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)
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 == NULL) {
1877 return "current locale";
1878 }
1879 if (sc->flag & SCONV_TO_CHARSET)
1880 return (sc->to_charset);
1881 else
1882 return (sc->from_charset);
1883 }
1884
1885 /*
1886 * Change the behavior of a string conversion.
1887 */
1888 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1889 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1890 {
1891 switch (opt) {
1892 /*
1893 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1894 * assumption that wchar_t was Unicode.
1895 * This option enables simulating the assumption in order to read
1896 * that filename correctly.
1897 */
1898 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1899 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1900 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1901 /*
1902 * Nothing to do for it since wchar_t on these platforms
1903 * is really Unicode.
1904 */
1905 (void)sc; /* UNUSED */
1906 #else
1907 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1908 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1909 /* Set up string converters. */
1910 setup_converter(sc);
1911 }
1912 #endif
1913 break;
1914 case SCONV_SET_OPT_NORMALIZATION_C:
1915 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1916 sc->flag |= SCONV_NORMALIZATION_C;
1917 sc->flag &= ~SCONV_NORMALIZATION_D;
1918 /* Set up string converters. */
1919 setup_converter(sc);
1920 }
1921 break;
1922 case SCONV_SET_OPT_NORMALIZATION_D:
1923 #if defined(HAVE_ICONV)
1924 /*
1925 * If iconv will take the string, do not change the
1926 * setting of the normalization.
1927 */
1928 if (!(sc->flag & SCONV_WIN_CP) &&
1929 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1930 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1931 break;
1932 #endif
1933 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1934 sc->flag |= SCONV_NORMALIZATION_D;
1935 sc->flag &= ~SCONV_NORMALIZATION_C;
1936 /* Set up string converters. */
1937 setup_converter(sc);
1938 }
1939 break;
1940 default:
1941 break;
1942 }
1943 }
1944
1945 /*
1946 *
1947 * Copy one archive_string to another in locale conversion.
1948 *
1949 * archive_strncat_l();
1950 * archive_strncpy_l();
1951 *
1952 */
1953
1954 static size_t
mbsnbytes(const void * _p,size_t n)1955 mbsnbytes(const void *_p, size_t n)
1956 {
1957 size_t s;
1958 const char *p, *pp;
1959
1960 if (_p == NULL)
1961 return (0);
1962 p = (const char *)_p;
1963
1964 /* Like strlen(p), except won't examine positions beyond p[n]. */
1965 s = 0;
1966 pp = p;
1967 while (s < n && *pp) {
1968 pp++;
1969 s++;
1970 }
1971 return (s);
1972 }
1973
1974 static size_t
utf16nbytes(const void * _p,size_t n)1975 utf16nbytes(const void *_p, size_t n)
1976 {
1977 size_t s;
1978 const char *p, *pp;
1979
1980 if (_p == NULL)
1981 return (0);
1982 p = (const char *)_p;
1983
1984 /* Like strlen(p), except won't examine positions beyond p[n]. */
1985 s = 0;
1986 pp = p;
1987 n >>= 1;
1988 while (s < n && (pp[0] || pp[1])) {
1989 pp += 2;
1990 s++;
1991 }
1992 return (s<<1);
1993 }
1994
1995 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1996 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1997 struct archive_string_conv *sc)
1998 {
1999 as->length = 0;
2000 return (archive_strncat_l(as, _p, n, sc));
2001 }
2002
2003 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2004 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
2005 struct archive_string_conv *sc)
2006 {
2007 const void *s;
2008 size_t length = 0;
2009 int i, r = 0, r2;
2010
2011 if (_p != NULL && n > 0) {
2012 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
2013 length = utf16nbytes(_p, n);
2014 else
2015 length = mbsnbytes(_p, n);
2016 }
2017
2018 /* We must allocate memory even if there is no data for conversion
2019 * or copy. This simulates archive_string_append behavior. */
2020 if (length == 0) {
2021 size_t tn = 1;
2022 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
2023 tn = 2;
2024 if (archive_string_ensure(as, as->length + tn) == NULL)
2025 return (-1);
2026 as->s[as->length] = 0;
2027 if (tn == 2)
2028 as->s[as->length+1] = 0;
2029 return (0);
2030 }
2031
2032 /*
2033 * If sc is NULL, we just make a copy.
2034 */
2035 if (sc == NULL) {
2036 if (archive_string_append(as, _p, length) == NULL)
2037 return (-1);/* No memory */
2038 return (0);
2039 }
2040
2041 s = _p;
2042 i = 0;
2043 if (sc->nconverter > 1) {
2044 sc->utftmp.length = 0;
2045 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2046 if (r2 != 0 && errno == ENOMEM)
2047 return (r2);
2048 if (r > r2)
2049 r = r2;
2050 s = sc->utftmp.s;
2051 length = sc->utftmp.length;
2052 ++i;
2053 }
2054 r2 = sc->converter[i](as, s, length, sc);
2055 if (r > r2)
2056 r = r2;
2057 return (r);
2058 }
2059
2060 struct archive_string *
archive_string_dirname(struct archive_string * as)2061 archive_string_dirname(struct archive_string *as)
2062 {
2063 /* strip trailing separators */
2064 while (as->length > 1 && as->s[as->length - 1] == '/')
2065 as->length--;
2066 /* strip final component */
2067 while (as->length > 0 && as->s[as->length - 1] != '/')
2068 as->length--;
2069 /* empty path -> cwd */
2070 if (as->length == 0)
2071 return (archive_strcat(as, "."));
2072 /* strip separator(s) */
2073 while (as->length > 1 && as->s[as->length - 1] == '/')
2074 as->length--;
2075 /* terminate */
2076 as->s[as->length] = '\0';
2077 return (as);
2078 }
2079
2080 #if HAVE_ICONV
2081
2082 /*
2083 * Return -1 if conversion fails.
2084 */
2085 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2086 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2087 size_t length, struct archive_string_conv *sc)
2088 {
2089 ICONV_CONST char *itp;
2090 size_t remaining;
2091 iconv_t cd;
2092 char *outp;
2093 size_t avail, bs;
2094 int return_value = 0; /* success */
2095 size_t to_size, from_size;
2096
2097 if (sc->flag & SCONV_TO_UTF16)
2098 to_size = 2;
2099 else
2100 to_size = 1;
2101 if (sc->flag & SCONV_FROM_UTF16)
2102 from_size = 2;
2103 else
2104 from_size = 1;
2105
2106 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2107 return (-1);
2108
2109 cd = sc->cd;
2110 itp = (char *)(uintptr_t)_p;
2111 remaining = length;
2112 outp = as->s + as->length;
2113 avail = as->buffer_length - as->length - to_size;
2114 while (remaining >= from_size) {
2115 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2116
2117 if (result != (size_t)-1)
2118 break; /* Conversion completed. */
2119
2120 if (errno == EILSEQ || errno == EINVAL) {
2121 /*
2122 * If an output charset is UTF-8 or UTF-16BE/LE,
2123 * unknown character should be U+FFFD
2124 * (replacement character).
2125 */
2126 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2127 size_t rbytes;
2128 if (sc->flag & SCONV_TO_UTF8)
2129 rbytes = sizeof(utf8_replacement_char);
2130 else
2131 rbytes = 2;
2132
2133 if (avail < rbytes) {
2134 as->length = outp - as->s;
2135 bs = as->buffer_length +
2136 (remaining * to_size) + rbytes;
2137 if (NULL ==
2138 archive_string_ensure(as, bs))
2139 return (-1);
2140 outp = as->s + as->length;
2141 avail = as->buffer_length
2142 - as->length - to_size;
2143 }
2144 if (sc->flag & SCONV_TO_UTF8)
2145 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2146 else if (sc->flag & SCONV_TO_UTF16BE)
2147 archive_be16enc(outp, UNICODE_R_CHAR);
2148 else
2149 archive_le16enc(outp, UNICODE_R_CHAR);
2150 outp += rbytes;
2151 avail -= rbytes;
2152 } else {
2153 /* Skip the illegal input bytes. */
2154 *outp++ = '?';
2155 avail--;
2156 }
2157 itp += from_size;
2158 remaining -= from_size;
2159 return_value = -1; /* failure */
2160 } else {
2161 /* E2BIG no output buffer,
2162 * Increase an output buffer. */
2163 as->length = outp - as->s;
2164 bs = as->buffer_length + remaining * 2;
2165 if (NULL == archive_string_ensure(as, bs))
2166 return (-1);
2167 outp = as->s + as->length;
2168 avail = as->buffer_length - as->length - to_size;
2169 }
2170 }
2171 as->length = outp - as->s;
2172 as->s[as->length] = 0;
2173 if (to_size == 2)
2174 as->s[as->length+1] = 0;
2175 return (return_value);
2176 }
2177
2178 #endif /* HAVE_ICONV */
2179
2180
2181 #if defined(_WIN32) && !defined(__CYGWIN__)
2182
2183 /*
2184 * Translate a string from a some CodePage to an another CodePage by
2185 * Windows APIs, and copy the result. Return -1 if conversion fails.
2186 */
2187 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2188 strncat_in_codepage(struct archive_string *as,
2189 const void *_p, size_t length, struct archive_string_conv *sc)
2190 {
2191 const char *s = (const char *)_p;
2192 struct archive_wstring aws;
2193 size_t l;
2194 int r, saved_flag;
2195
2196 archive_string_init(&aws);
2197 saved_flag = sc->flag;
2198 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2199 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2200 sc->flag = saved_flag;
2201 if (r != 0) {
2202 archive_wstring_free(&aws);
2203 if (errno != ENOMEM)
2204 archive_string_append(as, s, length);
2205 return (-1);
2206 }
2207
2208 l = as->length;
2209 r = archive_string_append_from_wcs_in_codepage(
2210 as, aws.s, aws.length, sc);
2211 if (r != 0 && errno != ENOMEM && l == as->length)
2212 archive_string_append(as, s, length);
2213 archive_wstring_free(&aws);
2214 return (r);
2215 }
2216
2217 /*
2218 * Test whether MBS ==> WCS is okay.
2219 */
2220 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2221 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2222 {
2223 const char *p = (const char *)_p;
2224 unsigned codepage;
2225 DWORD mbflag = MB_ERR_INVALID_CHARS;
2226
2227 if (sc->flag & SCONV_FROM_CHARSET)
2228 codepage = sc->to_cp;
2229 else
2230 codepage = sc->from_cp;
2231
2232 if (codepage == CP_C_LOCALE)
2233 return (0);
2234 if (codepage != CP_UTF8)
2235 mbflag |= MB_PRECOMPOSED;
2236
2237 if (n > (size_t)INT_MAX)
2238 return (-1); /* Invalid */
2239 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2240 return (-1); /* Invalid */
2241 return (0); /* Okay */
2242 }
2243
2244 #else
2245
2246 /*
2247 * Test whether MBS ==> WCS is okay.
2248 */
2249 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2250 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2251 {
2252 const char *p = (const char *)_p;
2253 size_t r;
2254
2255 #if HAVE_MBRTOWC
2256 mbstate_t shift_state;
2257
2258 memset(&shift_state, 0, sizeof(shift_state));
2259 #else
2260 /* Clear the shift state before starting. */
2261 mbtowc(NULL, NULL, 0);
2262 #endif
2263 while (n) {
2264 wchar_t wc;
2265
2266 #if HAVE_MBRTOWC
2267 r = mbrtowc(&wc, p, n, &shift_state);
2268 #else
2269 r = mbtowc(&wc, p, n);
2270 #endif
2271 if (r == (size_t)-1 || r == (size_t)-2)
2272 return (-1);/* Invalid. */
2273 if (r == 0)
2274 break;
2275 p += r;
2276 n -= r;
2277 }
2278 (void)sc; /* UNUSED */
2279 return (0); /* All Okey. */
2280 }
2281
2282 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2283
2284 /*
2285 * Basically returns -1 because we cannot make a conversion of charset
2286 * without iconv but in some cases this would return 0.
2287 * Returns 0 if all copied characters are ASCII.
2288 * Returns 0 if both from-locale and to-locale are the same and those
2289 * can be WCS with no error.
2290 */
2291 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2292 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2293 size_t length, struct archive_string_conv *sc)
2294 {
2295 size_t remaining;
2296 const uint8_t *itp;
2297 int return_value = 0; /* success */
2298
2299 /*
2300 * If both from-locale and to-locale is the same, this makes a copy.
2301 * And then this checks all copied MBS can be WCS if so returns 0.
2302 */
2303 if (sc->same) {
2304 if (archive_string_append(as, _p, length) == NULL)
2305 return (-1);/* No memory */
2306 return (invalid_mbs(_p, length, sc));
2307 }
2308
2309 /*
2310 * If a character is ASCII, this just copies it. If not, this
2311 * assigns '?' character instead but in UTF-8 locale this assigns
2312 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2313 * a Replacement Character in Unicode.
2314 */
2315
2316 remaining = length;
2317 itp = (const uint8_t *)_p;
2318 while (*itp && remaining > 0) {
2319 if (*itp > 127) {
2320 // Non-ASCII: Substitute with suitable replacement
2321 if (sc->flag & SCONV_TO_UTF8) {
2322 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2323 __archive_errx(1, "Out of memory");
2324 }
2325 } else {
2326 archive_strappend_char(as, '?');
2327 }
2328 return_value = -1;
2329 } else {
2330 archive_strappend_char(as, *itp);
2331 }
2332 ++itp;
2333 }
2334 return (return_value);
2335 }
2336
2337
2338 /*
2339 * Unicode conversion functions.
2340 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2341 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2342 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2343 * - UTF-16BE <===> UTF-8.
2344 *
2345 */
2346
2347 /*
2348 * Utility to convert a single UTF-8 sequence.
2349 *
2350 * Usually return used bytes, return used byte in negative value when
2351 * a unicode character is replaced with U+FFFD.
2352 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2353 * Recommended Practice for Replacement Characters.
2354 */
2355 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2356 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2357 {
2358 static const char utf8_count[256] = {
2359 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2360 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2361 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2362 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2363 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2364 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2365 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2366 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2367 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2368 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2369 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2370 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2371 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2372 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2373 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2374 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2375 };
2376 int ch, i;
2377 int cnt;
2378 uint32_t wc;
2379
2380 /* Sanity check. */
2381 if (n == 0)
2382 return (0);
2383 /*
2384 * Decode 1-4 bytes depending on the value of the first byte.
2385 */
2386 ch = (unsigned char)*s;
2387 if (ch == 0)
2388 return (0); /* Standard: return 0 for end-of-string. */
2389 cnt = utf8_count[ch];
2390
2391 /* Invalid sequence or there are not plenty bytes. */
2392 if (n < (size_t)cnt) {
2393 cnt = (int)n;
2394 for (i = 1; i < cnt; i++) {
2395 if ((s[i] & 0xc0) != 0x80) {
2396 cnt = i;
2397 break;
2398 }
2399 }
2400 goto invalid_sequence;
2401 }
2402
2403 /* Make a Unicode code point from a single UTF-8 sequence. */
2404 switch (cnt) {
2405 case 1: /* 1 byte sequence. */
2406 *pwc = ch & 0x7f;
2407 return (cnt);
2408 case 2: /* 2 bytes sequence. */
2409 if ((s[1] & 0xc0) != 0x80) {
2410 cnt = 1;
2411 goto invalid_sequence;
2412 }
2413 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2414 return (cnt);
2415 case 3: /* 3 bytes sequence. */
2416 if ((s[1] & 0xc0) != 0x80) {
2417 cnt = 1;
2418 goto invalid_sequence;
2419 }
2420 if ((s[2] & 0xc0) != 0x80) {
2421 cnt = 2;
2422 goto invalid_sequence;
2423 }
2424 wc = ((ch & 0x0f) << 12)
2425 | ((s[1] & 0x3f) << 6)
2426 | (s[2] & 0x3f);
2427 if (wc < 0x800)
2428 goto invalid_sequence;/* Overlong sequence. */
2429 break;
2430 case 4: /* 4 bytes sequence. */
2431 if ((s[1] & 0xc0) != 0x80) {
2432 cnt = 1;
2433 goto invalid_sequence;
2434 }
2435 if ((s[2] & 0xc0) != 0x80) {
2436 cnt = 2;
2437 goto invalid_sequence;
2438 }
2439 if ((s[3] & 0xc0) != 0x80) {
2440 cnt = 3;
2441 goto invalid_sequence;
2442 }
2443 wc = ((ch & 0x07) << 18)
2444 | ((s[1] & 0x3f) << 12)
2445 | ((s[2] & 0x3f) << 6)
2446 | (s[3] & 0x3f);
2447 if (wc < 0x10000)
2448 goto invalid_sequence;/* Overlong sequence. */
2449 break;
2450 default: /* Others are all invalid sequence. */
2451 if (ch == 0xc0 || ch == 0xc1)
2452 cnt = 2;
2453 else if (ch >= 0xf5 && ch <= 0xf7)
2454 cnt = 4;
2455 else if (ch >= 0xf8 && ch <= 0xfb)
2456 cnt = 5;
2457 else if (ch == 0xfc || ch == 0xfd)
2458 cnt = 6;
2459 else
2460 cnt = 1;
2461 if (n < (size_t)cnt)
2462 cnt = (int)n;
2463 for (i = 1; i < cnt; i++) {
2464 if ((s[i] & 0xc0) != 0x80) {
2465 cnt = i;
2466 break;
2467 }
2468 }
2469 goto invalid_sequence;
2470 }
2471
2472 /* The code point larger than 0x10FFFF is not legal
2473 * Unicode values. */
2474 if (wc > UNICODE_MAX)
2475 goto invalid_sequence;
2476 /* Correctly gets a Unicode, returns used bytes. */
2477 *pwc = wc;
2478 return (cnt);
2479 invalid_sequence:
2480 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2481 return (cnt * -1);
2482 }
2483
2484 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2485 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2486 {
2487 int cnt;
2488
2489 cnt = _utf8_to_unicode(pwc, s, n);
2490 /* Any of Surrogate pair is not legal Unicode values. */
2491 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2492 return (-3);
2493 return (cnt);
2494 }
2495
2496 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2497 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2498 {
2499 uc -= 0xD800;
2500 uc *= 0x400;
2501 uc += uc2 - 0xDC00;
2502 uc += 0x10000;
2503 return (uc);
2504 }
2505
2506 /*
2507 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2508 * removing surrogate pairs.
2509 *
2510 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2511 *
2512 * Usually return used bytes, return used byte in negative value when
2513 * a unicode character is replaced with U+FFFD.
2514 */
2515 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2516 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2517 {
2518 uint32_t wc = 0;
2519 int cnt;
2520
2521 cnt = _utf8_to_unicode(&wc, s, n);
2522 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2523 uint32_t wc2 = 0;
2524 if (n - 3 < 3) {
2525 /* Invalid byte sequence. */
2526 goto invalid_sequence;
2527 }
2528 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2529 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2530 /* Invalid byte sequence. */
2531 goto invalid_sequence;
2532 }
2533 wc = combine_surrogate_pair(wc, wc2);
2534 cnt = 6;
2535 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2536 /* Invalid byte sequence. */
2537 goto invalid_sequence;
2538 }
2539 *pwc = wc;
2540 return (cnt);
2541 invalid_sequence:
2542 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2543 if (cnt > 0)
2544 cnt *= -1;
2545 return (cnt);
2546 }
2547
2548 /*
2549 * Convert a Unicode code point to a single UTF-8 sequence.
2550 *
2551 * NOTE:This function does not check if the Unicode is legal or not.
2552 * Please you definitely check it before calling this.
2553 */
2554 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2555 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2556 {
2557 char *_p = p;
2558
2559 /* Invalid Unicode char maps to Replacement character */
2560 if (uc > UNICODE_MAX)
2561 uc = UNICODE_R_CHAR;
2562 /* Translate code point to UTF8 */
2563 if (uc <= 0x7f) {
2564 if (remaining == 0)
2565 return (0);
2566 *p++ = (char)uc;
2567 } else if (uc <= 0x7ff) {
2568 if (remaining < 2)
2569 return (0);
2570 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2571 *p++ = 0x80 | (uc & 0x3f);
2572 } else if (uc <= 0xffff) {
2573 if (remaining < 3)
2574 return (0);
2575 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2576 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2577 *p++ = 0x80 | (uc & 0x3f);
2578 } else {
2579 if (remaining < 4)
2580 return (0);
2581 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2582 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2583 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2584 *p++ = 0x80 | (uc & 0x3f);
2585 }
2586 return (p - _p);
2587 }
2588
2589 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2590 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2591 {
2592 return (utf16_to_unicode(pwc, s, n, 1));
2593 }
2594
2595 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2596 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2597 {
2598 return (utf16_to_unicode(pwc, s, n, 0));
2599 }
2600
2601 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2602 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2603 {
2604 const char *utf16 = s;
2605 unsigned uc;
2606
2607 if (n == 0)
2608 return (0);
2609 if (n == 1) {
2610 /* set the Replacement Character instead. */
2611 *pwc = UNICODE_R_CHAR;
2612 return (-1);
2613 }
2614
2615 if (be)
2616 uc = archive_be16dec(utf16);
2617 else
2618 uc = archive_le16dec(utf16);
2619 utf16 += 2;
2620
2621 /* If this is a surrogate pair, assemble the full code point.*/
2622 if (IS_HIGH_SURROGATE_LA(uc)) {
2623 unsigned uc2;
2624
2625 if (n >= 4) {
2626 if (be)
2627 uc2 = archive_be16dec(utf16);
2628 else
2629 uc2 = archive_le16dec(utf16);
2630 } else
2631 uc2 = 0;
2632 if (IS_LOW_SURROGATE_LA(uc2)) {
2633 uc = combine_surrogate_pair(uc, uc2);
2634 utf16 += 2;
2635 } else {
2636 /* Undescribed code point should be U+FFFD
2637 * (replacement character). */
2638 *pwc = UNICODE_R_CHAR;
2639 return (-2);
2640 }
2641 }
2642
2643 /*
2644 * Surrogate pair values(0xd800 through 0xdfff) are only
2645 * used by UTF-16, so, after above calculation, the code
2646 * must not be surrogate values, and Unicode has no codes
2647 * larger than 0x10ffff. Thus, those are not legal Unicode
2648 * values.
2649 */
2650 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2651 /* Undescribed code point should be U+FFFD
2652 * (replacement character). */
2653 *pwc = UNICODE_R_CHAR;
2654 return (((int)(utf16 - s)) * -1);
2655 }
2656 *pwc = uc;
2657 return ((int)(utf16 - s));
2658 }
2659
2660 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2661 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2662 {
2663 char *utf16 = p;
2664
2665 if (uc > 0xffff) {
2666 /* We have a code point that won't fit into a
2667 * wchar_t; convert it to a surrogate pair. */
2668 if (remaining < 4)
2669 return (0);
2670 uc -= 0x10000;
2671 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2672 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2673 return (4);
2674 } else {
2675 if (remaining < 2)
2676 return (0);
2677 archive_be16enc(utf16, (uint16_t)uc);
2678 return (2);
2679 }
2680 }
2681
2682 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2683 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2684 {
2685 char *utf16 = p;
2686
2687 if (uc > 0xffff) {
2688 /* We have a code point that won't fit into a
2689 * wchar_t; convert it to a surrogate pair. */
2690 if (remaining < 4)
2691 return (0);
2692 uc -= 0x10000;
2693 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2694 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2695 return (4);
2696 } else {
2697 if (remaining < 2)
2698 return (0);
2699 archive_le16enc(utf16, (uint16_t)uc);
2700 return (2);
2701 }
2702 }
2703
2704 /*
2705 * Append new UTF-8 string to existing UTF-8 string.
2706 * Existing string is assumed to already be in proper form;
2707 * the new string will have invalid sequences replaced and
2708 * surrogate pairs canonicalized.
2709 */
2710 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _src,size_t len,struct archive_string_conv * sc)2711 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2712 size_t len, struct archive_string_conv *sc)
2713 {
2714 int ret = 0;
2715 const char *src = _src;
2716 (void)sc; /* UNUSED */
2717
2718 /* Pre-extend the destination */
2719 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2720 return (-1);
2721
2722 /* Invariant: src points to the first UTF8 byte that hasn't
2723 * been copied to the destination `as`. */
2724 for (;;) {
2725 int n;
2726 uint32_t uc;
2727 const char *e = src;
2728
2729 /* Skip UTF-8 sequences until we reach end-of-string or
2730 * a code point that needs conversion. */
2731 while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2732 e += n;
2733 len -= n;
2734 }
2735 /* Copy the part that doesn't need conversion */
2736 if (e > src) {
2737 if (archive_string_append(as, src, e - src) == NULL)
2738 return (-1);
2739 src = e;
2740 }
2741
2742 if (n == 0) {
2743 /* We reached end-of-string */
2744 return (ret);
2745 } else {
2746 /* Next code point needs conversion */
2747 char t[4];
2748 size_t w;
2749
2750 /* Try decoding a surrogate pair */
2751 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2752 n = cesu8_to_unicode(&uc, src, len);
2753 }
2754 /* Not a (valid) surrogate, so use a replacement char */
2755 if (n < 0) {
2756 ret = -1; /* Return -1 if we used any replacement */
2757 n *= -1;
2758 }
2759 /* Consume converted code point */
2760 src += n;
2761 len -= n;
2762 /* Convert and append new UTF-8 sequence. */
2763 w = unicode_to_utf8(t, sizeof(t), uc);
2764 if (archive_string_append(as, t, w) == NULL)
2765 return (-1);
2766 }
2767 }
2768 }
2769
2770 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2771 archive_string_append_unicode(struct archive_string *as, const void *_p,
2772 size_t len, struct archive_string_conv *sc)
2773 {
2774 const char *s;
2775 char *p, *endp;
2776 uint32_t uc;
2777 size_t w;
2778 size_t ts, tm;
2779 int n, ret = 0;
2780 int (*parse)(uint32_t *, const char *, size_t);
2781 size_t (*unparse)(char *, size_t, uint32_t);
2782
2783 if (sc->flag & SCONV_TO_UTF16BE) {
2784 unparse = unicode_to_utf16be;
2785 ts = 2;
2786 } else if (sc->flag & SCONV_TO_UTF16LE) {
2787 unparse = unicode_to_utf16le;
2788 ts = 2;
2789 } else if (sc->flag & SCONV_TO_UTF8) {
2790 unparse = unicode_to_utf8;
2791 ts = 1;
2792 } else {
2793 /*
2794 * This case is going to be converted to another
2795 * character-set through iconv.
2796 */
2797 if (sc->flag & SCONV_FROM_UTF16BE) {
2798 unparse = unicode_to_utf16be;
2799 ts = 2;
2800 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2801 unparse = unicode_to_utf16le;
2802 ts = 2;
2803 } else {
2804 unparse = unicode_to_utf8;
2805 ts = 1;
2806 }
2807 }
2808
2809 if (sc->flag & SCONV_FROM_UTF16BE) {
2810 parse = utf16be_to_unicode;
2811 tm = 1;
2812 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2813 parse = utf16le_to_unicode;
2814 tm = 1;
2815 } else {
2816 parse = cesu8_to_unicode;
2817 tm = ts;
2818 }
2819
2820 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2821 return (-1);
2822
2823 s = (const char *)_p;
2824 p = as->s + as->length;
2825 endp = as->s + as->buffer_length - ts;
2826 while ((n = parse(&uc, s, len)) != 0) {
2827 if (n < 0) {
2828 /* Use a replaced unicode character. */
2829 n *= -1;
2830 ret = -1;
2831 }
2832 s += n;
2833 len -= n;
2834 while ((w = unparse(p, endp - p, uc)) == 0) {
2835 /* There is not enough output buffer so
2836 * we have to expand it. */
2837 as->length = p - as->s;
2838 if (archive_string_ensure(as,
2839 as->buffer_length + len * tm + ts) == NULL)
2840 return (-1);
2841 p = as->s + as->length;
2842 endp = as->s + as->buffer_length - ts;
2843 }
2844 p += w;
2845 }
2846 as->length = p - as->s;
2847 as->s[as->length] = '\0';
2848 if (ts == 2)
2849 as->s[as->length+1] = '\0';
2850 return (ret);
2851 }
2852
2853 /*
2854 * Following Constants for Hangul compositions this information comes from
2855 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2856 */
2857 #define HC_SBASE 0xAC00
2858 #define HC_LBASE 0x1100
2859 #define HC_VBASE 0x1161
2860 #define HC_TBASE 0x11A7
2861 #define HC_LCOUNT 19
2862 #define HC_VCOUNT 21
2863 #define HC_TCOUNT 28
2864 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2865 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2866
2867 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2868 get_nfc(uint32_t uc, uint32_t uc2)
2869 {
2870 int t, b;
2871
2872 t = 0;
2873 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2874 while (b >= t) {
2875 int m = (t + b) / 2;
2876 if (u_composition_table[m].cp1 < uc)
2877 t = m + 1;
2878 else if (u_composition_table[m].cp1 > uc)
2879 b = m - 1;
2880 else if (u_composition_table[m].cp2 < uc2)
2881 t = m + 1;
2882 else if (u_composition_table[m].cp2 > uc2)
2883 b = m - 1;
2884 else
2885 return (u_composition_table[m].nfc);
2886 }
2887 return (0);
2888 }
2889
2890 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2891 * Characters. */
2892
2893 /*
2894 * Update first code point.
2895 */
2896 #define UPDATE_UC(new_uc) do { \
2897 uc = new_uc; \
2898 ucptr = NULL; \
2899 } while (0)
2900
2901 /*
2902 * Replace first code point with second code point.
2903 */
2904 #define REPLACE_UC_WITH_UC2() do { \
2905 uc = uc2; \
2906 ucptr = uc2ptr; \
2907 n = n2; \
2908 } while (0)
2909
2910 #define EXPAND_BUFFER() do { \
2911 as->length = p - as->s; \
2912 if (archive_string_ensure(as, \
2913 as->buffer_length + len * tm + ts) == NULL)\
2914 return (-1); \
2915 p = as->s + as->length; \
2916 endp = as->s + as->buffer_length - ts; \
2917 } while (0)
2918
2919 #define UNPARSE(p, endp, uc) do { \
2920 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2921 EXPAND_BUFFER(); \
2922 } \
2923 p += w; \
2924 } while (0)
2925
2926 /*
2927 * Write first code point.
2928 * If the code point has not be changed from its original code,
2929 * this just copies it from its original buffer pointer.
2930 * If not, this converts it to UTF-8 byte sequence and copies it.
2931 */
2932 #define WRITE_UC() do { \
2933 if (ucptr) { \
2934 if (p + n > endp) \
2935 EXPAND_BUFFER(); \
2936 switch (n) { \
2937 case 4: \
2938 *p++ = *ucptr++; \
2939 /* FALL THROUGH */ \
2940 case 3: \
2941 *p++ = *ucptr++; \
2942 /* FALL THROUGH */ \
2943 case 2: \
2944 *p++ = *ucptr++; \
2945 /* FALL THROUGH */ \
2946 case 1: \
2947 *p++ = *ucptr; \
2948 break; \
2949 } \
2950 ucptr = NULL; \
2951 } else { \
2952 UNPARSE(p, endp, uc); \
2953 } \
2954 } while (0)
2955
2956 /*
2957 * Collect following decomposable code points.
2958 */
2959 #define COLLECT_CPS(start) do { \
2960 int _i; \
2961 for (_i = start; _i < FDC_MAX ; _i++) { \
2962 nx = parse(&ucx[_i], s, len); \
2963 if (nx <= 0) \
2964 break; \
2965 cx = CCC(ucx[_i]); \
2966 if (cl >= cx && cl != 228 && cx != 228)\
2967 break; \
2968 s += nx; \
2969 len -= nx; \
2970 cl = cx; \
2971 ccx[_i] = cx; \
2972 } \
2973 if (_i >= FDC_MAX) { \
2974 ret = -1; \
2975 ucx_size = FDC_MAX; \
2976 } else \
2977 ucx_size = _i; \
2978 } while (0)
2979
2980 /*
2981 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2982 *
2983 * TODO: Convert composition exclusions, which are never converted
2984 * from NFC,NFD,NFKC and NFKD, to Form C.
2985 */
2986 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2987 archive_string_normalize_C(struct archive_string *as, const void *_p,
2988 size_t len, struct archive_string_conv *sc)
2989 {
2990 const char *s = (const char *)_p;
2991 char *p, *endp;
2992 uint32_t uc, uc2;
2993 size_t w;
2994 int always_replace, n, n2, ret = 0, spair, ts, tm;
2995 int (*parse)(uint32_t *, const char *, size_t);
2996 size_t (*unparse)(char *, size_t, uint32_t);
2997
2998 always_replace = 1;
2999 ts = 1;/* text size. */
3000 if (sc->flag & SCONV_TO_UTF16BE) {
3001 unparse = unicode_to_utf16be;
3002 ts = 2;
3003 if (sc->flag & SCONV_FROM_UTF16BE)
3004 always_replace = 0;
3005 } else if (sc->flag & SCONV_TO_UTF16LE) {
3006 unparse = unicode_to_utf16le;
3007 ts = 2;
3008 if (sc->flag & SCONV_FROM_UTF16LE)
3009 always_replace = 0;
3010 } else if (sc->flag & SCONV_TO_UTF8) {
3011 unparse = unicode_to_utf8;
3012 if (sc->flag & SCONV_FROM_UTF8)
3013 always_replace = 0;
3014 } else {
3015 /*
3016 * This case is going to be converted to another
3017 * character-set through iconv.
3018 */
3019 always_replace = 0;
3020 if (sc->flag & SCONV_FROM_UTF16BE) {
3021 unparse = unicode_to_utf16be;
3022 ts = 2;
3023 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3024 unparse = unicode_to_utf16le;
3025 ts = 2;
3026 } else {
3027 unparse = unicode_to_utf8;
3028 }
3029 }
3030
3031 if (sc->flag & SCONV_FROM_UTF16BE) {
3032 parse = utf16be_to_unicode;
3033 tm = 1;
3034 spair = 4;/* surrogate pair size in UTF-16. */
3035 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3036 parse = utf16le_to_unicode;
3037 tm = 1;
3038 spair = 4;/* surrogate pair size in UTF-16. */
3039 } else {
3040 parse = cesu8_to_unicode;
3041 tm = ts;
3042 spair = 6;/* surrogate pair size in UTF-8. */
3043 }
3044
3045 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3046 return (-1);
3047
3048 p = as->s + as->length;
3049 endp = as->s + as->buffer_length - ts;
3050 while ((n = parse(&uc, s, len)) != 0) {
3051 const char *ucptr, *uc2ptr;
3052
3053 if (n < 0) {
3054 /* Use a replaced unicode character. */
3055 UNPARSE(p, endp, uc);
3056 s += n*-1;
3057 len -= n*-1;
3058 ret = -1;
3059 continue;
3060 } else if (n == spair || always_replace)
3061 /* uc is converted from a surrogate pair.
3062 * this should be treated as a changed code. */
3063 ucptr = NULL;
3064 else
3065 ucptr = s;
3066 s += n;
3067 len -= n;
3068
3069 /* Read second code point. */
3070 while ((n2 = parse(&uc2, s, len)) > 0) {
3071 uint32_t ucx[FDC_MAX];
3072 int ccx[FDC_MAX];
3073 int cl, cx, i, nx, ucx_size;
3074 int LIndex,SIndex;
3075 uint32_t nfc;
3076
3077 if (n2 == spair || always_replace)
3078 /* uc2 is converted from a surrogate pair.
3079 * this should be treated as a changed code. */
3080 uc2ptr = NULL;
3081 else
3082 uc2ptr = s;
3083 s += n2;
3084 len -= n2;
3085
3086 /*
3087 * If current second code point is out of decomposable
3088 * code points, finding compositions is unneeded.
3089 */
3090 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3091 WRITE_UC();
3092 REPLACE_UC_WITH_UC2();
3093 continue;
3094 }
3095
3096 /*
3097 * Try to combine current code points.
3098 */
3099 /*
3100 * We have to combine Hangul characters according to
3101 * http://uniicode.org/reports/tr15/#Hangul
3102 */
3103 if (0 <= (LIndex = uc - HC_LBASE) &&
3104 LIndex < HC_LCOUNT) {
3105 /*
3106 * Hangul Composition.
3107 * 1. Two current code points are L and V.
3108 */
3109 int VIndex = uc2 - HC_VBASE;
3110 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3111 /* Make syllable of form LV. */
3112 UPDATE_UC(HC_SBASE +
3113 (LIndex * HC_VCOUNT + VIndex) *
3114 HC_TCOUNT);
3115 } else {
3116 WRITE_UC();
3117 REPLACE_UC_WITH_UC2();
3118 }
3119 continue;
3120 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3121 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3122 /*
3123 * Hangul Composition.
3124 * 2. Two current code points are LV and T.
3125 */
3126 int TIndex = uc2 - HC_TBASE;
3127 if (0 < TIndex && TIndex < HC_TCOUNT) {
3128 /* Make syllable of form LVT. */
3129 UPDATE_UC(uc + TIndex);
3130 } else {
3131 WRITE_UC();
3132 REPLACE_UC_WITH_UC2();
3133 }
3134 continue;
3135 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3136 /* A composition to current code points
3137 * is found. */
3138 UPDATE_UC(nfc);
3139 continue;
3140 } else if ((cl = CCC(uc2)) == 0) {
3141 /* Clearly 'uc2' the second code point is not
3142 * a decomposable code. */
3143 WRITE_UC();
3144 REPLACE_UC_WITH_UC2();
3145 continue;
3146 }
3147
3148 /*
3149 * Collect following decomposable code points.
3150 */
3151 cx = 0;
3152 ucx[0] = uc2;
3153 ccx[0] = cl;
3154 COLLECT_CPS(1);
3155
3156 /*
3157 * Find a composed code in the collected code points.
3158 */
3159 i = 1;
3160 while (i < ucx_size) {
3161 int j;
3162
3163 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3164 i++;
3165 continue;
3166 }
3167
3168 /*
3169 * nfc is composed of uc and ucx[i].
3170 */
3171 UPDATE_UC(nfc);
3172
3173 /*
3174 * Remove ucx[i] by shifting
3175 * following code points.
3176 */
3177 for (j = i; j+1 < ucx_size; j++) {
3178 ucx[j] = ucx[j+1];
3179 ccx[j] = ccx[j+1];
3180 }
3181 ucx_size --;
3182
3183 /*
3184 * Collect following code points blocked
3185 * by ucx[i] the removed code point.
3186 */
3187 if (ucx_size > 0 && i == ucx_size &&
3188 nx > 0 && cx == cl) {
3189 cl = ccx[ucx_size-1];
3190 COLLECT_CPS(ucx_size);
3191 }
3192 /*
3193 * Restart finding a composed code with
3194 * the updated uc from the top of the
3195 * collected code points.
3196 */
3197 i = 0;
3198 }
3199
3200 /*
3201 * Apparently the current code points are not
3202 * decomposed characters or already composed.
3203 */
3204 WRITE_UC();
3205 for (i = 0; i < ucx_size; i++)
3206 UNPARSE(p, endp, ucx[i]);
3207
3208 /*
3209 * Flush out remaining canonical combining characters.
3210 */
3211 if (nx > 0 && cx == cl && len > 0) {
3212 while ((nx = parse(&ucx[0], s, len))
3213 > 0) {
3214 cx = CCC(ucx[0]);
3215 if (cl > cx)
3216 break;
3217 s += nx;
3218 len -= nx;
3219 cl = cx;
3220 UNPARSE(p, endp, ucx[0]);
3221 }
3222 }
3223 break;
3224 }
3225 if (n2 < 0) {
3226 WRITE_UC();
3227 /* Use a replaced unicode character. */
3228 UNPARSE(p, endp, uc2);
3229 s += n2*-1;
3230 len -= n2*-1;
3231 ret = -1;
3232 continue;
3233 } else if (n2 == 0) {
3234 WRITE_UC();
3235 break;
3236 }
3237 }
3238 as->length = p - as->s;
3239 as->s[as->length] = '\0';
3240 if (ts == 2)
3241 as->s[as->length+1] = '\0';
3242 return (ret);
3243 }
3244
3245 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3246 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3247 {
3248 int t, b;
3249
3250 /*
3251 * These are not converted to NFD on Mac OS.
3252 */
3253 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3254 (uc >= 0xF900 && uc <= 0xFAFF) ||
3255 (uc >= 0x2F800 && uc <= 0x2FAFF))
3256 return (0);
3257 /*
3258 * Those code points are not converted to NFD on Mac OS.
3259 * I do not know the reason because it is undocumented.
3260 * NFC NFD
3261 * 1109A ==> 11099 110BA
3262 * 1109C ==> 1109B 110BA
3263 * 110AB ==> 110A5 110BA
3264 */
3265 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3266 return (0);
3267
3268 t = 0;
3269 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3270 while (b >= t) {
3271 int m = (t + b) / 2;
3272 if (u_decomposition_table[m].nfc < uc)
3273 t = m + 1;
3274 else if (u_decomposition_table[m].nfc > uc)
3275 b = m - 1;
3276 else {
3277 *cp1 = u_decomposition_table[m].cp1;
3278 *cp2 = u_decomposition_table[m].cp2;
3279 return (1);
3280 }
3281 }
3282 return (0);
3283 }
3284
3285 #define REPLACE_UC_WITH(cp) do { \
3286 uc = cp; \
3287 ucptr = NULL; \
3288 } while (0)
3289
3290 /*
3291 * Normalize UTF-8 characters to Form D and copy the result.
3292 */
3293 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3294 archive_string_normalize_D(struct archive_string *as, const void *_p,
3295 size_t len, struct archive_string_conv *sc)
3296 {
3297 const char *s = (const char *)_p;
3298 char *p, *endp;
3299 uint32_t uc, uc2;
3300 size_t w;
3301 int always_replace, n, n2, ret = 0, spair, ts, tm;
3302 int (*parse)(uint32_t *, const char *, size_t);
3303 size_t (*unparse)(char *, size_t, uint32_t);
3304
3305 always_replace = 1;
3306 ts = 1;/* text size. */
3307 if (sc->flag & SCONV_TO_UTF16BE) {
3308 unparse = unicode_to_utf16be;
3309 ts = 2;
3310 if (sc->flag & SCONV_FROM_UTF16BE)
3311 always_replace = 0;
3312 } else if (sc->flag & SCONV_TO_UTF16LE) {
3313 unparse = unicode_to_utf16le;
3314 ts = 2;
3315 if (sc->flag & SCONV_FROM_UTF16LE)
3316 always_replace = 0;
3317 } else if (sc->flag & SCONV_TO_UTF8) {
3318 unparse = unicode_to_utf8;
3319 if (sc->flag & SCONV_FROM_UTF8)
3320 always_replace = 0;
3321 } else {
3322 /*
3323 * This case is going to be converted to another
3324 * character-set through iconv.
3325 */
3326 always_replace = 0;
3327 if (sc->flag & SCONV_FROM_UTF16BE) {
3328 unparse = unicode_to_utf16be;
3329 ts = 2;
3330 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3331 unparse = unicode_to_utf16le;
3332 ts = 2;
3333 } else {
3334 unparse = unicode_to_utf8;
3335 }
3336 }
3337
3338 if (sc->flag & SCONV_FROM_UTF16BE) {
3339 parse = utf16be_to_unicode;
3340 tm = 1;
3341 spair = 4;/* surrogate pair size in UTF-16. */
3342 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3343 parse = utf16le_to_unicode;
3344 tm = 1;
3345 spair = 4;/* surrogate pair size in UTF-16. */
3346 } else {
3347 parse = cesu8_to_unicode;
3348 tm = ts;
3349 spair = 6;/* surrogate pair size in UTF-8. */
3350 }
3351
3352 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3353 return (-1);
3354
3355 p = as->s + as->length;
3356 endp = as->s + as->buffer_length - ts;
3357 while ((n = parse(&uc, s, len)) != 0) {
3358 const char *ucptr;
3359 uint32_t cp1, cp2;
3360 int SIndex;
3361 struct {
3362 uint32_t uc;
3363 int ccc;
3364 } fdc[FDC_MAX];
3365 int fdi, fdj;
3366 int ccc;
3367
3368 check_first_code:
3369 if (n < 0) {
3370 /* Use a replaced unicode character. */
3371 UNPARSE(p, endp, uc);
3372 s += n*-1;
3373 len -= n*-1;
3374 ret = -1;
3375 continue;
3376 } else if (n == spair || always_replace)
3377 /* uc is converted from a surrogate pair.
3378 * this should be treated as a changed code. */
3379 ucptr = NULL;
3380 else
3381 ucptr = s;
3382 s += n;
3383 len -= n;
3384
3385 /* Hangul Decomposition. */
3386 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3387 int L = HC_LBASE + SIndex / HC_NCOUNT;
3388 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3389 int T = HC_TBASE + SIndex % HC_TCOUNT;
3390
3391 REPLACE_UC_WITH(L);
3392 WRITE_UC();
3393 REPLACE_UC_WITH(V);
3394 WRITE_UC();
3395 if (T != HC_TBASE) {
3396 REPLACE_UC_WITH(T);
3397 WRITE_UC();
3398 }
3399 continue;
3400 }
3401 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3402 WRITE_UC();
3403 continue;
3404 }
3405
3406 fdi = 0;
3407 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3408 int k;
3409
3410 for (k = fdi; k > 0; k--)
3411 fdc[k] = fdc[k-1];
3412 fdc[0].ccc = CCC(cp2);
3413 fdc[0].uc = cp2;
3414 fdi++;
3415 REPLACE_UC_WITH(cp1);
3416 }
3417
3418 /* Read following code points. */
3419 while ((n2 = parse(&uc2, s, len)) > 0 &&
3420 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3421 int j, k;
3422
3423 s += n2;
3424 len -= n2;
3425 for (j = 0; j < fdi; j++) {
3426 if (fdc[j].ccc > ccc)
3427 break;
3428 }
3429 if (j < fdi) {
3430 for (k = fdi; k > j; k--)
3431 fdc[k] = fdc[k-1];
3432 fdc[j].ccc = ccc;
3433 fdc[j].uc = uc2;
3434 } else {
3435 fdc[fdi].ccc = ccc;
3436 fdc[fdi].uc = uc2;
3437 }
3438 fdi++;
3439 }
3440
3441 WRITE_UC();
3442 for (fdj = 0; fdj < fdi; fdj++) {
3443 REPLACE_UC_WITH(fdc[fdj].uc);
3444 WRITE_UC();
3445 }
3446
3447 if (n2 == 0)
3448 break;
3449 REPLACE_UC_WITH(uc2);
3450 n = n2;
3451 goto check_first_code;
3452 }
3453 as->length = p - as->s;
3454 as->s[as->length] = '\0';
3455 if (ts == 2)
3456 as->s[as->length+1] = '\0';
3457 return (ret);
3458 }
3459
3460 /*
3461 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3462 * that WCS is Unicode. It is true for several platforms but some are false.
3463 * And then people who did not use UTF-8 locale on the non Unicode WCS
3464 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3465 * now cannot get right filename from libarchive 3.x and later since we
3466 * fixed the wrong assumption and it is incompatible to older its versions.
3467 * So we provide special option, "compat-2x.x", for resolving it.
3468 * That option enable the string conversion of libarchive 2.x.
3469 *
3470 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3471 * locale character set and appends to the archive_string.
3472 * Note: returns -1 if conversion fails.
3473 */
3474 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3475 strncat_from_utf8_libarchive2(struct archive_string *as,
3476 const void *_p, size_t len, struct archive_string_conv *sc)
3477 {
3478 const char *s;
3479 int n;
3480 char *p;
3481 char *end;
3482 uint32_t unicode;
3483 #if HAVE_WCRTOMB
3484 mbstate_t shift_state;
3485
3486 memset(&shift_state, 0, sizeof(shift_state));
3487 #else
3488 /* Clear the shift state before starting. */
3489 wctomb(NULL, L'\0');
3490 #endif
3491 (void)sc; /* UNUSED */
3492 /*
3493 * Allocate buffer for MBS.
3494 * We need this allocation here since it is possible that
3495 * as->s is still NULL.
3496 */
3497 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3498 return (-1);
3499
3500 s = (const char *)_p;
3501 p = as->s + as->length;
3502 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3503 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3504 wchar_t wc;
3505
3506 if (p >= end) {
3507 as->length = p - as->s;
3508 /* Re-allocate buffer for MBS. */
3509 if (archive_string_ensure(as,
3510 as->length + max(len * 2,
3511 (size_t)MB_CUR_MAX) + 1) == NULL)
3512 return (-1);
3513 p = as->s + as->length;
3514 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3515 }
3516
3517 /*
3518 * As libarchive 2.x, translates the UTF-8 characters into
3519 * wide-characters in the assumption that WCS is Unicode.
3520 */
3521 if (n < 0) {
3522 n *= -1;
3523 wc = L'?';
3524 } else
3525 wc = (wchar_t)unicode;
3526
3527 s += n;
3528 len -= n;
3529 /*
3530 * Translates the wide-character into the current locale MBS.
3531 */
3532 #if HAVE_WCRTOMB
3533 n = (int)wcrtomb(p, wc, &shift_state);
3534 #else
3535 n = (int)wctomb(p, wc);
3536 #endif
3537 if (n == -1)
3538 return (-1);
3539 p += n;
3540 }
3541 as->length = p - as->s;
3542 as->s[as->length] = '\0';
3543 return (0);
3544 }
3545
3546
3547 /*
3548 * Conversion functions between current locale dependent MBS and UTF-16BE.
3549 * strncat_from_utf16be() : UTF-16BE --> MBS
3550 * strncat_to_utf16be() : MBS --> UTF16BE
3551 */
3552
3553 #if defined(_WIN32) && !defined(__CYGWIN__)
3554
3555 /*
3556 * Convert a UTF-16BE/LE string to current locale and copy the result.
3557 * Return -1 if conversion fails.
3558 */
3559 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3560 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3561 struct archive_string_conv *sc, int be)
3562 {
3563 struct archive_string tmp;
3564 const char *u16;
3565 BOOL defchar;
3566 char *mbs;
3567 size_t mbs_size, b, ll;
3568 int ret = 0;
3569
3570 bytes &= ~1;
3571 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3572 return (-1);
3573
3574 mbs = as->s + as->length;
3575 mbs_size = as->buffer_length - as->length -1;
3576
3577 if (sc->to_cp == CP_C_LOCALE) {
3578 /*
3579 * "C" locale special processing.
3580 */
3581 u16 = _p;
3582 ll = 0;
3583 for (b = 0; b < bytes; b += 2) {
3584 uint16_t val;
3585 if (be)
3586 val = archive_be16dec(u16+b);
3587 else
3588 val = archive_le16dec(u16+b);
3589 if (val > 255) {
3590 *mbs++ = '?';
3591 ret = -1;
3592 } else
3593 *mbs++ = (char)(val&0xff);
3594 ll++;
3595 }
3596 as->length += ll;
3597 as->s[as->length] = '\0';
3598 return (ret);
3599 }
3600
3601 archive_string_init(&tmp);
3602 if (be) {
3603 if (IS_BIG_ENDIAN) {
3604 u16 = _p;
3605 } else {
3606 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3607 return (-1);
3608 memcpy(tmp.s, _p, bytes);
3609 for (b = 0; b < bytes; b += 2) {
3610 uint16_t val = archive_be16dec(tmp.s+b);
3611 archive_le16enc(tmp.s+b, val);
3612 }
3613 u16 = tmp.s;
3614 }
3615 } else {
3616 if (!IS_BIG_ENDIAN) {
3617 u16 = _p;
3618 } else {
3619 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3620 return (-1);
3621 memcpy(tmp.s, _p, bytes);
3622 for (b = 0; b < bytes; b += 2) {
3623 uint16_t val = archive_le16dec(tmp.s+b);
3624 archive_be16enc(tmp.s+b, val);
3625 }
3626 u16 = tmp.s;
3627 }
3628 }
3629
3630 do {
3631 int r;
3632 defchar = 0;
3633 /* WideCharToMultiByte is limited to int. */
3634 if (bytes > (size_t)INT_MAX || mbs_size > (size_t)INT_MAX)
3635 return (-1);
3636 r = WideCharToMultiByte(sc->to_cp, 0,
3637 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3638 NULL, &defchar);
3639 /* Exit loop if we succeeded */
3640 if (r != 0 ||
3641 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3642 ll = (size_t)r;
3643 break;
3644 }
3645 /* Else expand buffer and loop to try again. */
3646 r = WideCharToMultiByte(sc->to_cp, 0,
3647 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3648 ll = (size_t)r;
3649 if (archive_string_ensure(as, ll +1) == NULL)
3650 return (-1);
3651 mbs = as->s + as->length;
3652 mbs_size = as->buffer_length - as->length -1;
3653 } while (1);
3654 archive_string_free(&tmp);
3655 as->length += ll;
3656 as->s[as->length] = '\0';
3657 if (ll == 0 || defchar)
3658 ret = -1;
3659 return (ret);
3660 }
3661
3662 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3663 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3664 size_t bytes, struct archive_string_conv *sc)
3665 {
3666 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3667 }
3668
3669 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3670 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3671 size_t bytes, struct archive_string_conv *sc)
3672 {
3673 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3674 }
3675
3676 /*
3677 * Convert a current locale string to UTF-16BE/LE and copy the result.
3678 * Return -1 if conversion fails.
3679 */
3680 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3681 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3682 size_t length, struct archive_string_conv *sc, int bigendian)
3683 {
3684 const char *s = (const char *)_p;
3685 char *u16;
3686 size_t count, avail;
3687
3688 if (archive_string_ensure(as16,
3689 as16->length + (length + 1) * 2) == NULL)
3690 return (-1);
3691
3692 u16 = as16->s + as16->length;
3693 avail = as16->buffer_length - 2;
3694 if (sc->from_cp == CP_C_LOCALE) {
3695 /*
3696 * "C" locale special processing.
3697 */
3698 count = 0;
3699 while (count < length && *s) {
3700 if (bigendian)
3701 archive_be16enc(u16, *s);
3702 else
3703 archive_le16enc(u16, *s);
3704 u16 += 2;
3705 s++;
3706 count++;
3707 }
3708 as16->length += count << 1;
3709 as16->s[as16->length] = 0;
3710 as16->s[as16->length+1] = 0;
3711 return (0);
3712 }
3713 do {
3714 int r;
3715 if (length > (size_t)INT_MAX || (avail >> 1) > (size_t)INT_MAX)
3716 return (-1);
3717 r = MultiByteToWideChar(sc->from_cp,
3718 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3719 /* Exit loop if we succeeded */
3720 if (r != 0 ||
3721 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3722 count = (size_t)r;
3723 break;
3724 }
3725 /* Expand buffer and try again */
3726 r = MultiByteToWideChar(sc->from_cp,
3727 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3728 count = (size_t)r;
3729 if (archive_string_ensure(as16, (count +1) * 2)
3730 == NULL)
3731 return (-1);
3732 u16 = as16->s + as16->length;
3733 avail = as16->buffer_length - 2;
3734 } while (1);
3735 as16->length += count * 2;
3736 as16->s[as16->length] = 0;
3737 as16->s[as16->length+1] = 0;
3738 if (count == 0)
3739 return (-1);
3740
3741 if (IS_BIG_ENDIAN) {
3742 if (!bigendian) {
3743 while (count > 0) {
3744 uint16_t v = archive_be16dec(u16);
3745 archive_le16enc(u16, v);
3746 u16 += 2;
3747 count--;
3748 }
3749 }
3750 } else {
3751 if (bigendian) {
3752 while (count > 0) {
3753 uint16_t v = archive_le16dec(u16);
3754 archive_be16enc(u16, v);
3755 u16 += 2;
3756 count--;
3757 }
3758 }
3759 }
3760 return (0);
3761 }
3762
3763 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3764 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3765 size_t length, struct archive_string_conv *sc)
3766 {
3767 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3768 }
3769
3770 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3771 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3772 size_t length, struct archive_string_conv *sc)
3773 {
3774 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3775 }
3776
3777 #endif /* _WIN32 && !__CYGWIN__ */
3778
3779 /*
3780 * Do the best effort for conversions.
3781 * We cannot handle UTF-16BE character-set without such iconv,
3782 * but there is a chance if a string consists just ASCII code or
3783 * a current locale is UTF-8.
3784 */
3785
3786 /*
3787 * Convert a UTF-16BE string to current locale and copy the result.
3788 * Return -1 if conversion fails.
3789 */
3790 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3791 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3792 size_t bytes, struct archive_string_conv *sc, int be)
3793 {
3794 const char *utf16 = (const char *)_p;
3795 char *mbs;
3796 uint32_t uc;
3797 int n, ret;
3798
3799 (void)sc; /* UNUSED */
3800 /*
3801 * Other case, we should do the best effort.
3802 * If all character are ASCII(<0x7f), we can convert it.
3803 * if not , we set a alternative character and return -1.
3804 */
3805 ret = 0;
3806 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3807 return (-1);
3808 mbs = as->s + as->length;
3809
3810 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3811 if (n < 0) {
3812 n *= -1;
3813 ret = -1;
3814 }
3815 bytes -= n;
3816 utf16 += n;
3817
3818 if (uc > 127) {
3819 /* We cannot handle it. */
3820 *mbs++ = '?';
3821 ret = -1;
3822 } else
3823 *mbs++ = (char)uc;
3824 }
3825 as->length = mbs - as->s;
3826 as->s[as->length] = '\0';
3827 return (ret);
3828 }
3829
3830 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3831 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3832 size_t bytes, struct archive_string_conv *sc)
3833 {
3834 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3835 }
3836
3837 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3838 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3839 size_t bytes, struct archive_string_conv *sc)
3840 {
3841 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3842 }
3843
3844 /*
3845 * Convert a current locale string to UTF-16BE/LE and copy the result.
3846 * Return -1 if conversion fails.
3847 */
3848 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3849 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3850 size_t length, struct archive_string_conv *sc, int bigendian)
3851 {
3852 const char *s = (const char *)_p;
3853 char *utf16;
3854 size_t remaining;
3855 int ret;
3856
3857 (void)sc; /* UNUSED */
3858 /*
3859 * Other case, we should do the best effort.
3860 * If all character are ASCII(<0x7f), we can convert it.
3861 * if not , we set a alternative character and return -1.
3862 */
3863 ret = 0;
3864 remaining = length;
3865
3866 if (archive_string_ensure(as16,
3867 as16->length + (length + 1) * 2) == NULL)
3868 return (-1);
3869
3870 utf16 = as16->s + as16->length;
3871 while (remaining--) {
3872 unsigned c = *s++;
3873 if (c > 127) {
3874 /* We cannot handle it. */
3875 c = UNICODE_R_CHAR;
3876 ret = -1;
3877 }
3878 if (bigendian)
3879 archive_be16enc(utf16, (uint16_t)c);
3880 else
3881 archive_le16enc(utf16, (uint16_t)c);
3882 utf16 += 2;
3883 }
3884 as16->length = utf16 - as16->s;
3885 as16->s[as16->length] = 0;
3886 as16->s[as16->length+1] = 0;
3887 return (ret);
3888 }
3889
3890 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3891 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3892 size_t length, struct archive_string_conv *sc)
3893 {
3894 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3895 }
3896
3897 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3898 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3899 size_t length, struct archive_string_conv *sc)
3900 {
3901 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3902 }
3903
3904
3905 /*
3906 * Multistring operations.
3907 */
3908
3909 void
archive_mstring_clean(struct archive_mstring * aes)3910 archive_mstring_clean(struct archive_mstring *aes)
3911 {
3912 archive_wstring_free(&(aes->aes_wcs));
3913 archive_string_free(&(aes->aes_mbs));
3914 archive_string_free(&(aes->aes_utf8));
3915 archive_string_free(&(aes->aes_mbs_in_locale));
3916 aes->aes_set = 0;
3917 }
3918
3919 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3920 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3921 {
3922 dest->aes_set = src->aes_set;
3923 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3924 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3925 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3926 }
3927
3928 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3929 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3930 const char **p)
3931 {
3932 struct archive_string_conv *sc;
3933 int r;
3934
3935 /* If we already have a UTF8 form, return that immediately. */
3936 if (aes->aes_set & AES_SET_UTF8) {
3937 *p = aes->aes_utf8.s;
3938 return (0);
3939 }
3940
3941 *p = NULL;
3942 #if defined(_WIN32) && !defined(__CYGWIN__)
3943 /*
3944 * On Windows, first try converting from WCS because (1) there's no
3945 * guarantee that the conversion to MBS will succeed, e.g. when using
3946 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3947 * convert back to WCS again before finally converting to UTF-8
3948 */
3949 if ((aes->aes_set & AES_SET_WCS) != 0) {
3950 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3951 if (sc == NULL)
3952 return (-1);/* Couldn't allocate memory for sc. */
3953 archive_string_empty(&(aes->aes_utf8));
3954 r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3955 aes->aes_wcs.s, aes->aes_wcs.length, sc);
3956 if (a == NULL)
3957 free_sconv_object(sc);
3958 if (r == 0) {
3959 aes->aes_set |= AES_SET_UTF8;
3960 *p = aes->aes_utf8.s;
3961 return (0);/* success. */
3962 } else
3963 return (-1);/* failure. */
3964 }
3965 #endif
3966 /* Try converting WCS to MBS first if MBS does not exist yet. */
3967 if ((aes->aes_set & AES_SET_MBS) == 0) {
3968 const char *pm; /* unused */
3969 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3970 }
3971 if (aes->aes_set & AES_SET_MBS) {
3972 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3973 if (sc == NULL)
3974 return (-1);/* Couldn't allocate memory for sc. */
3975 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3976 aes->aes_mbs.length, sc);
3977 if (a == NULL)
3978 free_sconv_object(sc);
3979 if (r == 0) {
3980 aes->aes_set |= AES_SET_UTF8;
3981 *p = aes->aes_utf8.s;
3982 return (0);/* success. */
3983 } else
3984 return (-1);/* failure. */
3985 }
3986 return (0);/* success. */
3987 }
3988
3989 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3990 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3991 const char **p)
3992 {
3993 struct archive_string_conv *sc;
3994 int r, ret = 0;
3995
3996 /* If we already have an MBS form, return that immediately. */
3997 if (aes->aes_set & AES_SET_MBS) {
3998 *p = aes->aes_mbs.s;
3999 return (ret);
4000 }
4001
4002 *p = NULL;
4003 /* If there's a WCS form, try converting with the native locale. */
4004 if (aes->aes_set & AES_SET_WCS) {
4005 archive_string_empty(&(aes->aes_mbs));
4006 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4007 aes->aes_wcs.s, aes->aes_wcs.length);
4008 *p = aes->aes_mbs.s;
4009 if (r == 0) {
4010 aes->aes_set |= AES_SET_MBS;
4011 return (ret);
4012 } else
4013 ret = -1;
4014 }
4015
4016 /* If there's a UTF-8 form, try converting with the native locale. */
4017 if (aes->aes_set & AES_SET_UTF8) {
4018 archive_string_empty(&(aes->aes_mbs));
4019 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4020 if (sc == NULL)
4021 return (-1);/* Couldn't allocate memory for sc. */
4022 r = archive_strncpy_l(&(aes->aes_mbs),
4023 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4024 if (a == NULL)
4025 free_sconv_object(sc);
4026 *p = aes->aes_mbs.s;
4027 if (r == 0) {
4028 aes->aes_set |= AES_SET_MBS;
4029 ret = 0;/* success; overwrite previous error. */
4030 } else
4031 ret = -1;/* failure. */
4032 }
4033 return (ret);
4034 }
4035
4036 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)4037 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
4038 const wchar_t **wp)
4039 {
4040 int r, ret = 0;
4041
4042 (void)a;/* UNUSED */
4043 /* Return WCS form if we already have it. */
4044 if (aes->aes_set & AES_SET_WCS) {
4045 *wp = aes->aes_wcs.s;
4046 return (ret);
4047 }
4048
4049 *wp = NULL;
4050 #if defined(_WIN32) && !defined(__CYGWIN__)
4051 /*
4052 * On Windows, prefer converting from UTF-8 directly to WCS because:
4053 * (1) there's no guarantee that the string can be represented in MBS (e.g.
4054 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
4055 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
4056 * away that intermediate result
4057 */
4058 if (aes->aes_set & AES_SET_UTF8) {
4059 struct archive_string_conv *sc;
4060
4061 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4062 if (sc != NULL) {
4063 archive_wstring_empty((&aes->aes_wcs));
4064 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4065 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4066 if (a == NULL)
4067 free_sconv_object(sc);
4068 if (r == 0) {
4069 aes->aes_set |= AES_SET_WCS;
4070 *wp = aes->aes_wcs.s;
4071 return (0);
4072 }
4073 }
4074 }
4075 #endif
4076 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
4077 if ((aes->aes_set & AES_SET_MBS) == 0) {
4078 const char *p; /* unused */
4079 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4080 }
4081 /* Try converting MBS to WCS using native locale. */
4082 if (aes->aes_set & AES_SET_MBS) {
4083 archive_wstring_empty(&(aes->aes_wcs));
4084 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4085 aes->aes_mbs.s, aes->aes_mbs.length);
4086 if (r == 0) {
4087 aes->aes_set |= AES_SET_WCS;
4088 *wp = aes->aes_wcs.s;
4089 } else
4090 ret = -1;/* failure. */
4091 }
4092 return (ret);
4093 }
4094
4095 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)4096 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4097 const char **p, size_t *length, struct archive_string_conv *sc)
4098 {
4099 int ret = 0;
4100 #if defined(_WIN32) && !defined(__CYGWIN__)
4101 int r;
4102
4103 /*
4104 * Internationalization programming on Windows must use Wide
4105 * characters because Windows platform cannot make locale UTF-8.
4106 */
4107 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4108 archive_string_empty(&(aes->aes_mbs_in_locale));
4109 r = archive_string_append_from_wcs_in_codepage(
4110 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4111 aes->aes_wcs.length, sc);
4112 if (r == 0) {
4113 *p = aes->aes_mbs_in_locale.s;
4114 if (length != NULL)
4115 *length = aes->aes_mbs_in_locale.length;
4116 return (0);
4117 } else if (errno == ENOMEM)
4118 return (-1);
4119 else
4120 ret = -1;
4121 }
4122 #endif
4123
4124 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4125 * with the native locale to be used for translating it to specified
4126 * character-set. */
4127 if ((aes->aes_set & AES_SET_MBS) == 0) {
4128 const char *pm; /* unused */
4129 if (archive_mstring_get_mbs(a, aes, &pm) != 0) {
4130 /* We have another form, but failed to convert it to
4131 * the native locale. Transitively, we've failed to
4132 * convert it to the specified character set. */
4133 ret = -1;
4134 }
4135 }
4136 /* If we already have an MBS form, use it to be translated to
4137 * specified character-set. */
4138 if (aes->aes_set & AES_SET_MBS) {
4139 if (sc == NULL) {
4140 /* Conversion is unneeded. */
4141 *p = aes->aes_mbs.s;
4142 if (length != NULL)
4143 *length = aes->aes_mbs.length;
4144 return (0);
4145 }
4146 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4147 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4148 *p = aes->aes_mbs_in_locale.s;
4149 if (length != NULL)
4150 *length = aes->aes_mbs_in_locale.length;
4151 } else {
4152 /* Either we have no string in any form,
4153 * or conversion failed and set 'ret != 0'. */
4154 *p = NULL;
4155 if (length != NULL)
4156 *length = 0;
4157 }
4158 return (ret);
4159 }
4160
4161 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4162 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4163 {
4164 if (mbs == NULL) {
4165 aes->aes_set = 0;
4166 return (0);
4167 }
4168 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4169 }
4170
4171 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4172 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4173 size_t len)
4174 {
4175 if (mbs == NULL) {
4176 aes->aes_set = 0;
4177 return (0);
4178 }
4179 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4180 archive_strncpy(&(aes->aes_mbs), mbs, len);
4181 archive_string_empty(&(aes->aes_utf8));
4182 archive_wstring_empty(&(aes->aes_wcs));
4183 return (0);
4184 }
4185
4186 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4187 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4188 {
4189 return archive_mstring_copy_wcs_len(aes, wcs,
4190 wcs == NULL ? 0 : wcslen(wcs));
4191 }
4192
4193 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4194 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4195 {
4196 if (utf8 == NULL) {
4197 aes->aes_set = 0;
4198 return (0);
4199 }
4200 aes->aes_set = AES_SET_UTF8;
4201 archive_string_empty(&(aes->aes_mbs));
4202 archive_string_empty(&(aes->aes_wcs));
4203 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4204 return (int)strlen(utf8);
4205 }
4206
4207 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4208 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4209 size_t len)
4210 {
4211 if (wcs == NULL) {
4212 aes->aes_set = 0;
4213 return (0);
4214 }
4215 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4216 archive_string_empty(&(aes->aes_mbs));
4217 archive_string_empty(&(aes->aes_utf8));
4218 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4219 return (0);
4220 }
4221
4222 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4223 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4224 const char *mbs, size_t len, struct archive_string_conv *sc)
4225 {
4226 int r;
4227
4228 if (mbs == NULL) {
4229 aes->aes_set = 0;
4230 return (0);
4231 }
4232 archive_string_empty(&(aes->aes_mbs));
4233 archive_wstring_empty(&(aes->aes_wcs));
4234 archive_string_empty(&(aes->aes_utf8));
4235 #if defined(_WIN32) && !defined(__CYGWIN__)
4236 /*
4237 * Internationalization programming on Windows must use Wide
4238 * characters because Windows platform cannot make locale UTF-8.
4239 */
4240 if (sc == NULL) {
4241 if (archive_string_append(&(aes->aes_mbs),
4242 mbs, mbsnbytes(mbs, len)) == NULL) {
4243 aes->aes_set = 0;
4244 r = -1;
4245 } else {
4246 aes->aes_set = AES_SET_MBS;
4247 r = 0;
4248 }
4249 #if defined(HAVE_ICONV)
4250 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4251 /*
4252 * This case happens only when MultiByteToWideChar() cannot
4253 * handle sc->from_cp, and we have to iconv in order to
4254 * translate character-set to wchar_t,UTF-16.
4255 */
4256 iconv_t cd = sc->cd;
4257 unsigned from_cp;
4258 int flag;
4259
4260 /*
4261 * Translate multi-bytes from some character-set to UTF-8.
4262 */
4263 sc->cd = sc->cd_w;
4264 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4265 sc->cd = cd;
4266 if (r != 0) {
4267 aes->aes_set = 0;
4268 return (r);
4269 }
4270 aes->aes_set = AES_SET_UTF8;
4271
4272 /*
4273 * Append the UTF-8 string into wstring.
4274 */
4275 flag = sc->flag;
4276 sc->flag &= ~(SCONV_NORMALIZATION_C
4277 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4278 from_cp = sc->from_cp;
4279 sc->from_cp = CP_UTF8;
4280 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4281 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4282 sc->flag = flag;
4283 sc->from_cp = from_cp;
4284 if (r == 0)
4285 aes->aes_set |= AES_SET_WCS;
4286 #endif
4287 } else {
4288 r = archive_wstring_append_from_mbs_in_codepage(
4289 &(aes->aes_wcs), mbs, len, sc);
4290 if (r == 0)
4291 aes->aes_set = AES_SET_WCS;
4292 else
4293 aes->aes_set = 0;
4294 }
4295 #else
4296 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4297 if (r == 0)
4298 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4299 else
4300 aes->aes_set = 0;
4301 #endif
4302 return (r);
4303 }
4304
4305 /*
4306 * The 'update' form tries to proactively update all forms of
4307 * this string (WCS and MBS) and returns an error if any of
4308 * them fail. This is used by the 'pax' handler, for instance,
4309 * to detect and report character-conversion failures early while
4310 * still allowing clients to get potentially useful values from
4311 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4312 * strive to give the user something useful, so you can get hopefully
4313 * usable values even if some of the character conversions are failing.)
4314 */
4315 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4316 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4317 const char *utf8)
4318 {
4319 struct archive_string_conv *sc;
4320 int r;
4321
4322 if (utf8 == NULL) {
4323 aes->aes_set = 0;
4324 return (0); /* Succeeded in clearing everything. */
4325 }
4326
4327 /* Save the UTF8 string. */
4328 archive_strcpy(&(aes->aes_utf8), utf8);
4329
4330 /* Empty the mbs and wcs strings. */
4331 archive_string_empty(&(aes->aes_mbs));
4332 archive_wstring_empty(&(aes->aes_wcs));
4333
4334 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4335
4336 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4337 if (sc == NULL)
4338 return (-1);/* Couldn't allocate memory for sc. */
4339
4340 #if defined(_WIN32) && !defined(__CYGWIN__)
4341 /* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4342 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4343 * intermediate result, and (2) WCS will still be set even if we fail to
4344 * convert to MBS (e.g. with ACP that can't represent the characters) */
4345 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4346 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4347
4348 if (a == NULL)
4349 free_sconv_object(sc);
4350 if (r != 0)
4351 return (-1); /* This will guarantee we can't convert to MBS */
4352 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4353
4354 /* Try converting WCS to MBS, return false on failure. */
4355 if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4356 aes->aes_wcs.length))
4357 return (-1);
4358 #else
4359 /* Try converting UTF-8 to MBS, return false on failure. */
4360 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4361
4362 if (a == NULL)
4363 free_sconv_object(sc);
4364 if (r != 0)
4365 return (-1);
4366 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4367
4368 /* Try converting MBS to WCS, return false on failure. */
4369 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4370 aes->aes_mbs.length))
4371 return (-1);
4372 #endif
4373
4374 /* All conversions succeeded. */
4375 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4376
4377 return (0);
4378 }
4379