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(__FreeBSD__) && !defined(HAVE_LIBICONV)
1318 /*
1319 * FreeBSD's native iconv() by default returns the number of
1320 * invalid characters in the input string, as specified by
1321 * POSIX, but iconv_strncat_in_locale() assumes GNU iconv
1322 * semantics.
1323 */
1324 int v = 1;
1325
1326 (void)iconvctl(sc->cd, ICONV_SET_ILSEQ_INVALID, &v);
1327 #elif defined(_WIN32) && !defined(__CYGWIN__)
1328 /*
1329 * archive_mstring on Windows directly convert multi-bytes
1330 * into archive_wstring in order not to depend on locale
1331 * so that you can do a I18N programming. This will be
1332 * used only in archive_mstring_copy_mbs_len_l so far.
1333 */
1334 if (flag & SCONV_FROM_CHARSET) {
1335 sc->cd_w = iconv_open("UTF-8", fc);
1336 if (sc->cd_w == (iconv_t)-1 &&
1337 (sc->flag & SCONV_BEST_EFFORT)) {
1338 if (strcmp(fc, "CP932") == 0)
1339 sc->cd_w = iconv_open("UTF-8", "SJIS");
1340 }
1341 }
1342 #endif /* _WIN32 && !__CYGWIN__ */
1343 }
1344 #endif /* HAVE_ICONV */
1345
1346 sc->flag = flag;
1347
1348 /*
1349 * Set up converters.
1350 */
1351 setup_converter(sc);
1352
1353 return (sc);
1354 }
1355
1356 /*
1357 * Free a string conversion object.
1358 */
1359 static void
free_sconv_object(struct archive_string_conv * sc)1360 free_sconv_object(struct archive_string_conv *sc)
1361 {
1362 free(sc->from_charset);
1363 free(sc->to_charset);
1364 archive_string_free(&sc->utftmp);
1365 #if HAVE_ICONV
1366 if (sc->cd != (iconv_t)-1)
1367 iconv_close(sc->cd);
1368 if (sc->cd_w != (iconv_t)-1)
1369 iconv_close(sc->cd_w);
1370 #endif
1371 free(sc);
1372 }
1373
1374 #if defined(_WIN32) && !defined(__CYGWIN__)
1375 # if defined(WINAPI_FAMILY_PARTITION) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
1376 # define GetOEMCP() CP_OEMCP
1377 # endif
1378
1379 static unsigned
my_atoi(const char * p)1380 my_atoi(const char *p)
1381 {
1382 unsigned cp;
1383
1384 cp = 0;
1385 while (*p) {
1386 if (*p >= '0' && *p <= '9')
1387 cp = cp * 10 + (*p - '0');
1388 else
1389 return (-1);
1390 p++;
1391 }
1392 return (cp);
1393 }
1394
1395 /*
1396 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1397 * Return -1 if failed.
1398 *
1399 * Note: This translation code may be insufficient.
1400 */
1401 static struct charset {
1402 const char *name;
1403 unsigned cp;
1404 } charsets[] = {
1405 /* MUST BE SORTED! */
1406 {"ASCII", 1252},
1407 {"ASMO-708", 708},
1408 {"BIG5", 950},
1409 {"CHINESE", 936},
1410 {"CP367", 1252},
1411 {"CP819", 1252},
1412 {"CP1025", 21025},
1413 {"DOS-720", 720},
1414 {"DOS-862", 862},
1415 {"EUC-CN", 51936},
1416 {"EUC-JP", 51932},
1417 {"EUC-KR", 949},
1418 {"EUCCN", 51936},
1419 {"EUCJP", 51932},
1420 {"EUCKR", 949},
1421 {"GB18030", 54936},
1422 {"GB2312", 936},
1423 {"HEBREW", 1255},
1424 {"HZ-GB-2312", 52936},
1425 {"IBM273", 20273},
1426 {"IBM277", 20277},
1427 {"IBM278", 20278},
1428 {"IBM280", 20280},
1429 {"IBM284", 20284},
1430 {"IBM285", 20285},
1431 {"IBM290", 20290},
1432 {"IBM297", 20297},
1433 {"IBM367", 1252},
1434 {"IBM420", 20420},
1435 {"IBM423", 20423},
1436 {"IBM424", 20424},
1437 {"IBM819", 1252},
1438 {"IBM871", 20871},
1439 {"IBM880", 20880},
1440 {"IBM905", 20905},
1441 {"IBM924", 20924},
1442 {"ISO-8859-1", 28591},
1443 {"ISO-8859-13", 28603},
1444 {"ISO-8859-15", 28605},
1445 {"ISO-8859-2", 28592},
1446 {"ISO-8859-3", 28593},
1447 {"ISO-8859-4", 28594},
1448 {"ISO-8859-5", 28595},
1449 {"ISO-8859-6", 28596},
1450 {"ISO-8859-7", 28597},
1451 {"ISO-8859-8", 28598},
1452 {"ISO-8859-9", 28599},
1453 {"ISO8859-1", 28591},
1454 {"ISO8859-13", 28603},
1455 {"ISO8859-15", 28605},
1456 {"ISO8859-2", 28592},
1457 {"ISO8859-3", 28593},
1458 {"ISO8859-4", 28594},
1459 {"ISO8859-5", 28595},
1460 {"ISO8859-6", 28596},
1461 {"ISO8859-7", 28597},
1462 {"ISO8859-8", 28598},
1463 {"ISO8859-9", 28599},
1464 {"JOHAB", 1361},
1465 {"KOI8-R", 20866},
1466 {"KOI8-U", 21866},
1467 {"KS_C_5601-1987", 949},
1468 {"LATIN1", 1252},
1469 {"LATIN2", 28592},
1470 {"MACINTOSH", 10000},
1471 {"SHIFT-JIS", 932},
1472 {"SHIFT_JIS", 932},
1473 {"SJIS", 932},
1474 {"US", 1252},
1475 {"US-ASCII", 1252},
1476 {"UTF-16", 1200},
1477 {"UTF-16BE", 1201},
1478 {"UTF-16LE", 1200},
1479 {"UTF-8", CP_UTF8},
1480 {"X-EUROPA", 29001},
1481 {"X-MAC-ARABIC", 10004},
1482 {"X-MAC-CE", 10029},
1483 {"X-MAC-CHINESEIMP", 10008},
1484 {"X-MAC-CHINESETRAD", 10002},
1485 {"X-MAC-CROATIAN", 10082},
1486 {"X-MAC-CYRILLIC", 10007},
1487 {"X-MAC-GREEK", 10006},
1488 {"X-MAC-HEBREW", 10005},
1489 {"X-MAC-ICELANDIC", 10079},
1490 {"X-MAC-JAPANESE", 10001},
1491 {"X-MAC-KOREAN", 10003},
1492 {"X-MAC-ROMANIAN", 10010},
1493 {"X-MAC-THAI", 10021},
1494 {"X-MAC-TURKISH", 10081},
1495 {"X-MAC-UKRAINIAN", 10017},
1496 };
1497 static unsigned
make_codepage_from_charset(const char * charset)1498 make_codepage_from_charset(const char *charset)
1499 {
1500 char cs[16];
1501 char *p;
1502 unsigned cp;
1503 int a, b;
1504
1505 if (charset == NULL || strlen(charset) > 15)
1506 return -1;
1507
1508 /* Copy name to uppercase. */
1509 p = cs;
1510 while (*charset) {
1511 char c = *charset++;
1512 if (c >= 'a' && c <= 'z')
1513 c -= 'a' - 'A';
1514 *p++ = c;
1515 }
1516 *p++ = '\0';
1517 cp = -1;
1518
1519 /* Look it up in the table first, so that we can easily
1520 * override CP367, which we map to 1252 instead of 367. */
1521 a = 0;
1522 b = sizeof(charsets)/sizeof(charsets[0]);
1523 while (b > a) {
1524 int c = (b + a) / 2;
1525 int r = strcmp(charsets[c].name, cs);
1526 if (r < 0)
1527 a = c + 1;
1528 else if (r > 0)
1529 b = c;
1530 else
1531 return charsets[c].cp;
1532 }
1533
1534 /* If it's not in the table, try to parse it. */
1535 switch (*cs) {
1536 case 'C':
1537 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1538 cp = my_atoi(cs + 2);
1539 } else if (strcmp(cs, "CP_ACP") == 0)
1540 cp = get_current_codepage();
1541 else if (strcmp(cs, "CP_OEMCP") == 0)
1542 cp = get_current_oemcp();
1543 break;
1544 case 'I':
1545 if (cs[1] == 'B' && cs[2] == 'M' &&
1546 cs[3] >= '0' && cs[3] <= '9') {
1547 cp = my_atoi(cs + 3);
1548 }
1549 break;
1550 case 'W':
1551 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1552 cp = my_atoi(cs + 8);
1553 if (cp != 874 && (cp < 1250 || cp > 1258))
1554 cp = -1;/* This may invalid code. */
1555 }
1556 break;
1557 }
1558 return (cp);
1559 }
1560
1561 /*
1562 * Return ANSI Code Page of current locale set by setlocale().
1563 */
1564 static unsigned
get_current_codepage(void)1565 get_current_codepage(void)
1566 {
1567 char *locale, *p;
1568 unsigned cp;
1569
1570 locale = setlocale(LC_CTYPE, NULL);
1571 if (locale == NULL)
1572 return (GetACP());
1573 if (locale[0] == 'C' && locale[1] == '\0')
1574 return (CP_C_LOCALE);
1575 p = strrchr(locale, '.');
1576 if (p == NULL)
1577 return (GetACP());
1578 if ((strcmp(p+1, "utf8") == 0) || (strcmp(p+1, "UTF-8") == 0))
1579 return CP_UTF8;
1580 cp = my_atoi(p+1);
1581 if ((int)cp <= 0)
1582 return (GetACP());
1583 return (cp);
1584 }
1585
1586 /*
1587 * Translation table between Locale Name and ACP/OEMCP.
1588 */
1589 static struct {
1590 unsigned acp;
1591 unsigned ocp;
1592 const char *locale;
1593 } acp_ocp_map[] = {
1594 { 950, 950, "Chinese_Taiwan" },
1595 { 936, 936, "Chinese_People's Republic of China" },
1596 { 950, 950, "Chinese_Taiwan" },
1597 { 1250, 852, "Czech_Czech Republic" },
1598 { 1252, 850, "Danish_Denmark" },
1599 { 1252, 850, "Dutch_Netherlands" },
1600 { 1252, 850, "Dutch_Belgium" },
1601 { 1252, 437, "English_United States" },
1602 { 1252, 850, "English_Australia" },
1603 { 1252, 850, "English_Canada" },
1604 { 1252, 850, "English_New Zealand" },
1605 { 1252, 850, "English_United Kingdom" },
1606 { 1252, 437, "English_United States" },
1607 { 1252, 850, "Finnish_Finland" },
1608 { 1252, 850, "French_France" },
1609 { 1252, 850, "French_Belgium" },
1610 { 1252, 850, "French_Canada" },
1611 { 1252, 850, "French_Switzerland" },
1612 { 1252, 850, "German_Germany" },
1613 { 1252, 850, "German_Austria" },
1614 { 1252, 850, "German_Switzerland" },
1615 { 1253, 737, "Greek_Greece" },
1616 { 1250, 852, "Hungarian_Hungary" },
1617 { 1252, 850, "Icelandic_Iceland" },
1618 { 1252, 850, "Italian_Italy" },
1619 { 1252, 850, "Italian_Switzerland" },
1620 { 932, 932, "Japanese_Japan" },
1621 { 949, 949, "Korean_Korea" },
1622 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1623 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1624 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1625 { 1250, 852, "Polish_Poland" },
1626 { 1252, 850, "Portuguese_Portugal" },
1627 { 1252, 850, "Portuguese_Brazil" },
1628 { 1251, 866, "Russian_Russia" },
1629 { 1250, 852, "Slovak_Slovakia" },
1630 { 1252, 850, "Spanish_Spain" },
1631 { 1252, 850, "Spanish_Mexico" },
1632 { 1252, 850, "Spanish_Spain" },
1633 { 1252, 850, "Swedish_Sweden" },
1634 { 1254, 857, "Turkish_Turkey" },
1635 { 0, 0, NULL}
1636 };
1637
1638 /*
1639 * Return OEM Code Page of current locale set by setlocale().
1640 */
1641 static unsigned
get_current_oemcp(void)1642 get_current_oemcp(void)
1643 {
1644 int i;
1645 char *locale, *p;
1646 size_t len;
1647
1648 locale = setlocale(LC_CTYPE, NULL);
1649 if (locale == NULL)
1650 return (GetOEMCP());
1651 if (locale[0] == 'C' && locale[1] == '\0')
1652 return (CP_C_LOCALE);
1653
1654 p = strrchr(locale, '.');
1655 if (p == NULL)
1656 return (GetOEMCP());
1657 len = p - locale;
1658 for (i = 0; acp_ocp_map[i].acp; i++) {
1659 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1660 return (acp_ocp_map[i].ocp);
1661 }
1662 return (GetOEMCP());
1663 }
1664 #else
1665
1666 /*
1667 * POSIX platform does not use CodePage.
1668 */
1669
1670 static unsigned
get_current_codepage(void)1671 get_current_codepage(void)
1672 {
1673 return (-1);/* Unknown */
1674 }
1675 static unsigned
make_codepage_from_charset(const char * charset)1676 make_codepage_from_charset(const char *charset)
1677 {
1678 (void)charset; /* UNUSED */
1679 return (-1);/* Unknown */
1680 }
1681 static unsigned
get_current_oemcp(void)1682 get_current_oemcp(void)
1683 {
1684 return (-1);/* Unknown */
1685 }
1686
1687 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1688
1689 /*
1690 * Return a string conversion object.
1691 */
1692 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1693 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1694 {
1695 struct archive_string_conv *sc;
1696 unsigned current_codepage;
1697
1698 /* Check if we have made the sconv object. */
1699 sc = find_sconv_object(a, fc, tc);
1700 if (sc != NULL)
1701 return (sc);
1702
1703 if (a == NULL)
1704 current_codepage = get_current_codepage();
1705 else
1706 current_codepage = a->current_codepage;
1707
1708 sc = create_sconv_object(canonical_charset_name(fc),
1709 canonical_charset_name(tc), current_codepage, flag);
1710 if (sc == NULL) {
1711 if (a != NULL)
1712 archive_set_error(a, ENOMEM,
1713 "Could not allocate memory for "
1714 "a string conversion object");
1715 return (NULL);
1716 }
1717
1718 /*
1719 * If there is no converter for current string conversion object,
1720 * we cannot handle this conversion.
1721 */
1722 if (sc->nconverter == 0) {
1723 if (a != NULL) {
1724 #if HAVE_ICONV
1725 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1726 "iconv_open failed: Cannot handle ``%s''",
1727 (flag & SCONV_TO_CHARSET)?tc:fc);
1728 #else
1729 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1730 "A character-set conversion not fully supported "
1731 "on this platform");
1732 #endif
1733 }
1734 /* Failed; free a sconv object. */
1735 free_sconv_object(sc);
1736 return (NULL);
1737 }
1738
1739 /*
1740 * Success!
1741 */
1742 if (a != NULL)
1743 add_sconv_object(a, sc);
1744 return (sc);
1745 }
1746
1747 static const char *
get_current_charset(struct archive * a)1748 get_current_charset(struct archive *a)
1749 {
1750 const char *cur_charset;
1751
1752 if (a == NULL)
1753 cur_charset = default_iconv_charset("");
1754 else {
1755 cur_charset = default_iconv_charset(a->current_code);
1756 if (a->current_code == NULL) {
1757 a->current_code = strdup(cur_charset);
1758 a->current_codepage = get_current_codepage();
1759 a->current_oemcp = get_current_oemcp();
1760 }
1761 }
1762 return (cur_charset);
1763 }
1764
1765 /*
1766 * Make and Return a string conversion object.
1767 * Return NULL if the platform does not support the specified conversion
1768 * and best_effort is 0.
1769 * If best_effort is set, A string conversion object must be returned
1770 * unless memory allocation for the object fails, but the conversion
1771 * might fail when non-ASCII code is found.
1772 */
1773 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1774 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1775 int best_effort)
1776 {
1777 int flag = SCONV_TO_CHARSET;
1778
1779 if (best_effort)
1780 flag |= SCONV_BEST_EFFORT;
1781 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1782 }
1783
1784 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1785 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1786 int best_effort)
1787 {
1788 int flag = SCONV_FROM_CHARSET;
1789
1790 if (best_effort)
1791 flag |= SCONV_BEST_EFFORT;
1792 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1793 }
1794
1795 /*
1796 * archive_string_default_conversion_*_archive() are provided for Windows
1797 * platform because other archiver application use CP_OEMCP for
1798 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1799 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1800 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1801 * So we should make a string conversion between CP_ACP and CP_OEMCP
1802 * for compatibility.
1803 */
1804 #if defined(_WIN32) && !defined(__CYGWIN__)
1805 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1806 archive_string_default_conversion_for_read(struct archive *a)
1807 {
1808 const char *cur_charset = get_current_charset(a);
1809 char oemcp[16];
1810
1811 /* NOTE: a check of cur_charset is unneeded but we need
1812 * that get_current_charset() has been surely called at
1813 * this time whatever C compiler optimized. */
1814 if (cur_charset != NULL &&
1815 (a->current_codepage == CP_C_LOCALE ||
1816 a->current_codepage == a->current_oemcp))
1817 return (NULL);/* no conversion. */
1818
1819 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1820 /* Make sure a null termination must be set. */
1821 oemcp[sizeof(oemcp)-1] = '\0';
1822 return (get_sconv_object(a, oemcp, cur_charset,
1823 SCONV_FROM_CHARSET));
1824 }
1825
1826 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1827 archive_string_default_conversion_for_write(struct archive *a)
1828 {
1829 const char *cur_charset = get_current_charset(a);
1830 char oemcp[16];
1831
1832 /* NOTE: a check of cur_charset is unneeded but we need
1833 * that get_current_charset() has been surely called at
1834 * this time whatever C compiler optimized. */
1835 if (cur_charset != NULL &&
1836 (a->current_codepage == CP_C_LOCALE ||
1837 a->current_codepage == a->current_oemcp))
1838 return (NULL);/* no conversion. */
1839
1840 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1841 /* Make sure a null termination must be set. */
1842 oemcp[sizeof(oemcp)-1] = '\0';
1843 return (get_sconv_object(a, cur_charset, oemcp,
1844 SCONV_TO_CHARSET));
1845 }
1846 #else
1847 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1848 archive_string_default_conversion_for_read(struct archive *a)
1849 {
1850 (void)a; /* UNUSED */
1851 return (NULL);
1852 }
1853
1854 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1855 archive_string_default_conversion_for_write(struct archive *a)
1856 {
1857 (void)a; /* UNUSED */
1858 return (NULL);
1859 }
1860 #endif
1861
1862 /*
1863 * Dispose of all character conversion objects in the archive object.
1864 */
1865 void
archive_string_conversion_free(struct archive * a)1866 archive_string_conversion_free(struct archive *a)
1867 {
1868 struct archive_string_conv *sc;
1869 struct archive_string_conv *sc_next;
1870
1871 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1872 sc_next = sc->next;
1873 free_sconv_object(sc);
1874 }
1875 a->sconv = NULL;
1876 free(a->current_code);
1877 a->current_code = NULL;
1878 }
1879
1880 /*
1881 * Return a conversion charset name.
1882 */
1883 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1884 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1885 {
1886 if (sc == NULL) {
1887 return "current locale";
1888 }
1889 if (sc->flag & SCONV_TO_CHARSET)
1890 return (sc->to_charset);
1891 else
1892 return (sc->from_charset);
1893 }
1894
1895 /*
1896 * Change the behavior of a string conversion.
1897 */
1898 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1899 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1900 {
1901 switch (opt) {
1902 /*
1903 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1904 * assumption that wchar_t was Unicode.
1905 * This option enables simulating the assumption in order to read
1906 * that filename correctly.
1907 */
1908 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1909 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1910 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1911 /*
1912 * Nothing to do for it since wchar_t on these platforms
1913 * is really Unicode.
1914 */
1915 (void)sc; /* UNUSED */
1916 #else
1917 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1918 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1919 /* Set up string converters. */
1920 setup_converter(sc);
1921 }
1922 #endif
1923 break;
1924 case SCONV_SET_OPT_NORMALIZATION_C:
1925 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1926 sc->flag |= SCONV_NORMALIZATION_C;
1927 sc->flag &= ~SCONV_NORMALIZATION_D;
1928 /* Set up string converters. */
1929 setup_converter(sc);
1930 }
1931 break;
1932 case SCONV_SET_OPT_NORMALIZATION_D:
1933 #if defined(HAVE_ICONV)
1934 /*
1935 * If iconv will take the string, do not change the
1936 * setting of the normalization.
1937 */
1938 if (!(sc->flag & SCONV_WIN_CP) &&
1939 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1940 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1941 break;
1942 #endif
1943 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1944 sc->flag |= SCONV_NORMALIZATION_D;
1945 sc->flag &= ~SCONV_NORMALIZATION_C;
1946 /* Set up string converters. */
1947 setup_converter(sc);
1948 }
1949 break;
1950 default:
1951 break;
1952 }
1953 }
1954
1955 /*
1956 *
1957 * Copy one archive_string to another in locale conversion.
1958 *
1959 * archive_strncat_l();
1960 * archive_strncpy_l();
1961 *
1962 */
1963
1964 static size_t
mbsnbytes(const void * _p,size_t n)1965 mbsnbytes(const void *_p, size_t n)
1966 {
1967 size_t s;
1968 const char *p, *pp;
1969
1970 if (_p == NULL)
1971 return (0);
1972 p = (const char *)_p;
1973
1974 /* Like strlen(p), except won't examine positions beyond p[n]. */
1975 s = 0;
1976 pp = p;
1977 while (s < n && *pp) {
1978 pp++;
1979 s++;
1980 }
1981 return (s);
1982 }
1983
1984 static size_t
utf16nbytes(const void * _p,size_t n)1985 utf16nbytes(const void *_p, size_t n)
1986 {
1987 size_t s;
1988 const char *p, *pp;
1989
1990 if (_p == NULL)
1991 return (0);
1992 p = (const char *)_p;
1993
1994 /* Like strlen(p), except won't examine positions beyond p[n]. */
1995 s = 0;
1996 pp = p;
1997 n >>= 1;
1998 while (s < n && (pp[0] || pp[1])) {
1999 pp += 2;
2000 s++;
2001 }
2002 return (s<<1);
2003 }
2004
2005 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2006 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
2007 struct archive_string_conv *sc)
2008 {
2009 as->length = 0;
2010 return (archive_strncat_l(as, _p, n, sc));
2011 }
2012
2013 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2014 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
2015 struct archive_string_conv *sc)
2016 {
2017 const void *s;
2018 size_t length = 0;
2019 int i, r = 0, r2;
2020
2021 if (_p != NULL && n > 0) {
2022 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
2023 length = utf16nbytes(_p, n);
2024 else
2025 length = mbsnbytes(_p, n);
2026 }
2027
2028 /* We must allocate memory even if there is no data for conversion
2029 * or copy. This simulates archive_string_append behavior. */
2030 if (length == 0) {
2031 size_t tn = 1;
2032 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
2033 tn = 2;
2034 if (archive_string_ensure(as, as->length + tn) == NULL)
2035 return (-1);
2036 as->s[as->length] = 0;
2037 if (tn == 2)
2038 as->s[as->length+1] = 0;
2039 return (0);
2040 }
2041
2042 /*
2043 * If sc is NULL, we just make a copy.
2044 */
2045 if (sc == NULL) {
2046 if (archive_string_append(as, _p, length) == NULL)
2047 return (-1);/* No memory */
2048 return (0);
2049 }
2050
2051 s = _p;
2052 i = 0;
2053 if (sc->nconverter > 1) {
2054 sc->utftmp.length = 0;
2055 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2056 if (r2 != 0 && errno == ENOMEM)
2057 return (r2);
2058 if (r > r2)
2059 r = r2;
2060 s = sc->utftmp.s;
2061 length = sc->utftmp.length;
2062 ++i;
2063 }
2064 r2 = sc->converter[i](as, s, length, sc);
2065 if (r > r2)
2066 r = r2;
2067 return (r);
2068 }
2069
2070 struct archive_string *
archive_string_dirname(struct archive_string * as)2071 archive_string_dirname(struct archive_string *as)
2072 {
2073 /* strip trailing separators */
2074 while (as->length > 1 && as->s[as->length - 1] == '/')
2075 as->length--;
2076 /* strip final component */
2077 while (as->length > 0 && as->s[as->length - 1] != '/')
2078 as->length--;
2079 /* empty path -> cwd */
2080 if (as->length == 0)
2081 return (archive_strcat(as, "."));
2082 /* strip separator(s) */
2083 while (as->length > 1 && as->s[as->length - 1] == '/')
2084 as->length--;
2085 /* terminate */
2086 as->s[as->length] = '\0';
2087 return (as);
2088 }
2089
2090 #if HAVE_ICONV
2091
2092 /*
2093 * Return -1 if conversion fails.
2094 */
2095 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2096 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2097 size_t length, struct archive_string_conv *sc)
2098 {
2099 ICONV_CONST char *itp;
2100 size_t remaining;
2101 iconv_t cd;
2102 char *outp;
2103 size_t avail, bs;
2104 int return_value = 0; /* success */
2105 size_t to_size, from_size;
2106
2107 if (sc->flag & SCONV_TO_UTF16)
2108 to_size = 2;
2109 else
2110 to_size = 1;
2111 if (sc->flag & SCONV_FROM_UTF16)
2112 from_size = 2;
2113 else
2114 from_size = 1;
2115
2116 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2117 return (-1);
2118
2119 cd = sc->cd;
2120 itp = (char *)(uintptr_t)_p;
2121 remaining = length;
2122 outp = as->s + as->length;
2123 avail = as->buffer_length - as->length - to_size;
2124 while (remaining >= from_size) {
2125 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2126
2127 if (result != (size_t)-1)
2128 break; /* Conversion completed. */
2129
2130 if (errno == EILSEQ || errno == EINVAL) {
2131 /*
2132 * If an output charset is UTF-8 or UTF-16BE/LE,
2133 * unknown character should be U+FFFD
2134 * (replacement character).
2135 */
2136 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2137 size_t rbytes;
2138 if (sc->flag & SCONV_TO_UTF8)
2139 rbytes = sizeof(utf8_replacement_char);
2140 else
2141 rbytes = 2;
2142
2143 if (avail < rbytes) {
2144 as->length = outp - as->s;
2145 bs = as->buffer_length +
2146 (remaining * to_size) + rbytes;
2147 if (NULL ==
2148 archive_string_ensure(as, bs))
2149 return (-1);
2150 outp = as->s + as->length;
2151 avail = as->buffer_length
2152 - as->length - to_size;
2153 }
2154 if (sc->flag & SCONV_TO_UTF8)
2155 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2156 else if (sc->flag & SCONV_TO_UTF16BE)
2157 archive_be16enc(outp, UNICODE_R_CHAR);
2158 else
2159 archive_le16enc(outp, UNICODE_R_CHAR);
2160 outp += rbytes;
2161 avail -= rbytes;
2162 } else {
2163 /* Skip the illegal input bytes. */
2164 *outp++ = '?';
2165 avail--;
2166 }
2167 itp += from_size;
2168 remaining -= from_size;
2169 return_value = -1; /* failure */
2170 } else {
2171 /* E2BIG no output buffer,
2172 * Increase an output buffer. */
2173 as->length = outp - as->s;
2174 bs = as->buffer_length + remaining * 2;
2175 if (NULL == archive_string_ensure(as, bs))
2176 return (-1);
2177 outp = as->s + as->length;
2178 avail = as->buffer_length - as->length - to_size;
2179 }
2180 }
2181 as->length = outp - as->s;
2182 as->s[as->length] = 0;
2183 if (to_size == 2)
2184 as->s[as->length+1] = 0;
2185 return (return_value);
2186 }
2187
2188 #endif /* HAVE_ICONV */
2189
2190
2191 #if defined(_WIN32) && !defined(__CYGWIN__)
2192
2193 /*
2194 * Translate a string from a some CodePage to an another CodePage by
2195 * Windows APIs, and copy the result. Return -1 if conversion fails.
2196 */
2197 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2198 strncat_in_codepage(struct archive_string *as,
2199 const void *_p, size_t length, struct archive_string_conv *sc)
2200 {
2201 const char *s = (const char *)_p;
2202 struct archive_wstring aws;
2203 size_t l;
2204 int r, saved_flag;
2205
2206 archive_string_init(&aws);
2207 saved_flag = sc->flag;
2208 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2209 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2210 sc->flag = saved_flag;
2211 if (r != 0) {
2212 archive_wstring_free(&aws);
2213 if (errno != ENOMEM)
2214 archive_string_append(as, s, length);
2215 return (-1);
2216 }
2217
2218 l = as->length;
2219 r = archive_string_append_from_wcs_in_codepage(
2220 as, aws.s, aws.length, sc);
2221 if (r != 0 && errno != ENOMEM && l == as->length)
2222 archive_string_append(as, s, length);
2223 archive_wstring_free(&aws);
2224 return (r);
2225 }
2226
2227 /*
2228 * Test whether MBS ==> WCS is okay.
2229 */
2230 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2231 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2232 {
2233 const char *p = (const char *)_p;
2234 unsigned codepage;
2235 DWORD mbflag = MB_ERR_INVALID_CHARS;
2236
2237 if (sc->flag & SCONV_FROM_CHARSET)
2238 codepage = sc->to_cp;
2239 else
2240 codepage = sc->from_cp;
2241
2242 if (codepage == CP_C_LOCALE)
2243 return (0);
2244 if (codepage != CP_UTF8)
2245 mbflag |= MB_PRECOMPOSED;
2246
2247 if (n > (size_t)INT_MAX)
2248 return (-1); /* Invalid */
2249 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2250 return (-1); /* Invalid */
2251 return (0); /* Okay */
2252 }
2253
2254 #else
2255
2256 /*
2257 * Test whether MBS ==> WCS is okay.
2258 */
2259 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2260 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2261 {
2262 const char *p = (const char *)_p;
2263 size_t r;
2264
2265 #if HAVE_MBRTOWC
2266 mbstate_t shift_state;
2267
2268 memset(&shift_state, 0, sizeof(shift_state));
2269 #else
2270 /* Clear the shift state before starting. */
2271 mbtowc(NULL, NULL, 0);
2272 #endif
2273 while (n) {
2274 wchar_t wc;
2275
2276 #if HAVE_MBRTOWC
2277 r = mbrtowc(&wc, p, n, &shift_state);
2278 #else
2279 r = mbtowc(&wc, p, n);
2280 #endif
2281 if (r == (size_t)-1 || r == (size_t)-2)
2282 return (-1);/* Invalid. */
2283 if (r == 0)
2284 break;
2285 p += r;
2286 n -= r;
2287 }
2288 (void)sc; /* UNUSED */
2289 return (0); /* All Okey. */
2290 }
2291
2292 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2293
2294 /*
2295 * Basically returns -1 because we cannot make a conversion of charset
2296 * without iconv but in some cases this would return 0.
2297 * Returns 0 if all copied characters are ASCII.
2298 * Returns 0 if both from-locale and to-locale are the same and those
2299 * can be WCS with no error.
2300 */
2301 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2302 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2303 size_t length, struct archive_string_conv *sc)
2304 {
2305 size_t remaining;
2306 const uint8_t *itp;
2307 int return_value = 0; /* success */
2308
2309 /*
2310 * If both from-locale and to-locale is the same, this makes a copy.
2311 * And then this checks all copied MBS can be WCS if so returns 0.
2312 */
2313 if (sc->same) {
2314 if (archive_string_append(as, _p, length) == NULL)
2315 return (-1);/* No memory */
2316 return (invalid_mbs(_p, length, sc));
2317 }
2318
2319 /*
2320 * If a character is ASCII, this just copies it. If not, this
2321 * assigns '?' character instead but in UTF-8 locale this assigns
2322 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2323 * a Replacement Character in Unicode.
2324 */
2325
2326 remaining = length;
2327 itp = (const uint8_t *)_p;
2328 while (*itp && remaining > 0) {
2329 if (*itp > 127) {
2330 // Non-ASCII: Substitute with suitable replacement
2331 if (sc->flag & SCONV_TO_UTF8) {
2332 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2333 __archive_errx(1, "Out of memory");
2334 }
2335 } else {
2336 archive_strappend_char(as, '?');
2337 }
2338 return_value = -1;
2339 } else {
2340 archive_strappend_char(as, *itp);
2341 }
2342 ++itp;
2343 }
2344 return (return_value);
2345 }
2346
2347
2348 /*
2349 * Unicode conversion functions.
2350 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2351 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2352 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2353 * - UTF-16BE <===> UTF-8.
2354 *
2355 */
2356
2357 /*
2358 * Utility to convert a single UTF-8 sequence.
2359 *
2360 * Usually return used bytes, return used byte in negative value when
2361 * a unicode character is replaced with U+FFFD.
2362 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2363 * Recommended Practice for Replacement Characters.
2364 */
2365 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2366 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2367 {
2368 static const char utf8_count[256] = {
2369 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2370 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2371 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2372 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2373 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2374 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2375 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2376 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2377 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2378 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2379 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2380 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2381 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2382 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2383 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2384 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2385 };
2386 int ch, i;
2387 int cnt;
2388 uint32_t wc;
2389
2390 /* Sanity check. */
2391 if (n == 0)
2392 return (0);
2393 /*
2394 * Decode 1-4 bytes depending on the value of the first byte.
2395 */
2396 ch = (unsigned char)*s;
2397 if (ch == 0)
2398 return (0); /* Standard: return 0 for end-of-string. */
2399 cnt = utf8_count[ch];
2400
2401 /* Invalid sequence or there are not plenty bytes. */
2402 if (n < (size_t)cnt) {
2403 cnt = (int)n;
2404 for (i = 1; i < cnt; i++) {
2405 if ((s[i] & 0xc0) != 0x80) {
2406 cnt = i;
2407 break;
2408 }
2409 }
2410 goto invalid_sequence;
2411 }
2412
2413 /* Make a Unicode code point from a single UTF-8 sequence. */
2414 switch (cnt) {
2415 case 1: /* 1 byte sequence. */
2416 *pwc = ch & 0x7f;
2417 return (cnt);
2418 case 2: /* 2 bytes sequence. */
2419 if ((s[1] & 0xc0) != 0x80) {
2420 cnt = 1;
2421 goto invalid_sequence;
2422 }
2423 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2424 return (cnt);
2425 case 3: /* 3 bytes sequence. */
2426 if ((s[1] & 0xc0) != 0x80) {
2427 cnt = 1;
2428 goto invalid_sequence;
2429 }
2430 if ((s[2] & 0xc0) != 0x80) {
2431 cnt = 2;
2432 goto invalid_sequence;
2433 }
2434 wc = ((ch & 0x0f) << 12)
2435 | ((s[1] & 0x3f) << 6)
2436 | (s[2] & 0x3f);
2437 if (wc < 0x800)
2438 goto invalid_sequence;/* Overlong sequence. */
2439 break;
2440 case 4: /* 4 bytes sequence. */
2441 if ((s[1] & 0xc0) != 0x80) {
2442 cnt = 1;
2443 goto invalid_sequence;
2444 }
2445 if ((s[2] & 0xc0) != 0x80) {
2446 cnt = 2;
2447 goto invalid_sequence;
2448 }
2449 if ((s[3] & 0xc0) != 0x80) {
2450 cnt = 3;
2451 goto invalid_sequence;
2452 }
2453 wc = ((ch & 0x07) << 18)
2454 | ((s[1] & 0x3f) << 12)
2455 | ((s[2] & 0x3f) << 6)
2456 | (s[3] & 0x3f);
2457 if (wc < 0x10000)
2458 goto invalid_sequence;/* Overlong sequence. */
2459 break;
2460 default: /* Others are all invalid sequence. */
2461 if (ch == 0xc0 || ch == 0xc1)
2462 cnt = 2;
2463 else if (ch >= 0xf5 && ch <= 0xf7)
2464 cnt = 4;
2465 else if (ch >= 0xf8 && ch <= 0xfb)
2466 cnt = 5;
2467 else if (ch == 0xfc || ch == 0xfd)
2468 cnt = 6;
2469 else
2470 cnt = 1;
2471 if (n < (size_t)cnt)
2472 cnt = (int)n;
2473 for (i = 1; i < cnt; i++) {
2474 if ((s[i] & 0xc0) != 0x80) {
2475 cnt = i;
2476 break;
2477 }
2478 }
2479 goto invalid_sequence;
2480 }
2481
2482 /* The code point larger than 0x10FFFF is not legal
2483 * Unicode values. */
2484 if (wc > UNICODE_MAX)
2485 goto invalid_sequence;
2486 /* Correctly gets a Unicode, returns used bytes. */
2487 *pwc = wc;
2488 return (cnt);
2489 invalid_sequence:
2490 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2491 return (cnt * -1);
2492 }
2493
2494 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2495 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2496 {
2497 int cnt;
2498
2499 cnt = _utf8_to_unicode(pwc, s, n);
2500 /* Any of Surrogate pair is not legal Unicode values. */
2501 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2502 return (-3);
2503 return (cnt);
2504 }
2505
2506 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2507 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2508 {
2509 uc -= 0xD800;
2510 uc *= 0x400;
2511 uc += uc2 - 0xDC00;
2512 uc += 0x10000;
2513 return (uc);
2514 }
2515
2516 /*
2517 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2518 * removing surrogate pairs.
2519 *
2520 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2521 *
2522 * Usually return used bytes, return used byte in negative value when
2523 * a unicode character is replaced with U+FFFD.
2524 */
2525 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2526 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2527 {
2528 uint32_t wc = 0;
2529 int cnt;
2530
2531 cnt = _utf8_to_unicode(&wc, s, n);
2532 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2533 uint32_t wc2 = 0;
2534 if (n - 3 < 3) {
2535 /* Invalid byte sequence. */
2536 goto invalid_sequence;
2537 }
2538 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2539 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2540 /* Invalid byte sequence. */
2541 goto invalid_sequence;
2542 }
2543 wc = combine_surrogate_pair(wc, wc2);
2544 cnt = 6;
2545 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2546 /* Invalid byte sequence. */
2547 goto invalid_sequence;
2548 }
2549 *pwc = wc;
2550 return (cnt);
2551 invalid_sequence:
2552 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2553 if (cnt > 0)
2554 cnt *= -1;
2555 return (cnt);
2556 }
2557
2558 /*
2559 * Convert a Unicode code point to a single UTF-8 sequence.
2560 *
2561 * NOTE:This function does not check if the Unicode is legal or not.
2562 * Please you definitely check it before calling this.
2563 */
2564 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2565 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2566 {
2567 char *_p = p;
2568
2569 /* Invalid Unicode char maps to Replacement character */
2570 if (uc > UNICODE_MAX)
2571 uc = UNICODE_R_CHAR;
2572 /* Translate code point to UTF8 */
2573 if (uc <= 0x7f) {
2574 if (remaining == 0)
2575 return (0);
2576 *p++ = (char)uc;
2577 } else if (uc <= 0x7ff) {
2578 if (remaining < 2)
2579 return (0);
2580 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2581 *p++ = 0x80 | (uc & 0x3f);
2582 } else if (uc <= 0xffff) {
2583 if (remaining < 3)
2584 return (0);
2585 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2586 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2587 *p++ = 0x80 | (uc & 0x3f);
2588 } else {
2589 if (remaining < 4)
2590 return (0);
2591 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2592 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2593 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2594 *p++ = 0x80 | (uc & 0x3f);
2595 }
2596 return (p - _p);
2597 }
2598
2599 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2600 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2601 {
2602 return (utf16_to_unicode(pwc, s, n, 1));
2603 }
2604
2605 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2606 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2607 {
2608 return (utf16_to_unicode(pwc, s, n, 0));
2609 }
2610
2611 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2612 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2613 {
2614 const char *utf16 = s;
2615 unsigned uc;
2616
2617 if (n == 0)
2618 return (0);
2619 if (n == 1) {
2620 /* set the Replacement Character instead. */
2621 *pwc = UNICODE_R_CHAR;
2622 return (-1);
2623 }
2624
2625 if (be)
2626 uc = archive_be16dec(utf16);
2627 else
2628 uc = archive_le16dec(utf16);
2629 utf16 += 2;
2630
2631 /* If this is a surrogate pair, assemble the full code point.*/
2632 if (IS_HIGH_SURROGATE_LA(uc)) {
2633 unsigned uc2;
2634
2635 if (n >= 4) {
2636 if (be)
2637 uc2 = archive_be16dec(utf16);
2638 else
2639 uc2 = archive_le16dec(utf16);
2640 } else
2641 uc2 = 0;
2642 if (IS_LOW_SURROGATE_LA(uc2)) {
2643 uc = combine_surrogate_pair(uc, uc2);
2644 utf16 += 2;
2645 } else {
2646 /* Undescribed code point should be U+FFFD
2647 * (replacement character). */
2648 *pwc = UNICODE_R_CHAR;
2649 return (-2);
2650 }
2651 }
2652
2653 /*
2654 * Surrogate pair values(0xd800 through 0xdfff) are only
2655 * used by UTF-16, so, after above calculation, the code
2656 * must not be surrogate values, and Unicode has no codes
2657 * larger than 0x10ffff. Thus, those are not legal Unicode
2658 * values.
2659 */
2660 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2661 /* Undescribed code point should be U+FFFD
2662 * (replacement character). */
2663 *pwc = UNICODE_R_CHAR;
2664 return (((int)(utf16 - s)) * -1);
2665 }
2666 *pwc = uc;
2667 return ((int)(utf16 - s));
2668 }
2669
2670 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2671 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2672 {
2673 char *utf16 = p;
2674
2675 if (uc > 0xffff) {
2676 /* We have a code point that won't fit into a
2677 * wchar_t; convert it to a surrogate pair. */
2678 if (remaining < 4)
2679 return (0);
2680 uc -= 0x10000;
2681 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2682 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2683 return (4);
2684 } else {
2685 if (remaining < 2)
2686 return (0);
2687 archive_be16enc(utf16, (uint16_t)uc);
2688 return (2);
2689 }
2690 }
2691
2692 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2693 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2694 {
2695 char *utf16 = p;
2696
2697 if (uc > 0xffff) {
2698 /* We have a code point that won't fit into a
2699 * wchar_t; convert it to a surrogate pair. */
2700 if (remaining < 4)
2701 return (0);
2702 uc -= 0x10000;
2703 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2704 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2705 return (4);
2706 } else {
2707 if (remaining < 2)
2708 return (0);
2709 archive_le16enc(utf16, (uint16_t)uc);
2710 return (2);
2711 }
2712 }
2713
2714 /*
2715 * Append new UTF-8 string to existing UTF-8 string.
2716 * Existing string is assumed to already be in proper form;
2717 * the new string will have invalid sequences replaced and
2718 * surrogate pairs canonicalized.
2719 */
2720 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _src,size_t len,struct archive_string_conv * sc)2721 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2722 size_t len, struct archive_string_conv *sc)
2723 {
2724 int ret = 0;
2725 const char *src = _src;
2726 (void)sc; /* UNUSED */
2727
2728 /* Pre-extend the destination */
2729 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2730 return (-1);
2731
2732 /* Invariant: src points to the first UTF8 byte that hasn't
2733 * been copied to the destination `as`. */
2734 for (;;) {
2735 int n;
2736 uint32_t uc;
2737 const char *e = src;
2738
2739 /* Skip UTF-8 sequences until we reach end-of-string or
2740 * a code point that needs conversion. */
2741 while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2742 e += n;
2743 len -= n;
2744 }
2745 /* Copy the part that doesn't need conversion */
2746 if (e > src) {
2747 if (archive_string_append(as, src, e - src) == NULL)
2748 return (-1);
2749 src = e;
2750 }
2751
2752 if (n == 0) {
2753 /* We reached end-of-string */
2754 return (ret);
2755 } else {
2756 /* Next code point needs conversion */
2757 char t[4];
2758 size_t w;
2759
2760 /* Try decoding a surrogate pair */
2761 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2762 n = cesu8_to_unicode(&uc, src, len);
2763 }
2764 /* Not a (valid) surrogate, so use a replacement char */
2765 if (n < 0) {
2766 ret = -1; /* Return -1 if we used any replacement */
2767 n *= -1;
2768 }
2769 /* Consume converted code point */
2770 src += n;
2771 len -= n;
2772 /* Convert and append new UTF-8 sequence. */
2773 w = unicode_to_utf8(t, sizeof(t), uc);
2774 if (archive_string_append(as, t, w) == NULL)
2775 return (-1);
2776 }
2777 }
2778 }
2779
2780 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2781 archive_string_append_unicode(struct archive_string *as, const void *_p,
2782 size_t len, struct archive_string_conv *sc)
2783 {
2784 const char *s;
2785 char *p, *endp;
2786 uint32_t uc;
2787 size_t w;
2788 size_t ts, tm;
2789 int n, ret = 0;
2790 int (*parse)(uint32_t *, const char *, size_t);
2791 size_t (*unparse)(char *, size_t, uint32_t);
2792
2793 if (sc->flag & SCONV_TO_UTF16BE) {
2794 unparse = unicode_to_utf16be;
2795 ts = 2;
2796 } else if (sc->flag & SCONV_TO_UTF16LE) {
2797 unparse = unicode_to_utf16le;
2798 ts = 2;
2799 } else if (sc->flag & SCONV_TO_UTF8) {
2800 unparse = unicode_to_utf8;
2801 ts = 1;
2802 } else {
2803 /*
2804 * This case is going to be converted to another
2805 * character-set through iconv.
2806 */
2807 if (sc->flag & SCONV_FROM_UTF16BE) {
2808 unparse = unicode_to_utf16be;
2809 ts = 2;
2810 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2811 unparse = unicode_to_utf16le;
2812 ts = 2;
2813 } else {
2814 unparse = unicode_to_utf8;
2815 ts = 1;
2816 }
2817 }
2818
2819 if (sc->flag & SCONV_FROM_UTF16BE) {
2820 parse = utf16be_to_unicode;
2821 tm = 1;
2822 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2823 parse = utf16le_to_unicode;
2824 tm = 1;
2825 } else {
2826 parse = cesu8_to_unicode;
2827 tm = ts;
2828 }
2829
2830 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2831 return (-1);
2832
2833 s = (const char *)_p;
2834 p = as->s + as->length;
2835 endp = as->s + as->buffer_length - ts;
2836 while ((n = parse(&uc, s, len)) != 0) {
2837 if (n < 0) {
2838 /* Use a replaced unicode character. */
2839 n *= -1;
2840 ret = -1;
2841 }
2842 s += n;
2843 len -= n;
2844 while ((w = unparse(p, endp - p, uc)) == 0) {
2845 /* There is not enough output buffer so
2846 * we have to expand it. */
2847 as->length = p - as->s;
2848 if (archive_string_ensure(as,
2849 as->buffer_length + len * tm + ts) == NULL)
2850 return (-1);
2851 p = as->s + as->length;
2852 endp = as->s + as->buffer_length - ts;
2853 }
2854 p += w;
2855 }
2856 as->length = p - as->s;
2857 as->s[as->length] = '\0';
2858 if (ts == 2)
2859 as->s[as->length+1] = '\0';
2860 return (ret);
2861 }
2862
2863 /*
2864 * Following Constants for Hangul compositions this information comes from
2865 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2866 */
2867 #define HC_SBASE 0xAC00
2868 #define HC_LBASE 0x1100
2869 #define HC_VBASE 0x1161
2870 #define HC_TBASE 0x11A7
2871 #define HC_LCOUNT 19
2872 #define HC_VCOUNT 21
2873 #define HC_TCOUNT 28
2874 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2875 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2876
2877 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2878 get_nfc(uint32_t uc, uint32_t uc2)
2879 {
2880 int t, b;
2881
2882 t = 0;
2883 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2884 while (b >= t) {
2885 int m = (t + b) / 2;
2886 if (u_composition_table[m].cp1 < uc)
2887 t = m + 1;
2888 else if (u_composition_table[m].cp1 > uc)
2889 b = m - 1;
2890 else if (u_composition_table[m].cp2 < uc2)
2891 t = m + 1;
2892 else if (u_composition_table[m].cp2 > uc2)
2893 b = m - 1;
2894 else
2895 return (u_composition_table[m].nfc);
2896 }
2897 return (0);
2898 }
2899
2900 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2901 * Characters. */
2902
2903 /*
2904 * Update first code point.
2905 */
2906 #define UPDATE_UC(new_uc) do { \
2907 uc = new_uc; \
2908 ucptr = NULL; \
2909 } while (0)
2910
2911 /*
2912 * Replace first code point with second code point.
2913 */
2914 #define REPLACE_UC_WITH_UC2() do { \
2915 uc = uc2; \
2916 ucptr = uc2ptr; \
2917 n = n2; \
2918 } while (0)
2919
2920 #define EXPAND_BUFFER() do { \
2921 as->length = p - as->s; \
2922 if (archive_string_ensure(as, \
2923 as->buffer_length + len * tm + ts) == NULL)\
2924 return (-1); \
2925 p = as->s + as->length; \
2926 endp = as->s + as->buffer_length - ts; \
2927 } while (0)
2928
2929 #define UNPARSE(p, endp, uc) do { \
2930 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2931 EXPAND_BUFFER(); \
2932 } \
2933 p += w; \
2934 } while (0)
2935
2936 /*
2937 * Write first code point.
2938 * If the code point has not be changed from its original code,
2939 * this just copies it from its original buffer pointer.
2940 * If not, this converts it to UTF-8 byte sequence and copies it.
2941 */
2942 #define WRITE_UC() do { \
2943 if (ucptr) { \
2944 if (p + n > endp) \
2945 EXPAND_BUFFER(); \
2946 switch (n) { \
2947 case 4: \
2948 *p++ = *ucptr++; \
2949 /* FALL THROUGH */ \
2950 case 3: \
2951 *p++ = *ucptr++; \
2952 /* FALL THROUGH */ \
2953 case 2: \
2954 *p++ = *ucptr++; \
2955 /* FALL THROUGH */ \
2956 case 1: \
2957 *p++ = *ucptr; \
2958 break; \
2959 } \
2960 ucptr = NULL; \
2961 } else { \
2962 UNPARSE(p, endp, uc); \
2963 } \
2964 } while (0)
2965
2966 /*
2967 * Collect following decomposable code points.
2968 */
2969 #define COLLECT_CPS(start) do { \
2970 int _i; \
2971 for (_i = start; _i < FDC_MAX ; _i++) { \
2972 nx = parse(&ucx[_i], s, len); \
2973 if (nx <= 0) \
2974 break; \
2975 cx = CCC(ucx[_i]); \
2976 if (cl >= cx && cl != 228 && cx != 228)\
2977 break; \
2978 s += nx; \
2979 len -= nx; \
2980 cl = cx; \
2981 ccx[_i] = cx; \
2982 } \
2983 if (_i >= FDC_MAX) { \
2984 ret = -1; \
2985 ucx_size = FDC_MAX; \
2986 } else \
2987 ucx_size = _i; \
2988 } while (0)
2989
2990 /*
2991 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2992 *
2993 * TODO: Convert composition exclusions, which are never converted
2994 * from NFC,NFD,NFKC and NFKD, to Form C.
2995 */
2996 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2997 archive_string_normalize_C(struct archive_string *as, const void *_p,
2998 size_t len, struct archive_string_conv *sc)
2999 {
3000 const char *s = (const char *)_p;
3001 char *p, *endp;
3002 uint32_t uc, uc2;
3003 size_t w;
3004 int always_replace, n, n2, ret = 0, spair, ts, tm;
3005 int (*parse)(uint32_t *, const char *, size_t);
3006 size_t (*unparse)(char *, size_t, uint32_t);
3007
3008 always_replace = 1;
3009 ts = 1;/* text size. */
3010 if (sc->flag & SCONV_TO_UTF16BE) {
3011 unparse = unicode_to_utf16be;
3012 ts = 2;
3013 if (sc->flag & SCONV_FROM_UTF16BE)
3014 always_replace = 0;
3015 } else if (sc->flag & SCONV_TO_UTF16LE) {
3016 unparse = unicode_to_utf16le;
3017 ts = 2;
3018 if (sc->flag & SCONV_FROM_UTF16LE)
3019 always_replace = 0;
3020 } else if (sc->flag & SCONV_TO_UTF8) {
3021 unparse = unicode_to_utf8;
3022 if (sc->flag & SCONV_FROM_UTF8)
3023 always_replace = 0;
3024 } else {
3025 /*
3026 * This case is going to be converted to another
3027 * character-set through iconv.
3028 */
3029 always_replace = 0;
3030 if (sc->flag & SCONV_FROM_UTF16BE) {
3031 unparse = unicode_to_utf16be;
3032 ts = 2;
3033 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3034 unparse = unicode_to_utf16le;
3035 ts = 2;
3036 } else {
3037 unparse = unicode_to_utf8;
3038 }
3039 }
3040
3041 if (sc->flag & SCONV_FROM_UTF16BE) {
3042 parse = utf16be_to_unicode;
3043 tm = 1;
3044 spair = 4;/* surrogate pair size in UTF-16. */
3045 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3046 parse = utf16le_to_unicode;
3047 tm = 1;
3048 spair = 4;/* surrogate pair size in UTF-16. */
3049 } else {
3050 parse = cesu8_to_unicode;
3051 tm = ts;
3052 spair = 6;/* surrogate pair size in UTF-8. */
3053 }
3054
3055 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3056 return (-1);
3057
3058 p = as->s + as->length;
3059 endp = as->s + as->buffer_length - ts;
3060 while ((n = parse(&uc, s, len)) != 0) {
3061 const char *ucptr, *uc2ptr;
3062
3063 if (n < 0) {
3064 /* Use a replaced unicode character. */
3065 UNPARSE(p, endp, uc);
3066 s += n*-1;
3067 len -= n*-1;
3068 ret = -1;
3069 continue;
3070 } else if (n == spair || always_replace)
3071 /* uc is converted from a surrogate pair.
3072 * this should be treated as a changed code. */
3073 ucptr = NULL;
3074 else
3075 ucptr = s;
3076 s += n;
3077 len -= n;
3078
3079 /* Read second code point. */
3080 while ((n2 = parse(&uc2, s, len)) > 0) {
3081 uint32_t ucx[FDC_MAX];
3082 int ccx[FDC_MAX];
3083 int cl, cx, i, nx, ucx_size;
3084 int LIndex,SIndex;
3085 uint32_t nfc;
3086
3087 if (n2 == spair || always_replace)
3088 /* uc2 is converted from a surrogate pair.
3089 * this should be treated as a changed code. */
3090 uc2ptr = NULL;
3091 else
3092 uc2ptr = s;
3093 s += n2;
3094 len -= n2;
3095
3096 /*
3097 * If current second code point is out of decomposable
3098 * code points, finding compositions is unneeded.
3099 */
3100 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3101 WRITE_UC();
3102 REPLACE_UC_WITH_UC2();
3103 continue;
3104 }
3105
3106 /*
3107 * Try to combine current code points.
3108 */
3109 /*
3110 * We have to combine Hangul characters according to
3111 * http://uniicode.org/reports/tr15/#Hangul
3112 */
3113 if (0 <= (LIndex = uc - HC_LBASE) &&
3114 LIndex < HC_LCOUNT) {
3115 /*
3116 * Hangul Composition.
3117 * 1. Two current code points are L and V.
3118 */
3119 int VIndex = uc2 - HC_VBASE;
3120 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3121 /* Make syllable of form LV. */
3122 UPDATE_UC(HC_SBASE +
3123 (LIndex * HC_VCOUNT + VIndex) *
3124 HC_TCOUNT);
3125 } else {
3126 WRITE_UC();
3127 REPLACE_UC_WITH_UC2();
3128 }
3129 continue;
3130 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3131 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3132 /*
3133 * Hangul Composition.
3134 * 2. Two current code points are LV and T.
3135 */
3136 int TIndex = uc2 - HC_TBASE;
3137 if (0 < TIndex && TIndex < HC_TCOUNT) {
3138 /* Make syllable of form LVT. */
3139 UPDATE_UC(uc + TIndex);
3140 } else {
3141 WRITE_UC();
3142 REPLACE_UC_WITH_UC2();
3143 }
3144 continue;
3145 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3146 /* A composition to current code points
3147 * is found. */
3148 UPDATE_UC(nfc);
3149 continue;
3150 } else if ((cl = CCC(uc2)) == 0) {
3151 /* Clearly 'uc2' the second code point is not
3152 * a decomposable code. */
3153 WRITE_UC();
3154 REPLACE_UC_WITH_UC2();
3155 continue;
3156 }
3157
3158 /*
3159 * Collect following decomposable code points.
3160 */
3161 cx = 0;
3162 ucx[0] = uc2;
3163 ccx[0] = cl;
3164 COLLECT_CPS(1);
3165
3166 /*
3167 * Find a composed code in the collected code points.
3168 */
3169 i = 1;
3170 while (i < ucx_size) {
3171 int j;
3172
3173 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3174 i++;
3175 continue;
3176 }
3177
3178 /*
3179 * nfc is composed of uc and ucx[i].
3180 */
3181 UPDATE_UC(nfc);
3182
3183 /*
3184 * Remove ucx[i] by shifting
3185 * following code points.
3186 */
3187 for (j = i; j+1 < ucx_size; j++) {
3188 ucx[j] = ucx[j+1];
3189 ccx[j] = ccx[j+1];
3190 }
3191 ucx_size --;
3192
3193 /*
3194 * Collect following code points blocked
3195 * by ucx[i] the removed code point.
3196 */
3197 if (ucx_size > 0 && i == ucx_size &&
3198 nx > 0 && cx == cl) {
3199 cl = ccx[ucx_size-1];
3200 COLLECT_CPS(ucx_size);
3201 }
3202 /*
3203 * Restart finding a composed code with
3204 * the updated uc from the top of the
3205 * collected code points.
3206 */
3207 i = 0;
3208 }
3209
3210 /*
3211 * Apparently the current code points are not
3212 * decomposed characters or already composed.
3213 */
3214 WRITE_UC();
3215 for (i = 0; i < ucx_size; i++)
3216 UNPARSE(p, endp, ucx[i]);
3217
3218 /*
3219 * Flush out remaining canonical combining characters.
3220 */
3221 if (nx > 0 && cx == cl && len > 0) {
3222 while ((nx = parse(&ucx[0], s, len))
3223 > 0) {
3224 cx = CCC(ucx[0]);
3225 if (cl > cx)
3226 break;
3227 s += nx;
3228 len -= nx;
3229 cl = cx;
3230 UNPARSE(p, endp, ucx[0]);
3231 }
3232 }
3233 break;
3234 }
3235 if (n2 < 0) {
3236 WRITE_UC();
3237 /* Use a replaced unicode character. */
3238 UNPARSE(p, endp, uc2);
3239 s += n2*-1;
3240 len -= n2*-1;
3241 ret = -1;
3242 continue;
3243 } else if (n2 == 0) {
3244 WRITE_UC();
3245 break;
3246 }
3247 }
3248 as->length = p - as->s;
3249 as->s[as->length] = '\0';
3250 if (ts == 2)
3251 as->s[as->length+1] = '\0';
3252 return (ret);
3253 }
3254
3255 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3256 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3257 {
3258 int t, b;
3259
3260 /*
3261 * These are not converted to NFD on Mac OS.
3262 */
3263 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3264 (uc >= 0xF900 && uc <= 0xFAFF) ||
3265 (uc >= 0x2F800 && uc <= 0x2FAFF))
3266 return (0);
3267 /*
3268 * Those code points are not converted to NFD on Mac OS.
3269 * I do not know the reason because it is undocumented.
3270 * NFC NFD
3271 * 1109A ==> 11099 110BA
3272 * 1109C ==> 1109B 110BA
3273 * 110AB ==> 110A5 110BA
3274 */
3275 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3276 return (0);
3277
3278 t = 0;
3279 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3280 while (b >= t) {
3281 int m = (t + b) / 2;
3282 if (u_decomposition_table[m].nfc < uc)
3283 t = m + 1;
3284 else if (u_decomposition_table[m].nfc > uc)
3285 b = m - 1;
3286 else {
3287 *cp1 = u_decomposition_table[m].cp1;
3288 *cp2 = u_decomposition_table[m].cp2;
3289 return (1);
3290 }
3291 }
3292 return (0);
3293 }
3294
3295 #define REPLACE_UC_WITH(cp) do { \
3296 uc = cp; \
3297 ucptr = NULL; \
3298 } while (0)
3299
3300 /*
3301 * Normalize UTF-8 characters to Form D and copy the result.
3302 */
3303 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3304 archive_string_normalize_D(struct archive_string *as, const void *_p,
3305 size_t len, struct archive_string_conv *sc)
3306 {
3307 const char *s = (const char *)_p;
3308 char *p, *endp;
3309 uint32_t uc, uc2;
3310 size_t w;
3311 int always_replace, n, n2, ret = 0, spair, ts, tm;
3312 int (*parse)(uint32_t *, const char *, size_t);
3313 size_t (*unparse)(char *, size_t, uint32_t);
3314
3315 always_replace = 1;
3316 ts = 1;/* text size. */
3317 if (sc->flag & SCONV_TO_UTF16BE) {
3318 unparse = unicode_to_utf16be;
3319 ts = 2;
3320 if (sc->flag & SCONV_FROM_UTF16BE)
3321 always_replace = 0;
3322 } else if (sc->flag & SCONV_TO_UTF16LE) {
3323 unparse = unicode_to_utf16le;
3324 ts = 2;
3325 if (sc->flag & SCONV_FROM_UTF16LE)
3326 always_replace = 0;
3327 } else if (sc->flag & SCONV_TO_UTF8) {
3328 unparse = unicode_to_utf8;
3329 if (sc->flag & SCONV_FROM_UTF8)
3330 always_replace = 0;
3331 } else {
3332 /*
3333 * This case is going to be converted to another
3334 * character-set through iconv.
3335 */
3336 always_replace = 0;
3337 if (sc->flag & SCONV_FROM_UTF16BE) {
3338 unparse = unicode_to_utf16be;
3339 ts = 2;
3340 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3341 unparse = unicode_to_utf16le;
3342 ts = 2;
3343 } else {
3344 unparse = unicode_to_utf8;
3345 }
3346 }
3347
3348 if (sc->flag & SCONV_FROM_UTF16BE) {
3349 parse = utf16be_to_unicode;
3350 tm = 1;
3351 spair = 4;/* surrogate pair size in UTF-16. */
3352 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3353 parse = utf16le_to_unicode;
3354 tm = 1;
3355 spair = 4;/* surrogate pair size in UTF-16. */
3356 } else {
3357 parse = cesu8_to_unicode;
3358 tm = ts;
3359 spair = 6;/* surrogate pair size in UTF-8. */
3360 }
3361
3362 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3363 return (-1);
3364
3365 p = as->s + as->length;
3366 endp = as->s + as->buffer_length - ts;
3367 while ((n = parse(&uc, s, len)) != 0) {
3368 const char *ucptr;
3369 uint32_t cp1, cp2;
3370 int SIndex;
3371 struct {
3372 uint32_t uc;
3373 int ccc;
3374 } fdc[FDC_MAX];
3375 int fdi, fdj;
3376 int ccc;
3377
3378 check_first_code:
3379 if (n < 0) {
3380 /* Use a replaced unicode character. */
3381 UNPARSE(p, endp, uc);
3382 s += n*-1;
3383 len -= n*-1;
3384 ret = -1;
3385 continue;
3386 } else if (n == spair || always_replace)
3387 /* uc is converted from a surrogate pair.
3388 * this should be treated as a changed code. */
3389 ucptr = NULL;
3390 else
3391 ucptr = s;
3392 s += n;
3393 len -= n;
3394
3395 /* Hangul Decomposition. */
3396 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3397 int L = HC_LBASE + SIndex / HC_NCOUNT;
3398 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3399 int T = HC_TBASE + SIndex % HC_TCOUNT;
3400
3401 REPLACE_UC_WITH(L);
3402 WRITE_UC();
3403 REPLACE_UC_WITH(V);
3404 WRITE_UC();
3405 if (T != HC_TBASE) {
3406 REPLACE_UC_WITH(T);
3407 WRITE_UC();
3408 }
3409 continue;
3410 }
3411 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3412 WRITE_UC();
3413 continue;
3414 }
3415
3416 fdi = 0;
3417 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3418 int k;
3419
3420 for (k = fdi; k > 0; k--)
3421 fdc[k] = fdc[k-1];
3422 fdc[0].ccc = CCC(cp2);
3423 fdc[0].uc = cp2;
3424 fdi++;
3425 REPLACE_UC_WITH(cp1);
3426 }
3427
3428 /* Read following code points. */
3429 while ((n2 = parse(&uc2, s, len)) > 0 &&
3430 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3431 int j, k;
3432
3433 s += n2;
3434 len -= n2;
3435 for (j = 0; j < fdi; j++) {
3436 if (fdc[j].ccc > ccc)
3437 break;
3438 }
3439 if (j < fdi) {
3440 for (k = fdi; k > j; k--)
3441 fdc[k] = fdc[k-1];
3442 fdc[j].ccc = ccc;
3443 fdc[j].uc = uc2;
3444 } else {
3445 fdc[fdi].ccc = ccc;
3446 fdc[fdi].uc = uc2;
3447 }
3448 fdi++;
3449 }
3450
3451 WRITE_UC();
3452 for (fdj = 0; fdj < fdi; fdj++) {
3453 REPLACE_UC_WITH(fdc[fdj].uc);
3454 WRITE_UC();
3455 }
3456
3457 if (n2 == 0)
3458 break;
3459 REPLACE_UC_WITH(uc2);
3460 n = n2;
3461 goto check_first_code;
3462 }
3463 as->length = p - as->s;
3464 as->s[as->length] = '\0';
3465 if (ts == 2)
3466 as->s[as->length+1] = '\0';
3467 return (ret);
3468 }
3469
3470 /*
3471 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3472 * that WCS is Unicode. It is true for several platforms but some are false.
3473 * And then people who did not use UTF-8 locale on the non Unicode WCS
3474 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3475 * now cannot get right filename from libarchive 3.x and later since we
3476 * fixed the wrong assumption and it is incompatible to older its versions.
3477 * So we provide special option, "compat-2x.x", for resolving it.
3478 * That option enable the string conversion of libarchive 2.x.
3479 *
3480 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3481 * locale character set and appends to the archive_string.
3482 * Note: returns -1 if conversion fails.
3483 */
3484 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3485 strncat_from_utf8_libarchive2(struct archive_string *as,
3486 const void *_p, size_t len, struct archive_string_conv *sc)
3487 {
3488 const char *s;
3489 int n;
3490 char *p;
3491 char *end;
3492 uint32_t unicode;
3493 #if HAVE_WCRTOMB
3494 mbstate_t shift_state;
3495
3496 memset(&shift_state, 0, sizeof(shift_state));
3497 #else
3498 /* Clear the shift state before starting. */
3499 wctomb(NULL, L'\0');
3500 #endif
3501 (void)sc; /* UNUSED */
3502 /*
3503 * Allocate buffer for MBS.
3504 * We need this allocation here since it is possible that
3505 * as->s is still NULL.
3506 */
3507 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3508 return (-1);
3509
3510 s = (const char *)_p;
3511 p = as->s + as->length;
3512 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3513 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3514 wchar_t wc;
3515
3516 if (p >= end) {
3517 as->length = p - as->s;
3518 /* Re-allocate buffer for MBS. */
3519 if (archive_string_ensure(as,
3520 as->length + max(len * 2,
3521 (size_t)MB_CUR_MAX) + 1) == NULL)
3522 return (-1);
3523 p = as->s + as->length;
3524 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3525 }
3526
3527 /*
3528 * As libarchive 2.x, translates the UTF-8 characters into
3529 * wide-characters in the assumption that WCS is Unicode.
3530 */
3531 if (n < 0) {
3532 n *= -1;
3533 wc = L'?';
3534 } else
3535 wc = (wchar_t)unicode;
3536
3537 s += n;
3538 len -= n;
3539 /*
3540 * Translates the wide-character into the current locale MBS.
3541 */
3542 #if HAVE_WCRTOMB
3543 n = (int)wcrtomb(p, wc, &shift_state);
3544 #else
3545 n = (int)wctomb(p, wc);
3546 #endif
3547 if (n == -1)
3548 return (-1);
3549 p += n;
3550 }
3551 as->length = p - as->s;
3552 as->s[as->length] = '\0';
3553 return (0);
3554 }
3555
3556
3557 /*
3558 * Conversion functions between current locale dependent MBS and UTF-16BE.
3559 * strncat_from_utf16be() : UTF-16BE --> MBS
3560 * strncat_to_utf16be() : MBS --> UTF16BE
3561 */
3562
3563 #if defined(_WIN32) && !defined(__CYGWIN__)
3564
3565 /*
3566 * Convert a UTF-16BE/LE string to current locale and copy the result.
3567 * Return -1 if conversion fails.
3568 */
3569 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3570 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3571 struct archive_string_conv *sc, int be)
3572 {
3573 struct archive_string tmp;
3574 const char *u16;
3575 BOOL defchar;
3576 char *mbs;
3577 size_t mbs_size, b, ll;
3578 int ret = 0;
3579
3580 bytes &= ~1;
3581 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3582 return (-1);
3583
3584 mbs = as->s + as->length;
3585 mbs_size = as->buffer_length - as->length -1;
3586
3587 if (sc->to_cp == CP_C_LOCALE) {
3588 /*
3589 * "C" locale special processing.
3590 */
3591 u16 = _p;
3592 ll = 0;
3593 for (b = 0; b < bytes; b += 2) {
3594 uint16_t val;
3595 if (be)
3596 val = archive_be16dec(u16+b);
3597 else
3598 val = archive_le16dec(u16+b);
3599 if (val > 255) {
3600 *mbs++ = '?';
3601 ret = -1;
3602 } else
3603 *mbs++ = (char)(val&0xff);
3604 ll++;
3605 }
3606 as->length += ll;
3607 as->s[as->length] = '\0';
3608 return (ret);
3609 }
3610
3611 archive_string_init(&tmp);
3612 if (be) {
3613 if (IS_BIG_ENDIAN) {
3614 u16 = _p;
3615 } else {
3616 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3617 return (-1);
3618 memcpy(tmp.s, _p, bytes);
3619 for (b = 0; b < bytes; b += 2) {
3620 uint16_t val = archive_be16dec(tmp.s+b);
3621 archive_le16enc(tmp.s+b, val);
3622 }
3623 u16 = tmp.s;
3624 }
3625 } else {
3626 if (!IS_BIG_ENDIAN) {
3627 u16 = _p;
3628 } else {
3629 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3630 return (-1);
3631 memcpy(tmp.s, _p, bytes);
3632 for (b = 0; b < bytes; b += 2) {
3633 uint16_t val = archive_le16dec(tmp.s+b);
3634 archive_be16enc(tmp.s+b, val);
3635 }
3636 u16 = tmp.s;
3637 }
3638 }
3639
3640 do {
3641 int r;
3642 defchar = 0;
3643 /* WideCharToMultiByte is limited to int. */
3644 if (bytes > (size_t)INT_MAX || mbs_size > (size_t)INT_MAX)
3645 return (-1);
3646 r = WideCharToMultiByte(sc->to_cp, 0,
3647 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3648 NULL, &defchar);
3649 /* Exit loop if we succeeded */
3650 if (r != 0 ||
3651 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3652 ll = (size_t)r;
3653 break;
3654 }
3655 /* Else expand buffer and loop to try again. */
3656 r = WideCharToMultiByte(sc->to_cp, 0,
3657 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3658 ll = (size_t)r;
3659 if (archive_string_ensure(as, ll +1) == NULL)
3660 return (-1);
3661 mbs = as->s + as->length;
3662 mbs_size = as->buffer_length - as->length -1;
3663 } while (1);
3664 archive_string_free(&tmp);
3665 as->length += ll;
3666 as->s[as->length] = '\0';
3667 if (ll == 0 || defchar)
3668 ret = -1;
3669 return (ret);
3670 }
3671
3672 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3673 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3674 size_t bytes, struct archive_string_conv *sc)
3675 {
3676 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3677 }
3678
3679 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3680 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3681 size_t bytes, struct archive_string_conv *sc)
3682 {
3683 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3684 }
3685
3686 /*
3687 * Convert a current locale string to UTF-16BE/LE and copy the result.
3688 * Return -1 if conversion fails.
3689 */
3690 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3691 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3692 size_t length, struct archive_string_conv *sc, int bigendian)
3693 {
3694 const char *s = (const char *)_p;
3695 char *u16;
3696 size_t count, avail;
3697
3698 if (archive_string_ensure(as16,
3699 as16->length + (length + 1) * 2) == NULL)
3700 return (-1);
3701
3702 u16 = as16->s + as16->length;
3703 avail = as16->buffer_length - 2;
3704 if (sc->from_cp == CP_C_LOCALE) {
3705 /*
3706 * "C" locale special processing.
3707 */
3708 count = 0;
3709 while (count < length && *s) {
3710 if (bigendian)
3711 archive_be16enc(u16, *s);
3712 else
3713 archive_le16enc(u16, *s);
3714 u16 += 2;
3715 s++;
3716 count++;
3717 }
3718 as16->length += count << 1;
3719 as16->s[as16->length] = 0;
3720 as16->s[as16->length+1] = 0;
3721 return (0);
3722 }
3723 do {
3724 int r;
3725 if (length > (size_t)INT_MAX || (avail >> 1) > (size_t)INT_MAX)
3726 return (-1);
3727 r = MultiByteToWideChar(sc->from_cp,
3728 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3729 /* Exit loop if we succeeded */
3730 if (r != 0 ||
3731 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3732 count = (size_t)r;
3733 break;
3734 }
3735 /* Expand buffer and try again */
3736 r = MultiByteToWideChar(sc->from_cp,
3737 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3738 count = (size_t)r;
3739 if (archive_string_ensure(as16, (count +1) * 2)
3740 == NULL)
3741 return (-1);
3742 u16 = as16->s + as16->length;
3743 avail = as16->buffer_length - 2;
3744 } while (1);
3745 as16->length += count * 2;
3746 as16->s[as16->length] = 0;
3747 as16->s[as16->length+1] = 0;
3748 if (count == 0)
3749 return (-1);
3750
3751 if (IS_BIG_ENDIAN) {
3752 if (!bigendian) {
3753 while (count > 0) {
3754 uint16_t v = archive_be16dec(u16);
3755 archive_le16enc(u16, v);
3756 u16 += 2;
3757 count--;
3758 }
3759 }
3760 } else {
3761 if (bigendian) {
3762 while (count > 0) {
3763 uint16_t v = archive_le16dec(u16);
3764 archive_be16enc(u16, v);
3765 u16 += 2;
3766 count--;
3767 }
3768 }
3769 }
3770 return (0);
3771 }
3772
3773 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3774 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3775 size_t length, struct archive_string_conv *sc)
3776 {
3777 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3778 }
3779
3780 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3781 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3782 size_t length, struct archive_string_conv *sc)
3783 {
3784 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3785 }
3786
3787 #endif /* _WIN32 && !__CYGWIN__ */
3788
3789 /*
3790 * Do the best effort for conversions.
3791 * We cannot handle UTF-16BE character-set without such iconv,
3792 * but there is a chance if a string consists just ASCII code or
3793 * a current locale is UTF-8.
3794 */
3795
3796 /*
3797 * Convert a UTF-16BE string to current locale and copy the result.
3798 * Return -1 if conversion fails.
3799 */
3800 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3801 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3802 size_t bytes, struct archive_string_conv *sc, int be)
3803 {
3804 const char *utf16 = (const char *)_p;
3805 char *mbs;
3806 uint32_t uc;
3807 int n, ret;
3808
3809 (void)sc; /* UNUSED */
3810 /*
3811 * Other case, we should do the best effort.
3812 * If all character are ASCII(<0x7f), we can convert it.
3813 * if not , we set a alternative character and return -1.
3814 */
3815 ret = 0;
3816 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3817 return (-1);
3818 mbs = as->s + as->length;
3819
3820 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3821 if (n < 0) {
3822 n *= -1;
3823 ret = -1;
3824 }
3825 bytes -= n;
3826 utf16 += n;
3827
3828 if (uc > 127) {
3829 /* We cannot handle it. */
3830 *mbs++ = '?';
3831 ret = -1;
3832 } else
3833 *mbs++ = (char)uc;
3834 }
3835 as->length = mbs - as->s;
3836 as->s[as->length] = '\0';
3837 return (ret);
3838 }
3839
3840 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3841 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3842 size_t bytes, struct archive_string_conv *sc)
3843 {
3844 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3845 }
3846
3847 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3848 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3849 size_t bytes, struct archive_string_conv *sc)
3850 {
3851 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3852 }
3853
3854 /*
3855 * Convert a current locale string to UTF-16BE/LE and copy the result.
3856 * Return -1 if conversion fails.
3857 */
3858 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3859 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3860 size_t length, struct archive_string_conv *sc, int bigendian)
3861 {
3862 const char *s = (const char *)_p;
3863 char *utf16;
3864 size_t remaining;
3865 int ret;
3866
3867 (void)sc; /* UNUSED */
3868 /*
3869 * Other case, we should do the best effort.
3870 * If all character are ASCII(<0x7f), we can convert it.
3871 * if not , we set a alternative character and return -1.
3872 */
3873 ret = 0;
3874 remaining = length;
3875
3876 if (archive_string_ensure(as16,
3877 as16->length + (length + 1) * 2) == NULL)
3878 return (-1);
3879
3880 utf16 = as16->s + as16->length;
3881 while (remaining--) {
3882 unsigned c = *s++;
3883 if (c > 127) {
3884 /* We cannot handle it. */
3885 c = UNICODE_R_CHAR;
3886 ret = -1;
3887 }
3888 if (bigendian)
3889 archive_be16enc(utf16, (uint16_t)c);
3890 else
3891 archive_le16enc(utf16, (uint16_t)c);
3892 utf16 += 2;
3893 }
3894 as16->length = utf16 - as16->s;
3895 as16->s[as16->length] = 0;
3896 as16->s[as16->length+1] = 0;
3897 return (ret);
3898 }
3899
3900 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3901 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3902 size_t length, struct archive_string_conv *sc)
3903 {
3904 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3905 }
3906
3907 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3908 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3909 size_t length, struct archive_string_conv *sc)
3910 {
3911 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3912 }
3913
3914
3915 /*
3916 * Multistring operations.
3917 */
3918
3919 void
archive_mstring_clean(struct archive_mstring * aes)3920 archive_mstring_clean(struct archive_mstring *aes)
3921 {
3922 archive_wstring_free(&(aes->aes_wcs));
3923 archive_string_free(&(aes->aes_mbs));
3924 archive_string_free(&(aes->aes_utf8));
3925 archive_string_free(&(aes->aes_mbs_in_locale));
3926 aes->aes_set = 0;
3927 }
3928
3929 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3930 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3931 {
3932 dest->aes_set = src->aes_set;
3933 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3934 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3935 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3936 }
3937
3938 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3939 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3940 const char **p)
3941 {
3942 struct archive_string_conv *sc;
3943 int r;
3944
3945 /* If we already have a UTF8 form, return that immediately. */
3946 if (aes->aes_set & AES_SET_UTF8) {
3947 *p = aes->aes_utf8.s;
3948 return (0);
3949 }
3950
3951 *p = NULL;
3952 #if defined(_WIN32) && !defined(__CYGWIN__)
3953 /*
3954 * On Windows, first try converting from WCS because (1) there's no
3955 * guarantee that the conversion to MBS will succeed, e.g. when using
3956 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3957 * convert back to WCS again before finally converting to UTF-8
3958 */
3959 if ((aes->aes_set & AES_SET_WCS) != 0) {
3960 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3961 if (sc == NULL)
3962 return (-1);/* Couldn't allocate memory for sc. */
3963 archive_string_empty(&(aes->aes_utf8));
3964 r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3965 aes->aes_wcs.s, aes->aes_wcs.length, sc);
3966 if (a == NULL)
3967 free_sconv_object(sc);
3968 if (r == 0) {
3969 aes->aes_set |= AES_SET_UTF8;
3970 *p = aes->aes_utf8.s;
3971 return (0);/* success. */
3972 } else
3973 return (-1);/* failure. */
3974 }
3975 #endif
3976 /* Try converting WCS to MBS first if MBS does not exist yet. */
3977 if ((aes->aes_set & AES_SET_MBS) == 0) {
3978 const char *pm; /* unused */
3979 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3980 }
3981 if (aes->aes_set & AES_SET_MBS) {
3982 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3983 if (sc == NULL)
3984 return (-1);/* Couldn't allocate memory for sc. */
3985 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3986 aes->aes_mbs.length, sc);
3987 if (a == NULL)
3988 free_sconv_object(sc);
3989 if (r == 0) {
3990 aes->aes_set |= AES_SET_UTF8;
3991 *p = aes->aes_utf8.s;
3992 return (0);/* success. */
3993 } else
3994 return (-1);/* failure. */
3995 }
3996 return (0);/* success. */
3997 }
3998
3999 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)4000 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
4001 const char **p)
4002 {
4003 struct archive_string_conv *sc;
4004 int r, ret = 0;
4005
4006 /* If we already have an MBS form, return that immediately. */
4007 if (aes->aes_set & AES_SET_MBS) {
4008 *p = aes->aes_mbs.s;
4009 return (ret);
4010 }
4011
4012 *p = NULL;
4013 /* If there's a WCS form, try converting with the native locale. */
4014 if (aes->aes_set & AES_SET_WCS) {
4015 archive_string_empty(&(aes->aes_mbs));
4016 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4017 aes->aes_wcs.s, aes->aes_wcs.length);
4018 *p = aes->aes_mbs.s;
4019 if (r == 0) {
4020 aes->aes_set |= AES_SET_MBS;
4021 return (ret);
4022 } else
4023 ret = -1;
4024 }
4025
4026 /* If there's a UTF-8 form, try converting with the native locale. */
4027 if (aes->aes_set & AES_SET_UTF8) {
4028 archive_string_empty(&(aes->aes_mbs));
4029 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4030 if (sc == NULL)
4031 return (-1);/* Couldn't allocate memory for sc. */
4032 r = archive_strncpy_l(&(aes->aes_mbs),
4033 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4034 if (a == NULL)
4035 free_sconv_object(sc);
4036 *p = aes->aes_mbs.s;
4037 if (r == 0) {
4038 aes->aes_set |= AES_SET_MBS;
4039 ret = 0;/* success; overwrite previous error. */
4040 } else
4041 ret = -1;/* failure. */
4042 }
4043 return (ret);
4044 }
4045
4046 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)4047 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
4048 const wchar_t **wp)
4049 {
4050 int r, ret = 0;
4051
4052 (void)a;/* UNUSED */
4053 /* Return WCS form if we already have it. */
4054 if (aes->aes_set & AES_SET_WCS) {
4055 *wp = aes->aes_wcs.s;
4056 return (ret);
4057 }
4058
4059 *wp = NULL;
4060 #if defined(_WIN32) && !defined(__CYGWIN__)
4061 /*
4062 * On Windows, prefer converting from UTF-8 directly to WCS because:
4063 * (1) there's no guarantee that the string can be represented in MBS (e.g.
4064 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
4065 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
4066 * away that intermediate result
4067 */
4068 if (aes->aes_set & AES_SET_UTF8) {
4069 struct archive_string_conv *sc;
4070
4071 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4072 if (sc != NULL) {
4073 archive_wstring_empty((&aes->aes_wcs));
4074 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4075 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4076 if (a == NULL)
4077 free_sconv_object(sc);
4078 if (r == 0) {
4079 aes->aes_set |= AES_SET_WCS;
4080 *wp = aes->aes_wcs.s;
4081 return (0);
4082 }
4083 }
4084 }
4085 #endif
4086 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
4087 if ((aes->aes_set & AES_SET_MBS) == 0) {
4088 const char *p; /* unused */
4089 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4090 }
4091 /* Try converting MBS to WCS using native locale. */
4092 if (aes->aes_set & AES_SET_MBS) {
4093 archive_wstring_empty(&(aes->aes_wcs));
4094 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4095 aes->aes_mbs.s, aes->aes_mbs.length);
4096 if (r == 0) {
4097 aes->aes_set |= AES_SET_WCS;
4098 *wp = aes->aes_wcs.s;
4099 } else
4100 ret = -1;/* failure. */
4101 }
4102 return (ret);
4103 }
4104
4105 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)4106 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4107 const char **p, size_t *length, struct archive_string_conv *sc)
4108 {
4109 int ret = 0;
4110 #if defined(_WIN32) && !defined(__CYGWIN__)
4111 int r;
4112
4113 /*
4114 * Internationalization programming on Windows must use Wide
4115 * characters because Windows platform cannot make locale UTF-8.
4116 */
4117 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4118 archive_string_empty(&(aes->aes_mbs_in_locale));
4119 r = archive_string_append_from_wcs_in_codepage(
4120 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4121 aes->aes_wcs.length, sc);
4122 if (r == 0) {
4123 *p = aes->aes_mbs_in_locale.s;
4124 if (length != NULL)
4125 *length = aes->aes_mbs_in_locale.length;
4126 return (0);
4127 } else if (errno == ENOMEM)
4128 return (-1);
4129 else
4130 ret = -1;
4131 }
4132 #endif
4133
4134 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4135 * with the native locale to be used for translating it to specified
4136 * character-set. */
4137 if ((aes->aes_set & AES_SET_MBS) == 0) {
4138 const char *pm; /* unused */
4139 if (archive_mstring_get_mbs(a, aes, &pm) != 0) {
4140 /* We have another form, but failed to convert it to
4141 * the native locale. Transitively, we've failed to
4142 * convert it to the specified character set. */
4143 ret = -1;
4144 }
4145 }
4146 /* If we already have an MBS form, use it to be translated to
4147 * specified character-set. */
4148 if (aes->aes_set & AES_SET_MBS) {
4149 if (sc == NULL) {
4150 /* Conversion is unneeded. */
4151 *p = aes->aes_mbs.s;
4152 if (length != NULL)
4153 *length = aes->aes_mbs.length;
4154 return (0);
4155 }
4156 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4157 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4158 *p = aes->aes_mbs_in_locale.s;
4159 if (length != NULL)
4160 *length = aes->aes_mbs_in_locale.length;
4161 } else {
4162 /* Either we have no string in any form,
4163 * or conversion failed and set 'ret != 0'. */
4164 *p = NULL;
4165 if (length != NULL)
4166 *length = 0;
4167 }
4168 return (ret);
4169 }
4170
4171 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4172 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4173 {
4174 if (mbs == NULL) {
4175 aes->aes_set = 0;
4176 return (0);
4177 }
4178 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4179 }
4180
4181 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4182 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4183 size_t len)
4184 {
4185 if (mbs == NULL) {
4186 aes->aes_set = 0;
4187 return (0);
4188 }
4189 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4190 archive_strncpy(&(aes->aes_mbs), mbs, len);
4191 archive_string_empty(&(aes->aes_utf8));
4192 archive_wstring_empty(&(aes->aes_wcs));
4193 return (0);
4194 }
4195
4196 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4197 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4198 {
4199 return archive_mstring_copy_wcs_len(aes, wcs,
4200 wcs == NULL ? 0 : wcslen(wcs));
4201 }
4202
4203 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4204 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4205 {
4206 if (utf8 == NULL) {
4207 aes->aes_set = 0;
4208 return (0);
4209 }
4210 aes->aes_set = AES_SET_UTF8;
4211 archive_string_empty(&(aes->aes_mbs));
4212 archive_string_empty(&(aes->aes_wcs));
4213 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4214 return (int)strlen(utf8);
4215 }
4216
4217 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4218 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4219 size_t len)
4220 {
4221 if (wcs == NULL) {
4222 aes->aes_set = 0;
4223 return (0);
4224 }
4225 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4226 archive_string_empty(&(aes->aes_mbs));
4227 archive_string_empty(&(aes->aes_utf8));
4228 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4229 return (0);
4230 }
4231
4232 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4233 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4234 const char *mbs, size_t len, struct archive_string_conv *sc)
4235 {
4236 int r;
4237
4238 if (mbs == NULL) {
4239 aes->aes_set = 0;
4240 return (0);
4241 }
4242 archive_string_empty(&(aes->aes_mbs));
4243 archive_wstring_empty(&(aes->aes_wcs));
4244 archive_string_empty(&(aes->aes_utf8));
4245 #if defined(_WIN32) && !defined(__CYGWIN__)
4246 /*
4247 * Internationalization programming on Windows must use Wide
4248 * characters because Windows platform cannot make locale UTF-8.
4249 */
4250 if (sc == NULL) {
4251 if (archive_string_append(&(aes->aes_mbs),
4252 mbs, mbsnbytes(mbs, len)) == NULL) {
4253 aes->aes_set = 0;
4254 r = -1;
4255 } else {
4256 aes->aes_set = AES_SET_MBS;
4257 r = 0;
4258 }
4259 #if defined(HAVE_ICONV)
4260 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4261 /*
4262 * This case happens only when MultiByteToWideChar() cannot
4263 * handle sc->from_cp, and we have to iconv in order to
4264 * translate character-set to wchar_t,UTF-16.
4265 */
4266 iconv_t cd = sc->cd;
4267 unsigned from_cp;
4268 int flag;
4269
4270 /*
4271 * Translate multi-bytes from some character-set to UTF-8.
4272 */
4273 sc->cd = sc->cd_w;
4274 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4275 sc->cd = cd;
4276 if (r != 0) {
4277 aes->aes_set = 0;
4278 return (r);
4279 }
4280 aes->aes_set = AES_SET_UTF8;
4281
4282 /*
4283 * Append the UTF-8 string into wstring.
4284 */
4285 flag = sc->flag;
4286 sc->flag &= ~(SCONV_NORMALIZATION_C
4287 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4288 from_cp = sc->from_cp;
4289 sc->from_cp = CP_UTF8;
4290 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4291 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4292 sc->flag = flag;
4293 sc->from_cp = from_cp;
4294 if (r == 0)
4295 aes->aes_set |= AES_SET_WCS;
4296 #endif
4297 } else {
4298 r = archive_wstring_append_from_mbs_in_codepage(
4299 &(aes->aes_wcs), mbs, len, sc);
4300 if (r == 0)
4301 aes->aes_set = AES_SET_WCS;
4302 else
4303 aes->aes_set = 0;
4304 }
4305 #else
4306 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4307 if (r == 0)
4308 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4309 else
4310 aes->aes_set = 0;
4311 #endif
4312 return (r);
4313 }
4314
4315 /*
4316 * The 'update' form tries to proactively update all forms of
4317 * this string (WCS and MBS) and returns an error if any of
4318 * them fail. This is used by the 'pax' handler, for instance,
4319 * to detect and report character-conversion failures early while
4320 * still allowing clients to get potentially useful values from
4321 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4322 * strive to give the user something useful, so you can get hopefully
4323 * usable values even if some of the character conversions are failing.)
4324 */
4325 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4326 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4327 const char *utf8)
4328 {
4329 struct archive_string_conv *sc;
4330 int r;
4331
4332 if (utf8 == NULL) {
4333 aes->aes_set = 0;
4334 return (0); /* Succeeded in clearing everything. */
4335 }
4336
4337 /* Save the UTF8 string. */
4338 archive_strcpy(&(aes->aes_utf8), utf8);
4339
4340 /* Empty the mbs and wcs strings. */
4341 archive_string_empty(&(aes->aes_mbs));
4342 archive_wstring_empty(&(aes->aes_wcs));
4343
4344 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4345
4346 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4347 if (sc == NULL)
4348 return (-1);/* Couldn't allocate memory for sc. */
4349
4350 #if defined(_WIN32) && !defined(__CYGWIN__)
4351 /* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4352 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4353 * intermediate result, and (2) WCS will still be set even if we fail to
4354 * convert to MBS (e.g. with ACP that can't represent the characters) */
4355 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4356 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4357
4358 if (a == NULL)
4359 free_sconv_object(sc);
4360 if (r != 0)
4361 return (-1); /* This will guarantee we can't convert to MBS */
4362 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4363
4364 /* Try converting WCS to MBS, return false on failure. */
4365 if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4366 aes->aes_wcs.length))
4367 return (-1);
4368 #else
4369 /* Try converting UTF-8 to MBS, return false on failure. */
4370 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4371
4372 if (a == NULL)
4373 free_sconv_object(sc);
4374 if (r != 0)
4375 return (-1);
4376 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4377
4378 /* Try converting MBS to WCS, return false on failure. */
4379 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4380 aes->aes_mbs.length))
4381 return (-1);
4382 #endif
4383
4384 /* All conversions succeeded. */
4385 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4386
4387 return (0);
4388 }
4389