struct archive_wstring * as,const wchar_t * p)409 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
410 {
411 /* Ditto. */
412 return archive_wstrncat(as, p, 0x1000000);
413 }
414
415 struct archive_string *
archive_strappend_char(struct archive_string * as,char c)416 archive_strappend_char(struct archive_string *as, char c)
417 {
418 if ((as = archive_string_append(as, &c, 1)) == NULL)
419 __archive_errx(1, "Out of memory");
420 return (as);
421 }
422
423 struct archive_wstring *
archive_wstrappend_wchar(struct archive_wstring * as,wchar_t c)424 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
425 {
426 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
427 __archive_errx(1, "Out of memory");
428 return (as);
429 }
430
431 /*
432 * Get the "current character set" name to use with iconv.
433 * On FreeBSD, the empty character set name "" chooses
434 * the correct character encoding for the current locale,
435 * so this isn't necessary.
436 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
437 * on that system, we have to explicitly call nl_langinfo()
438 * to get the right name. Not sure about other platforms.
439 *
440 * NOTE: GNU libiconv does not recognize the character-set name
441 * which some platform nl_langinfo(CODESET) returns, so we should
442 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
443 */
444 static const char *
default_iconv_charset(const char * charset)445 default_iconv_charset(const char *charset) {
446 if (charset != NULL && charset[0] != '\0')
447 return charset;
448 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
449 /* locale_charset() is broken on Mac OS */
450 return locale_charset();
451 #elif HAVE_NL_LANGINFO
452 return nl_langinfo(CODESET);
453 #else
454 return "";
455 #endif
456 }
457
458 #if defined(_WIN32) && !defined(__CYGWIN__)
459
460 /*
461 * Convert MBS to WCS.
462 * Note: returns -1 if conversion fails.
463 */
464 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)465 archive_wstring_append_from_mbs(struct archive_wstring *dest,
466 const char *p, size_t len)
467 {
468 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
469 }
470
471 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)472 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
473 const char *s, size_t length, struct archive_string_conv *sc)
474 {
475 int ret = 0;
476 size_t count;
477 UINT from_cp;
478
479 if (sc != NULL)
480 from_cp = sc->from_cp;
481 else
482 from_cp = get_current_codepage();
483
484 if (from_cp == CP_C_LOCALE) {
485 /*
486 * "C" locale special processing.
487 */
488 wchar_t *ws;
489 const unsigned char *mp;
490
491 if (NULL == archive_wstring_ensure(dest,
492 dest->length + length + 1))
493 return (-1);
494
495 ws = dest->s + dest->length;
496 mp = (const unsigned char *)s;
497 count = 0;
498 while (count < length && *mp) {
499 *ws++ = (wchar_t)*mp++;
500 count++;
501 }
502 } else if (sc != NULL &&
503 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
504 /*
505 * Normalize UTF-8 and UTF-16BE and convert it directly
506 * to UTF-16 as wchar_t.
507 */
508 struct archive_string u16;
509 int saved_flag = sc->flag;/* save current flag. */
510
511 if (IS_BIG_ENDIAN)
512 sc->flag |= SCONV_TO_UTF16BE;
513 else
514 sc->flag |= SCONV_TO_UTF16LE;
515
516 if (sc->flag & SCONV_FROM_UTF16) {
517 /*
518 * UTF-16BE/LE NFD ===> UTF-16 NFC
519 * UTF-16BE/LE NFC ===> UTF-16 NFD
520 */
521 count = utf16nbytes(s, length);
522 } else {
523 /*
524 * UTF-8 NFD ===> UTF-16 NFC
525 * UTF-8 NFC ===> UTF-16 NFD
526 */
527 count = mbsnbytes(s, length);
528 }
529 u16.s = (char *)dest->s;
530 u16.length = dest->length << 1;
531 u16.buffer_length = dest->buffer_length;
532 if (sc->flag & SCONV_NORMALIZATION_C)
533 ret = archive_string_normalize_C(&u16, s, count, sc);
534 else
535 ret = archive_string_normalize_D(&u16, s, count, sc);
536 dest->s = (wchar_t *)u16.s;
537 dest->length = u16.length >> 1;
538 dest->buffer_length = u16.buffer_length;
539 sc->flag = saved_flag;/* restore the saved flag. */
540 return (ret);
541 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
542 count = utf16nbytes(s, length);
543 count >>= 1; /* to be WCS length */
544 /* Allocate memory for WCS. */
545 if (NULL == archive_wstring_ensure(dest,
546 dest->length + count + 1))
547 return (-1);
548 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
549 if ((sc->flag & SCONV_FROM_UTF16BE) && !IS_BIG_ENDIAN) {
550 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
551 size_t b;
552 for (b = 0; b < count; b++) {
553 uint16_t val = archive_le16dec(u16+b);
554 archive_be16enc(u16+b, val);
555 }
556 } else if ((sc->flag & SCONV_FROM_UTF16LE) && IS_BIG_ENDIAN) {
557 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
558 size_t b;
559 for (b = 0; b < count; b++) {
560 uint16_t val = archive_be16dec(u16+b);
561 archive_le16enc(u16+b, val);
562 }
563 }
564 } else {
565 DWORD mbflag;
566 size_t buffsize;
567
568 if (sc == NULL)
569 mbflag = 0;
570 else if (sc->flag & SCONV_FROM_CHARSET) {
571 /* Do not trust the length which comes from
572 * an archive file. */
573 length = mbsnbytes(s, length);
574 mbflag = 0;
575 } else
576 mbflag = MB_PRECOMPOSED;
577
578 mbflag |= MB_ERR_INVALID_CHARS;
579
580 buffsize = dest->length + length + 1;
581 do {
582 int r;
583
584 /* MultiByteToWideChar is limited to int. */
585 if (length > (size_t)INT_MAX ||
586 (dest->buffer_length >> 1) > (size_t)INT_MAX)
587 return (-1);
588 /* Allocate memory for WCS. */
589 if (NULL == archive_wstring_ensure(dest, buffsize))
590 return (-1);
591 /* Convert MBS to WCS. */
592 r = MultiByteToWideChar(from_cp,
593 mbflag, s, (int)length, dest->s + dest->length,
594 (int)(dest->buffer_length >> 1) -1);
595 if (r == 0 &&
596 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
597 /* Expand the WCS buffer. */
598 buffsize = dest->buffer_length << 1;
599 continue;
600 }
601 if (r == 0 && length != 0)
602 ret = -1;
603 count = (size_t)r;
604 break;
605 } while (1);
606 }
607 dest->length += count;
608 dest->s[dest->length] = L'\0';
609 return (ret);
610 }
611
612 #else
613
614 /*
615 * Convert MBS to WCS.
616 * Note: returns -1 if conversion fails.
617 */
618 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)619 archive_wstring_append_from_mbs(struct archive_wstring *dest,
620 const char *p, size_t len)
621 {
622 size_t r;
623 int ret_val = 0;
624 /*
625 * No single byte will be more than one wide character,
626 * so this length estimate will always be big enough.
627 */
628 // size_t wcs_length = len;
629 size_t mbs_length = len;
630 const char *mbs = p;
631 wchar_t *wcs;
632 #if HAVE_MBRTOWC
633 mbstate_t shift_state;
634
635 memset(&shift_state, 0, sizeof(shift_state));
636 #endif
637 /*
638 * As we decided to have wcs_length == mbs_length == len
639 * we can use len here instead of wcs_length
640 */
641 if (NULL == archive_wstring_ensure(dest, dest->length + len + 1))
642 return (-1);
643 wcs = dest->s + dest->length;
644 /*
645 * We cannot use mbsrtowcs/mbstowcs here because those may convert
646 * extra MBS when strlen(p) > len and one wide character consists of
647 * multi bytes.
648 */
649 while (*mbs && mbs_length > 0) {
650 /*
651 * The buffer we allocated is always big enough.
652 * Keep this code path in a comment if we decide to choose
653 * smaller wcs_length in the future
654 */
655 /*
656 if (wcs_length == 0) {
657 dest->length = wcs - dest->s;
658 dest->s[dest->length] = L'\0';
659 wcs_length = mbs_length;
660 if (NULL == archive_wstring_ensure(dest,
661 dest->length + wcs_length + 1))
662 return (-1);
663 wcs = dest->s + dest->length;
664 }
665 */
666 #if HAVE_MBRTOWC
667 r = mbrtowc(wcs, mbs, mbs_length, &shift_state);
668 #else
669 r = mbtowc(wcs, mbs, mbs_length);
670 #endif
671 if (r == (size_t)-1 || r == (size_t)-2) {
672 ret_val = -1;
673 break;
674 }
675 if (r == 0 || r > mbs_length)
676 break;
677 wcs++;
678 // wcs_length--;
679 mbs += r;
680 mbs_length -= r;
681 }
682 dest->length = wcs - dest->s;
683 dest->s[dest->length] = L'\0';
684 return (ret_val);
685 }
686
687 #endif
688
689 #if defined(_WIN32) && !defined(__CYGWIN__)
690
691 /*
692 * WCS ==> MBS.
693 * Note: returns -1 if conversion fails.
694 *
695 * Win32 builds use WideCharToMultiByte from the Windows API.
696 * (Maybe Cygwin should too? WideCharToMultiByte will know a
697 * lot more about local character encodings than the wcrtomb()
698 * wrapper is going to know.)
699 */
700 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)701 archive_string_append_from_wcs(struct archive_string *as,
702 const wchar_t *w, size_t len)
703 {
704 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
705 }
706
707 static int
archive_string_append_from_wcs_in_codepage(struct archive_string * as,const wchar_t * ws,size_t len,struct archive_string_conv * sc)708 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
709 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
710 {
711 BOOL defchar_used, *dp;
712 int ret = 0;
713 UINT to_cp;
714 size_t count, wslen = len;
715
716 if (sc != NULL)
717 to_cp = sc->to_cp;
718 else
719 to_cp = get_current_codepage();
720
721 if (to_cp == CP_C_LOCALE) {
722 /*
723 * "C" locale special processing.
724 */
725 const wchar_t *wp = ws;
726 char *p;
727
728 if (NULL == archive_string_ensure(as,
729 as->length + wslen +1))
730 return (-1);
731 p = as->s + as->length;
732 count = 0;
733 defchar_used = 0;
734 while (count < wslen && *wp) {
735 if (*wp > 255) {
736 *p++ = '?';
737 wp++;
738 defchar_used = 1;
739 } else
740 *p++ = (char)*wp++;
741 count++;
742 }
743 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
744 uint16_t *u16;
745
746 if (NULL ==
747 archive_string_ensure(as, as->length + len * 2 + 2))
748 return (-1);
749 u16 = (uint16_t *)(as->s + as->length);
750 count = 0;
751 defchar_used = 0;
752 if (sc->flag & SCONV_TO_UTF16BE) {
753 while (count < len && *ws) {
754 archive_be16enc(u16+count, *ws);
755 ws++;
756 count++;
757 }
758 } else {
759 while (count < len && *ws) {
760 archive_le16enc(u16+count, *ws);
761 ws++;
762 count++;
763 }
764 }
765 count <<= 1; /* to be byte size */
766 } else {
767 /* Make sure the MBS buffer has plenty to set. */
768 if (NULL ==
769 archive_string_ensure(as, as->length + len * 2 + 1))
770 return (-1);
771 do {
772 int r;
773
774 defchar_used = 0;
775 if (to_cp == CP_UTF8 || sc == NULL)
776 dp = NULL;
777 else
778 dp = &defchar_used;
779 /* WideCharToMultiByte is limited to int. */
780 if (as->buffer_length - as->length - 1 > (size_t)INT_MAX ||
781 wslen > (size_t)INT_MAX)
782 return (-1);
783 r = WideCharToMultiByte(to_cp, 0, ws, (int)wslen,
784 as->s + as->length,
785 (int)(as->buffer_length - as->length - 1), NULL, dp);
786 if (r == 0 &&
787 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
788 /* Expand the MBS buffer and retry. */
789 if (NULL == archive_string_ensure(as,
790 as->buffer_length + len))
791 return (-1);
792 continue;
793 }
794 if (r == 0)
795 ret = -1;
796 count = (size_t)r;
797 break;
798 } while (1);
799 }
800 as->length += count;
801 as->s[as->length] = '\0';
802 return (defchar_used?-1:ret);
803 }
804
805 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
806
807 /*
808 * Translates a wide character string into current locale character set
809 * and appends to the archive_string. Note: returns -1 if conversion
810 * fails.
811 */
812 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)813 archive_string_append_from_wcs(struct archive_string *as,
814 const wchar_t *w, size_t len)
815 {
816 /* We cannot use the standard wcstombs() here because it
817 * cannot tell us how big the output buffer should be. So
818 * I've built a loop around wcrtomb() or wctomb() that
819 * converts a character at a time and resizes the string as
820 * needed. We prefer wcrtomb() when it's available because
821 * it's thread-safe. */
822 int n, ret_val = 0;
823 char *p;
824 char *end;
825 #if HAVE_WCRTOMB
826 mbstate_t shift_state;
827
828 memset(&shift_state, 0, sizeof(shift_state));
829 #else
830 /* Clear the shift state before starting. */
831 wctomb(NULL, L'\0');
832 #endif
833 /*
834 * Allocate buffer for MBS.
835 * We need this allocation here since it is possible that
836 * as->s is still NULL.
837 */
838 if (archive_string_ensure(as, as->length + len + 1) == NULL)
839 return (-1);
840
841 p = as->s + as->length;
842 end = as->s + as->buffer_length - MB_CUR_MAX -1;
843 while (*w != L'\0' && len > 0) {
844 if (p >= end) {
845 as->length = p - as->s;
846 as->s[as->length] = '\0';
847 /* Re-allocate buffer for MBS. */
848 if (archive_string_ensure(as,
849 as->length + max(len * 2,
850 (size_t)MB_CUR_MAX) + 1) == NULL)
851 return (-1);
852 p = as->s + as->length;
853 end = as->s + as->buffer_length - MB_CUR_MAX -1;
854 }
855 #if HAVE_WCRTOMB
856 n = wcrtomb(p, *w++, &shift_state);
857 #else
858 n = wctomb(p, *w++);
859 #endif
860 if (n == -1) {
861 if (errno == EILSEQ) {
862 /* Skip an illegal wide char. */
863 *p++ = '?';
864 ret_val = -1;
865 } else {
866 ret_val = -1;
867 break;
868 }
869 } else
870 p += n;
871 len--;
872 }
873 as->length = p - as->s;
874 as->s[as->length] = '\0';
875 return (ret_val);
876 }
877
878 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
879
880 /*
881 * TODO: Test if __STDC_ISO_10646__ is defined.
882 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
883 * one character at a time. If a non-Windows platform doesn't have
884 * either of these, fall back to the built-in UTF8 conversion.
885 */
886 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)887 archive_string_append_from_wcs(struct archive_string *as,
888 const wchar_t *w, size_t len)
889 {
890 (void)as;/* UNUSED */
891 (void)w;/* UNUSED */
892 (void)len;/* UNUSED */
893 errno = ENOSYS;
894 return (-1);
895 }
896
897 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
898
899 /*
900 * Find a string conversion object by a pair of 'from' charset name
901 * and 'to' charset name from an archive object.
902 * Return NULL if not found.
903 */
904 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)905 find_sconv_object(struct archive *a, const char *fc, const char *tc)
906 {
907 struct archive_string_conv *sc;
908
909 if (a == NULL)
910 return (NULL);
911
912 for (sc = a->sconv; sc != NULL; sc = sc->next) {
913 if (strcmp(sc->from_charset, fc) == 0 &&
914 strcmp(sc->to_charset, tc) == 0)
915 break;
916 }
917 return (sc);
918 }
919
920 /*
921 * Register a string object to an archive object.
922 */
923 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)924 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
925 {
926 struct archive_string_conv **psc;
927
928 /* Add a new sconv to sconv list. */
929 psc = &(a->sconv);
930 while (*psc != NULL)
931 psc = &((*psc)->next);
932 *psc = sc;
933 }
934
935 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))936 add_converter(struct archive_string_conv *sc, int (*converter)
937 (struct archive_string *, const void *, size_t,
938 struct archive_string_conv *))
939 {
940 if (sc == NULL || sc->nconverter >= 2)
941 __archive_errx(1, "Programming error");
942 sc->converter[sc->nconverter++] = converter;
943 }
944
945 static void
setup_converter(struct archive_string_conv * sc)946 setup_converter(struct archive_string_conv *sc)
947 {
948
949 /* Reset. */
950 sc->nconverter = 0;
951
952 /*
953 * Perform special sequence for the incorrect UTF-8 filenames
954 * made by libarchive2.x.
955 */
956 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
957 add_converter(sc, strncat_from_utf8_libarchive2);
958 return;
959 }
960
961 /*
962 * Convert a string to UTF-16BE/LE.
963 */
964 if (sc->flag & SCONV_TO_UTF16) {
965 /*
966 * If the current locale is UTF-8, we can translate
967 * a UTF-8 string into a UTF-16BE string.
968 */
969 if (sc->flag & SCONV_FROM_UTF8) {
970 add_converter(sc, archive_string_append_unicode);
971 return;
972 }
973
974 #if defined(_WIN32) && !defined(__CYGWIN__)
975 if (sc->flag & SCONV_WIN_CP) {
976 if (sc->flag & SCONV_TO_UTF16BE)
977 add_converter(sc, win_strncat_to_utf16be);
978 else
979 add_converter(sc, win_strncat_to_utf16le);
980 return;
981 }
982 #endif
983
984 #if defined(HAVE_ICONV)
985 if (sc->cd != (iconv_t)-1) {
986 add_converter(sc, iconv_strncat_in_locale);
987 return;
988 }
989 #endif
990
991 if (sc->flag & SCONV_BEST_EFFORT) {
992 if (sc->flag & SCONV_TO_UTF16BE)
993 add_converter(sc,
994 best_effort_strncat_to_utf16be);
995 else
996 add_converter(sc,
997 best_effort_strncat_to_utf16le);
998 } else
999 /* Make sure we have no converter. */
1000 sc->nconverter = 0;
1001 return;
1002 }
1003
1004 /*
1005 * Convert a string from UTF-16BE/LE.
1006 */
1007 if (sc->flag & SCONV_FROM_UTF16) {
1008 /*
1009 * At least we should normalize a UTF-16BE string.
1010 */
1011 if (sc->flag & SCONV_NORMALIZATION_D)
1012 add_converter(sc,archive_string_normalize_D);
1013 else if (sc->flag & SCONV_NORMALIZATION_C)
1014 add_converter(sc, archive_string_normalize_C);
1015
1016 if (sc->flag & SCONV_TO_UTF8) {
1017 /*
1018 * If the current locale is UTF-8, we can translate
1019 * a UTF-16BE/LE string into a UTF-8 string directly.
1020 */
1021 if (!(sc->flag &
1022 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1023 add_converter(sc,
1024 archive_string_append_unicode);
1025 return;
1026 }
1027
1028 #if defined(_WIN32) && !defined(__CYGWIN__)
1029 if (sc->flag & SCONV_WIN_CP) {
1030 if (sc->flag & SCONV_FROM_UTF16BE)
1031 add_converter(sc, win_strncat_from_utf16be);
1032 else
1033 add_converter(sc, win_strncat_from_utf16le);
1034 return;
1035 }
1036 #endif
1037
1038 #if defined(HAVE_ICONV)
1039 if (sc->cd != (iconv_t)-1) {
1040 add_converter(sc, iconv_strncat_in_locale);
1041 return;
1042 }
1043 #endif
1044
1045 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1046 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1047 add_converter(sc, best_effort_strncat_from_utf16be);
1048 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1049 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1050 add_converter(sc, best_effort_strncat_from_utf16le);
1051 else
1052 /* Make sure we have no converter. */
1053 sc->nconverter = 0;
1054 return;
1055 }
1056
1057 if (sc->flag & SCONV_FROM_UTF8) {
1058 /*
1059 * At least we should normalize a UTF-8 string.
1060 */
1061 if (sc->flag & SCONV_NORMALIZATION_D)
1062 add_converter(sc,archive_string_normalize_D);
1063 else if (sc->flag & SCONV_NORMALIZATION_C)
1064 add_converter(sc, archive_string_normalize_C);
1065
1066 /*
1067 * Copy UTF-8 string with a check of CESU-8.
1068 * Apparently, iconv does not check surrogate pairs in UTF-8
1069 * when both from-charset and to-charset are UTF-8, and then
1070 * we use our UTF-8 copy code.
1071 */
1072 if (sc->flag & SCONV_TO_UTF8) {
1073 /*
1074 * If the current locale is UTF-8, we can translate
1075 * a UTF-16BE string into a UTF-8 string directly.
1076 */
1077 if (!(sc->flag &
1078 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1079 add_converter(sc, strncat_from_utf8_to_utf8);
1080 return;
1081 }
1082 }
1083
1084 #if defined(_WIN32) && !defined(__CYGWIN__)
1085 /*
1086 * On Windows we can use Windows API for a string conversion.
1087 */
1088 if (sc->flag & SCONV_WIN_CP) {
1089 add_converter(sc, strncat_in_codepage);
1090 return;
1091 }
1092 #endif
1093
1094 #if HAVE_ICONV
1095 if (sc->cd != (iconv_t)-1) {
1096 add_converter(sc, iconv_strncat_in_locale);
1097 /*
1098 * iconv generally does not support UTF-8-MAC and so
1099 * we have to the output of iconv from NFC to NFD if
1100 * need.
1101 */
1102 if ((sc->flag & SCONV_FROM_CHARSET) &&
1103 (sc->flag & SCONV_TO_UTF8)) {
1104 if (sc->flag & SCONV_NORMALIZATION_D)
1105 add_converter(sc, archive_string_normalize_D);
1106 }
1107 return;
1108 }
1109 #endif
1110
1111 /*
1112 * Try conversion in the best effort or no conversion.
1113 */
1114 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1115 add_converter(sc, best_effort_strncat_in_locale);
1116 else
1117 /* Make sure we have no converter. */
1118 sc->nconverter = 0;
1119 }
1120
1121 /*
1122 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1123 * and CP932 which are referenced in create_sconv_object().
1124 */
1125 static const char *
canonical_charset_name(const char * charset)1126 canonical_charset_name(const char *charset)
1127 {
1128 char cs[16];
1129 char *p;
1130 const char *s;
1131
1132 if (charset == NULL || charset[0] == '\0'
1133 || strlen(charset) > 15)
1134 return (charset);
1135
1136 /* Copy name to uppercase. */
1137 p = cs;
1138 s = charset;
1139 while (*s) {
1140 char c = *s++;
1141 if (c >= 'a' && c <= 'z')
1142 c -= 'a' - 'A';
1143 *p++ = c;
1144 }
1145 *p++ = '\0';
1146
1147 if (strcmp(cs, "UTF-8") == 0 ||
1148 strcmp(cs, "UTF8") == 0)
1149 return ("UTF-8");
1150 if (strcmp(cs, "UTF-16BE") == 0 ||
1151 strcmp(cs, "UTF16BE") == 0)
1152 return ("UTF-16BE");
1153 if (strcmp(cs, "UTF-16LE") == 0 ||
1154 strcmp(cs, "UTF16LE") == 0)
1155 return ("UTF-16LE");
1156 if (strcmp(cs, "CP932") == 0)
1157 return ("CP932");
1158 return (charset);
1159 }
1160
1161 /*
1162 * Create a string conversion object.
1163 */
1164 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1165 create_sconv_object(const char *fc, const char *tc,
1166 unsigned current_codepage, int flag)
1167 {
1168 struct archive_string_conv *sc;
1169
1170 sc = calloc(1, sizeof(*sc));
1171 if (sc == NULL)
1172 return (NULL);
1173 sc->next = NULL;
1174 sc->from_charset = strdup(fc);
1175 if (sc->from_charset == NULL) {
1176 free(sc);
1177 return (NULL);
1178 }
1179 sc->to_charset = strdup(tc);
1180 if (sc->to_charset == NULL) {
1181 free(sc->from_charset);
1182 free(sc);
1183 return (NULL);
1184 }
1185 archive_string_init(&sc->utftmp);
1186
1187 if (flag & SCONV_TO_CHARSET) {
1188 /*
1189 * Convert characters from the current locale charset to
1190 * a specified charset.
1191 */
1192 sc->from_cp = current_codepage;
1193 sc->to_cp = make_codepage_from_charset(tc);
1194 #if defined(_WIN32) && !defined(__CYGWIN__)
1195 if (IsValidCodePage(sc->to_cp))
1196 flag |= SCONV_WIN_CP;
1197 #endif
1198 } else if (flag & SCONV_FROM_CHARSET) {
1199 /*
1200 * Convert characters from a specified charset to
1201 * the current locale charset.
1202 */
1203 sc->to_cp = current_codepage;
1204 sc->from_cp = make_codepage_from_charset(fc);
1205 #if defined(_WIN32) && !defined(__CYGWIN__)
1206 if (IsValidCodePage(sc->from_cp))
1207 flag |= SCONV_WIN_CP;
1208 #endif
1209 }
1210
1211 /*
1212 * Check if "from charset" and "to charset" are the same.
1213 */
1214 if (strcmp(fc, tc) == 0 ||
1215 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1216 sc->same = 1;
1217 else
1218 sc->same = 0;
1219
1220 /*
1221 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1222 */
1223 if (strcmp(tc, "UTF-8") == 0)
1224 flag |= SCONV_TO_UTF8;
1225 else if (strcmp(tc, "UTF-16BE") == 0)
1226 flag |= SCONV_TO_UTF16BE;
1227 else if (strcmp(tc, "UTF-16LE") == 0)
1228 flag |= SCONV_TO_UTF16LE;
1229 if (strcmp(fc, "UTF-8") == 0)
1230 flag |= SCONV_FROM_UTF8;
1231 else if (strcmp(fc, "UTF-16BE") == 0)
1232 flag |= SCONV_FROM_UTF16BE;
1233 else if (strcmp(fc, "UTF-16LE") == 0)
1234 flag |= SCONV_FROM_UTF16LE;
1235 #if defined(_WIN32) && !defined(__CYGWIN__)
1236 if (sc->to_cp == CP_UTF8)
1237 flag |= SCONV_TO_UTF8;
1238 else if (sc->to_cp == CP_UTF16BE)
1239 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1240 else if (sc->to_cp == CP_UTF16LE)
1241 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1242 if (sc->from_cp == CP_UTF8)
1243 flag |= SCONV_FROM_UTF8;
1244 else if (sc->from_cp == CP_UTF16BE)
1245 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1246 else if (sc->from_cp == CP_UTF16LE)
1247 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1248 #endif
1249
1250 /*
1251 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1252 * handle it. So we have to translate NFD characters to NFC ones
1253 * ourselves before iconv handles. Another reason is to prevent
1254 * that the same sight of two filenames, one is NFC and other
1255 * is NFD, would be in its directory.
1256 * On Mac OS X, although its filesystem layer automatically
1257 * convert filenames to NFD, it would be useful for filename
1258 * comparing to find out the same filenames that we normalize
1259 * that to be NFD ourselves.
1260 */
1261 if ((flag & SCONV_FROM_CHARSET) &&
1262 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1263 #if defined(__APPLE__)
1264 if (flag & SCONV_TO_UTF8)
1265 flag |= SCONV_NORMALIZATION_D;
1266 else
1267 #endif
1268 flag |= SCONV_NORMALIZATION_C;
1269 }
1270 #if defined(__APPLE__)
1271 /*
1272 * In case writing an archive file, make sure that a filename
1273 * going to be passed to iconv is a Unicode NFC string since
1274 * a filename in HFS Plus filesystem is a Unicode NFD one and
1275 * iconv cannot handle it with "UTF-8" charset. It is simpler
1276 * than a use of "UTF-8-MAC" charset.
1277 */
1278 if ((flag & SCONV_TO_CHARSET) &&
1279 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1280 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1281 flag |= SCONV_NORMALIZATION_C;
1282 /*
1283 * In case reading an archive file. make sure that a filename
1284 * will be passed to users is a Unicode NFD string in order to
1285 * correctly compare the filename with other one which comes
1286 * from HFS Plus filesystem.
1287 */
1288 if ((flag & SCONV_FROM_CHARSET) &&
1289 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1290 (flag & SCONV_TO_UTF8))
1291 flag |= SCONV_NORMALIZATION_D;
1292 #endif
1293
1294 #if defined(HAVE_ICONV)
1295 sc->cd_w = (iconv_t)-1;
1296 /*
1297 * Create an iconv object.
1298 */
1299 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1300 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1301 (flag & SCONV_WIN_CP)) {
1302 /* This case we won't use iconv. */
1303 sc->cd = (iconv_t)-1;
1304 } else {
1305 sc->cd = iconv_open(tc, fc);
1306 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1307 /*
1308 * Unfortunately, all of iconv implements do support
1309 * "CP932" character-set, so we should use "SJIS"
1310 * instead if iconv_open failed.
1311 */
1312 if (strcmp(tc, "CP932") == 0)
1313 sc->cd = iconv_open("SJIS", fc);
1314 else if (strcmp(fc, "CP932") == 0)
1315 sc->cd = iconv_open(tc, "SJIS");
1316 }
1317 #if defined(_WIN32) && !defined(__CYGWIN__)
1318 /*
1319 * archive_mstring on Windows directly convert multi-bytes
1320 * into archive_wstring in order not to depend on locale
1321 * so that you can do a I18N programming. This will be
1322 * used only in archive_mstring_copy_mbs_len_l so far.
1323 */
1324 if (flag & SCONV_FROM_CHARSET) {
1325 sc->cd_w = iconv_open("UTF-8", fc);
1326 if (sc->cd_w == (iconv_t)-1 &&
1327 (sc->flag & SCONV_BEST_EFFORT)) {
1328 if (strcmp(fc, "CP932") == 0)
1329 sc->cd_w = iconv_open("UTF-8", "SJIS");
1330 }
1331 }
1332 #endif /* _WIN32 && !__CYGWIN__ */
1333 }
1334 #endif /* HAVE_ICONV */
1335
1336 sc->flag = flag;
1337
1338 /*
1339 * Set up converters.
1340 */
1341 setup_converter(sc);
1342
1343 return (sc);
1344 }
1345
1346 /*
1347 * Free a string conversion object.
1348 */
1349 static void
free_sconv_object(struct archive_string_conv * sc)1350 free_sconv_object(struct archive_string_conv *sc)
1351 {
1352 free(sc->from_charset);
1353 free(sc->to_charset);
1354 archive_string_free(&sc->utftmp);
1355 #if HAVE_ICONV
1356 if (sc->cd != (iconv_t)-1)
1357 iconv_close(sc->cd);
1358 if (sc->cd_w != (iconv_t)-1)
1359 iconv_close(sc->cd_w);
1360 #endif
1361 free(sc);
1362 }
1363
1364 #if defined(_WIN32) && !defined(__CYGWIN__)
1365 # if defined(WINAPI_FAMILY_PARTITION) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
1366 # define GetOEMCP() CP_OEMCP
1367 # endif
1368
1369 static unsigned
my_atoi(const char * p)1370 my_atoi(const char *p)
1371 {
1372 unsigned cp;
1373
1374 cp = 0;
1375 while (*p) {
1376 if (*p >= '0' && *p <= '9')
1377 cp = cp * 10 + (*p - '0');
1378 else
1379 return (-1);
1380 p++;
1381 }
1382 return (cp);
1383 }
1384
1385 /*
1386 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1387 * Return -1 if failed.
1388 *
1389 * Note: This translation code may be insufficient.
1390 */
1391 static struct charset {
1392 const char *name;
1393 unsigned cp;
1394 } charsets[] = {
1395 /* MUST BE SORTED! */
1396 {"ASCII", 1252},
1397 {"ASMO-708", 708},
1398 {"BIG5", 950},
1399 {"CHINESE", 936},
1400 {"CP367", 1252},
1401 {"CP819", 1252},
1402 {"CP1025", 21025},
1403 {"DOS-720", 720},
1404 {"DOS-862", 862},
1405 {"EUC-CN", 51936},
1406 {"EUC-JP", 51932},
1407 {"EUC-KR", 949},
1408 {"EUCCN", 51936},
1409 {"EUCJP", 51932},
1410 {"EUCKR", 949},
1411 {"GB18030", 54936},
1412 {"GB2312", 936},
1413 {"HEBREW", 1255},
1414 {"HZ-GB-2312", 52936},
1415 {"IBM273", 20273},
1416 {"IBM277", 20277},
1417 {"IBM278", 20278},
1418 {"IBM280", 20280},
1419 {"IBM284", 20284},
1420 {"IBM285", 20285},
1421 {"IBM290", 20290},
1422 {"IBM297", 20297},
1423 {"IBM367", 1252},
1424 {"IBM420", 20420},
1425 {"IBM423", 20423},
1426 {"IBM424", 20424},
1427 {"IBM819", 1252},
1428 {"IBM871", 20871},
1429 {"IBM880", 20880},
1430 {"IBM905", 20905},
1431 {"IBM924", 20924},
1432 {"ISO-8859-1", 28591},
1433 {"ISO-8859-13", 28603},
1434 {"ISO-8859-15", 28605},
1435 {"ISO-8859-2", 28592},
1436 {"ISO-8859-3", 28593},
1437 {"ISO-8859-4", 28594},
1438 {"ISO-8859-5", 28595},
1439 {"ISO-8859-6", 28596},
1440 {"ISO-8859-7", 28597},
1441 {"ISO-8859-8", 28598},
1442 {"ISO-8859-9", 28599},
1443 {"ISO8859-1", 28591},
1444 {"ISO8859-13", 28603},
1445 {"ISO8859-15", 28605},
1446 {"ISO8859-2", 28592},
1447 {"ISO8859-3", 28593},
1448 {"ISO8859-4", 28594},
1449 {"ISO8859-5", 28595},
1450 {"ISO8859-6", 28596},
1451 {"ISO8859-7", 28597},
1452 {"ISO8859-8", 28598},
1453 {"ISO8859-9", 28599},
1454 {"JOHAB", 1361},
1455 {"KOI8-R", 20866},
1456 {"KOI8-U", 21866},
1457 {"KS_C_5601-1987", 949},
1458 {"LATIN1", 1252},
1459 {"LATIN2", 28592},
1460 {"MACINTOSH", 10000},
1461 {"SHIFT-JIS", 932},
1462 {"SHIFT_JIS", 932},
1463 {"SJIS", 932},
1464 {"US", 1252},
1465 {"US-ASCII", 1252},
1466 {"UTF-16", 1200},
1467 {"UTF-16BE", 1201},
1468 {"UTF-16LE", 1200},
1469 {"UTF-8", CP_UTF8},
1470 {"X-EUROPA", 29001},
1471 {"X-MAC-ARABIC", 10004},
1472 {"X-MAC-CE", 10029},
1473 {"X-MAC-CHINESEIMP", 10008},
1474 {"X-MAC-CHINESETRAD", 10002},
1475 {"X-MAC-CROATIAN", 10082},
1476 {"X-MAC-CYRILLIC", 10007},
1477 {"X-MAC-GREEK", 10006},
1478 {"X-MAC-HEBREW", 10005},
1479 {"X-MAC-ICELANDIC", 10079},
1480 {"X-MAC-JAPANESE", 10001},
1481 {"X-MAC-KOREAN", 10003},
1482 {"X-MAC-ROMANIAN", 10010},
1483 {"X-MAC-THAI", 10021},
1484 {"X-MAC-TURKISH", 10081},
1485 {"X-MAC-UKRAINIAN", 10017},
1486 };
1487 static unsigned
make_codepage_from_charset(const char * charset)1488 make_codepage_from_charset(const char *charset)
1489 {
1490 char cs[16];
1491 char *p;
1492 unsigned cp;
1493 int a, b;
1494
1495 if (charset == NULL || strlen(charset) > 15)
1496 return -1;
1497
1498 /* Copy name to uppercase. */
1499 p = cs;
1500 while (*charset) {
1501 char c = *charset++;
1502 if (c >= 'a' && c <= 'z')
1503 c -= 'a' - 'A';
1504 *p++ = c;
1505 }
1506 *p++ = '\0';
1507 cp = -1;
1508
1509 /* Look it up in the table first, so that we can easily
1510 * override CP367, which we map to 1252 instead of 367. */
1511 a = 0;
1512 b = sizeof(charsets)/sizeof(charsets[0]);
1513 while (b > a) {
1514 int c = (b + a) / 2;
1515 int r = strcmp(charsets[c].name, cs);
1516 if (r < 0)
1517 a = c + 1;
1518 else if (r > 0)
1519 b = c;
1520 else
1521 return charsets[c].cp;
1522 }
1523
1524 /* If it's not in the table, try to parse it. */
1525 switch (*cs) {
1526 case 'C':
1527 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1528 cp = my_atoi(cs + 2);
1529 } else if (strcmp(cs, "CP_ACP") == 0)
1530 cp = get_current_codepage();
1531 else if (strcmp(cs, "CP_OEMCP") == 0)
1532 cp = get_current_oemcp();
1533 break;
1534 case 'I':
1535 if (cs[1] == 'B' && cs[2] == 'M' &&
1536 cs[3] >= '0' && cs[3] <= '9') {
1537 cp = my_atoi(cs + 3);
1538 }
1539 break;
1540 case 'W':
1541 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1542 cp = my_atoi(cs + 8);
1543 if (cp != 874 && (cp < 1250 || cp > 1258))
1544 cp = -1;/* This may invalid code. */
1545 }
1546 break;
1547 }
1548 return (cp);
1549 }
1550
1551 /*
1552 * Return ANSI Code Page of current locale set by setlocale().
1553 */
1554 static unsigned
get_current_codepage(void)1555 get_current_codepage(void)
1556 {
1557 char *locale, *p;
1558 unsigned cp;
1559
1560 locale = setlocale(LC_CTYPE, NULL);
1561 if (locale == NULL)
1562 return (GetACP());
1563 if (locale[0] == 'C' && locale[1] == '\0')
1564 return (CP_C_LOCALE);
1565 p = strrchr(locale, '.');
1566 if (p == NULL)
1567 return (GetACP());
1568 if ((strcmp(p+1, "utf8") == 0) || (strcmp(p+1, "UTF-8") == 0))
1569 return CP_UTF8;
1570 cp = my_atoi(p+1);
1571 if ((int)cp <= 0)
1572 return (GetACP());
1573 return (cp);
1574 }
1575
1576 /*
1577 * Translation table between Locale Name and ACP/OEMCP.
1578 */
1579 static struct {
1580 unsigned acp;
1581 unsigned ocp;
1582 const char *locale;
1583 } acp_ocp_map[] = {
1584 { 950, 950, "Chinese_Taiwan" },
1585 { 936, 936, "Chinese_People's Republic of China" },
1586 { 950, 950, "Chinese_Taiwan" },
1587 { 1250, 852, "Czech_Czech Republic" },
1588 { 1252, 850, "Danish_Denmark" },
1589 { 1252, 850, "Dutch_Netherlands" },
1590 { 1252, 850, "Dutch_Belgium" },
1591 { 1252, 437, "English_United States" },
1592 { 1252, 850, "English_Australia" },
1593 { 1252, 850, "English_Canada" },
1594 { 1252, 850, "English_New Zealand" },
1595 { 1252, 850, "English_United Kingdom" },
1596 { 1252, 437, "English_United States" },
1597 { 1252, 850, "Finnish_Finland" },
1598 { 1252, 850, "French_France" },
1599 { 1252, 850, "French_Belgium" },
1600 { 1252, 850, "French_Canada" },
1601 { 1252, 850, "French_Switzerland" },
1602 { 1252, 850, "German_Germany" },
1603 { 1252, 850, "German_Austria" },
1604 { 1252, 850, "German_Switzerland" },
1605 { 1253, 737, "Greek_Greece" },
1606 { 1250, 852, "Hungarian_Hungary" },
1607 { 1252, 850, "Icelandic_Iceland" },
1608 { 1252, 850, "Italian_Italy" },
1609 { 1252, 850, "Italian_Switzerland" },
1610 { 932, 932, "Japanese_Japan" },
1611 { 949, 949, "Korean_Korea" },
1612 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1613 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1614 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1615 { 1250, 852, "Polish_Poland" },
1616 { 1252, 850, "Portuguese_Portugal" },
1617 { 1252, 850, "Portuguese_Brazil" },
1618 { 1251, 866, "Russian_Russia" },
1619 { 1250, 852, "Slovak_Slovakia" },
1620 { 1252, 850, "Spanish_Spain" },
1621 { 1252, 850, "Spanish_Mexico" },
1622 { 1252, 850, "Spanish_Spain" },
1623 { 1252, 850, "Swedish_Sweden" },
1624 { 1254, 857, "Turkish_Turkey" },
1625 { 0, 0, NULL}
1626 };
1627
1628 /*
1629 * Return OEM Code Page of current locale set by setlocale().
1630 */
1631 static unsigned
get_current_oemcp(void)1632 get_current_oemcp(void)
1633 {
1634 int i;
1635 char *locale, *p;
1636 size_t len;
1637
1638 locale = setlocale(LC_CTYPE, NULL);
1639 if (locale == NULL)
1640 return (GetOEMCP());
1641 if (locale[0] == 'C' && locale[1] == '\0')
1642 return (CP_C_LOCALE);
1643
1644 p = strrchr(locale, '.');
1645 if (p == NULL)
1646 return (GetOEMCP());
1647 len = p - locale;
1648 for (i = 0; acp_ocp_map[i].acp; i++) {
1649 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1650 return (acp_ocp_map[i].ocp);
1651 }
1652 return (GetOEMCP());
1653 }
1654 #else
1655
1656 /*
1657 * POSIX platform does not use CodePage.
1658 */
1659
1660 static unsigned
get_current_codepage(void)1661 get_current_codepage(void)
1662 {
1663 return (-1);/* Unknown */
1664 }
1665 static unsigned
make_codepage_from_charset(const char * charset)1666 make_codepage_from_charset(const char *charset)
1667 {
1668 (void)charset; /* UNUSED */
1669 return (-1);/* Unknown */
1670 }
1671 static unsigned
get_current_oemcp(void)1672 get_current_oemcp(void)
1673 {
1674 return (-1);/* Unknown */
1675 }
1676
1677 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1678
1679 /*
1680 * Return a string conversion object.
1681 */
1682 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1683 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1684 {
1685 struct archive_string_conv *sc;
1686 unsigned current_codepage;
1687
1688 /* Check if we have made the sconv object. */
1689 sc = find_sconv_object(a, fc, tc);
1690 if (sc != NULL)
1691 return (sc);
1692
1693 if (a == NULL)
1694 current_codepage = get_current_codepage();
1695 else
1696 current_codepage = a->current_codepage;
1697
1698 sc = create_sconv_object(canonical_charset_name(fc),
1699 canonical_charset_name(tc), current_codepage, flag);
1700 if (sc == NULL) {
1701 if (a != NULL)
1702 archive_set_error(a, ENOMEM,
1703 "Could not allocate memory for "
1704 "a string conversion object");
1705 return (NULL);
1706 }
1707
1708 /*
1709 * If there is no converter for current string conversion object,
1710 * we cannot handle this conversion.
1711 */
1712 if (sc->nconverter == 0) {
1713 if (a != NULL) {
1714 #if HAVE_ICONV
1715 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1716 "iconv_open failed : Cannot handle ``%s''",
1717 (flag & SCONV_TO_CHARSET)?tc:fc);
1718 #else
1719 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1720 "A character-set conversion not fully supported "
1721 "on this platform");
1722 #endif
1723 }
1724 /* Failed; free a sconv object. */
1725 free_sconv_object(sc);
1726 return (NULL);
1727 }
1728
1729 /*
1730 * Success!
1731 */
1732 if (a != NULL)
1733 add_sconv_object(a, sc);
1734 return (sc);
1735 }
1736
1737 static const char *
get_current_charset(struct archive * a)1738 get_current_charset(struct archive *a)
1739 {
1740 const char *cur_charset;
1741
1742 if (a == NULL)
1743 cur_charset = default_iconv_charset("");
1744 else {
1745 cur_charset = default_iconv_charset(a->current_code);
1746 if (a->current_code == NULL) {
1747 a->current_code = strdup(cur_charset);
1748 a->current_codepage = get_current_codepage();
1749 a->current_oemcp = get_current_oemcp();
1750 }
1751 }
1752 return (cur_charset);
1753 }
1754
1755 /*
1756 * Make and Return a string conversion object.
1757 * Return NULL if the platform does not support the specified conversion
1758 * and best_effort is 0.
1759 * If best_effort is set, A string conversion object must be returned
1760 * unless memory allocation for the object fails, but the conversion
1761 * might fail when non-ASCII code is found.
1762 */
1763 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1764 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1765 int best_effort)
1766 {
1767 int flag = SCONV_TO_CHARSET;
1768
1769 if (best_effort)
1770 flag |= SCONV_BEST_EFFORT;
1771 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1772 }
1773
1774 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1775 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1776 int best_effort)
1777 {
1778 int flag = SCONV_FROM_CHARSET;
1779
1780 if (best_effort)
1781 flag |= SCONV_BEST_EFFORT;
1782 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1783 }
1784
1785 /*
1786 * archive_string_default_conversion_*_archive() are provided for Windows
1787 * platform because other archiver application use CP_OEMCP for
1788 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1789 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1790 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1791 * So we should make a string conversion between CP_ACP and CP_OEMCP
1792 * for compatibility.
1793 */
1794 #if defined(_WIN32) && !defined(__CYGWIN__)
1795 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1796 archive_string_default_conversion_for_read(struct archive *a)
1797 {
1798 const char *cur_charset = get_current_charset(a);
1799 char oemcp[16];
1800
1801 /* NOTE: a check of cur_charset is unneeded but we need
1802 * that get_current_charset() has been surely called at
1803 * this time whatever C compiler optimized. */
1804 if (cur_charset != NULL &&
1805 (a->current_codepage == CP_C_LOCALE ||
1806 a->current_codepage == a->current_oemcp))
1807 return (NULL);/* no conversion. */
1808
1809 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1810 /* Make sure a null termination must be set. */
1811 oemcp[sizeof(oemcp)-1] = '\0';
1812 return (get_sconv_object(a, oemcp, cur_charset,
1813 SCONV_FROM_CHARSET));
1814 }
1815
1816 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1817 archive_string_default_conversion_for_write(struct archive *a)
1818 {
1819 const char *cur_charset = get_current_charset(a);
1820 char oemcp[16];
1821
1822 /* NOTE: a check of cur_charset is unneeded but we need
1823 * that get_current_charset() has been surely called at
1824 * this time whatever C compiler optimized. */
1825 if (cur_charset != NULL &&
1826 (a->current_codepage == CP_C_LOCALE ||
1827 a->current_codepage == a->current_oemcp))
1828 return (NULL);/* no conversion. */
1829
1830 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1831 /* Make sure a null termination must be set. */
1832 oemcp[sizeof(oemcp)-1] = '\0';
1833 return (get_sconv_object(a, cur_charset, oemcp,
1834 SCONV_TO_CHARSET));
1835 }
1836 #else
1837 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1838 archive_string_default_conversion_for_read(struct archive *a)
1839 {
1840 (void)a; /* UNUSED */
1841 return (NULL);
1842 }
1843
1844 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1845 archive_string_default_conversion_for_write(struct archive *a)
1846 {
1847 (void)a; /* UNUSED */
1848 return (NULL);
1849 }
1850 #endif
1851
1852 /*
1853 * Dispose of all character conversion objects in the archive object.
1854 */
1855 void
archive_string_conversion_free(struct archive * a)1856 archive_string_conversion_free(struct archive *a)
1857 {
1858 struct archive_string_conv *sc;
1859 struct archive_string_conv *sc_next;
1860
1861 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1862 sc_next = sc->next;
1863 free_sconv_object(sc);
1864 }
1865 a->sconv = NULL;
1866 free(a->current_code);
1867 a->current_code = NULL;
1868 }
1869
1870 /*
1871 * Return a conversion charset name.
1872 */
1873 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1874 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1875 {
1876 if (sc->flag & SCONV_TO_CHARSET)
1877 return (sc->to_charset);
1878 else
1879 return (sc->from_charset);
1880 }
1881
1882 /*
1883 * Change the behavior of a string conversion.
1884 */
1885 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1886 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1887 {
1888 switch (opt) {
1889 /*
1890 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1891 * assumption that wchar_t was Unicode.
1892 * This option enables simulating the assumption in order to read
1893 * that filename correctly.
1894 */
1895 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1896 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1897 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1898 /*
1899 * Nothing to do for it since wchar_t on these platforms
1900 * is really Unicode.
1901 */
1902 (void)sc; /* UNUSED */
1903 #else
1904 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1905 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1906 /* Set up string converters. */
1907 setup_converter(sc);
1908 }
1909 #endif
1910 break;
1911 case SCONV_SET_OPT_NORMALIZATION_C:
1912 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1913 sc->flag |= SCONV_NORMALIZATION_C;
1914 sc->flag &= ~SCONV_NORMALIZATION_D;
1915 /* Set up string converters. */
1916 setup_converter(sc);
1917 }
1918 break;
1919 case SCONV_SET_OPT_NORMALIZATION_D:
1920 #if defined(HAVE_ICONV)
1921 /*
1922 * If iconv will take the string, do not change the
1923 * setting of the normalization.
1924 */
1925 if (!(sc->flag & SCONV_WIN_CP) &&
1926 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1927 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1928 break;
1929 #endif
1930 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1931 sc->flag |= SCONV_NORMALIZATION_D;
1932 sc->flag &= ~SCONV_NORMALIZATION_C;
1933 /* Set up string converters. */
1934 setup_converter(sc);
1935 }
1936 break;
1937 default:
1938 break;
1939 }
1940 }
1941
1942 /*
1943 *
1944 * Copy one archive_string to another in locale conversion.
1945 *
1946 * archive_strncat_l();
1947 * archive_strncpy_l();
1948 *
1949 */
1950
1951 static size_t
mbsnbytes(const void * _p,size_t n)1952 mbsnbytes(const void *_p, size_t n)
1953 {
1954 size_t s;
1955 const char *p, *pp;
1956
1957 if (_p == NULL)
1958 return (0);
1959 p = (const char *)_p;
1960
1961 /* Like strlen(p), except won't examine positions beyond p[n]. */
1962 s = 0;
1963 pp = p;
1964 while (s < n && *pp) {
1965 pp++;
1966 s++;
1967 }
1968 return (s);
1969 }
1970
1971 static size_t
utf16nbytes(const void * _p,size_t n)1972 utf16nbytes(const void *_p, size_t n)
1973 {
1974 size_t s;
1975 const char *p, *pp;
1976
1977 if (_p == NULL)
1978 return (0);
1979 p = (const char *)_p;
1980
1981 /* Like strlen(p), except won't examine positions beyond p[n]. */
1982 s = 0;
1983 pp = p;
1984 n >>= 1;
1985 while (s < n && (pp[0] || pp[1])) {
1986 pp += 2;
1987 s++;
1988 }
1989 return (s<<1);
1990 }
1991
1992 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1993 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1994 struct archive_string_conv *sc)
1995 {
1996 as->length = 0;
1997 return (archive_strncat_l(as, _p, n, sc));
1998 }
1999
2000 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2001 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
2002 struct archive_string_conv *sc)
2003 {
2004 const void *s;
2005 size_t length = 0;
2006 int i, r = 0, r2;
2007
2008 if (_p != NULL && n > 0) {
2009 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
2010 length = utf16nbytes(_p, n);
2011 else
2012 length = mbsnbytes(_p, n);
2013 }
2014
2015 /* We must allocate memory even if there is no data for conversion
2016 * or copy. This simulates archive_string_append behavior. */
2017 if (length == 0) {
2018 size_t tn = 1;
2019 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
2020 tn = 2;
2021 if (archive_string_ensure(as, as->length + tn) == NULL)
2022 return (-1);
2023 as->s[as->length] = 0;
2024 if (tn == 2)
2025 as->s[as->length+1] = 0;
2026 return (0);
2027 }
2028
2029 /*
2030 * If sc is NULL, we just make a copy.
2031 */
2032 if (sc == NULL) {
2033 if (archive_string_append(as, _p, length) == NULL)
2034 return (-1);/* No memory */
2035 return (0);
2036 }
2037
2038 s = _p;
2039 i = 0;
2040 if (sc->nconverter > 1) {
2041 sc->utftmp.length = 0;
2042 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2043 if (r2 != 0 && errno == ENOMEM)
2044 return (r2);
2045 if (r > r2)
2046 r = r2;
2047 s = sc->utftmp.s;
2048 length = sc->utftmp.length;
2049 ++i;
2050 }
2051 r2 = sc->converter[i](as, s, length, sc);
2052 if (r > r2)
2053 r = r2;
2054 return (r);
2055 }
2056
2057 struct archive_string *
archive_string_dirname(struct archive_string * as)2058 archive_string_dirname(struct archive_string *as)
2059 {
2060 /* strip trailing separators */
2061 while (as->length > 1 && as->s[as->length - 1] == '/')
2062 as->length--;
2063 /* strip final component */
2064 while (as->length > 0 && as->s[as->length - 1] != '/')
2065 as->length--;
2066 /* empty path -> cwd */
2067 if (as->length == 0)
2068 return (archive_strcat(as, "."));
2069 /* strip separator(s) */
2070 while (as->length > 1 && as->s[as->length - 1] == '/')
2071 as->length--;
2072 /* terminate */
2073 as->s[as->length] = '\0';
2074 return (as);
2075 }
2076
2077 #if HAVE_ICONV
2078
2079 /*
2080 * Return -1 if conversion fails.
2081 */
2082 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2083 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2084 size_t length, struct archive_string_conv *sc)
2085 {
2086 ICONV_CONST char *itp;
2087 size_t remaining;
2088 iconv_t cd;
2089 char *outp;
2090 size_t avail, bs;
2091 int return_value = 0; /* success */
2092 size_t to_size, from_size;
2093
2094 if (sc->flag & SCONV_TO_UTF16)
2095 to_size = 2;
2096 else
2097 to_size = 1;
2098 if (sc->flag & SCONV_FROM_UTF16)
2099 from_size = 2;
2100 else
2101 from_size = 1;
2102
2103 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2104 return (-1);
2105
2106 cd = sc->cd;
2107 itp = (char *)(uintptr_t)_p;
2108 remaining = length;
2109 outp = as->s + as->length;
2110 avail = as->buffer_length - as->length - to_size;
2111 while (remaining >= from_size) {
2112 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2113
2114 if (result != (size_t)-1)
2115 break; /* Conversion completed. */
2116
2117 if (errno == EILSEQ || errno == EINVAL) {
2118 /*
2119 * If an output charset is UTF-8 or UTF-16BE/LE,
2120 * unknown character should be U+FFFD
2121 * (replacement character).
2122 */
2123 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2124 size_t rbytes;
2125 if (sc->flag & SCONV_TO_UTF8)
2126 rbytes = sizeof(utf8_replacement_char);
2127 else
2128 rbytes = 2;
2129
2130 if (avail < rbytes) {
2131 as->length = outp - as->s;
2132 bs = as->buffer_length +
2133 (remaining * to_size) + rbytes;
2134 if (NULL ==
2135 archive_string_ensure(as, bs))
2136 return (-1);
2137 outp = as->s + as->length;
2138 avail = as->buffer_length
2139 - as->length - to_size;
2140 }
2141 if (sc->flag & SCONV_TO_UTF8)
2142 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2143 else if (sc->flag & SCONV_TO_UTF16BE)
2144 archive_be16enc(outp, UNICODE_R_CHAR);
2145 else
2146 archive_le16enc(outp, UNICODE_R_CHAR);
2147 outp += rbytes;
2148 avail -= rbytes;
2149 } else {
2150 /* Skip the illegal input bytes. */
2151 *outp++ = '?';
2152 avail--;
2153 }
2154 itp += from_size;
2155 remaining -= from_size;
2156 return_value = -1; /* failure */
2157 } else {
2158 /* E2BIG no output buffer,
2159 * Increase an output buffer. */
2160 as->length = outp - as->s;
2161 bs = as->buffer_length + remaining * 2;
2162 if (NULL == archive_string_ensure(as, bs))
2163 return (-1);
2164 outp = as->s + as->length;
2165 avail = as->buffer_length - as->length - to_size;
2166 }
2167 }
2168 as->length = outp - as->s;
2169 as->s[as->length] = 0;
2170 if (to_size == 2)
2171 as->s[as->length+1] = 0;
2172 return (return_value);
2173 }
2174
2175 #endif /* HAVE_ICONV */
2176
2177
2178 #if defined(_WIN32) && !defined(__CYGWIN__)
2179
2180 /*
2181 * Translate a string from a some CodePage to an another CodePage by
2182 * Windows APIs, and copy the result. Return -1 if conversion fails.
2183 */
2184 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2185 strncat_in_codepage(struct archive_string *as,
2186 const void *_p, size_t length, struct archive_string_conv *sc)
2187 {
2188 const char *s = (const char *)_p;
2189 struct archive_wstring aws;
2190 size_t l;
2191 int r, saved_flag;
2192
2193 archive_string_init(&aws);
2194 saved_flag = sc->flag;
2195 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2196 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2197 sc->flag = saved_flag;
2198 if (r != 0) {
2199 archive_wstring_free(&aws);
2200 if (errno != ENOMEM)
2201 archive_string_append(as, s, length);
2202 return (-1);
2203 }
2204
2205 l = as->length;
2206 r = archive_string_append_from_wcs_in_codepage(
2207 as, aws.s, aws.length, sc);
2208 if (r != 0 && errno != ENOMEM && l == as->length)
2209 archive_string_append(as, s, length);
2210 archive_wstring_free(&aws);
2211 return (r);
2212 }
2213
2214 /*
2215 * Test whether MBS ==> WCS is okay.
2216 */
2217 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2218 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2219 {
2220 const char *p = (const char *)_p;
2221 unsigned codepage;
2222 DWORD mbflag = MB_ERR_INVALID_CHARS;
2223
2224 if (sc->flag & SCONV_FROM_CHARSET)
2225 codepage = sc->to_cp;
2226 else
2227 codepage = sc->from_cp;
2228
2229 if (codepage == CP_C_LOCALE)
2230 return (0);
2231 if (codepage != CP_UTF8)
2232 mbflag |= MB_PRECOMPOSED;
2233
2234 if (n > (size_t)INT_MAX)
2235 return (-1); /* Invalid */
2236 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2237 return (-1); /* Invalid */
2238 return (0); /* Okay */
2239 }
2240
2241 #else
2242
2243 /*
2244 * Test whether MBS ==> WCS is okay.
2245 */
2246 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2247 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2248 {
2249 const char *p = (const char *)_p;
2250 size_t r;
2251
2252 #if HAVE_MBRTOWC
2253 mbstate_t shift_state;
2254
2255 memset(&shift_state, 0, sizeof(shift_state));
2256 #else
2257 /* Clear the shift state before starting. */
2258 mbtowc(NULL, NULL, 0);
2259 #endif
2260 while (n) {
2261 wchar_t wc;
2262
2263 #if HAVE_MBRTOWC
2264 r = mbrtowc(&wc, p, n, &shift_state);
2265 #else
2266 r = mbtowc(&wc, p, n);
2267 #endif
2268 if (r == (size_t)-1 || r == (size_t)-2)
2269 return (-1);/* Invalid. */
2270 if (r == 0)
2271 break;
2272 p += r;
2273 n -= r;
2274 }
2275 (void)sc; /* UNUSED */
2276 return (0); /* All Okey. */
2277 }
2278
2279 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2280
2281 /*
2282 * Basically returns -1 because we cannot make a conversion of charset
2283 * without iconv but in some cases this would return 0.
2284 * Returns 0 if all copied characters are ASCII.
2285 * Returns 0 if both from-locale and to-locale are the same and those
2286 * can be WCS with no error.
2287 */
2288 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2289 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2290 size_t length, struct archive_string_conv *sc)
2291 {
2292 size_t remaining;
2293 const uint8_t *itp;
2294 int return_value = 0; /* success */
2295
2296 /*
2297 * If both from-locale and to-locale is the same, this makes a copy.
2298 * And then this checks all copied MBS can be WCS if so returns 0.
2299 */
2300 if (sc->same) {
2301 if (archive_string_append(as, _p, length) == NULL)
2302 return (-1);/* No memory */
2303 return (invalid_mbs(_p, length, sc));
2304 }
2305
2306 /*
2307 * If a character is ASCII, this just copies it. If not, this
2308 * assigns '?' character instead but in UTF-8 locale this assigns
2309 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2310 * a Replacement Character in Unicode.
2311 */
2312
2313 remaining = length;
2314 itp = (const uint8_t *)_p;
2315 while (*itp && remaining > 0) {
2316 if (*itp > 127) {
2317 // Non-ASCII: Substitute with suitable replacement
2318 if (sc->flag & SCONV_TO_UTF8) {
2319 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2320 __archive_errx(1, "Out of memory");
2321 }
2322 } else {
2323 archive_strappend_char(as, '?');
2324 }
2325 return_value = -1;
2326 } else {
2327 archive_strappend_char(as, *itp);
2328 }
2329 ++itp;
2330 }
2331 return (return_value);
2332 }
2333
2334
2335 /*
2336 * Unicode conversion functions.
2337 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2338 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2339 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2340 * - UTF-16BE <===> UTF-8.
2341 *
2342 */
2343
2344 /*
2345 * Utility to convert a single UTF-8 sequence.
2346 *
2347 * Usually return used bytes, return used byte in negative value when
2348 * a unicode character is replaced with U+FFFD.
2349 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2350 * Recommended Practice for Replacement Characters.
2351 */
2352 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2353 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2354 {
2355 static const char utf8_count[256] = {
2356 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2357 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2358 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2359 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2360 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2361 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2362 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2363 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2364 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2365 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2366 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2367 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2368 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2369 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2370 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2371 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2372 };
2373 int ch, i;
2374 int cnt;
2375 uint32_t wc;
2376
2377 /* Sanity check. */
2378 if (n == 0)
2379 return (0);
2380 /*
2381 * Decode 1-4 bytes depending on the value of the first byte.
2382 */
2383 ch = (unsigned char)*s;
2384 if (ch == 0)
2385 return (0); /* Standard: return 0 for end-of-string. */
2386 cnt = utf8_count[ch];
2387
2388 /* Invalid sequence or there are not plenty bytes. */
2389 if (n < (size_t)cnt) {
2390 cnt = (int)n;
2391 for (i = 1; i < cnt; i++) {
2392 if ((s[i] & 0xc0) != 0x80) {
2393 cnt = i;
2394 break;
2395 }
2396 }
2397 goto invalid_sequence;
2398 }
2399
2400 /* Make a Unicode code point from a single UTF-8 sequence. */
2401 switch (cnt) {
2402 case 1: /* 1 byte sequence. */
2403 *pwc = ch & 0x7f;
2404 return (cnt);
2405 case 2: /* 2 bytes sequence. */
2406 if ((s[1] & 0xc0) != 0x80) {
2407 cnt = 1;
2408 goto invalid_sequence;
2409 }
2410 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2411 return (cnt);
2412 case 3: /* 3 bytes sequence. */
2413 if ((s[1] & 0xc0) != 0x80) {
2414 cnt = 1;
2415 goto invalid_sequence;
2416 }
2417 if ((s[2] & 0xc0) != 0x80) {
2418 cnt = 2;
2419 goto invalid_sequence;
2420 }
2421 wc = ((ch & 0x0f) << 12)
2422 | ((s[1] & 0x3f) << 6)
2423 | (s[2] & 0x3f);
2424 if (wc < 0x800)
2425 goto invalid_sequence;/* Overlong sequence. */
2426 break;
2427 case 4: /* 4 bytes sequence. */
2428 if ((s[1] & 0xc0) != 0x80) {
2429 cnt = 1;
2430 goto invalid_sequence;
2431 }
2432 if ((s[2] & 0xc0) != 0x80) {
2433 cnt = 2;
2434 goto invalid_sequence;
2435 }
2436 if ((s[3] & 0xc0) != 0x80) {
2437 cnt = 3;
2438 goto invalid_sequence;
2439 }
2440 wc = ((ch & 0x07) << 18)
2441 | ((s[1] & 0x3f) << 12)
2442 | ((s[2] & 0x3f) << 6)
2443 | (s[3] & 0x3f);
2444 if (wc < 0x10000)
2445 goto invalid_sequence;/* Overlong sequence. */
2446 break;
2447 default: /* Others are all invalid sequence. */
2448 if (ch == 0xc0 || ch == 0xc1)
2449 cnt = 2;
2450 else if (ch >= 0xf5 && ch <= 0xf7)
2451 cnt = 4;
2452 else if (ch >= 0xf8 && ch <= 0xfb)
2453 cnt = 5;
2454 else if (ch == 0xfc || ch == 0xfd)
2455 cnt = 6;
2456 else
2457 cnt = 1;
2458 if (n < (size_t)cnt)
2459 cnt = (int)n;
2460 for (i = 1; i < cnt; i++) {
2461 if ((s[i] & 0xc0) != 0x80) {
2462 cnt = i;
2463 break;
2464 }
2465 }
2466 goto invalid_sequence;
2467 }
2468
2469 /* The code point larger than 0x10FFFF is not legal
2470 * Unicode values. */
2471 if (wc > UNICODE_MAX)
2472 goto invalid_sequence;
2473 /* Correctly gets a Unicode, returns used bytes. */
2474 *pwc = wc;
2475 return (cnt);
2476 invalid_sequence:
2477 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2478 return (cnt * -1);
2479 }
2480
2481 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2482 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2483 {
2484 int cnt;
2485
2486 cnt = _utf8_to_unicode(pwc, s, n);
2487 /* Any of Surrogate pair is not legal Unicode values. */
2488 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2489 return (-3);
2490 return (cnt);
2491 }
2492
2493 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2494 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2495 {
2496 uc -= 0xD800;
2497 uc *= 0x400;
2498 uc += uc2 - 0xDC00;
2499 uc += 0x10000;
2500 return (uc);
2501 }
2502
2503 /*
2504 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2505 * removing surrogate pairs.
2506 *
2507 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2508 *
2509 * Usually return used bytes, return used byte in negative value when
2510 * a unicode character is replaced with U+FFFD.
2511 */
2512 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2513 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2514 {
2515 uint32_t wc = 0;
2516 int cnt;
2517
2518 cnt = _utf8_to_unicode(&wc, s, n);
2519 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2520 uint32_t wc2 = 0;
2521 if (n - 3 < 3) {
2522 /* Invalid byte sequence. */
2523 goto invalid_sequence;
2524 }
2525 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2526 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2527 /* Invalid byte sequence. */
2528 goto invalid_sequence;
2529 }
2530 wc = combine_surrogate_pair(wc, wc2);
2531 cnt = 6;
2532 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2533 /* Invalid byte sequence. */
2534 goto invalid_sequence;
2535 }
2536 *pwc = wc;
2537 return (cnt);
2538 invalid_sequence:
2539 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2540 if (cnt > 0)
2541 cnt *= -1;
2542 return (cnt);
2543 }
2544
2545 /*
2546 * Convert a Unicode code point to a single UTF-8 sequence.
2547 *
2548 * NOTE:This function does not check if the Unicode is legal or not.
2549 * Please you definitely check it before calling this.
2550 */
2551 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2552 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2553 {
2554 char *_p = p;
2555
2556 /* Invalid Unicode char maps to Replacement character */
2557 if (uc > UNICODE_MAX)
2558 uc = UNICODE_R_CHAR;
2559 /* Translate code point to UTF8 */
2560 if (uc <= 0x7f) {
2561 if (remaining == 0)
2562 return (0);
2563 *p++ = (char)uc;
2564 } else if (uc <= 0x7ff) {
2565 if (remaining < 2)
2566 return (0);
2567 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2568 *p++ = 0x80 | (uc & 0x3f);
2569 } else if (uc <= 0xffff) {
2570 if (remaining < 3)
2571 return (0);
2572 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2573 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2574 *p++ = 0x80 | (uc & 0x3f);
2575 } else {
2576 if (remaining < 4)
2577 return (0);
2578 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2579 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2580 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2581 *p++ = 0x80 | (uc & 0x3f);
2582 }
2583 return (p - _p);
2584 }
2585
2586 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2587 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2588 {
2589 return (utf16_to_unicode(pwc, s, n, 1));
2590 }
2591
2592 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2593 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2594 {
2595 return (utf16_to_unicode(pwc, s, n, 0));
2596 }
2597
2598 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2599 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2600 {
2601 const char *utf16 = s;
2602 unsigned uc;
2603
2604 if (n == 0)
2605 return (0);
2606 if (n == 1) {
2607 /* set the Replacement Character instead. */
2608 *pwc = UNICODE_R_CHAR;
2609 return (-1);
2610 }
2611
2612 if (be)
2613 uc = archive_be16dec(utf16);
2614 else
2615 uc = archive_le16dec(utf16);
2616 utf16 += 2;
2617
2618 /* If this is a surrogate pair, assemble the full code point.*/
2619 if (IS_HIGH_SURROGATE_LA(uc)) {
2620 unsigned uc2;
2621
2622 if (n >= 4) {
2623 if (be)
2624 uc2 = archive_be16dec(utf16);
2625 else
2626 uc2 = archive_le16dec(utf16);
2627 } else
2628 uc2 = 0;
2629 if (IS_LOW_SURROGATE_LA(uc2)) {
2630 uc = combine_surrogate_pair(uc, uc2);
2631 utf16 += 2;
2632 } else {
2633 /* Undescribed code point should be U+FFFD
2634 * (replacement character). */
2635 *pwc = UNICODE_R_CHAR;
2636 return (-2);
2637 }
2638 }
2639
2640 /*
2641 * Surrogate pair values(0xd800 through 0xdfff) are only
2642 * used by UTF-16, so, after above calculation, the code
2643 * must not be surrogate values, and Unicode has no codes
2644 * larger than 0x10ffff. Thus, those are not legal Unicode
2645 * values.
2646 */
2647 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2648 /* Undescribed code point should be U+FFFD
2649 * (replacement character). */
2650 *pwc = UNICODE_R_CHAR;
2651 return (((int)(utf16 - s)) * -1);
2652 }
2653 *pwc = uc;
2654 return ((int)(utf16 - s));
2655 }
2656
2657 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2658 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2659 {
2660 char *utf16 = p;
2661
2662 if (uc > 0xffff) {
2663 /* We have a code point that won't fit into a
2664 * wchar_t; convert it to a surrogate pair. */
2665 if (remaining < 4)
2666 return (0);
2667 uc -= 0x10000;
2668 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2669 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2670 return (4);
2671 } else {
2672 if (remaining < 2)
2673 return (0);
2674 archive_be16enc(utf16, (uint16_t)uc);
2675 return (2);
2676 }
2677 }
2678
2679 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2680 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2681 {
2682 char *utf16 = p;
2683
2684 if (uc > 0xffff) {
2685 /* We have a code point that won't fit into a
2686 * wchar_t; convert it to a surrogate pair. */
2687 if (remaining < 4)
2688 return (0);
2689 uc -= 0x10000;
2690 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2691 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2692 return (4);
2693 } else {
2694 if (remaining < 2)
2695 return (0);
2696 archive_le16enc(utf16, (uint16_t)uc);
2697 return (2);
2698 }
2699 }
2700
2701 /*
2702 * Append new UTF-8 string to existing UTF-8 string.
2703 * Existing string is assumed to already be in proper form;
2704 * the new string will have invalid sequences replaced and
2705 * surrogate pairs canonicalized.
2706 */
2707 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _src,size_t len,struct archive_string_conv * sc)2708 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2709 size_t len, struct archive_string_conv *sc)
2710 {
2711 int ret = 0;
2712 const char *src = _src;
2713 (void)sc; /* UNUSED */
2714
2715 /* Pre-extend the destination */
2716 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2717 return (-1);
2718
2719 /* Invariant: src points to the first UTF8 byte that hasn't
2720 * been copied to the destination `as`. */
2721 for (;;) {
2722 int n;
2723 uint32_t uc;
2724 const char *e = src;
2725
2726 /* Skip UTF-8 sequences until we reach end-of-string or
2727 * a code point that needs conversion. */
2728 while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2729 e += n;
2730 len -= n;
2731 }
2732 /* Copy the part that doesn't need conversion */
2733 if (e > src) {
2734 if (archive_string_append(as, src, e - src) == NULL)
2735 return (-1);
2736 src = e;
2737 }
2738
2739 if (n == 0) {
2740 /* We reached end-of-string */
2741 return (ret);
2742 } else {
2743 /* Next code point needs conversion */
2744 char t[4];
2745 size_t w;
2746
2747 /* Try decoding a surrogate pair */
2748 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2749 n = cesu8_to_unicode(&uc, src, len);
2750 }
2751 /* Not a (valid) surrogate, so use a replacement char */
2752 if (n < 0) {
2753 ret = -1; /* Return -1 if we used any replacement */
2754 n *= -1;
2755 }
2756 /* Consume converted code point */
2757 src += n;
2758 len -= n;
2759 /* Convert and append new UTF-8 sequence. */
2760 w = unicode_to_utf8(t, sizeof(t), uc);
2761 if (archive_string_append(as, t, w) == NULL)
2762 return (-1);
2763 }
2764 }
2765 }
2766
2767 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2768 archive_string_append_unicode(struct archive_string *as, const void *_p,
2769 size_t len, struct archive_string_conv *sc)
2770 {
2771 const char *s;
2772 char *p, *endp;
2773 uint32_t uc;
2774 size_t w;
2775 size_t ts, tm;
2776 int n, ret = 0;
2777 int (*parse)(uint32_t *, const char *, size_t);
2778 size_t (*unparse)(char *, size_t, uint32_t);
2779
2780 if (sc->flag & SCONV_TO_UTF16BE) {
2781 unparse = unicode_to_utf16be;
2782 ts = 2;
2783 } else if (sc->flag & SCONV_TO_UTF16LE) {
2784 unparse = unicode_to_utf16le;
2785 ts = 2;
2786 } else if (sc->flag & SCONV_TO_UTF8) {
2787 unparse = unicode_to_utf8;
2788 ts = 1;
2789 } else {
2790 /*
2791 * This case is going to be converted to another
2792 * character-set through iconv.
2793 */
2794 if (sc->flag & SCONV_FROM_UTF16BE) {
2795 unparse = unicode_to_utf16be;
2796 ts = 2;
2797 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2798 unparse = unicode_to_utf16le;
2799 ts = 2;
2800 } else {
2801 unparse = unicode_to_utf8;
2802 ts = 1;
2803 }
2804 }
2805
2806 if (sc->flag & SCONV_FROM_UTF16BE) {
2807 parse = utf16be_to_unicode;
2808 tm = 1;
2809 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2810 parse = utf16le_to_unicode;
2811 tm = 1;
2812 } else {
2813 parse = cesu8_to_unicode;
2814 tm = ts;
2815 }
2816
2817 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2818 return (-1);
2819
2820 s = (const char *)_p;
2821 p = as->s + as->length;
2822 endp = as->s + as->buffer_length - ts;
2823 while ((n = parse(&uc, s, len)) != 0) {
2824 if (n < 0) {
2825 /* Use a replaced unicode character. */
2826 n *= -1;
2827 ret = -1;
2828 }
2829 s += n;
2830 len -= n;
2831 while ((w = unparse(p, endp - p, uc)) == 0) {
2832 /* There is not enough output buffer so
2833 * we have to expand it. */
2834 as->length = p - as->s;
2835 if (archive_string_ensure(as,
2836 as->buffer_length + len * tm + ts) == NULL)
2837 return (-1);
2838 p = as->s + as->length;
2839 endp = as->s + as->buffer_length - ts;
2840 }
2841 p += w;
2842 }
2843 as->length = p - as->s;
2844 as->s[as->length] = '\0';
2845 if (ts == 2)
2846 as->s[as->length+1] = '\0';
2847 return (ret);
2848 }
2849
2850 /*
2851 * Following Constants for Hangul compositions this information comes from
2852 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2853 */
2854 #define HC_SBASE 0xAC00
2855 #define HC_LBASE 0x1100
2856 #define HC_VBASE 0x1161
2857 #define HC_TBASE 0x11A7
2858 #define HC_LCOUNT 19
2859 #define HC_VCOUNT 21
2860 #define HC_TCOUNT 28
2861 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2862 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2863
2864 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2865 get_nfc(uint32_t uc, uint32_t uc2)
2866 {
2867 int t, b;
2868
2869 t = 0;
2870 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2871 while (b >= t) {
2872 int m = (t + b) / 2;
2873 if (u_composition_table[m].cp1 < uc)
2874 t = m + 1;
2875 else if (u_composition_table[m].cp1 > uc)
2876 b = m - 1;
2877 else if (u_composition_table[m].cp2 < uc2)
2878 t = m + 1;
2879 else if (u_composition_table[m].cp2 > uc2)
2880 b = m - 1;
2881 else
2882 return (u_composition_table[m].nfc);
2883 }
2884 return (0);
2885 }
2886
2887 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2888 * Characters. */
2889
2890 /*
2891 * Update first code point.
2892 */
2893 #define UPDATE_UC(new_uc) do { \
2894 uc = new_uc; \
2895 ucptr = NULL; \
2896 } while (0)
2897
2898 /*
2899 * Replace first code point with second code point.
2900 */
2901 #define REPLACE_UC_WITH_UC2() do { \
2902 uc = uc2; \
2903 ucptr = uc2ptr; \
2904 n = n2; \
2905 } while (0)
2906
2907 #define EXPAND_BUFFER() do { \
2908 as->length = p - as->s; \
2909 if (archive_string_ensure(as, \
2910 as->buffer_length + len * tm + ts) == NULL)\
2911 return (-1); \
2912 p = as->s + as->length; \
2913 endp = as->s + as->buffer_length - ts; \
2914 } while (0)
2915
2916 #define UNPARSE(p, endp, uc) do { \
2917 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2918 EXPAND_BUFFER(); \
2919 } \
2920 p += w; \
2921 } while (0)
2922
2923 /*
2924 * Write first code point.
2925 * If the code point has not be changed from its original code,
2926 * this just copies it from its original buffer pointer.
2927 * If not, this converts it to UTF-8 byte sequence and copies it.
2928 */
2929 #define WRITE_UC() do { \
2930 if (ucptr) { \
2931 if (p + n > endp) \
2932 EXPAND_BUFFER(); \
2933 switch (n) { \
2934 case 4: \
2935 *p++ = *ucptr++; \
2936 /* FALL THROUGH */ \
2937 case 3: \
2938 *p++ = *ucptr++; \
2939 /* FALL THROUGH */ \
2940 case 2: \
2941 *p++ = *ucptr++; \
2942 /* FALL THROUGH */ \
2943 case 1: \
2944 *p++ = *ucptr; \
2945 break; \
2946 } \
2947 ucptr = NULL; \
2948 } else { \
2949 UNPARSE(p, endp, uc); \
2950 } \
2951 } while (0)
2952
2953 /*
2954 * Collect following decomposable code points.
2955 */
2956 #define COLLECT_CPS(start) do { \
2957 int _i; \
2958 for (_i = start; _i < FDC_MAX ; _i++) { \
2959 nx = parse(&ucx[_i], s, len); \
2960 if (nx <= 0) \
2961 break; \
2962 cx = CCC(ucx[_i]); \
2963 if (cl >= cx && cl != 228 && cx != 228)\
2964 break; \
2965 s += nx; \
2966 len -= nx; \
2967 cl = cx; \
2968 ccx[_i] = cx; \
2969 } \
2970 if (_i >= FDC_MAX) { \
2971 ret = -1; \
2972 ucx_size = FDC_MAX; \
2973 } else \
2974 ucx_size = _i; \
2975 } while (0)
2976
2977 /*
2978 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2979 *
2980 * TODO: Convert composition exclusions, which are never converted
2981 * from NFC,NFD,NFKC and NFKD, to Form C.
2982 */
2983 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2984 archive_string_normalize_C(struct archive_string *as, const void *_p,
2985 size_t len, struct archive_string_conv *sc)
2986 {
2987 const char *s = (const char *)_p;
2988 char *p, *endp;
2989 uint32_t uc, uc2;
2990 size_t w;
2991 int always_replace, n, n2, ret = 0, spair, ts, tm;
2992 int (*parse)(uint32_t *, const char *, size_t);
2993 size_t (*unparse)(char *, size_t, uint32_t);
2994
2995 always_replace = 1;
2996 ts = 1;/* text size. */
2997 if (sc->flag & SCONV_TO_UTF16BE) {
2998 unparse = unicode_to_utf16be;
2999 ts = 2;
3000 if (sc->flag & SCONV_FROM_UTF16BE)
3001 always_replace = 0;
3002 } else if (sc->flag & SCONV_TO_UTF16LE) {
3003 unparse = unicode_to_utf16le;
3004 ts = 2;
3005 if (sc->flag & SCONV_FROM_UTF16LE)
3006 always_replace = 0;
3007 } else if (sc->flag & SCONV_TO_UTF8) {
3008 unparse = unicode_to_utf8;
3009 if (sc->flag & SCONV_FROM_UTF8)
3010 always_replace = 0;
3011 } else {
3012 /*
3013 * This case is going to be converted to another
3014 * character-set through iconv.
3015 */
3016 always_replace = 0;
3017 if (sc->flag & SCONV_FROM_UTF16BE) {
3018 unparse = unicode_to_utf16be;
3019 ts = 2;
3020 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3021 unparse = unicode_to_utf16le;
3022 ts = 2;
3023 } else {
3024 unparse = unicode_to_utf8;
3025 }
3026 }
3027
3028 if (sc->flag & SCONV_FROM_UTF16BE) {
3029 parse = utf16be_to_unicode;
3030 tm = 1;
3031 spair = 4;/* surrogate pair size in UTF-16. */
3032 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3033 parse = utf16le_to_unicode;
3034 tm = 1;
3035 spair = 4;/* surrogate pair size in UTF-16. */
3036 } else {
3037 parse = cesu8_to_unicode;
3038 tm = ts;
3039 spair = 6;/* surrogate pair size in UTF-8. */
3040 }
3041
3042 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3043 return (-1);
3044
3045 p = as->s + as->length;
3046 endp = as->s + as->buffer_length - ts;
3047 while ((n = parse(&uc, s, len)) != 0) {
3048 const char *ucptr, *uc2ptr;
3049
3050 if (n < 0) {
3051 /* Use a replaced unicode character. */
3052 UNPARSE(p, endp, uc);
3053 s += n*-1;
3054 len -= n*-1;
3055 ret = -1;
3056 continue;
3057 } else if (n == spair || always_replace)
3058 /* uc is converted from a surrogate pair.
3059 * this should be treated as a changed code. */
3060 ucptr = NULL;
3061 else
3062 ucptr = s;
3063 s += n;
3064 len -= n;
3065
3066 /* Read second code point. */
3067 while ((n2 = parse(&uc2, s, len)) > 0) {
3068 uint32_t ucx[FDC_MAX];
3069 int ccx[FDC_MAX];
3070 int cl, cx, i, nx, ucx_size;
3071 int LIndex,SIndex;
3072 uint32_t nfc;
3073
3074 if (n2 == spair || always_replace)
3075 /* uc2 is converted from a surrogate pair.
3076 * this should be treated as a changed code. */
3077 uc2ptr = NULL;
3078 else
3079 uc2ptr = s;
3080 s += n2;
3081 len -= n2;
3082
3083 /*
3084 * If current second code point is out of decomposable
3085 * code points, finding compositions is unneeded.
3086 */
3087 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3088 WRITE_UC();
3089 REPLACE_UC_WITH_UC2();
3090 continue;
3091 }
3092
3093 /*
3094 * Try to combine current code points.
3095 */
3096 /*
3097 * We have to combine Hangul characters according to
3098 * http://uniicode.org/reports/tr15/#Hangul
3099 */
3100 if (0 <= (LIndex = uc - HC_LBASE) &&
3101 LIndex < HC_LCOUNT) {
3102 /*
3103 * Hangul Composition.
3104 * 1. Two current code points are L and V.
3105 */
3106 int VIndex = uc2 - HC_VBASE;
3107 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3108 /* Make syllable of form LV. */
3109 UPDATE_UC(HC_SBASE +
3110 (LIndex * HC_VCOUNT + VIndex) *
3111 HC_TCOUNT);
3112 } else {
3113 WRITE_UC();
3114 REPLACE_UC_WITH_UC2();
3115 }
3116 continue;
3117 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3118 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3119 /*
3120 * Hangul Composition.
3121 * 2. Two current code points are LV and T.
3122 */
3123 int TIndex = uc2 - HC_TBASE;
3124 if (0 < TIndex && TIndex < HC_TCOUNT) {
3125 /* Make syllable of form LVT. */
3126 UPDATE_UC(uc + TIndex);
3127 } else {
3128 WRITE_UC();
3129 REPLACE_UC_WITH_UC2();
3130 }
3131 continue;
3132 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3133 /* A composition to current code points
3134 * is found. */
3135 UPDATE_UC(nfc);
3136 continue;
3137 } else if ((cl = CCC(uc2)) == 0) {
3138 /* Clearly 'uc2' the second code point is not
3139 * a decomposable code. */
3140 WRITE_UC();
3141 REPLACE_UC_WITH_UC2();
3142 continue;
3143 }
3144
3145 /*
3146 * Collect following decomposable code points.
3147 */
3148 cx = 0;
3149 ucx[0] = uc2;
3150 ccx[0] = cl;
3151 COLLECT_CPS(1);
3152
3153 /*
3154 * Find a composed code in the collected code points.
3155 */
3156 i = 1;
3157 while (i < ucx_size) {
3158 int j;
3159
3160 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3161 i++;
3162 continue;
3163 }
3164
3165 /*
3166 * nfc is composed of uc and ucx[i].
3167 */
3168 UPDATE_UC(nfc);
3169
3170 /*
3171 * Remove ucx[i] by shifting
3172 * following code points.
3173 */
3174 for (j = i; j+1 < ucx_size; j++) {
3175 ucx[j] = ucx[j+1];
3176 ccx[j] = ccx[j+1];
3177 }
3178 ucx_size --;
3179
3180 /*
3181 * Collect following code points blocked
3182 * by ucx[i] the removed code point.
3183 */
3184 if (ucx_size > 0 && i == ucx_size &&
3185 nx > 0 && cx == cl) {
3186 cl = ccx[ucx_size-1];
3187 COLLECT_CPS(ucx_size);
3188 }
3189 /*
3190 * Restart finding a composed code with
3191 * the updated uc from the top of the
3192 * collected code points.
3193 */
3194 i = 0;
3195 }
3196
3197 /*
3198 * Apparently the current code points are not
3199 * decomposed characters or already composed.
3200 */
3201 WRITE_UC();
3202 for (i = 0; i < ucx_size; i++)
3203 UNPARSE(p, endp, ucx[i]);
3204
3205 /*
3206 * Flush out remaining canonical combining characters.
3207 */
3208 if (nx > 0 && cx == cl && len > 0) {
3209 while ((nx = parse(&ucx[0], s, len))
3210 > 0) {
3211 cx = CCC(ucx[0]);
3212 if (cl > cx)
3213 break;
3214 s += nx;
3215 len -= nx;
3216 cl = cx;
3217 UNPARSE(p, endp, ucx[0]);
3218 }
3219 }
3220 break;
3221 }
3222 if (n2 < 0) {
3223 WRITE_UC();
3224 /* Use a replaced unicode character. */
3225 UNPARSE(p, endp, uc2);
3226 s += n2*-1;
3227 len -= n2*-1;
3228 ret = -1;
3229 continue;
3230 } else if (n2 == 0) {
3231 WRITE_UC();
3232 break;
3233 }
3234 }
3235 as->length = p - as->s;
3236 as->s[as->length] = '\0';
3237 if (ts == 2)
3238 as->s[as->length+1] = '\0';
3239 return (ret);
3240 }
3241
3242 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3243 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3244 {
3245 int t, b;
3246
3247 /*
3248 * These are not converted to NFD on Mac OS.
3249 */
3250 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3251 (uc >= 0xF900 && uc <= 0xFAFF) ||
3252 (uc >= 0x2F800 && uc <= 0x2FAFF))
3253 return (0);
3254 /*
3255 * Those code points are not converted to NFD on Mac OS.
3256 * I do not know the reason because it is undocumented.
3257 * NFC NFD
3258 * 1109A ==> 11099 110BA
3259 * 1109C ==> 1109B 110BA
3260 * 110AB ==> 110A5 110BA
3261 */
3262 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3263 return (0);
3264
3265 t = 0;
3266 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3267 while (b >= t) {
3268 int m = (t + b) / 2;
3269 if (u_decomposition_table[m].nfc < uc)
3270 t = m + 1;
3271 else if (u_decomposition_table[m].nfc > uc)
3272 b = m - 1;
3273 else {
3274 *cp1 = u_decomposition_table[m].cp1;
3275 *cp2 = u_decomposition_table[m].cp2;
3276 return (1);
3277 }
3278 }
3279 return (0);
3280 }
3281
3282 #define REPLACE_UC_WITH(cp) do { \
3283 uc = cp; \
3284 ucptr = NULL; \
3285 } while (0)
3286
3287 /*
3288 * Normalize UTF-8 characters to Form D and copy the result.
3289 */
3290 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3291 archive_string_normalize_D(struct archive_string *as, const void *_p,
3292 size_t len, struct archive_string_conv *sc)
3293 {
3294 const char *s = (const char *)_p;
3295 char *p, *endp;
3296 uint32_t uc, uc2;
3297 size_t w;
3298 int always_replace, n, n2, ret = 0, spair, ts, tm;
3299 int (*parse)(uint32_t *, const char *, size_t);
3300 size_t (*unparse)(char *, size_t, uint32_t);
3301
3302 always_replace = 1;
3303 ts = 1;/* text size. */
3304 if (sc->flag & SCONV_TO_UTF16BE) {
3305 unparse = unicode_to_utf16be;
3306 ts = 2;
3307 if (sc->flag & SCONV_FROM_UTF16BE)
3308 always_replace = 0;
3309 } else if (sc->flag & SCONV_TO_UTF16LE) {
3310 unparse = unicode_to_utf16le;
3311 ts = 2;
3312 if (sc->flag & SCONV_FROM_UTF16LE)
3313 always_replace = 0;
3314 } else if (sc->flag & SCONV_TO_UTF8) {
3315 unparse = unicode_to_utf8;
3316 if (sc->flag & SCONV_FROM_UTF8)
3317 always_replace = 0;
3318 } else {
3319 /*
3320 * This case is going to be converted to another
3321 * character-set through iconv.
3322 */
3323 always_replace = 0;
3324 if (sc->flag & SCONV_FROM_UTF16BE) {
3325 unparse = unicode_to_utf16be;
3326 ts = 2;
3327 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3328 unparse = unicode_to_utf16le;
3329 ts = 2;
3330 } else {
3331 unparse = unicode_to_utf8;
3332 }
3333 }
3334
3335 if (sc->flag & SCONV_FROM_UTF16BE) {
3336 parse = utf16be_to_unicode;
3337 tm = 1;
3338 spair = 4;/* surrogate pair size in UTF-16. */
3339 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3340 parse = utf16le_to_unicode;
3341 tm = 1;
3342 spair = 4;/* surrogate pair size in UTF-16. */
3343 } else {
3344 parse = cesu8_to_unicode;
3345 tm = ts;
3346 spair = 6;/* surrogate pair size in UTF-8. */
3347 }
3348
3349 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3350 return (-1);
3351
3352 p = as->s + as->length;
3353 endp = as->s + as->buffer_length - ts;
3354 while ((n = parse(&uc, s, len)) != 0) {
3355 const char *ucptr;
3356 uint32_t cp1, cp2;
3357 int SIndex;
3358 struct {
3359 uint32_t uc;
3360 int ccc;
3361 } fdc[FDC_MAX];
3362 int fdi, fdj;
3363 int ccc;
3364
3365 check_first_code:
3366 if (n < 0) {
3367 /* Use a replaced unicode character. */
3368 UNPARSE(p, endp, uc);
3369 s += n*-1;
3370 len -= n*-1;
3371 ret = -1;
3372 continue;
3373 } else if (n == spair || always_replace)
3374 /* uc is converted from a surrogate pair.
3375 * this should be treated as a changed code. */
3376 ucptr = NULL;
3377 else
3378 ucptr = s;
3379 s += n;
3380 len -= n;
3381
3382 /* Hangul Decomposition. */
3383 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3384 int L = HC_LBASE + SIndex / HC_NCOUNT;
3385 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3386 int T = HC_TBASE + SIndex % HC_TCOUNT;
3387
3388 REPLACE_UC_WITH(L);
3389 WRITE_UC();
3390 REPLACE_UC_WITH(V);
3391 WRITE_UC();
3392 if (T != HC_TBASE) {
3393 REPLACE_UC_WITH(T);
3394 WRITE_UC();
3395 }
3396 continue;
3397 }
3398 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3399 WRITE_UC();
3400 continue;
3401 }
3402
3403 fdi = 0;
3404 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3405 int k;
3406
3407 for (k = fdi; k > 0; k--)
3408 fdc[k] = fdc[k-1];
3409 fdc[0].ccc = CCC(cp2);
3410 fdc[0].uc = cp2;
3411 fdi++;
3412 REPLACE_UC_WITH(cp1);
3413 }
3414
3415 /* Read following code points. */
3416 while ((n2 = parse(&uc2, s, len)) > 0 &&
3417 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3418 int j, k;
3419
3420 s += n2;
3421 len -= n2;
3422 for (j = 0; j < fdi; j++) {
3423 if (fdc[j].ccc > ccc)
3424 break;
3425 }
3426 if (j < fdi) {
3427 for (k = fdi; k > j; k--)
3428 fdc[k] = fdc[k-1];
3429 fdc[j].ccc = ccc;
3430 fdc[j].uc = uc2;
3431 } else {
3432 fdc[fdi].ccc = ccc;
3433 fdc[fdi].uc = uc2;
3434 }
3435 fdi++;
3436 }
3437
3438 WRITE_UC();
3439 for (fdj = 0; fdj < fdi; fdj++) {
3440 REPLACE_UC_WITH(fdc[fdj].uc);
3441 WRITE_UC();
3442 }
3443
3444 if (n2 == 0)
3445 break;
3446 REPLACE_UC_WITH(uc2);
3447 n = n2;
3448 goto check_first_code;
3449 }
3450 as->length = p - as->s;
3451 as->s[as->length] = '\0';
3452 if (ts == 2)
3453 as->s[as->length+1] = '\0';
3454 return (ret);
3455 }
3456
3457 /*
3458 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3459 * that WCS is Unicode. It is true for several platforms but some are false.
3460 * And then people who did not use UTF-8 locale on the non Unicode WCS
3461 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3462 * now cannot get right filename from libarchive 3.x and later since we
3463 * fixed the wrong assumption and it is incompatible to older its versions.
3464 * So we provide special option, "compat-2x.x", for resolving it.
3465 * That option enable the string conversion of libarchive 2.x.
3466 *
3467 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3468 * locale character set and appends to the archive_string.
3469 * Note: returns -1 if conversion fails.
3470 */
3471 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3472 strncat_from_utf8_libarchive2(struct archive_string *as,
3473 const void *_p, size_t len, struct archive_string_conv *sc)
3474 {
3475 const char *s;
3476 int n;
3477 char *p;
3478 char *end;
3479 uint32_t unicode;
3480 #if HAVE_WCRTOMB
3481 mbstate_t shift_state;
3482
3483 memset(&shift_state, 0, sizeof(shift_state));
3484 #else
3485 /* Clear the shift state before starting. */
3486 wctomb(NULL, L'\0');
3487 #endif
3488 (void)sc; /* UNUSED */
3489 /*
3490 * Allocate buffer for MBS.
3491 * We need this allocation here since it is possible that
3492 * as->s is still NULL.
3493 */
3494 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3495 return (-1);
3496
3497 s = (const char *)_p;
3498 p = as->s + as->length;
3499 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3500 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3501 wchar_t wc;
3502
3503 if (p >= end) {
3504 as->length = p - as->s;
3505 /* Re-allocate buffer for MBS. */
3506 if (archive_string_ensure(as,
3507 as->length + max(len * 2,
3508 (size_t)MB_CUR_MAX) + 1) == NULL)
3509 return (-1);
3510 p = as->s + as->length;
3511 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3512 }
3513
3514 /*
3515 * As libarchive 2.x, translates the UTF-8 characters into
3516 * wide-characters in the assumption that WCS is Unicode.
3517 */
3518 if (n < 0) {
3519 n *= -1;
3520 wc = L'?';
3521 } else
3522 wc = (wchar_t)unicode;
3523
3524 s += n;
3525 len -= n;
3526 /*
3527 * Translates the wide-character into the current locale MBS.
3528 */
3529 #if HAVE_WCRTOMB
3530 n = (int)wcrtomb(p, wc, &shift_state);
3531 #else
3532 n = (int)wctomb(p, wc);
3533 #endif
3534 if (n == -1)
3535 return (-1);
3536 p += n;
3537 }
3538 as->length = p - as->s;
3539 as->s[as->length] = '\0';
3540 return (0);
3541 }
3542
3543
3544 /*
3545 * Conversion functions between current locale dependent MBS and UTF-16BE.
3546 * strncat_from_utf16be() : UTF-16BE --> MBS
3547 * strncat_to_utf16be() : MBS --> UTF16BE
3548 */
3549
3550 #if defined(_WIN32) && !defined(__CYGWIN__)
3551
3552 /*
3553 * Convert a UTF-16BE/LE string to current locale and copy the result.
3554 * Return -1 if conversion fails.
3555 */
3556 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3557 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3558 struct archive_string_conv *sc, int be)
3559 {
3560 struct archive_string tmp;
3561 const char *u16;
3562 BOOL defchar;
3563 char *mbs;
3564 size_t mbs_size, b, ll;
3565 int ret = 0;
3566
3567 bytes &= ~1;
3568 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3569 return (-1);
3570
3571 mbs = as->s + as->length;
3572 mbs_size = as->buffer_length - as->length -1;
3573
3574 if (sc->to_cp == CP_C_LOCALE) {
3575 /*
3576 * "C" locale special process.
3577 */
3578 u16 = _p;
3579 ll = 0;
3580 for (b = 0; b < bytes; b += 2) {
3581 uint16_t val;
3582 if (be)
3583 val = archive_be16dec(u16+b);
3584 else
3585 val = archive_le16dec(u16+b);
3586 if (val > 255) {
3587 *mbs++ = '?';
3588 ret = -1;
3589 } else
3590 *mbs++ = (char)(val&0xff);
3591 ll++;
3592 }
3593 as->length += ll;
3594 as->s[as->length] = '\0';
3595 return (ret);
3596 }
3597
3598 archive_string_init(&tmp);
3599 if (be) {
3600 if (IS_BIG_ENDIAN) {
3601 u16 = _p;
3602 } else {
3603 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3604 return (-1);
3605 memcpy(tmp.s, _p, bytes);
3606 for (b = 0; b < bytes; b += 2) {
3607 uint16_t val = archive_be16dec(tmp.s+b);
3608 archive_le16enc(tmp.s+b, val);
3609 }
3610 u16 = tmp.s;
3611 }
3612 } else {
3613 if (!IS_BIG_ENDIAN) {
3614 u16 = _p;
3615 } else {
3616 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3617 return (-1);
3618 memcpy(tmp.s, _p, bytes);
3619 for (b = 0; b < bytes; b += 2) {
3620 uint16_t val = archive_le16dec(tmp.s+b);
3621 archive_be16enc(tmp.s+b, val);
3622 }
3623 u16 = tmp.s;
3624 }
3625 }
3626
3627 do {
3628 int r;
3629 defchar = 0;
3630 /* WideCharToMultiByte is limited to int. */
3631 if (bytes > (size_t)INT_MAX || mbs_size > (size_t)INT_MAX)
3632 return (-1);
3633 r = WideCharToMultiByte(sc->to_cp, 0,
3634 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3635 NULL, &defchar);
3636 /* Exit loop if we succeeded */
3637 if (r != 0 ||
3638 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3639 ll = (size_t)r;
3640 break;
3641 }
3642 /* Else expand buffer and loop to try again. */
3643 r = WideCharToMultiByte(sc->to_cp, 0,
3644 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3645 ll = (size_t)r;
3646 if (archive_string_ensure(as, ll +1) == NULL)
3647 return (-1);
3648 mbs = as->s + as->length;
3649 mbs_size = as->buffer_length - as->length -1;
3650 } while (1);
3651 archive_string_free(&tmp);
3652 as->length += ll;
3653 as->s[as->length] = '\0';
3654 if (ll == 0 || defchar)
3655 ret = -1;
3656 return (ret);
3657 }
3658
3659 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3660 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3661 size_t bytes, struct archive_string_conv *sc)
3662 {
3663 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3664 }
3665
3666 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3667 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3668 size_t bytes, struct archive_string_conv *sc)
3669 {
3670 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3671 }
3672
3673 /*
3674 * Convert a current locale string to UTF-16BE/LE and copy the result.
3675 * Return -1 if conversion fails.
3676 */
3677 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3678 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3679 size_t length, struct archive_string_conv *sc, int bigendian)
3680 {
3681 const char *s = (const char *)_p;
3682 char *u16;
3683 size_t count, avail;
3684
3685 if (archive_string_ensure(as16,
3686 as16->length + (length + 1) * 2) == NULL)
3687 return (-1);
3688
3689 u16 = as16->s + as16->length;
3690 avail = as16->buffer_length - 2;
3691 if (sc->from_cp == CP_C_LOCALE) {
3692 /*
3693 * "C" locale special process.
3694 */
3695 count = 0;
3696 while (count < length && *s) {
3697 if (bigendian)
3698 archive_be16enc(u16, *s);
3699 else
3700 archive_le16enc(u16, *s);
3701 u16 += 2;
3702 s++;
3703 count++;
3704 }
3705 as16->length += count << 1;
3706 as16->s[as16->length] = 0;
3707 as16->s[as16->length+1] = 0;
3708 return (0);
3709 }
3710 do {
3711 int r;
3712 if (length > (size_t)INT_MAX || (avail >> 1) > (size_t)INT_MAX)
3713 return (-1);
3714 r = MultiByteToWideChar(sc->from_cp,
3715 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3716 /* Exit loop if we succeeded */
3717 if (r != 0 ||
3718 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3719 count = (size_t)r;
3720 break;
3721 }
3722 /* Expand buffer and try again */
3723 r = MultiByteToWideChar(sc->from_cp,
3724 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3725 count = (size_t)r;
3726 if (archive_string_ensure(as16, (count +1) * 2)
3727 == NULL)
3728 return (-1);
3729 u16 = as16->s + as16->length;
3730 avail = as16->buffer_length - 2;
3731 } while (1);
3732 as16->length += count * 2;
3733 as16->s[as16->length] = 0;
3734 as16->s[as16->length+1] = 0;
3735 if (count == 0)
3736 return (-1);
3737
3738 if (IS_BIG_ENDIAN) {
3739 if (!bigendian) {
3740 while (count > 0) {
3741 uint16_t v = archive_be16dec(u16);
3742 archive_le16enc(u16, v);
3743 u16 += 2;
3744 count--;
3745 }
3746 }
3747 } else {
3748 if (bigendian) {
3749 while (count > 0) {
3750 uint16_t v = archive_le16dec(u16);
3751 archive_be16enc(u16, v);
3752 u16 += 2;
3753 count--;
3754 }
3755 }
3756 }
3757 return (0);
3758 }
3759
3760 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3761 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3762 size_t length, struct archive_string_conv *sc)
3763 {
3764 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3765 }
3766
3767 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3768 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3769 size_t length, struct archive_string_conv *sc)
3770 {
3771 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3772 }
3773
3774 #endif /* _WIN32 && !__CYGWIN__ */
3775
3776 /*
3777 * Do the best effort for conversions.
3778 * We cannot handle UTF-16BE character-set without such iconv,
3779 * but there is a chance if a string consists just ASCII code or
3780 * a current locale is UTF-8.
3781 */
3782
3783 /*
3784 * Convert a UTF-16BE string to current locale and copy the result.
3785 * Return -1 if conversion fails.
3786 */
3787 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3788 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3789 size_t bytes, struct archive_string_conv *sc, int be)
3790 {
3791 const char *utf16 = (const char *)_p;
3792 char *mbs;
3793 uint32_t uc;
3794 int n, ret;
3795
3796 (void)sc; /* UNUSED */
3797 /*
3798 * Other case, we should do the best effort.
3799 * If all character are ASCII(<0x7f), we can convert it.
3800 * if not , we set a alternative character and return -1.
3801 */
3802 ret = 0;
3803 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3804 return (-1);
3805 mbs = as->s + as->length;
3806
3807 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3808 if (n < 0) {
3809 n *= -1;
3810 ret = -1;
3811 }
3812 bytes -= n;
3813 utf16 += n;
3814
3815 if (uc > 127) {
3816 /* We cannot handle it. */
3817 *mbs++ = '?';
3818 ret = -1;
3819 } else
3820 *mbs++ = (char)uc;
3821 }
3822 as->length = mbs - as->s;
3823 as->s[as->length] = '\0';
3824 return (ret);
3825 }
3826
3827 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3828 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3829 size_t bytes, struct archive_string_conv *sc)
3830 {
3831 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3832 }
3833
3834 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3835 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3836 size_t bytes, struct archive_string_conv *sc)
3837 {
3838 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3839 }
3840
3841 /*
3842 * Convert a current locale string to UTF-16BE/LE and copy the result.
3843 * Return -1 if conversion fails.
3844 */
3845 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3846 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3847 size_t length, struct archive_string_conv *sc, int bigendian)
3848 {
3849 const char *s = (const char *)_p;
3850 char *utf16;
3851 size_t remaining;
3852 int ret;
3853
3854 (void)sc; /* UNUSED */
3855 /*
3856 * Other case, we should do the best effort.
3857 * If all character are ASCII(<0x7f), we can convert it.
3858 * if not , we set a alternative character and return -1.
3859 */
3860 ret = 0;
3861 remaining = length;
3862
3863 if (archive_string_ensure(as16,
3864 as16->length + (length + 1) * 2) == NULL)
3865 return (-1);
3866
3867 utf16 = as16->s + as16->length;
3868 while (remaining--) {
3869 unsigned c = *s++;
3870 if (c > 127) {
3871 /* We cannot handle it. */
3872 c = UNICODE_R_CHAR;
3873 ret = -1;
3874 }
3875 if (bigendian)
3876 archive_be16enc(utf16, (uint16_t)c);
3877 else
3878 archive_le16enc(utf16, (uint16_t)c);
3879 utf16 += 2;
3880 }
3881 as16->length = utf16 - as16->s;
3882 as16->s[as16->length] = 0;
3883 as16->s[as16->length+1] = 0;
3884 return (ret);
3885 }
3886
3887 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3888 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3889 size_t length, struct archive_string_conv *sc)
3890 {
3891 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3892 }
3893
3894 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3895 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3896 size_t length, struct archive_string_conv *sc)
3897 {
3898 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3899 }
3900
3901
3902 /*
3903 * Multistring operations.
3904 */
3905
3906 void
archive_mstring_clean(struct archive_mstring * aes)3907 archive_mstring_clean(struct archive_mstring *aes)
3908 {
3909 archive_wstring_free(&(aes->aes_wcs));
3910 archive_string_free(&(aes->aes_mbs));
3911 archive_string_free(&(aes->aes_utf8));
3912 archive_string_free(&(aes->aes_mbs_in_locale));
3913 aes->aes_set = 0;
3914 }
3915
3916 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3917 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3918 {
3919 dest->aes_set = src->aes_set;
3920 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3921 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3922 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3923 }
3924
3925 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3926 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3927 const char **p)
3928 {
3929 struct archive_string_conv *sc;
3930 int r;
3931
3932 /* If we already have a UTF8 form, return that immediately. */
3933 if (aes->aes_set & AES_SET_UTF8) {
3934 *p = aes->aes_utf8.s;
3935 return (0);
3936 }
3937
3938 *p = NULL;
3939 #if defined(_WIN32) && !defined(__CYGWIN__)
3940 /*
3941 * On Windows, first try converting from WCS because (1) there's no
3942 * guarantee that the conversion to MBS will succeed, e.g. when using
3943 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3944 * convert back to WCS again before finally converting to UTF-8
3945 */
3946 if ((aes->aes_set & AES_SET_WCS) != 0) {
3947 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3948 if (sc == NULL)
3949 return (-1);/* Couldn't allocate memory for sc. */
3950 archive_string_empty(&(aes->aes_utf8));
3951 r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3952 aes->aes_wcs.s, aes->aes_wcs.length, sc);
3953 if (a == NULL)
3954 free_sconv_object(sc);
3955 if (r == 0) {
3956 aes->aes_set |= AES_SET_UTF8;
3957 *p = aes->aes_utf8.s;
3958 return (0);/* success. */
3959 } else
3960 return (-1);/* failure. */
3961 }
3962 #endif
3963 /* Try converting WCS to MBS first if MBS does not exist yet. */
3964 if ((aes->aes_set & AES_SET_MBS) == 0) {
3965 const char *pm; /* unused */
3966 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3967 }
3968 if (aes->aes_set & AES_SET_MBS) {
3969 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3970 if (sc == NULL)
3971 return (-1);/* Couldn't allocate memory for sc. */
3972 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3973 aes->aes_mbs.length, sc);
3974 if (a == NULL)
3975 free_sconv_object(sc);
3976 if (r == 0) {
3977 aes->aes_set |= AES_SET_UTF8;
3978 *p = aes->aes_utf8.s;
3979 return (0);/* success. */
3980 } else
3981 return (-1);/* failure. */
3982 }
3983 return (0);/* success. */
3984 }
3985
3986 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3987 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3988 const char **p)
3989 {
3990 struct archive_string_conv *sc;
3991 int r, ret = 0;
3992
3993 /* If we already have an MBS form, return that immediately. */
3994 if (aes->aes_set & AES_SET_MBS) {
3995 *p = aes->aes_mbs.s;
3996 return (ret);
3997 }
3998
3999 *p = NULL;
4000 /* If there's a WCS form, try converting with the native locale. */
4001 if (aes->aes_set & AES_SET_WCS) {
4002 archive_string_empty(&(aes->aes_mbs));
4003 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4004 aes->aes_wcs.s, aes->aes_wcs.length);
4005 *p = aes->aes_mbs.s;
4006 if (r == 0) {
4007 aes->aes_set |= AES_SET_MBS;
4008 return (ret);
4009 } else
4010 ret = -1;
4011 }
4012
4013 /* If there's a UTF-8 form, try converting with the native locale. */
4014 if (aes->aes_set & AES_SET_UTF8) {
4015 archive_string_empty(&(aes->aes_mbs));
4016 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4017 if (sc == NULL)
4018 return (-1);/* Couldn't allocate memory for sc. */
4019 r = archive_strncpy_l(&(aes->aes_mbs),
4020 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4021 if (a == NULL)
4022 free_sconv_object(sc);
4023 *p = aes->aes_mbs.s;
4024 if (r == 0) {
4025 aes->aes_set |= AES_SET_MBS;
4026 ret = 0;/* success; overwrite previous error. */
4027 } else
4028 ret = -1;/* failure. */
4029 }
4030 return (ret);
4031 }
4032
4033 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)4034 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
4035 const wchar_t **wp)
4036 {
4037 int r, ret = 0;
4038
4039 (void)a;/* UNUSED */
4040 /* Return WCS form if we already have it. */
4041 if (aes->aes_set & AES_SET_WCS) {
4042 *wp = aes->aes_wcs.s;
4043 return (ret);
4044 }
4045
4046 *wp = NULL;
4047 #if defined(_WIN32) && !defined(__CYGWIN__)
4048 /*
4049 * On Windows, prefer converting from UTF-8 directly to WCS because:
4050 * (1) there's no guarantee that the string can be represented in MBS (e.g.
4051 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
4052 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
4053 * away that intermediate result
4054 */
4055 if (aes->aes_set & AES_SET_UTF8) {
4056 struct archive_string_conv *sc;
4057
4058 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4059 if (sc != NULL) {
4060 archive_wstring_empty((&aes->aes_wcs));
4061 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4062 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4063 if (a == NULL)
4064 free_sconv_object(sc);
4065 if (r == 0) {
4066 aes->aes_set |= AES_SET_WCS;
4067 *wp = aes->aes_wcs.s;
4068 return (0);
4069 }
4070 }
4071 }
4072 #endif
4073 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
4074 if ((aes->aes_set & AES_SET_MBS) == 0) {
4075 const char *p; /* unused */
4076 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4077 }
4078 /* Try converting MBS to WCS using native locale. */
4079 if (aes->aes_set & AES_SET_MBS) {
4080 archive_wstring_empty(&(aes->aes_wcs));
4081 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4082 aes->aes_mbs.s, aes->aes_mbs.length);
4083 if (r == 0) {
4084 aes->aes_set |= AES_SET_WCS;
4085 *wp = aes->aes_wcs.s;
4086 } else
4087 ret = -1;/* failure. */
4088 }
4089 return (ret);
4090 }
4091
4092 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)4093 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4094 const char **p, size_t *length, struct archive_string_conv *sc)
4095 {
4096 int ret = 0;
4097 #if defined(_WIN32) && !defined(__CYGWIN__)
4098 int r;
4099
4100 /*
4101 * Internationalization programming on Windows must use Wide
4102 * characters because Windows platform cannot make locale UTF-8.
4103 */
4104 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4105 archive_string_empty(&(aes->aes_mbs_in_locale));
4106 r = archive_string_append_from_wcs_in_codepage(
4107 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4108 aes->aes_wcs.length, sc);
4109 if (r == 0) {
4110 *p = aes->aes_mbs_in_locale.s;
4111 if (length != NULL)
4112 *length = aes->aes_mbs_in_locale.length;
4113 return (0);
4114 } else if (errno == ENOMEM)
4115 return (-1);
4116 else
4117 ret = -1;
4118 }
4119 #endif
4120
4121 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4122 * with the native locale to be used for translating it to specified
4123 * character-set. */
4124 if ((aes->aes_set & AES_SET_MBS) == 0) {
4125 const char *pm; /* unused */
4126 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
4127 }
4128 /* If we already have an MBS form, use it to be translated to
4129 * specified character-set. */
4130 if (aes->aes_set & AES_SET_MBS) {
4131 if (sc == NULL) {
4132 /* Conversion is unneeded. */
4133 *p = aes->aes_mbs.s;
4134 if (length != NULL)
4135 *length = aes->aes_mbs.length;
4136 return (0);
4137 }
4138 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4139 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4140 *p = aes->aes_mbs_in_locale.s;
4141 if (length != NULL)
4142 *length = aes->aes_mbs_in_locale.length;
4143 } else {
4144 *p = NULL;
4145 if (length != NULL)
4146 *length = 0;
4147 }
4148 return (ret);
4149 }
4150
4151 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4152 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4153 {
4154 if (mbs == NULL) {
4155 aes->aes_set = 0;
4156 return (0);
4157 }
4158 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4159 }
4160
4161 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4162 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4163 size_t len)
4164 {
4165 if (mbs == NULL) {
4166 aes->aes_set = 0;
4167 return (0);
4168 }
4169 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4170 archive_strncpy(&(aes->aes_mbs), mbs, len);
4171 archive_string_empty(&(aes->aes_utf8));
4172 archive_wstring_empty(&(aes->aes_wcs));
4173 return (0);
4174 }
4175
4176 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4177 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4178 {
4179 return archive_mstring_copy_wcs_len(aes, wcs,
4180 wcs == NULL ? 0 : wcslen(wcs));
4181 }
4182
4183 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4184 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4185 {
4186 if (utf8 == NULL) {
4187 aes->aes_set = 0;
4188 return (0);
4189 }
4190 aes->aes_set = AES_SET_UTF8;
4191 archive_string_empty(&(aes->aes_mbs));
4192 archive_string_empty(&(aes->aes_wcs));
4193 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4194 return (int)strlen(utf8);
4195 }
4196
4197 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4198 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4199 size_t len)
4200 {
4201 if (wcs == NULL) {
4202 aes->aes_set = 0;
4203 return (0);
4204 }
4205 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4206 archive_string_empty(&(aes->aes_mbs));
4207 archive_string_empty(&(aes->aes_utf8));
4208 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4209 return (0);
4210 }
4211
4212 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4213 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4214 const char *mbs, size_t len, struct archive_string_conv *sc)
4215 {
4216 int r;
4217
4218 if (mbs == NULL) {
4219 aes->aes_set = 0;
4220 return (0);
4221 }
4222 archive_string_empty(&(aes->aes_mbs));
4223 archive_wstring_empty(&(aes->aes_wcs));
4224 archive_string_empty(&(aes->aes_utf8));
4225 #if defined(_WIN32) && !defined(__CYGWIN__)
4226 /*
4227 * Internationalization programming on Windows must use Wide
4228 * characters because Windows platform cannot make locale UTF-8.
4229 */
4230 if (sc == NULL) {
4231 if (archive_string_append(&(aes->aes_mbs),
4232 mbs, mbsnbytes(mbs, len)) == NULL) {
4233 aes->aes_set = 0;
4234 r = -1;
4235 } else {
4236 aes->aes_set = AES_SET_MBS;
4237 r = 0;
4238 }
4239 #if defined(HAVE_ICONV)
4240 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4241 /*
4242 * This case happens only when MultiByteToWideChar() cannot
4243 * handle sc->from_cp, and we have to iconv in order to
4244 * translate character-set to wchar_t,UTF-16.
4245 */
4246 iconv_t cd = sc->cd;
4247 unsigned from_cp;
4248 int flag;
4249
4250 /*
4251 * Translate multi-bytes from some character-set to UTF-8.
4252 */
4253 sc->cd = sc->cd_w;
4254 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4255 sc->cd = cd;
4256 if (r != 0) {
4257 aes->aes_set = 0;
4258 return (r);
4259 }
4260 aes->aes_set = AES_SET_UTF8;
4261
4262 /*
4263 * Append the UTF-8 string into wstring.
4264 */
4265 flag = sc->flag;
4266 sc->flag &= ~(SCONV_NORMALIZATION_C
4267 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4268 from_cp = sc->from_cp;
4269 sc->from_cp = CP_UTF8;
4270 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4271 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4272 sc->flag = flag;
4273 sc->from_cp = from_cp;
4274 if (r == 0)
4275 aes->aes_set |= AES_SET_WCS;
4276 #endif
4277 } else {
4278 r = archive_wstring_append_from_mbs_in_codepage(
4279 &(aes->aes_wcs), mbs, len, sc);
4280 if (r == 0)
4281 aes->aes_set = AES_SET_WCS;
4282 else
4283 aes->aes_set = 0;
4284 }
4285 #else
4286 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4287 if (r == 0)
4288 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4289 else
4290 aes->aes_set = 0;
4291 #endif
4292 return (r);
4293 }
4294
4295 /*
4296 * The 'update' form tries to proactively update all forms of
4297 * this string (WCS and MBS) and returns an error if any of
4298 * them fail. This is used by the 'pax' handler, for instance,
4299 * to detect and report character-conversion failures early while
4300 * still allowing clients to get potentially useful values from
4301 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4302 * strive to give the user something useful, so you can get hopefully
4303 * usable values even if some of the character conversions are failing.)
4304 */
4305 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4306 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4307 const char *utf8)
4308 {
4309 struct archive_string_conv *sc;
4310 int r;
4311
4312 if (utf8 == NULL) {
4313 aes->aes_set = 0;
4314 return (0); /* Succeeded in clearing everything. */
4315 }
4316
4317 /* Save the UTF8 string. */
4318 archive_strcpy(&(aes->aes_utf8), utf8);
4319
4320 /* Empty the mbs and wcs strings. */
4321 archive_string_empty(&(aes->aes_mbs));
4322 archive_wstring_empty(&(aes->aes_wcs));
4323
4324 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4325
4326 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4327 if (sc == NULL)
4328 return (-1);/* Couldn't allocate memory for sc. */
4329
4330 #if defined(_WIN32) && !defined(__CYGWIN__)
4331 /* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4332 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4333 * intermediate result, and (2) WCS will still be set even if we fail to
4334 * convert to MBS (e.g. with ACP that can't represent the characters) */
4335 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4336 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4337
4338 if (a == NULL)
4339 free_sconv_object(sc);
4340 if (r != 0)
4341 return (-1); /* This will guarantee we can't convert to MBS */
4342 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4343
4344 /* Try converting WCS to MBS, return false on failure. */
4345 if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4346 aes->aes_wcs.length))
4347 return (-1);
4348 #else
4349 /* Try converting UTF-8 to MBS, return false on failure. */
4350 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4351
4352 if (a == NULL)
4353 free_sconv_object(sc);
4354 if (r != 0)
4355 return (-1);
4356 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4357
4358 /* Try converting MBS to WCS, return false on failure. */
4359 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4360 aes->aes_mbs.length))
4361 return (-1);
4362 #endif
4363
4364 /* All conversions succeeded. */
4365 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4366
4367 return (0);
4368 }
4369