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 #elif defined(__BIONIC__)
454 return "UTF-8";
455 #else
456 return "";
457 #endif
458 }
459
460 #if defined(_WIN32) && !defined(__CYGWIN__)
461
462 /*
463 * Convert MBS to WCS.
464 * Note: returns -1 if conversion fails.
465 */
466 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)467 archive_wstring_append_from_mbs(struct archive_wstring *dest,
468 const char *p, size_t len)
469 {
470 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
471 }
472
473 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)474 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
475 const char *s, size_t length, struct archive_string_conv *sc)
476 {
477 int ret = 0;
478 size_t count;
479 UINT from_cp;
480
481 if (sc != NULL)
482 from_cp = sc->from_cp;
483 else
484 from_cp = get_current_codepage();
485
486 if (from_cp == CP_C_LOCALE) {
487 /*
488 * "C" locale special processing.
489 */
490 wchar_t *ws;
491 const unsigned char *mp;
492
493 if (NULL == archive_wstring_ensure(dest,
494 dest->length + length + 1))
495 return (-1);
496
497 ws = dest->s + dest->length;
498 mp = (const unsigned char *)s;
499 count = 0;
500 while (count < length && *mp) {
501 *ws++ = (wchar_t)*mp++;
502 count++;
503 }
504 } else if (sc != NULL &&
505 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
506 /*
507 * Normalize UTF-8 and UTF-16BE and convert it directly
508 * to UTF-16 as wchar_t.
509 */
510 struct archive_string u16;
511 int saved_flag = sc->flag;/* save current flag. */
512
513 if (IS_BIG_ENDIAN)
514 sc->flag |= SCONV_TO_UTF16BE;
515 else
516 sc->flag |= SCONV_TO_UTF16LE;
517
518 if (sc->flag & SCONV_FROM_UTF16) {
519 /*
520 * UTF-16BE/LE NFD ===> UTF-16 NFC
521 * UTF-16BE/LE NFC ===> UTF-16 NFD
522 */
523 count = utf16nbytes(s, length);
524 } else {
525 /*
526 * UTF-8 NFD ===> UTF-16 NFC
527 * UTF-8 NFC ===> UTF-16 NFD
528 */
529 count = mbsnbytes(s, length);
530 }
531 u16.s = (char *)dest->s;
532 u16.length = dest->length << 1;
533 u16.buffer_length = dest->buffer_length;
534 if (sc->flag & SCONV_NORMALIZATION_C)
535 ret = archive_string_normalize_C(&u16, s, count, sc);
536 else
537 ret = archive_string_normalize_D(&u16, s, count, sc);
538 dest->s = (wchar_t *)u16.s;
539 dest->length = u16.length >> 1;
540 dest->buffer_length = u16.buffer_length;
541 sc->flag = saved_flag;/* restore the saved flag. */
542 return (ret);
543 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
544 count = utf16nbytes(s, length);
545 count >>= 1; /* to be WCS length */
546 /* Allocate memory for WCS. */
547 if (NULL == archive_wstring_ensure(dest,
548 dest->length + count + 1))
549 return (-1);
550 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
551 if ((sc->flag & SCONV_FROM_UTF16BE) && !IS_BIG_ENDIAN) {
552 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
553 size_t b;
554 for (b = 0; b < count; b++) {
555 uint16_t val = archive_le16dec(u16+b);
556 archive_be16enc(u16+b, val);
557 }
558 } else if ((sc->flag & SCONV_FROM_UTF16LE) && IS_BIG_ENDIAN) {
559 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
560 size_t b;
561 for (b = 0; b < count; b++) {
562 uint16_t val = archive_be16dec(u16+b);
563 archive_le16enc(u16+b, val);
564 }
565 }
566 } else {
567 DWORD mbflag;
568 size_t buffsize;
569
570 if (sc == NULL)
571 mbflag = 0;
572 else if (sc->flag & SCONV_FROM_CHARSET) {
573 /* Do not trust the length which comes from
574 * an archive file. */
575 length = mbsnbytes(s, length);
576 mbflag = 0;
577 } else
578 mbflag = MB_PRECOMPOSED;
579
580 mbflag |= MB_ERR_INVALID_CHARS;
581
582 buffsize = dest->length + length + 1;
583 do {
584 int r;
585
586 /* MultiByteToWideChar is limited to int. */
587 if (length > (size_t)INT_MAX ||
588 (dest->buffer_length >> 1) > (size_t)INT_MAX)
589 return (-1);
590 /* Allocate memory for WCS. */
591 if (NULL == archive_wstring_ensure(dest, buffsize))
592 return (-1);
593 /* Convert MBS to WCS. */
594 r = MultiByteToWideChar(from_cp,
595 mbflag, s, (int)length, dest->s + dest->length,
596 (int)(dest->buffer_length >> 1) -1);
597 if (r == 0 &&
598 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
599 /* Expand the WCS buffer. */
600 buffsize = dest->buffer_length << 1;
601 continue;
602 }
603 if (r == 0 && length != 0)
604 ret = -1;
605 count = (size_t)r;
606 break;
607 } while (1);
608 }
609 dest->length += count;
610 dest->s[dest->length] = L'\0';
611 return (ret);
612 }
613
614 #else
615
616 /*
617 * Convert MBS to WCS.
618 * Note: returns -1 if conversion fails.
619 */
620 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)621 archive_wstring_append_from_mbs(struct archive_wstring *dest,
622 const char *p, size_t len)
623 {
624 size_t r;
625 int ret_val = 0;
626 /*
627 * No single byte will be more than one wide character,
628 * so this length estimate will always be big enough.
629 */
630 // size_t wcs_length = len;
631 size_t mbs_length = len;
632 const char *mbs = p;
633 wchar_t *wcs;
634 #if HAVE_MBRTOWC
635 mbstate_t shift_state;
636
637 memset(&shift_state, 0, sizeof(shift_state));
638 #endif
639 /*
640 * As we decided to have wcs_length == mbs_length == len
641 * we can use len here instead of wcs_length
642 */
643 if (NULL == archive_wstring_ensure(dest, dest->length + len + 1))
644 return (-1);
645 wcs = dest->s + dest->length;
646 /*
647 * We cannot use mbsrtowcs/mbstowcs here because those may convert
648 * extra MBS when strlen(p) > len and one wide character consists of
649 * multi bytes.
650 */
651 while (*mbs && mbs_length > 0) {
652 /*
653 * The buffer we allocated is always big enough.
654 * Keep this code path in a comment if we decide to choose
655 * smaller wcs_length in the future
656 */
657 /*
658 if (wcs_length == 0) {
659 dest->length = wcs - dest->s;
660 dest->s[dest->length] = L'\0';
661 wcs_length = mbs_length;
662 if (NULL == archive_wstring_ensure(dest,
663 dest->length + wcs_length + 1))
664 return (-1);
665 wcs = dest->s + dest->length;
666 }
667 */
668 #if HAVE_MBRTOWC
669 r = mbrtowc(wcs, mbs, mbs_length, &shift_state);
670 #else
671 r = mbtowc(wcs, mbs, mbs_length);
672 #endif
673 if (r == (size_t)-1 || r == (size_t)-2) {
674 ret_val = -1;
675 break;
676 }
677 if (r == 0 || r > mbs_length)
678 break;
679 wcs++;
680 // wcs_length--;
681 mbs += r;
682 mbs_length -= r;
683 }
684 dest->length = wcs - dest->s;
685 dest->s[dest->length] = L'\0';
686 return (ret_val);
687 }
688
689 #endif
690
691 #if defined(_WIN32) && !defined(__CYGWIN__)
692
693 /*
694 * WCS ==> MBS.
695 * Note: returns -1 if conversion fails.
696 *
697 * Win32 builds use WideCharToMultiByte from the Windows API.
698 * (Maybe Cygwin should too? WideCharToMultiByte will know a
699 * lot more about local character encodings than the wcrtomb()
700 * wrapper is going to know.)
701 */
702 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)703 archive_string_append_from_wcs(struct archive_string *as,
704 const wchar_t *w, size_t len)
705 {
706 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
707 }
708
709 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)710 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
711 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
712 {
713 BOOL defchar_used, *dp;
714 int ret = 0;
715 UINT to_cp;
716 size_t count, wslen = len;
717
718 if (sc != NULL)
719 to_cp = sc->to_cp;
720 else
721 to_cp = get_current_codepage();
722
723 if (to_cp == CP_C_LOCALE) {
724 /*
725 * "C" locale special processing.
726 */
727 const wchar_t *wp = ws;
728 char *p;
729
730 if (NULL == archive_string_ensure(as,
731 as->length + wslen +1))
732 return (-1);
733 p = as->s + as->length;
734 count = 0;
735 defchar_used = 0;
736 while (count < wslen && *wp) {
737 if (*wp > 255) {
738 *p++ = '?';
739 wp++;
740 defchar_used = 1;
741 } else
742 *p++ = (char)*wp++;
743 count++;
744 }
745 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
746 uint16_t *u16;
747
748 if (NULL ==
749 archive_string_ensure(as, as->length + len * 2 + 2))
750 return (-1);
751 u16 = (uint16_t *)(as->s + as->length);
752 count = 0;
753 defchar_used = 0;
754 if (sc->flag & SCONV_TO_UTF16BE) {
755 while (count < len && *ws) {
756 archive_be16enc(u16+count, *ws);
757 ws++;
758 count++;
759 }
760 } else {
761 while (count < len && *ws) {
762 archive_le16enc(u16+count, *ws);
763 ws++;
764 count++;
765 }
766 }
767 count <<= 1; /* to be byte size */
768 } else {
769 /* Make sure the MBS buffer has plenty to set. */
770 if (NULL ==
771 archive_string_ensure(as, as->length + len * 2 + 1))
772 return (-1);
773 do {
774 int r;
775
776 defchar_used = 0;
777 if (to_cp == CP_UTF8)
778 dp = NULL;
779 else
780 dp = &defchar_used;
781 /* WideCharToMultiByte is limited to int. */
782 if (as->buffer_length - as->length - 1 > (size_t)INT_MAX ||
783 wslen > (size_t)INT_MAX)
784 return (-1);
785 r = WideCharToMultiByte(to_cp, 0, ws, (int)wslen,
786 as->s + as->length,
787 (int)(as->buffer_length - as->length - 1), NULL, dp);
788 if (r == 0 &&
789 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
790 /* Expand the MBS buffer and retry. */
791 if (NULL == archive_string_ensure(as,
792 as->buffer_length + len))
793 return (-1);
794 continue;
795 }
796 if (r == 0)
797 ret = -1;
798 count = (size_t)r;
799 break;
800 } while (1);
801 }
802 as->length += count;
803 as->s[as->length] = '\0';
804 return (defchar_used?-1:ret);
805 }
806
807 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
808
809 /*
810 * Translates a wide character string into current locale character set
811 * and appends to the archive_string. Note: returns -1 if conversion
812 * fails.
813 */
814 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)815 archive_string_append_from_wcs(struct archive_string *as,
816 const wchar_t *w, size_t len)
817 {
818 /* We cannot use the standard wcstombs() here because it
819 * cannot tell us how big the output buffer should be. So
820 * I've built a loop around wcrtomb() or wctomb() that
821 * converts a character at a time and resizes the string as
822 * needed. We prefer wcrtomb() when it's available because
823 * it's thread-safe. */
824 int n, ret_val = 0;
825 char *p;
826 char *end;
827 #if HAVE_WCRTOMB
828 mbstate_t shift_state;
829
830 memset(&shift_state, 0, sizeof(shift_state));
831 #else
832 /* Clear the shift state before starting. */
833 wctomb(NULL, L'\0');
834 #endif
835 /*
836 * Allocate buffer for MBS.
837 * We need this allocation here since it is possible that
838 * as->s is still NULL.
839 */
840 if (archive_string_ensure(as, as->length + len + 1) == NULL)
841 return (-1);
842
843 p = as->s + as->length;
844 end = as->s + as->buffer_length - MB_CUR_MAX -1;
845 while (*w != L'\0' && len > 0) {
846 if (p >= end) {
847 as->length = p - as->s;
848 as->s[as->length] = '\0';
849 /* Re-allocate buffer for MBS. */
850 if (archive_string_ensure(as,
851 as->length + max(len * 2,
852 (size_t)MB_CUR_MAX) + 1) == NULL)
853 return (-1);
854 p = as->s + as->length;
855 end = as->s + as->buffer_length - MB_CUR_MAX -1;
856 }
857 #if HAVE_WCRTOMB
858 n = wcrtomb(p, *w++, &shift_state);
859 #else
860 n = wctomb(p, *w++);
861 #endif
862 if (n == -1) {
863 if (errno == EILSEQ) {
864 /* Skip an illegal wide char. */
865 *p++ = '?';
866 ret_val = -1;
867 } else {
868 ret_val = -1;
869 break;
870 }
871 } else
872 p += n;
873 len--;
874 }
875 as->length = p - as->s;
876 as->s[as->length] = '\0';
877 return (ret_val);
878 }
879
880 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
881
882 /*
883 * TODO: Test if __STDC_ISO_10646__ is defined.
884 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
885 * one character at a time. If a non-Windows platform doesn't have
886 * either of these, fall back to the built-in UTF8 conversion.
887 */
888 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)889 archive_string_append_from_wcs(struct archive_string *as,
890 const wchar_t *w, size_t len)
891 {
892 (void)as;/* UNUSED */
893 (void)w;/* UNUSED */
894 (void)len;/* UNUSED */
895 errno = ENOSYS;
896 return (-1);
897 }
898
899 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
900
901 /*
902 * Find a string conversion object by a pair of 'from' charset name
903 * and 'to' charset name from an archive object.
904 * Return NULL if not found.
905 */
906 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)907 find_sconv_object(struct archive *a, const char *fc, const char *tc)
908 {
909 struct archive_string_conv *sc;
910
911 if (a == NULL)
912 return (NULL);
913
914 for (sc = a->sconv; sc != NULL; sc = sc->next) {
915 if (strcmp(sc->from_charset, fc) == 0 &&
916 strcmp(sc->to_charset, tc) == 0)
917 break;
918 }
919 return (sc);
920 }
921
922 /*
923 * Register a string object to an archive object.
924 */
925 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)926 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
927 {
928 struct archive_string_conv **psc;
929
930 /* Add a new sconv to sconv list. */
931 psc = &(a->sconv);
932 while (*psc != NULL)
933 psc = &((*psc)->next);
934 *psc = sc;
935 }
936
937 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))938 add_converter(struct archive_string_conv *sc, int (*converter)
939 (struct archive_string *, const void *, size_t,
940 struct archive_string_conv *))
941 {
942 if (sc == NULL || sc->nconverter >= 2)
943 __archive_errx(1, "Programming error");
944 sc->converter[sc->nconverter++] = converter;
945 }
946
947 static void
setup_converter(struct archive_string_conv * sc)948 setup_converter(struct archive_string_conv *sc)
949 {
950
951 /* Reset. */
952 sc->nconverter = 0;
953
954 /*
955 * Perform special sequence for the incorrect UTF-8 filenames
956 * made by libarchive2.x.
957 */
958 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
959 add_converter(sc, strncat_from_utf8_libarchive2);
960 return;
961 }
962
963 /*
964 * Convert a string to UTF-16BE/LE.
965 */
966 if (sc->flag & SCONV_TO_UTF16) {
967 /*
968 * If the current locale is UTF-8, we can translate
969 * a UTF-8 string into a UTF-16BE string.
970 */
971 if (sc->flag & SCONV_FROM_UTF8) {
972 add_converter(sc, archive_string_append_unicode);
973 return;
974 }
975
976 #if defined(_WIN32) && !defined(__CYGWIN__)
977 if (sc->flag & SCONV_WIN_CP) {
978 if (sc->flag & SCONV_TO_UTF16BE)
979 add_converter(sc, win_strncat_to_utf16be);
980 else
981 add_converter(sc, win_strncat_to_utf16le);
982 return;
983 }
984 #endif
985
986 #if defined(HAVE_ICONV)
987 if (sc->cd != (iconv_t)-1) {
988 add_converter(sc, iconv_strncat_in_locale);
989 return;
990 }
991 #endif
992
993 if (sc->flag & SCONV_BEST_EFFORT) {
994 if (sc->flag & SCONV_TO_UTF16BE)
995 add_converter(sc,
996 best_effort_strncat_to_utf16be);
997 else
998 add_converter(sc,
999 best_effort_strncat_to_utf16le);
1000 } else
1001 /* Make sure we have no converter. */
1002 sc->nconverter = 0;
1003 return;
1004 }
1005
1006 /*
1007 * Convert a string from UTF-16BE/LE.
1008 */
1009 if (sc->flag & SCONV_FROM_UTF16) {
1010 /*
1011 * At least we should normalize a UTF-16BE string.
1012 */
1013 if (sc->flag & SCONV_NORMALIZATION_D)
1014 add_converter(sc,archive_string_normalize_D);
1015 else if (sc->flag & SCONV_NORMALIZATION_C)
1016 add_converter(sc, archive_string_normalize_C);
1017
1018 if (sc->flag & SCONV_TO_UTF8) {
1019 /*
1020 * If the current locale is UTF-8, we can translate
1021 * a UTF-16BE/LE string into a UTF-8 string directly.
1022 */
1023 if (!(sc->flag &
1024 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1025 add_converter(sc,
1026 archive_string_append_unicode);
1027 return;
1028 }
1029
1030 #if defined(_WIN32) && !defined(__CYGWIN__)
1031 if (sc->flag & SCONV_WIN_CP) {
1032 if (sc->flag & SCONV_FROM_UTF16BE)
1033 add_converter(sc, win_strncat_from_utf16be);
1034 else
1035 add_converter(sc, win_strncat_from_utf16le);
1036 return;
1037 }
1038 #endif
1039
1040 #if defined(HAVE_ICONV)
1041 if (sc->cd != (iconv_t)-1) {
1042 add_converter(sc, iconv_strncat_in_locale);
1043 return;
1044 }
1045 #endif
1046
1047 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1048 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1049 add_converter(sc, best_effort_strncat_from_utf16be);
1050 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1051 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1052 add_converter(sc, best_effort_strncat_from_utf16le);
1053 else
1054 /* Make sure we have no converter. */
1055 sc->nconverter = 0;
1056 return;
1057 }
1058
1059 if (sc->flag & SCONV_FROM_UTF8) {
1060 /*
1061 * At least we should normalize a UTF-8 string.
1062 */
1063 if (sc->flag & SCONV_NORMALIZATION_D)
1064 add_converter(sc,archive_string_normalize_D);
1065 else if (sc->flag & SCONV_NORMALIZATION_C)
1066 add_converter(sc, archive_string_normalize_C);
1067
1068 /*
1069 * Copy UTF-8 string with a check of CESU-8.
1070 * Apparently, iconv does not check surrogate pairs in UTF-8
1071 * when both from-charset and to-charset are UTF-8, and then
1072 * we use our UTF-8 copy code.
1073 */
1074 if (sc->flag & SCONV_TO_UTF8) {
1075 /*
1076 * If the current locale is UTF-8, we can translate
1077 * a UTF-16BE string into a UTF-8 string directly.
1078 */
1079 if (!(sc->flag &
1080 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1081 add_converter(sc, strncat_from_utf8_to_utf8);
1082 return;
1083 }
1084 }
1085
1086 #if defined(_WIN32) && !defined(__CYGWIN__)
1087 /*
1088 * On Windows we can use Windows API for a string conversion.
1089 */
1090 if (sc->flag & SCONV_WIN_CP) {
1091 add_converter(sc, strncat_in_codepage);
1092 return;
1093 }
1094 #endif
1095
1096 #if HAVE_ICONV
1097 if (sc->cd != (iconv_t)-1) {
1098 add_converter(sc, iconv_strncat_in_locale);
1099 /*
1100 * iconv generally does not support UTF-8-MAC and so
1101 * we have to the output of iconv from NFC to NFD if
1102 * need.
1103 */
1104 if ((sc->flag & SCONV_FROM_CHARSET) &&
1105 (sc->flag & SCONV_TO_UTF8)) {
1106 if (sc->flag & SCONV_NORMALIZATION_D)
1107 add_converter(sc, archive_string_normalize_D);
1108 }
1109 return;
1110 }
1111 #endif
1112
1113 /*
1114 * Try conversion in the best effort or no conversion.
1115 */
1116 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1117 add_converter(sc, best_effort_strncat_in_locale);
1118 else
1119 /* Make sure we have no converter. */
1120 sc->nconverter = 0;
1121 }
1122
1123 /*
1124 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1125 * and CP932 which are referenced in create_sconv_object().
1126 */
1127 static const char *
canonical_charset_name(const char * charset)1128 canonical_charset_name(const char *charset)
1129 {
1130 char cs[16];
1131 char *p;
1132 const char *s;
1133
1134 if (charset == NULL || charset[0] == '\0'
1135 || strlen(charset) > 15)
1136 return (charset);
1137
1138 /* Copy name to uppercase. */
1139 p = cs;
1140 s = charset;
1141 while (*s) {
1142 char c = *s++;
1143 if (c >= 'a' && c <= 'z')
1144 c -= 'a' - 'A';
1145 *p++ = c;
1146 }
1147 *p++ = '\0';
1148
1149 if (strcmp(cs, "UTF-8") == 0 ||
1150 strcmp(cs, "UTF8") == 0)
1151 return ("UTF-8");
1152 if (strcmp(cs, "UTF-16BE") == 0 ||
1153 strcmp(cs, "UTF16BE") == 0)
1154 return ("UTF-16BE");
1155 if (strcmp(cs, "UTF-16LE") == 0 ||
1156 strcmp(cs, "UTF16LE") == 0)
1157 return ("UTF-16LE");
1158 if (strcmp(cs, "CP932") == 0)
1159 return ("CP932");
1160 return (charset);
1161 }
1162
1163 /*
1164 * Create a string conversion object.
1165 */
1166 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1167 create_sconv_object(const char *fc, const char *tc,
1168 unsigned current_codepage, int flag)
1169 {
1170 struct archive_string_conv *sc;
1171
1172 sc = calloc(1, sizeof(*sc));
1173 if (sc == NULL)
1174 return (NULL);
1175 sc->next = NULL;
1176 sc->from_charset = strdup(fc);
1177 if (sc->from_charset == NULL) {
1178 free(sc);
1179 return (NULL);
1180 }
1181 sc->to_charset = strdup(tc);
1182 if (sc->to_charset == NULL) {
1183 free(sc->from_charset);
1184 free(sc);
1185 return (NULL);
1186 }
1187 archive_string_init(&sc->utftmp);
1188
1189 if (flag & SCONV_TO_CHARSET) {
1190 /*
1191 * Convert characters from the current locale charset to
1192 * a specified charset.
1193 */
1194 sc->from_cp = current_codepage;
1195 sc->to_cp = make_codepage_from_charset(tc);
1196 #if defined(_WIN32) && !defined(__CYGWIN__)
1197 if (IsValidCodePage(sc->to_cp))
1198 flag |= SCONV_WIN_CP;
1199 #endif
1200 } else if (flag & SCONV_FROM_CHARSET) {
1201 /*
1202 * Convert characters from a specified charset to
1203 * the current locale charset.
1204 */
1205 sc->to_cp = current_codepage;
1206 sc->from_cp = make_codepage_from_charset(fc);
1207 #if defined(_WIN32) && !defined(__CYGWIN__)
1208 if (IsValidCodePage(sc->from_cp))
1209 flag |= SCONV_WIN_CP;
1210 #endif
1211 }
1212
1213 /*
1214 * Check if "from charset" and "to charset" are the same.
1215 */
1216 if (strcmp(fc, tc) == 0 ||
1217 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1218 sc->same = 1;
1219 else
1220 sc->same = 0;
1221
1222 /*
1223 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1224 */
1225 if (strcmp(tc, "UTF-8") == 0)
1226 flag |= SCONV_TO_UTF8;
1227 else if (strcmp(tc, "UTF-16BE") == 0)
1228 flag |= SCONV_TO_UTF16BE;
1229 else if (strcmp(tc, "UTF-16LE") == 0)
1230 flag |= SCONV_TO_UTF16LE;
1231 if (strcmp(fc, "UTF-8") == 0)
1232 flag |= SCONV_FROM_UTF8;
1233 else if (strcmp(fc, "UTF-16BE") == 0)
1234 flag |= SCONV_FROM_UTF16BE;
1235 else if (strcmp(fc, "UTF-16LE") == 0)
1236 flag |= SCONV_FROM_UTF16LE;
1237 #if defined(_WIN32) && !defined(__CYGWIN__)
1238 if (sc->to_cp == CP_UTF8)
1239 flag |= SCONV_TO_UTF8;
1240 else if (sc->to_cp == CP_UTF16BE)
1241 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1242 else if (sc->to_cp == CP_UTF16LE)
1243 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1244 if (sc->from_cp == CP_UTF8)
1245 flag |= SCONV_FROM_UTF8;
1246 else if (sc->from_cp == CP_UTF16BE)
1247 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1248 else if (sc->from_cp == CP_UTF16LE)
1249 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1250 #endif
1251
1252 /*
1253 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1254 * handle it. So we have to translate NFD characters to NFC ones
1255 * ourselves before iconv handles. Another reason is to prevent
1256 * that the same sight of two filenames, one is NFC and other
1257 * is NFD, would be in its directory.
1258 * On Mac OS X, although its filesystem layer automatically
1259 * convert filenames to NFD, it would be useful for filename
1260 * comparing to find out the same filenames that we normalize
1261 * that to be NFD ourselves.
1262 */
1263 if ((flag & SCONV_FROM_CHARSET) &&
1264 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1265 #if defined(__APPLE__)
1266 if (flag & SCONV_TO_UTF8)
1267 flag |= SCONV_NORMALIZATION_D;
1268 else
1269 #endif
1270 flag |= SCONV_NORMALIZATION_C;
1271 }
1272 #if defined(__APPLE__)
1273 /*
1274 * In case writing an archive file, make sure that a filename
1275 * going to be passed to iconv is a Unicode NFC string since
1276 * a filename in HFS Plus filesystem is a Unicode NFD one and
1277 * iconv cannot handle it with "UTF-8" charset. It is simpler
1278 * than a use of "UTF-8-MAC" charset.
1279 */
1280 if ((flag & SCONV_TO_CHARSET) &&
1281 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1282 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1283 flag |= SCONV_NORMALIZATION_C;
1284 /*
1285 * In case reading an archive file. make sure that a filename
1286 * will be passed to users is a Unicode NFD string in order to
1287 * correctly compare the filename with other one which comes
1288 * from HFS Plus filesystem.
1289 */
1290 if ((flag & SCONV_FROM_CHARSET) &&
1291 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1292 (flag & SCONV_TO_UTF8))
1293 flag |= SCONV_NORMALIZATION_D;
1294 #endif
1295
1296 #if defined(HAVE_ICONV)
1297 sc->cd_w = (iconv_t)-1;
1298 /*
1299 * Create an iconv object.
1300 */
1301 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1302 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1303 (flag & SCONV_WIN_CP)) {
1304 /* This case we won't use iconv. */
1305 sc->cd = (iconv_t)-1;
1306 } else {
1307 sc->cd = iconv_open(tc, fc);
1308 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1309 /*
1310 * Unfortunately, all of iconv implements do support
1311 * "CP932" character-set, so we should use "SJIS"
1312 * instead if iconv_open failed.
1313 */
1314 if (strcmp(tc, "CP932") == 0)
1315 sc->cd = iconv_open("SJIS", fc);
1316 else if (strcmp(fc, "CP932") == 0)
1317 sc->cd = iconv_open(tc, "SJIS");
1318 }
1319 #if defined(__FreeBSD__) && !defined(HAVE_LIBICONV)
1320 /*
1321 * FreeBSD's native iconv() by default returns the number of
1322 * invalid characters in the input string, as specified by
1323 * POSIX, but iconv_strncat_in_locale() assumes GNU iconv
1324 * semantics.
1325 */
1326 int v = 1;
1327
1328 (void)iconvctl(sc->cd, ICONV_SET_ILSEQ_INVALID, &v);
1329 #elif defined(_WIN32) && !defined(__CYGWIN__)
1330 /*
1331 * archive_mstring on Windows directly convert multi-bytes
1332 * into archive_wstring in order not to depend on locale
1333 * so that you can do a I18N programming. This will be
1334 * used only in archive_mstring_copy_mbs_len_l so far.
1335 */
1336 if (flag & SCONV_FROM_CHARSET) {
1337 sc->cd_w = iconv_open("UTF-8", fc);
1338 if (sc->cd_w == (iconv_t)-1 &&
1339 (sc->flag & SCONV_BEST_EFFORT)) {
1340 if (strcmp(fc, "CP932") == 0)
1341 sc->cd_w = iconv_open("UTF-8", "SJIS");
1342 }
1343 }
1344 #endif /* _WIN32 && !__CYGWIN__ */
1345 }
1346 #endif /* HAVE_ICONV */
1347
1348 sc->flag = flag;
1349
1350 /*
1351 * Set up converters.
1352 */
1353 setup_converter(sc);
1354
1355 return (sc);
1356 }
1357
1358 /*
1359 * Free a string conversion object.
1360 */
1361 static void
free_sconv_object(struct archive_string_conv * sc)1362 free_sconv_object(struct archive_string_conv *sc)
1363 {
1364 free(sc->from_charset);
1365 free(sc->to_charset);
1366 archive_string_free(&sc->utftmp);
1367 #if HAVE_ICONV
1368 if (sc->cd != (iconv_t)-1)
1369 iconv_close(sc->cd);
1370 if (sc->cd_w != (iconv_t)-1)
1371 iconv_close(sc->cd_w);
1372 #endif
1373 free(sc);
1374 }
1375
1376 #if defined(_WIN32) && !defined(__CYGWIN__)
1377 # if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
1378 # define GetOEMCP() CP_OEMCP
1379 # endif
1380
1381 static unsigned
my_atoi(const char * p)1382 my_atoi(const char *p)
1383 {
1384 unsigned cp;
1385
1386 cp = 0;
1387 while (*p) {
1388 if (*p >= '0' && *p <= '9')
1389 cp = cp * 10 + (*p - '0');
1390 else
1391 return (-1);
1392 p++;
1393 }
1394 return (cp);
1395 }
1396
1397 /*
1398 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1399 * Return -1 if failed.
1400 *
1401 * Note: This translation code may be insufficient.
1402 */
1403 static struct charset {
1404 const char *name;
1405 unsigned cp;
1406 } charsets[] = {
1407 /* MUST BE SORTED! */
1408 {"ASCII", 1252},
1409 {"ASMO-708", 708},
1410 {"BIG5", 950},
1411 {"CHINESE", 936},
1412 {"CP367", 1252},
1413 {"CP819", 1252},
1414 {"CP1025", 21025},
1415 {"DOS-720", 720},
1416 {"DOS-862", 862},
1417 {"EUC-CN", 51936},
1418 {"EUC-JP", 51932},
1419 {"EUC-KR", 949},
1420 {"EUCCN", 51936},
1421 {"EUCJP", 51932},
1422 {"EUCKR", 949},
1423 {"GB18030", 54936},
1424 {"GB2312", 936},
1425 {"HEBREW", 1255},
1426 {"HZ-GB-2312", 52936},
1427 {"IBM273", 20273},
1428 {"IBM277", 20277},
1429 {"IBM278", 20278},
1430 {"IBM280", 20280},
1431 {"IBM284", 20284},
1432 {"IBM285", 20285},
1433 {"IBM290", 20290},
1434 {"IBM297", 20297},
1435 {"IBM367", 1252},
1436 {"IBM420", 20420},
1437 {"IBM423", 20423},
1438 {"IBM424", 20424},
1439 {"IBM819", 1252},
1440 {"IBM871", 20871},
1441 {"IBM880", 20880},
1442 {"IBM905", 20905},
1443 {"IBM924", 20924},
1444 {"ISO-8859-1", 28591},
1445 {"ISO-8859-13", 28603},
1446 {"ISO-8859-15", 28605},
1447 {"ISO-8859-2", 28592},
1448 {"ISO-8859-3", 28593},
1449 {"ISO-8859-4", 28594},
1450 {"ISO-8859-5", 28595},
1451 {"ISO-8859-6", 28596},
1452 {"ISO-8859-7", 28597},
1453 {"ISO-8859-8", 28598},
1454 {"ISO-8859-9", 28599},
1455 {"ISO8859-1", 28591},
1456 {"ISO8859-13", 28603},
1457 {"ISO8859-15", 28605},
1458 {"ISO8859-2", 28592},
1459 {"ISO8859-3", 28593},
1460 {"ISO8859-4", 28594},
1461 {"ISO8859-5", 28595},
1462 {"ISO8859-6", 28596},
1463 {"ISO8859-7", 28597},
1464 {"ISO8859-8", 28598},
1465 {"ISO8859-9", 28599},
1466 {"JOHAB", 1361},
1467 {"KOI8-R", 20866},
1468 {"KOI8-U", 21866},
1469 {"KS_C_5601-1987", 949},
1470 {"LATIN1", 1252},
1471 {"LATIN2", 28592},
1472 {"MACINTOSH", 10000},
1473 {"SHIFT-JIS", 932},
1474 {"SHIFT_JIS", 932},
1475 {"SJIS", 932},
1476 {"US", 1252},
1477 {"US-ASCII", 1252},
1478 {"UTF-16", 1200},
1479 {"UTF-16BE", 1201},
1480 {"UTF-16LE", 1200},
1481 {"UTF-8", CP_UTF8},
1482 {"X-EUROPA", 29001},
1483 {"X-MAC-ARABIC", 10004},
1484 {"X-MAC-CE", 10029},
1485 {"X-MAC-CHINESEIMP", 10008},
1486 {"X-MAC-CHINESETRAD", 10002},
1487 {"X-MAC-CROATIAN", 10082},
1488 {"X-MAC-CYRILLIC", 10007},
1489 {"X-MAC-GREEK", 10006},
1490 {"X-MAC-HEBREW", 10005},
1491 {"X-MAC-ICELANDIC", 10079},
1492 {"X-MAC-JAPANESE", 10001},
1493 {"X-MAC-KOREAN", 10003},
1494 {"X-MAC-ROMANIAN", 10010},
1495 {"X-MAC-THAI", 10021},
1496 {"X-MAC-TURKISH", 10081},
1497 {"X-MAC-UKRAINIAN", 10017},
1498 };
1499 static unsigned
make_codepage_from_charset(const char * charset)1500 make_codepage_from_charset(const char *charset)
1501 {
1502 char cs[16];
1503 char *p;
1504 unsigned cp;
1505 int a, b;
1506
1507 if (charset == NULL || strlen(charset) > 15)
1508 return -1;
1509
1510 /* Copy name to uppercase. */
1511 p = cs;
1512 while (*charset) {
1513 char c = *charset++;
1514 if (c >= 'a' && c <= 'z')
1515 c -= 'a' - 'A';
1516 *p++ = c;
1517 }
1518 *p++ = '\0';
1519 cp = -1;
1520
1521 /* Look it up in the table first, so that we can easily
1522 * override CP367, which we map to 1252 instead of 367. */
1523 a = 0;
1524 b = sizeof(charsets)/sizeof(charsets[0]);
1525 while (b > a) {
1526 int c = (b + a) / 2;
1527 int r = strcmp(charsets[c].name, cs);
1528 if (r < 0)
1529 a = c + 1;
1530 else if (r > 0)
1531 b = c;
1532 else
1533 return charsets[c].cp;
1534 }
1535
1536 /* If it's not in the table, try to parse it. */
1537 switch (*cs) {
1538 case 'C':
1539 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1540 cp = my_atoi(cs + 2);
1541 } else if (strcmp(cs, "CP_ACP") == 0)
1542 cp = get_current_codepage();
1543 else if (strcmp(cs, "CP_OEMCP") == 0)
1544 cp = get_current_oemcp();
1545 break;
1546 case 'I':
1547 if (cs[1] == 'B' && cs[2] == 'M' &&
1548 cs[3] >= '0' && cs[3] <= '9') {
1549 cp = my_atoi(cs + 3);
1550 }
1551 break;
1552 case 'W':
1553 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1554 cp = my_atoi(cs + 8);
1555 if (cp != 874 && (cp < 1250 || cp > 1258))
1556 cp = -1;/* This may invalid code. */
1557 }
1558 break;
1559 }
1560 return (cp);
1561 }
1562
1563 /*
1564 * Return ANSI Code Page of current locale set by setlocale().
1565 */
1566 static unsigned
get_current_codepage(void)1567 get_current_codepage(void)
1568 {
1569 char *locale, *p;
1570 unsigned cp;
1571
1572 locale = setlocale(LC_CTYPE, NULL);
1573 if (locale == NULL)
1574 return (GetACP());
1575 if (locale[0] == 'C' && locale[1] == '\0')
1576 return (CP_C_LOCALE);
1577 p = strrchr(locale, '.');
1578 if (p == NULL)
1579 return (GetACP());
1580 if ((strcmp(p+1, "utf8") == 0) || (strcmp(p+1, "UTF-8") == 0))
1581 return CP_UTF8;
1582 cp = my_atoi(p+1);
1583 if ((int)cp <= 0)
1584 return (GetACP());
1585 return (cp);
1586 }
1587
1588 /*
1589 * Translation table between Locale Name and ACP/OEMCP.
1590 */
1591 static struct {
1592 unsigned acp;
1593 unsigned ocp;
1594 const char *locale;
1595 } acp_ocp_map[] = {
1596 { 950, 950, "Chinese_Taiwan" },
1597 { 936, 936, "Chinese_People's Republic of China" },
1598 { 950, 950, "Chinese_Taiwan" },
1599 { 1250, 852, "Czech_Czech Republic" },
1600 { 1252, 850, "Danish_Denmark" },
1601 { 1252, 850, "Dutch_Netherlands" },
1602 { 1252, 850, "Dutch_Belgium" },
1603 { 1252, 437, "English_United States" },
1604 { 1252, 850, "English_Australia" },
1605 { 1252, 850, "English_Canada" },
1606 { 1252, 850, "English_New Zealand" },
1607 { 1252, 850, "English_United Kingdom" },
1608 { 1252, 437, "English_United States" },
1609 { 1252, 850, "Finnish_Finland" },
1610 { 1252, 850, "French_France" },
1611 { 1252, 850, "French_Belgium" },
1612 { 1252, 850, "French_Canada" },
1613 { 1252, 850, "French_Switzerland" },
1614 { 1252, 850, "German_Germany" },
1615 { 1252, 850, "German_Austria" },
1616 { 1252, 850, "German_Switzerland" },
1617 { 1253, 737, "Greek_Greece" },
1618 { 1250, 852, "Hungarian_Hungary" },
1619 { 1252, 850, "Icelandic_Iceland" },
1620 { 1252, 850, "Italian_Italy" },
1621 { 1252, 850, "Italian_Switzerland" },
1622 { 932, 932, "Japanese_Japan" },
1623 { 949, 949, "Korean_Korea" },
1624 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1625 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1626 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1627 { 1250, 852, "Polish_Poland" },
1628 { 1252, 850, "Portuguese_Portugal" },
1629 { 1252, 850, "Portuguese_Brazil" },
1630 { 1251, 866, "Russian_Russia" },
1631 { 1250, 852, "Slovak_Slovakia" },
1632 { 1252, 850, "Spanish_Spain" },
1633 { 1252, 850, "Spanish_Mexico" },
1634 { 1252, 850, "Spanish_Spain" },
1635 { 1252, 850, "Swedish_Sweden" },
1636 { 1254, 857, "Turkish_Turkey" },
1637 { 0, 0, NULL}
1638 };
1639
1640 /*
1641 * Return OEM Code Page of current locale set by setlocale().
1642 */
1643 static unsigned
get_current_oemcp(void)1644 get_current_oemcp(void)
1645 {
1646 int i;
1647 char *locale, *p;
1648 size_t len;
1649
1650 locale = setlocale(LC_CTYPE, NULL);
1651 if (locale == NULL)
1652 return (GetOEMCP());
1653 if (locale[0] == 'C' && locale[1] == '\0')
1654 return (CP_C_LOCALE);
1655
1656 p = strrchr(locale, '.');
1657 if (p == NULL)
1658 return (GetOEMCP());
1659 len = p - locale;
1660 for (i = 0; acp_ocp_map[i].acp; i++) {
1661 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1662 return (acp_ocp_map[i].ocp);
1663 }
1664 return (GetOEMCP());
1665 }
1666 #else
1667
1668 /*
1669 * POSIX platform does not use CodePage.
1670 */
1671
1672 static unsigned
get_current_codepage(void)1673 get_current_codepage(void)
1674 {
1675 return (-1);/* Unknown */
1676 }
1677 static unsigned
make_codepage_from_charset(const char * charset)1678 make_codepage_from_charset(const char *charset)
1679 {
1680 (void)charset; /* UNUSED */
1681 return (-1);/* Unknown */
1682 }
1683 static unsigned
get_current_oemcp(void)1684 get_current_oemcp(void)
1685 {
1686 return (-1);/* Unknown */
1687 }
1688
1689 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1690
1691 /*
1692 * Return a string conversion object.
1693 */
1694 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1695 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1696 {
1697 struct archive_string_conv *sc;
1698 unsigned current_codepage;
1699
1700 /* Check if we have made the sconv object. */
1701 sc = find_sconv_object(a, fc, tc);
1702 if (sc != NULL)
1703 return (sc);
1704
1705 if (a == NULL)
1706 current_codepage = get_current_codepage();
1707 else
1708 current_codepage = a->current_codepage;
1709
1710 sc = create_sconv_object(canonical_charset_name(fc),
1711 canonical_charset_name(tc), current_codepage, flag);
1712 if (sc == NULL) {
1713 if (a != NULL)
1714 archive_set_error(a, ENOMEM,
1715 "Could not allocate memory for "
1716 "a string conversion object");
1717 return (NULL);
1718 }
1719
1720 /*
1721 * If there is no converter for current string conversion object,
1722 * we cannot handle this conversion.
1723 */
1724 if (sc->nconverter == 0) {
1725 if (a != NULL) {
1726 #if HAVE_ICONV
1727 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1728 "iconv_open failed: Cannot handle ``%s''",
1729 (flag & SCONV_TO_CHARSET)?tc:fc);
1730 #else
1731 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1732 "A character-set conversion not fully supported "
1733 "on this platform");
1734 #endif
1735 }
1736 /* Failed; free a sconv object. */
1737 free_sconv_object(sc);
1738 return (NULL);
1739 }
1740
1741 /*
1742 * Success!
1743 */
1744 if (a != NULL)
1745 add_sconv_object(a, sc);
1746 return (sc);
1747 }
1748
1749 static const char *
get_current_charset(struct archive * a)1750 get_current_charset(struct archive *a)
1751 {
1752 const char *cur_charset;
1753
1754 if (a == NULL)
1755 cur_charset = default_iconv_charset("");
1756 else {
1757 cur_charset = default_iconv_charset(a->current_code);
1758 if (a->current_code == NULL) {
1759 a->current_code = strdup(cur_charset);
1760 a->current_codepage = get_current_codepage();
1761 a->current_oemcp = get_current_oemcp();
1762 }
1763 }
1764 return (cur_charset);
1765 }
1766
1767 /*
1768 * Make and Return a string conversion object.
1769 * Return NULL if the platform does not support the specified conversion
1770 * and best_effort is 0.
1771 * If best_effort is set, A string conversion object must be returned
1772 * unless memory allocation for the object fails, but the conversion
1773 * might fail when non-ASCII code is found.
1774 */
1775 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1776 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1777 int best_effort)
1778 {
1779 int flag = SCONV_TO_CHARSET;
1780
1781 if (best_effort)
1782 flag |= SCONV_BEST_EFFORT;
1783 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1784 }
1785
1786 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1787 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1788 int best_effort)
1789 {
1790 int flag = SCONV_FROM_CHARSET;
1791
1792 if (best_effort)
1793 flag |= SCONV_BEST_EFFORT;
1794 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1795 }
1796
1797 /*
1798 * archive_string_default_conversion_*_archive() are provided for Windows
1799 * platform because other archiver application use CP_OEMCP for
1800 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1801 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1802 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1803 * So we should make a string conversion between CP_ACP and CP_OEMCP
1804 * for compatibility.
1805 */
1806 #if defined(_WIN32) && !defined(__CYGWIN__)
1807 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1808 archive_string_default_conversion_for_read(struct archive *a)
1809 {
1810 const char *cur_charset = get_current_charset(a);
1811 char oemcp[16];
1812
1813 /* NOTE: a check of cur_charset is unneeded but we need
1814 * that get_current_charset() has been surely called at
1815 * this time whatever C compiler optimized. */
1816 if (cur_charset != NULL &&
1817 (a->current_codepage == CP_C_LOCALE ||
1818 a->current_codepage == a->current_oemcp))
1819 return (NULL);/* no conversion. */
1820
1821 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1822 /* Make sure a null termination must be set. */
1823 oemcp[sizeof(oemcp)-1] = '\0';
1824 return (get_sconv_object(a, oemcp, cur_charset,
1825 SCONV_FROM_CHARSET));
1826 }
1827
1828 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1829 archive_string_default_conversion_for_write(struct archive *a)
1830 {
1831 const char *cur_charset = get_current_charset(a);
1832 char oemcp[16];
1833
1834 /* NOTE: a check of cur_charset is unneeded but we need
1835 * that get_current_charset() has been surely called at
1836 * this time whatever C compiler optimized. */
1837 if (cur_charset != NULL &&
1838 (a->current_codepage == CP_C_LOCALE ||
1839 a->current_codepage == a->current_oemcp))
1840 return (NULL);/* no conversion. */
1841
1842 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1843 /* Make sure a null termination must be set. */
1844 oemcp[sizeof(oemcp)-1] = '\0';
1845 return (get_sconv_object(a, cur_charset, oemcp,
1846 SCONV_TO_CHARSET));
1847 }
1848 #else
1849 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1850 archive_string_default_conversion_for_read(struct archive *a)
1851 {
1852 (void)a; /* UNUSED */
1853 return (NULL);
1854 }
1855
1856 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1857 archive_string_default_conversion_for_write(struct archive *a)
1858 {
1859 (void)a; /* UNUSED */
1860 return (NULL);
1861 }
1862 #endif
1863
1864 /*
1865 * Dispose of all character conversion objects in the archive object.
1866 */
1867 void
archive_string_conversion_free(struct archive * a)1868 archive_string_conversion_free(struct archive *a)
1869 {
1870 struct archive_string_conv *sc;
1871 struct archive_string_conv *sc_next;
1872
1873 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1874 sc_next = sc->next;
1875 free_sconv_object(sc);
1876 }
1877 a->sconv = NULL;
1878 free(a->current_code);
1879 a->current_code = NULL;
1880 }
1881
1882 /*
1883 * Return a conversion charset name.
1884 */
1885 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1886 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1887 {
1888 if (sc == NULL) {
1889 return "current locale";
1890 }
1891 if (sc->flag & SCONV_TO_CHARSET)
1892 return (sc->to_charset);
1893 else
1894 return (sc->from_charset);
1895 }
1896
1897 /*
1898 * Change the behavior of a string conversion.
1899 */
1900 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1901 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1902 {
1903 switch (opt) {
1904 /*
1905 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1906 * assumption that wchar_t was Unicode.
1907 * This option enables simulating the assumption in order to read
1908 * that filename correctly.
1909 */
1910 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1911 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1912 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1913 /*
1914 * Nothing to do for it since wchar_t on these platforms
1915 * is really Unicode.
1916 */
1917 (void)sc; /* UNUSED */
1918 #else
1919 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1920 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1921 /* Set up string converters. */
1922 setup_converter(sc);
1923 }
1924 #endif
1925 break;
1926 case SCONV_SET_OPT_NORMALIZATION_C:
1927 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1928 sc->flag |= SCONV_NORMALIZATION_C;
1929 sc->flag &= ~SCONV_NORMALIZATION_D;
1930 /* Set up string converters. */
1931 setup_converter(sc);
1932 }
1933 break;
1934 case SCONV_SET_OPT_NORMALIZATION_D:
1935 #if defined(HAVE_ICONV)
1936 /*
1937 * If iconv will take the string, do not change the
1938 * setting of the normalization.
1939 */
1940 if (!(sc->flag & SCONV_WIN_CP) &&
1941 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1942 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1943 break;
1944 #endif
1945 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1946 sc->flag |= SCONV_NORMALIZATION_D;
1947 sc->flag &= ~SCONV_NORMALIZATION_C;
1948 /* Set up string converters. */
1949 setup_converter(sc);
1950 }
1951 break;
1952 default:
1953 break;
1954 }
1955 }
1956
1957 /*
1958 *
1959 * Copy one archive_string to another in locale conversion.
1960 *
1961 * archive_strncat_l();
1962 * archive_strncpy_l();
1963 *
1964 */
1965
1966 static size_t
mbsnbytes(const void * _p,size_t n)1967 mbsnbytes(const void *_p, size_t n)
1968 {
1969 size_t s;
1970 const char *p, *pp;
1971
1972 if (_p == NULL)
1973 return (0);
1974 p = (const char *)_p;
1975
1976 /* Like strlen(p), except won't examine positions beyond p[n]. */
1977 s = 0;
1978 pp = p;
1979 while (s < n && *pp) {
1980 pp++;
1981 s++;
1982 }
1983 return (s);
1984 }
1985
1986 static size_t
utf16nbytes(const void * _p,size_t n)1987 utf16nbytes(const void *_p, size_t n)
1988 {
1989 size_t s;
1990 const char *p, *pp;
1991
1992 if (_p == NULL)
1993 return (0);
1994 p = (const char *)_p;
1995
1996 /* Like strlen(p), except won't examine positions beyond p[n]. */
1997 s = 0;
1998 pp = p;
1999 n >>= 1;
2000 while (s < n && (pp[0] || pp[1])) {
2001 pp += 2;
2002 s++;
2003 }
2004 return (s<<1);
2005 }
2006
2007 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2008 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
2009 struct archive_string_conv *sc)
2010 {
2011 as->length = 0;
2012 return (archive_strncat_l(as, _p, n, sc));
2013 }
2014
2015 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)2016 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
2017 struct archive_string_conv *sc)
2018 {
2019 const void *s;
2020 size_t length = 0;
2021 int i, r = 0, r2;
2022
2023 if (_p != NULL && n > 0) {
2024 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
2025 length = utf16nbytes(_p, n);
2026 else
2027 length = mbsnbytes(_p, n);
2028 }
2029
2030 /* We must allocate memory even if there is no data for conversion
2031 * or copy. This simulates archive_string_append behavior. */
2032 if (length == 0) {
2033 size_t tn = 1;
2034 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
2035 tn = 2;
2036 if (archive_string_ensure(as, as->length + tn) == NULL)
2037 return (-1);
2038 as->s[as->length] = 0;
2039 if (tn == 2)
2040 as->s[as->length+1] = 0;
2041 return (0);
2042 }
2043
2044 /*
2045 * If sc is NULL, we just make a copy.
2046 */
2047 if (sc == NULL) {
2048 if (archive_string_append(as, _p, length) == NULL)
2049 return (-1);/* No memory */
2050 return (0);
2051 }
2052
2053 s = _p;
2054 i = 0;
2055 if (sc->nconverter > 1) {
2056 sc->utftmp.length = 0;
2057 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2058 if (r2 != 0 && errno == ENOMEM)
2059 return (r2);
2060 if (r > r2)
2061 r = r2;
2062 s = sc->utftmp.s;
2063 length = sc->utftmp.length;
2064 ++i;
2065 }
2066 r2 = sc->converter[i](as, s, length, sc);
2067 if (r > r2)
2068 r = r2;
2069 return (r);
2070 }
2071
2072 struct archive_string *
archive_string_dirname(struct archive_string * as)2073 archive_string_dirname(struct archive_string *as)
2074 {
2075 /* strip trailing separators */
2076 while (as->length > 1 && as->s[as->length - 1] == '/')
2077 as->length--;
2078 /* strip final component */
2079 while (as->length > 0 && as->s[as->length - 1] != '/')
2080 as->length--;
2081 /* empty path -> cwd */
2082 if (as->length == 0)
2083 return (archive_strcat(as, "."));
2084 /* strip separator(s) */
2085 while (as->length > 1 && as->s[as->length - 1] == '/')
2086 as->length--;
2087 /* terminate */
2088 as->s[as->length] = '\0';
2089 return (as);
2090 }
2091
2092 #if HAVE_ICONV
2093
2094 /*
2095 * Return -1 if conversion fails.
2096 */
2097 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2098 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2099 size_t length, struct archive_string_conv *sc)
2100 {
2101 ICONV_CONST char *itp;
2102 size_t remaining;
2103 iconv_t cd;
2104 char *outp;
2105 size_t avail, bs;
2106 int return_value = 0; /* success */
2107 size_t to_size, from_size;
2108
2109 if (sc->flag & SCONV_TO_UTF16)
2110 to_size = 2;
2111 else
2112 to_size = 1;
2113 if (sc->flag & SCONV_FROM_UTF16)
2114 from_size = 2;
2115 else
2116 from_size = 1;
2117
2118 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2119 return (-1);
2120
2121 cd = sc->cd;
2122 itp = (char *)(uintptr_t)_p;
2123 remaining = length;
2124 outp = as->s + as->length;
2125 avail = as->buffer_length - as->length - to_size;
2126 while (remaining >= from_size) {
2127 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2128
2129 if (result != (size_t)-1)
2130 break; /* Conversion completed. */
2131
2132 if (errno == EILSEQ || errno == EINVAL) {
2133 /*
2134 * If an output charset is UTF-8 or UTF-16BE/LE,
2135 * unknown character should be U+FFFD
2136 * (replacement character).
2137 */
2138 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2139 size_t rbytes;
2140 if (sc->flag & SCONV_TO_UTF8)
2141 rbytes = sizeof(utf8_replacement_char);
2142 else
2143 rbytes = 2;
2144
2145 if (avail < rbytes) {
2146 as->length = outp - as->s;
2147 bs = as->buffer_length +
2148 (remaining * to_size) + rbytes;
2149 if (NULL ==
2150 archive_string_ensure(as, bs))
2151 return (-1);
2152 outp = as->s + as->length;
2153 avail = as->buffer_length
2154 - as->length - to_size;
2155 }
2156 if (sc->flag & SCONV_TO_UTF8)
2157 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2158 else if (sc->flag & SCONV_TO_UTF16BE)
2159 archive_be16enc(outp, UNICODE_R_CHAR);
2160 else
2161 archive_le16enc(outp, UNICODE_R_CHAR);
2162 outp += rbytes;
2163 avail -= rbytes;
2164 } else {
2165 /* Skip the illegal input bytes. */
2166 *outp++ = '?';
2167 avail--;
2168 }
2169 itp += from_size;
2170 remaining -= from_size;
2171 return_value = -1; /* failure */
2172 } else {
2173 /* E2BIG no output buffer,
2174 * Increase an output buffer. */
2175 as->length = outp - as->s;
2176 bs = as->buffer_length + remaining * 2;
2177 if (NULL == archive_string_ensure(as, bs))
2178 return (-1);
2179 outp = as->s + as->length;
2180 avail = as->buffer_length - as->length - to_size;
2181 }
2182 }
2183 as->length = outp - as->s;
2184 as->s[as->length] = 0;
2185 if (to_size == 2)
2186 as->s[as->length+1] = 0;
2187 return (return_value);
2188 }
2189
2190 #endif /* HAVE_ICONV */
2191
2192
2193 #if defined(_WIN32) && !defined(__CYGWIN__)
2194
2195 /*
2196 * Translate a string from a some CodePage to an another CodePage by
2197 * Windows APIs, and copy the result. Return -1 if conversion fails.
2198 */
2199 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2200 strncat_in_codepage(struct archive_string *as,
2201 const void *_p, size_t length, struct archive_string_conv *sc)
2202 {
2203 const char *s = (const char *)_p;
2204 struct archive_wstring aws;
2205 size_t l;
2206 int r, saved_flag;
2207
2208 archive_string_init(&aws);
2209 saved_flag = sc->flag;
2210 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2211 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2212 sc->flag = saved_flag;
2213 if (r != 0) {
2214 archive_wstring_free(&aws);
2215 if (errno != ENOMEM)
2216 archive_string_append(as, s, length);
2217 return (-1);
2218 }
2219
2220 l = as->length;
2221 r = archive_string_append_from_wcs_in_codepage(
2222 as, aws.s, aws.length, sc);
2223 if (r != 0 && errno != ENOMEM && l == as->length)
2224 archive_string_append(as, s, length);
2225 archive_wstring_free(&aws);
2226 return (r);
2227 }
2228
2229 /*
2230 * Test whether MBS ==> WCS is okay.
2231 */
2232 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2233 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2234 {
2235 const char *p = (const char *)_p;
2236 unsigned codepage;
2237 DWORD mbflag = MB_ERR_INVALID_CHARS;
2238
2239 if (sc->flag & SCONV_FROM_CHARSET)
2240 codepage = sc->to_cp;
2241 else
2242 codepage = sc->from_cp;
2243
2244 if (codepage == CP_C_LOCALE)
2245 return (0);
2246 if (codepage != CP_UTF8)
2247 mbflag |= MB_PRECOMPOSED;
2248
2249 if (n > (size_t)INT_MAX)
2250 return (-1); /* Invalid */
2251 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2252 return (-1); /* Invalid */
2253 return (0); /* Okay */
2254 }
2255
2256 #else
2257
2258 /*
2259 * Test whether MBS ==> WCS is okay.
2260 */
2261 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2262 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2263 {
2264 const char *p = (const char *)_p;
2265 size_t r;
2266
2267 #if HAVE_MBRTOWC
2268 mbstate_t shift_state;
2269
2270 memset(&shift_state, 0, sizeof(shift_state));
2271 #else
2272 /* Clear the shift state before starting. */
2273 mbtowc(NULL, NULL, 0);
2274 #endif
2275 while (n) {
2276 wchar_t wc;
2277
2278 #if HAVE_MBRTOWC
2279 r = mbrtowc(&wc, p, n, &shift_state);
2280 #else
2281 r = mbtowc(&wc, p, n);
2282 #endif
2283 if (r == (size_t)-1 || r == (size_t)-2)
2284 return (-1);/* Invalid. */
2285 if (r == 0)
2286 break;
2287 p += r;
2288 n -= r;
2289 }
2290 (void)sc; /* UNUSED */
2291 return (0); /* All Okey. */
2292 }
2293
2294 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2295
2296 /*
2297 * Basically returns -1 because we cannot make a conversion of charset
2298 * without iconv but in some cases this would return 0.
2299 * Returns 0 if all copied characters are ASCII.
2300 * Returns 0 if both from-locale and to-locale are the same and those
2301 * can be WCS with no error.
2302 */
2303 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2304 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2305 size_t length, struct archive_string_conv *sc)
2306 {
2307 size_t remaining;
2308 const uint8_t *itp;
2309 int return_value = 0; /* success */
2310
2311 /*
2312 * If both from-locale and to-locale is the same, this makes a copy.
2313 * And then this checks all copied MBS can be WCS if so returns 0.
2314 */
2315 if (sc->same) {
2316 if (archive_string_append(as, _p, length) == NULL)
2317 return (-1);/* No memory */
2318 return (invalid_mbs(_p, length, sc));
2319 }
2320
2321 /*
2322 * If a character is ASCII, this just copies it. If not, this
2323 * assigns '?' character instead but in UTF-8 locale this assigns
2324 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2325 * a Replacement Character in Unicode.
2326 */
2327
2328 remaining = length;
2329 itp = (const uint8_t *)_p;
2330 while (remaining > 0 && *itp) {
2331 if (*itp > 127) {
2332 // Non-ASCII: Substitute with suitable replacement
2333 if (sc->flag & SCONV_TO_UTF8) {
2334 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2335 __archive_errx(1, "Out of memory");
2336 }
2337 } else {
2338 archive_strappend_char(as, '?');
2339 }
2340 return_value = -1;
2341 } else {
2342 archive_strappend_char(as, *itp);
2343 }
2344 ++itp;
2345 --remaining;
2346 }
2347 return (return_value);
2348 }
2349
2350
2351 /*
2352 * Unicode conversion functions.
2353 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2354 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2355 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2356 * - UTF-16BE <===> UTF-8.
2357 *
2358 */
2359
2360 /*
2361 * Utility to convert a single UTF-8 sequence.
2362 *
2363 * Usually return used bytes, return used byte in negative value when
2364 * a unicode character is replaced with U+FFFD.
2365 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2366 * Recommended Practice for Replacement Characters.
2367 */
2368 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2369 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2370 {
2371 static const char utf8_count[256] = {
2372 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2373 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2374 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2375 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2376 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2377 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2378 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2379 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2380 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2381 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2382 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2383 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2384 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2385 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2386 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2387 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2388 };
2389 int ch, i;
2390 int cnt;
2391 uint32_t wc;
2392
2393 /* Sanity check. */
2394 if (n == 0)
2395 return (0);
2396 /*
2397 * Decode 1-4 bytes depending on the value of the first byte.
2398 */
2399 ch = (unsigned char)*s;
2400 if (ch == 0)
2401 return (0); /* Standard: return 0 for end-of-string. */
2402 cnt = utf8_count[ch];
2403
2404 /* Invalid sequence or there are not plenty bytes. */
2405 if (n < (size_t)cnt) {
2406 cnt = (int)n;
2407 for (i = 1; i < cnt; i++) {
2408 if ((s[i] & 0xc0) != 0x80) {
2409 cnt = i;
2410 break;
2411 }
2412 }
2413 goto invalid_sequence;
2414 }
2415
2416 /* Make a Unicode code point from a single UTF-8 sequence. */
2417 switch (cnt) {
2418 case 1: /* 1 byte sequence. */
2419 *pwc = ch & 0x7f;
2420 return (cnt);
2421 case 2: /* 2 bytes sequence. */
2422 if ((s[1] & 0xc0) != 0x80) {
2423 cnt = 1;
2424 goto invalid_sequence;
2425 }
2426 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2427 return (cnt);
2428 case 3: /* 3 bytes sequence. */
2429 if ((s[1] & 0xc0) != 0x80) {
2430 cnt = 1;
2431 goto invalid_sequence;
2432 }
2433 if ((s[2] & 0xc0) != 0x80) {
2434 cnt = 2;
2435 goto invalid_sequence;
2436 }
2437 wc = ((ch & 0x0f) << 12)
2438 | ((s[1] & 0x3f) << 6)
2439 | (s[2] & 0x3f);
2440 if (wc < 0x800)
2441 goto invalid_sequence;/* Overlong sequence. */
2442 break;
2443 case 4: /* 4 bytes sequence. */
2444 if ((s[1] & 0xc0) != 0x80) {
2445 cnt = 1;
2446 goto invalid_sequence;
2447 }
2448 if ((s[2] & 0xc0) != 0x80) {
2449 cnt = 2;
2450 goto invalid_sequence;
2451 }
2452 if ((s[3] & 0xc0) != 0x80) {
2453 cnt = 3;
2454 goto invalid_sequence;
2455 }
2456 wc = ((ch & 0x07) << 18)
2457 | ((s[1] & 0x3f) << 12)
2458 | ((s[2] & 0x3f) << 6)
2459 | (s[3] & 0x3f);
2460 if (wc < 0x10000)
2461 goto invalid_sequence;/* Overlong sequence. */
2462 break;
2463 default: /* Others are all invalid sequence. */
2464 if (ch == 0xc0 || ch == 0xc1)
2465 cnt = 2;
2466 else if (ch >= 0xf5 && ch <= 0xf7)
2467 cnt = 4;
2468 else if (ch >= 0xf8 && ch <= 0xfb)
2469 cnt = 5;
2470 else if (ch == 0xfc || ch == 0xfd)
2471 cnt = 6;
2472 else
2473 cnt = 1;
2474 if (n < (size_t)cnt)
2475 cnt = (int)n;
2476 for (i = 1; i < cnt; i++) {
2477 if ((s[i] & 0xc0) != 0x80) {
2478 cnt = i;
2479 break;
2480 }
2481 }
2482 goto invalid_sequence;
2483 }
2484
2485 /* The code point larger than 0x10FFFF is not legal
2486 * Unicode values. */
2487 if (wc > UNICODE_MAX)
2488 goto invalid_sequence;
2489 /* Correctly gets a Unicode, returns used bytes. */
2490 *pwc = wc;
2491 return (cnt);
2492 invalid_sequence:
2493 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2494 return (cnt * -1);
2495 }
2496
2497 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2498 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2499 {
2500 int cnt;
2501
2502 cnt = _utf8_to_unicode(pwc, s, n);
2503 /* Any of Surrogate pair is not legal Unicode values. */
2504 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2505 return (-3);
2506 return (cnt);
2507 }
2508
2509 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2510 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2511 {
2512 uc -= 0xD800;
2513 uc *= 0x400;
2514 uc += uc2 - 0xDC00;
2515 uc += 0x10000;
2516 return (uc);
2517 }
2518
2519 /*
2520 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2521 * removing surrogate pairs.
2522 *
2523 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2524 *
2525 * Usually return used bytes, return used byte in negative value when
2526 * a unicode character is replaced with U+FFFD.
2527 */
2528 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2529 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2530 {
2531 uint32_t wc = 0;
2532 int cnt;
2533
2534 cnt = _utf8_to_unicode(&wc, s, n);
2535 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2536 uint32_t wc2 = 0;
2537 if (n - 3 < 3) {
2538 /* Invalid byte sequence. */
2539 goto invalid_sequence;
2540 }
2541 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2542 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2543 /* Invalid byte sequence. */
2544 goto invalid_sequence;
2545 }
2546 wc = combine_surrogate_pair(wc, wc2);
2547 cnt = 6;
2548 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2549 /* Invalid byte sequence. */
2550 goto invalid_sequence;
2551 }
2552 *pwc = wc;
2553 return (cnt);
2554 invalid_sequence:
2555 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2556 if (cnt > 0)
2557 cnt *= -1;
2558 return (cnt);
2559 }
2560
2561 /*
2562 * Convert a Unicode code point to a single UTF-8 sequence.
2563 *
2564 * NOTE:This function does not check if the Unicode is legal or not.
2565 * Please you definitely check it before calling this.
2566 */
2567 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2568 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2569 {
2570 char *_p = p;
2571
2572 /* Invalid Unicode char maps to Replacement character */
2573 if (uc > UNICODE_MAX)
2574 uc = UNICODE_R_CHAR;
2575 /* Translate code point to UTF8 */
2576 if (uc <= 0x7f) {
2577 if (remaining == 0)
2578 return (0);
2579 *p++ = (char)uc;
2580 } else if (uc <= 0x7ff) {
2581 if (remaining < 2)
2582 return (0);
2583 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2584 *p++ = 0x80 | (uc & 0x3f);
2585 } else if (uc <= 0xffff) {
2586 if (remaining < 3)
2587 return (0);
2588 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2589 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2590 *p++ = 0x80 | (uc & 0x3f);
2591 } else {
2592 if (remaining < 4)
2593 return (0);
2594 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2595 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2596 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2597 *p++ = 0x80 | (uc & 0x3f);
2598 }
2599 return (p - _p);
2600 }
2601
2602 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2603 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2604 {
2605 return (utf16_to_unicode(pwc, s, n, 1));
2606 }
2607
2608 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2609 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2610 {
2611 return (utf16_to_unicode(pwc, s, n, 0));
2612 }
2613
2614 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2615 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2616 {
2617 const char *utf16 = s;
2618 unsigned uc;
2619
2620 if (n == 0)
2621 return (0);
2622 if (n == 1) {
2623 /* set the Replacement Character instead. */
2624 *pwc = UNICODE_R_CHAR;
2625 return (-1);
2626 }
2627
2628 if (be)
2629 uc = archive_be16dec(utf16);
2630 else
2631 uc = archive_le16dec(utf16);
2632 utf16 += 2;
2633
2634 /* If this is a surrogate pair, assemble the full code point.*/
2635 if (IS_HIGH_SURROGATE_LA(uc)) {
2636 unsigned uc2;
2637
2638 if (n >= 4) {
2639 if (be)
2640 uc2 = archive_be16dec(utf16);
2641 else
2642 uc2 = archive_le16dec(utf16);
2643 } else
2644 uc2 = 0;
2645 if (IS_LOW_SURROGATE_LA(uc2)) {
2646 uc = combine_surrogate_pair(uc, uc2);
2647 utf16 += 2;
2648 } else {
2649 /* Undescribed code point should be U+FFFD
2650 * (replacement character). */
2651 *pwc = UNICODE_R_CHAR;
2652 return (-2);
2653 }
2654 }
2655
2656 /*
2657 * Surrogate pair values(0xd800 through 0xdfff) are only
2658 * used by UTF-16, so, after above calculation, the code
2659 * must not be surrogate values, and Unicode has no codes
2660 * larger than 0x10ffff. Thus, those are not legal Unicode
2661 * values.
2662 */
2663 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2664 /* Undescribed code point should be U+FFFD
2665 * (replacement character). */
2666 *pwc = UNICODE_R_CHAR;
2667 return (((int)(utf16 - s)) * -1);
2668 }
2669 *pwc = uc;
2670 return ((int)(utf16 - s));
2671 }
2672
2673 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2674 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2675 {
2676 char *utf16 = p;
2677
2678 if (uc > 0xffff) {
2679 /* We have a code point that won't fit into a
2680 * wchar_t; convert it to a surrogate pair. */
2681 if (remaining < 4)
2682 return (0);
2683 uc -= 0x10000;
2684 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2685 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2686 return (4);
2687 } else {
2688 if (remaining < 2)
2689 return (0);
2690 archive_be16enc(utf16, (uint16_t)uc);
2691 return (2);
2692 }
2693 }
2694
2695 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2696 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2697 {
2698 char *utf16 = p;
2699
2700 if (uc > 0xffff) {
2701 /* We have a code point that won't fit into a
2702 * wchar_t; convert it to a surrogate pair. */
2703 if (remaining < 4)
2704 return (0);
2705 uc -= 0x10000;
2706 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2707 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2708 return (4);
2709 } else {
2710 if (remaining < 2)
2711 return (0);
2712 archive_le16enc(utf16, (uint16_t)uc);
2713 return (2);
2714 }
2715 }
2716
2717 /*
2718 * Append new UTF-8 string to existing UTF-8 string.
2719 * Existing string is assumed to already be in proper form;
2720 * the new string will have invalid sequences replaced and
2721 * surrogate pairs canonicalized.
2722 */
2723 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _src,size_t len,struct archive_string_conv * sc)2724 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2725 size_t len, struct archive_string_conv *sc)
2726 {
2727 int ret = 0;
2728 const char *src = _src;
2729 (void)sc; /* UNUSED */
2730
2731 /* Pre-extend the destination */
2732 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2733 return (-1);
2734
2735 /* Invariant: src points to the first UTF8 byte that hasn't
2736 * been copied to the destination `as`. */
2737 for (;;) {
2738 int n;
2739 uint32_t uc;
2740 const char *e = src;
2741
2742 /* Skip UTF-8 sequences until we reach end-of-string or
2743 * a code point that needs conversion. */
2744 while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2745 e += n;
2746 len -= n;
2747 }
2748 /* Copy the part that doesn't need conversion */
2749 if (e > src) {
2750 if (archive_string_append(as, src, e - src) == NULL)
2751 return (-1);
2752 src = e;
2753 }
2754
2755 if (n == 0) {
2756 /* We reached end-of-string */
2757 return (ret);
2758 } else {
2759 /* Next code point needs conversion */
2760 char t[4];
2761 size_t w;
2762
2763 /* Try decoding a surrogate pair */
2764 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2765 n = cesu8_to_unicode(&uc, src, len);
2766 }
2767 /* Not a (valid) surrogate, so use a replacement char */
2768 if (n < 0) {
2769 ret = -1; /* Return -1 if we used any replacement */
2770 n *= -1;
2771 }
2772 /* Consume converted code point */
2773 src += n;
2774 len -= n;
2775 /* Convert and append new UTF-8 sequence. */
2776 w = unicode_to_utf8(t, sizeof(t), uc);
2777 if (archive_string_append(as, t, w) == NULL)
2778 return (-1);
2779 }
2780 }
2781 }
2782
2783 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2784 archive_string_append_unicode(struct archive_string *as, const void *_p,
2785 size_t len, struct archive_string_conv *sc)
2786 {
2787 const char *s;
2788 char *p, *endp;
2789 uint32_t uc;
2790 size_t w;
2791 size_t ts, tm;
2792 int n, ret = 0;
2793 int (*parse)(uint32_t *, const char *, size_t);
2794 size_t (*unparse)(char *, size_t, uint32_t);
2795
2796 if (sc->flag & SCONV_TO_UTF16BE) {
2797 unparse = unicode_to_utf16be;
2798 ts = 2;
2799 } else if (sc->flag & SCONV_TO_UTF16LE) {
2800 unparse = unicode_to_utf16le;
2801 ts = 2;
2802 } else if (sc->flag & SCONV_TO_UTF8) {
2803 unparse = unicode_to_utf8;
2804 ts = 1;
2805 } else {
2806 /*
2807 * This case is going to be converted to another
2808 * character-set through iconv.
2809 */
2810 if (sc->flag & SCONV_FROM_UTF16BE) {
2811 unparse = unicode_to_utf16be;
2812 ts = 2;
2813 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2814 unparse = unicode_to_utf16le;
2815 ts = 2;
2816 } else {
2817 unparse = unicode_to_utf8;
2818 ts = 1;
2819 }
2820 }
2821
2822 if (sc->flag & SCONV_FROM_UTF16BE) {
2823 parse = utf16be_to_unicode;
2824 tm = 1;
2825 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2826 parse = utf16le_to_unicode;
2827 tm = 1;
2828 } else {
2829 parse = cesu8_to_unicode;
2830 tm = ts;
2831 }
2832
2833 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2834 return (-1);
2835
2836 s = (const char *)_p;
2837 p = as->s + as->length;
2838 endp = as->s + as->buffer_length - ts;
2839 while ((n = parse(&uc, s, len)) != 0) {
2840 if (n < 0) {
2841 /* Use a replaced unicode character. */
2842 n *= -1;
2843 ret = -1;
2844 }
2845 s += n;
2846 len -= n;
2847 while ((w = unparse(p, endp - p, uc)) == 0) {
2848 /* There is not enough output buffer so
2849 * we have to expand it. */
2850 as->length = p - as->s;
2851 if (archive_string_ensure(as,
2852 as->buffer_length + len * tm + ts) == NULL)
2853 return (-1);
2854 p = as->s + as->length;
2855 endp = as->s + as->buffer_length - ts;
2856 }
2857 p += w;
2858 }
2859 as->length = p - as->s;
2860 as->s[as->length] = '\0';
2861 if (ts == 2)
2862 as->s[as->length+1] = '\0';
2863 return (ret);
2864 }
2865
2866 /*
2867 * Following Constants for Hangul compositions this information comes from
2868 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2869 */
2870 #define HC_SBASE 0xAC00
2871 #define HC_LBASE 0x1100
2872 #define HC_VBASE 0x1161
2873 #define HC_TBASE 0x11A7
2874 #define HC_LCOUNT 19
2875 #define HC_VCOUNT 21
2876 #define HC_TCOUNT 28
2877 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2878 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2879
2880 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2881 get_nfc(uint32_t uc, uint32_t uc2)
2882 {
2883 int t, b;
2884
2885 t = 0;
2886 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2887 while (b >= t) {
2888 int m = (t + b) / 2;
2889 if (u_composition_table[m].cp1 < uc)
2890 t = m + 1;
2891 else if (u_composition_table[m].cp1 > uc)
2892 b = m - 1;
2893 else if (u_composition_table[m].cp2 < uc2)
2894 t = m + 1;
2895 else if (u_composition_table[m].cp2 > uc2)
2896 b = m - 1;
2897 else
2898 return (u_composition_table[m].nfc);
2899 }
2900 return (0);
2901 }
2902
2903 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2904 * Characters. */
2905
2906 /*
2907 * Update first code point.
2908 */
2909 #define UPDATE_UC(new_uc) do { \
2910 uc = new_uc; \
2911 ucptr = NULL; \
2912 } while (0)
2913
2914 /*
2915 * Replace first code point with second code point.
2916 */
2917 #define REPLACE_UC_WITH_UC2() do { \
2918 uc = uc2; \
2919 ucptr = uc2ptr; \
2920 n = n2; \
2921 } while (0)
2922
2923 #define EXPAND_BUFFER() do { \
2924 as->length = p - as->s; \
2925 if (archive_string_ensure(as, \
2926 as->buffer_length + len * tm + ts) == NULL)\
2927 return (-1); \
2928 p = as->s + as->length; \
2929 endp = as->s + as->buffer_length - ts; \
2930 } while (0)
2931
2932 #define UNPARSE(p, endp, uc) do { \
2933 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2934 EXPAND_BUFFER(); \
2935 } \
2936 p += w; \
2937 } while (0)
2938
2939 /*
2940 * Write first code point.
2941 * If the code point has not be changed from its original code,
2942 * this just copies it from its original buffer pointer.
2943 * If not, this converts it to UTF-8 byte sequence and copies it.
2944 */
2945 #define WRITE_UC() do { \
2946 if (ucptr) { \
2947 if (p + n > endp) \
2948 EXPAND_BUFFER(); \
2949 switch (n) { \
2950 case 4: \
2951 *p++ = *ucptr++; \
2952 /* FALL THROUGH */ \
2953 case 3: \
2954 *p++ = *ucptr++; \
2955 /* FALL THROUGH */ \
2956 case 2: \
2957 *p++ = *ucptr++; \
2958 /* FALL THROUGH */ \
2959 case 1: \
2960 *p++ = *ucptr; \
2961 break; \
2962 } \
2963 ucptr = NULL; \
2964 } else { \
2965 UNPARSE(p, endp, uc); \
2966 } \
2967 } while (0)
2968
2969 /*
2970 * Collect following decomposable code points.
2971 */
2972 #define COLLECT_CPS(start) do { \
2973 int _i; \
2974 for (_i = start; _i < FDC_MAX ; _i++) { \
2975 nx = parse(&ucx[_i], s, len); \
2976 if (nx <= 0) \
2977 break; \
2978 cx = CCC(ucx[_i]); \
2979 if (cl >= cx && cl != 228 && cx != 228)\
2980 break; \
2981 s += nx; \
2982 len -= nx; \
2983 cl = cx; \
2984 ccx[_i] = cx; \
2985 } \
2986 if (_i >= FDC_MAX) { \
2987 ret = -1; \
2988 ucx_size = FDC_MAX; \
2989 } else \
2990 ucx_size = _i; \
2991 } while (0)
2992
2993 /*
2994 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2995 *
2996 * TODO: Convert composition exclusions, which are never converted
2997 * from NFC,NFD,NFKC and NFKD, to Form C.
2998 */
2999 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3000 archive_string_normalize_C(struct archive_string *as, const void *_p,
3001 size_t len, struct archive_string_conv *sc)
3002 {
3003 const char *s = (const char *)_p;
3004 char *p, *endp;
3005 uint32_t uc, uc2;
3006 size_t w;
3007 int always_replace, n, n2, ret = 0, spair, ts, tm;
3008 int (*parse)(uint32_t *, const char *, size_t);
3009 size_t (*unparse)(char *, size_t, uint32_t);
3010
3011 always_replace = 1;
3012 ts = 1;/* text size. */
3013 if (sc->flag & SCONV_TO_UTF16BE) {
3014 unparse = unicode_to_utf16be;
3015 ts = 2;
3016 if (sc->flag & SCONV_FROM_UTF16BE)
3017 always_replace = 0;
3018 } else if (sc->flag & SCONV_TO_UTF16LE) {
3019 unparse = unicode_to_utf16le;
3020 ts = 2;
3021 if (sc->flag & SCONV_FROM_UTF16LE)
3022 always_replace = 0;
3023 } else if (sc->flag & SCONV_TO_UTF8) {
3024 unparse = unicode_to_utf8;
3025 if (sc->flag & SCONV_FROM_UTF8)
3026 always_replace = 0;
3027 } else {
3028 /*
3029 * This case is going to be converted to another
3030 * character-set through iconv.
3031 */
3032 always_replace = 0;
3033 if (sc->flag & SCONV_FROM_UTF16BE) {
3034 unparse = unicode_to_utf16be;
3035 ts = 2;
3036 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3037 unparse = unicode_to_utf16le;
3038 ts = 2;
3039 } else {
3040 unparse = unicode_to_utf8;
3041 }
3042 }
3043
3044 if (sc->flag & SCONV_FROM_UTF16BE) {
3045 parse = utf16be_to_unicode;
3046 tm = 1;
3047 spair = 4;/* surrogate pair size in UTF-16. */
3048 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3049 parse = utf16le_to_unicode;
3050 tm = 1;
3051 spair = 4;/* surrogate pair size in UTF-16. */
3052 } else {
3053 parse = cesu8_to_unicode;
3054 tm = ts;
3055 spair = 6;/* surrogate pair size in UTF-8. */
3056 }
3057
3058 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3059 return (-1);
3060
3061 p = as->s + as->length;
3062 endp = as->s + as->buffer_length - ts;
3063 while ((n = parse(&uc, s, len)) != 0) {
3064 const char *ucptr, *uc2ptr;
3065
3066 if (n < 0) {
3067 /* Use a replaced unicode character. */
3068 UNPARSE(p, endp, uc);
3069 s += n*-1;
3070 len -= n*-1;
3071 ret = -1;
3072 continue;
3073 } else if (n == spair || always_replace)
3074 /* uc is converted from a surrogate pair.
3075 * this should be treated as a changed code. */
3076 ucptr = NULL;
3077 else
3078 ucptr = s;
3079 s += n;
3080 len -= n;
3081
3082 /* Read second code point. */
3083 while ((n2 = parse(&uc2, s, len)) > 0) {
3084 uint32_t ucx[FDC_MAX];
3085 int ccx[FDC_MAX];
3086 int cl, cx, i, nx, ucx_size;
3087 int LIndex,SIndex;
3088 uint32_t nfc;
3089
3090 if (n2 == spair || always_replace)
3091 /* uc2 is converted from a surrogate pair.
3092 * this should be treated as a changed code. */
3093 uc2ptr = NULL;
3094 else
3095 uc2ptr = s;
3096 s += n2;
3097 len -= n2;
3098
3099 /*
3100 * If current second code point is out of decomposable
3101 * code points, finding compositions is unneeded.
3102 */
3103 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3104 WRITE_UC();
3105 REPLACE_UC_WITH_UC2();
3106 continue;
3107 }
3108
3109 /*
3110 * Try to combine current code points.
3111 */
3112 /*
3113 * We have to combine Hangul characters according to
3114 * http://uniicode.org/reports/tr15/#Hangul
3115 */
3116 if (0 <= (LIndex = uc - HC_LBASE) &&
3117 LIndex < HC_LCOUNT) {
3118 /*
3119 * Hangul Composition.
3120 * 1. Two current code points are L and V.
3121 */
3122 int VIndex = uc2 - HC_VBASE;
3123 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3124 /* Make syllable of form LV. */
3125 UPDATE_UC(HC_SBASE +
3126 (LIndex * HC_VCOUNT + VIndex) *
3127 HC_TCOUNT);
3128 } else {
3129 WRITE_UC();
3130 REPLACE_UC_WITH_UC2();
3131 }
3132 continue;
3133 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3134 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3135 /*
3136 * Hangul Composition.
3137 * 2. Two current code points are LV and T.
3138 */
3139 int TIndex = uc2 - HC_TBASE;
3140 if (0 < TIndex && TIndex < HC_TCOUNT) {
3141 /* Make syllable of form LVT. */
3142 UPDATE_UC(uc + TIndex);
3143 } else {
3144 WRITE_UC();
3145 REPLACE_UC_WITH_UC2();
3146 }
3147 continue;
3148 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3149 /* A composition to current code points
3150 * is found. */
3151 UPDATE_UC(nfc);
3152 continue;
3153 } else if ((cl = CCC(uc2)) == 0) {
3154 /* Clearly 'uc2' the second code point is not
3155 * a decomposable code. */
3156 WRITE_UC();
3157 REPLACE_UC_WITH_UC2();
3158 continue;
3159 }
3160
3161 /*
3162 * Collect following decomposable code points.
3163 */
3164 cx = 0;
3165 ucx[0] = uc2;
3166 ccx[0] = cl;
3167 COLLECT_CPS(1);
3168
3169 /*
3170 * Find a composed code in the collected code points.
3171 */
3172 i = 1;
3173 while (i < ucx_size) {
3174 int j;
3175
3176 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3177 i++;
3178 continue;
3179 }
3180
3181 /*
3182 * nfc is composed of uc and ucx[i].
3183 */
3184 UPDATE_UC(nfc);
3185
3186 /*
3187 * Remove ucx[i] by shifting
3188 * following code points.
3189 */
3190 for (j = i; j+1 < ucx_size; j++) {
3191 ucx[j] = ucx[j+1];
3192 ccx[j] = ccx[j+1];
3193 }
3194 ucx_size --;
3195
3196 /*
3197 * Collect following code points blocked
3198 * by ucx[i] the removed code point.
3199 */
3200 if (ucx_size > 0 && i == ucx_size &&
3201 nx > 0 && cx == cl) {
3202 cl = ccx[ucx_size-1];
3203 COLLECT_CPS(ucx_size);
3204 }
3205 /*
3206 * Restart finding a composed code with
3207 * the updated uc from the top of the
3208 * collected code points.
3209 */
3210 i = 0;
3211 }
3212
3213 /*
3214 * Apparently the current code points are not
3215 * decomposed characters or already composed.
3216 */
3217 WRITE_UC();
3218 for (i = 0; i < ucx_size; i++)
3219 UNPARSE(p, endp, ucx[i]);
3220
3221 /*
3222 * Flush out remaining canonical combining characters.
3223 */
3224 if (nx > 0 && cx == cl && len > 0) {
3225 while ((nx = parse(&ucx[0], s, len))
3226 > 0) {
3227 cx = CCC(ucx[0]);
3228 if (cl > cx)
3229 break;
3230 s += nx;
3231 len -= nx;
3232 cl = cx;
3233 UNPARSE(p, endp, ucx[0]);
3234 }
3235 }
3236 break;
3237 }
3238 if (n2 < 0) {
3239 WRITE_UC();
3240 /* Use a replaced unicode character. */
3241 UNPARSE(p, endp, uc2);
3242 s += n2*-1;
3243 len -= n2*-1;
3244 ret = -1;
3245 continue;
3246 } else if (n2 == 0) {
3247 WRITE_UC();
3248 break;
3249 }
3250 }
3251 as->length = p - as->s;
3252 as->s[as->length] = '\0';
3253 if (ts == 2)
3254 as->s[as->length+1] = '\0';
3255 return (ret);
3256 }
3257
3258 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3259 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3260 {
3261 int t, b;
3262
3263 /*
3264 * These are not converted to NFD on Mac OS.
3265 */
3266 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3267 (uc >= 0xF900 && uc <= 0xFAFF) ||
3268 (uc >= 0x2F800 && uc <= 0x2FAFF))
3269 return (0);
3270 /*
3271 * Those code points are not converted to NFD on Mac OS.
3272 * I do not know the reason because it is undocumented.
3273 * NFC NFD
3274 * 1109A ==> 11099 110BA
3275 * 1109C ==> 1109B 110BA
3276 * 110AB ==> 110A5 110BA
3277 */
3278 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3279 return (0);
3280
3281 t = 0;
3282 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3283 while (b >= t) {
3284 int m = (t + b) / 2;
3285 if (u_decomposition_table[m].nfc < uc)
3286 t = m + 1;
3287 else if (u_decomposition_table[m].nfc > uc)
3288 b = m - 1;
3289 else {
3290 *cp1 = u_decomposition_table[m].cp1;
3291 *cp2 = u_decomposition_table[m].cp2;
3292 return (1);
3293 }
3294 }
3295 return (0);
3296 }
3297
3298 #define REPLACE_UC_WITH(cp) do { \
3299 uc = cp; \
3300 ucptr = NULL; \
3301 } while (0)
3302
3303 /*
3304 * Normalize UTF-8 characters to Form D and copy the result.
3305 */
3306 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3307 archive_string_normalize_D(struct archive_string *as, const void *_p,
3308 size_t len, struct archive_string_conv *sc)
3309 {
3310 const char *s = (const char *)_p;
3311 char *p, *endp;
3312 uint32_t uc, uc2;
3313 size_t w;
3314 int always_replace, n, n2, ret = 0, spair, ts, tm;
3315 int (*parse)(uint32_t *, const char *, size_t);
3316 size_t (*unparse)(char *, size_t, uint32_t);
3317
3318 always_replace = 1;
3319 ts = 1;/* text size. */
3320 if (sc->flag & SCONV_TO_UTF16BE) {
3321 unparse = unicode_to_utf16be;
3322 ts = 2;
3323 if (sc->flag & SCONV_FROM_UTF16BE)
3324 always_replace = 0;
3325 } else if (sc->flag & SCONV_TO_UTF16LE) {
3326 unparse = unicode_to_utf16le;
3327 ts = 2;
3328 if (sc->flag & SCONV_FROM_UTF16LE)
3329 always_replace = 0;
3330 } else if (sc->flag & SCONV_TO_UTF8) {
3331 unparse = unicode_to_utf8;
3332 if (sc->flag & SCONV_FROM_UTF8)
3333 always_replace = 0;
3334 } else {
3335 /*
3336 * This case is going to be converted to another
3337 * character-set through iconv.
3338 */
3339 always_replace = 0;
3340 if (sc->flag & SCONV_FROM_UTF16BE) {
3341 unparse = unicode_to_utf16be;
3342 ts = 2;
3343 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3344 unparse = unicode_to_utf16le;
3345 ts = 2;
3346 } else {
3347 unparse = unicode_to_utf8;
3348 }
3349 }
3350
3351 if (sc->flag & SCONV_FROM_UTF16BE) {
3352 parse = utf16be_to_unicode;
3353 tm = 1;
3354 spair = 4;/* surrogate pair size in UTF-16. */
3355 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3356 parse = utf16le_to_unicode;
3357 tm = 1;
3358 spair = 4;/* surrogate pair size in UTF-16. */
3359 } else {
3360 parse = cesu8_to_unicode;
3361 tm = ts;
3362 spair = 6;/* surrogate pair size in UTF-8. */
3363 }
3364
3365 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3366 return (-1);
3367
3368 p = as->s + as->length;
3369 endp = as->s + as->buffer_length - ts;
3370 while ((n = parse(&uc, s, len)) != 0) {
3371 const char *ucptr;
3372 uint32_t cp1, cp2;
3373 int SIndex;
3374 struct {
3375 uint32_t uc;
3376 int ccc;
3377 } fdc[FDC_MAX];
3378 int fdi, fdj;
3379 int ccc;
3380
3381 check_first_code:
3382 if (n < 0) {
3383 /* Use a replaced unicode character. */
3384 UNPARSE(p, endp, uc);
3385 s += n*-1;
3386 len -= n*-1;
3387 ret = -1;
3388 continue;
3389 } else if (n == spair || always_replace)
3390 /* uc is converted from a surrogate pair.
3391 * this should be treated as a changed code. */
3392 ucptr = NULL;
3393 else
3394 ucptr = s;
3395 s += n;
3396 len -= n;
3397
3398 /* Hangul Decomposition. */
3399 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3400 int L = HC_LBASE + SIndex / HC_NCOUNT;
3401 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3402 int T = HC_TBASE + SIndex % HC_TCOUNT;
3403
3404 REPLACE_UC_WITH(L);
3405 WRITE_UC();
3406 REPLACE_UC_WITH(V);
3407 WRITE_UC();
3408 if (T != HC_TBASE) {
3409 REPLACE_UC_WITH(T);
3410 WRITE_UC();
3411 }
3412 continue;
3413 }
3414 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3415 WRITE_UC();
3416 continue;
3417 }
3418
3419 fdi = 0;
3420 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3421 int k;
3422
3423 for (k = fdi; k > 0; k--)
3424 fdc[k] = fdc[k-1];
3425 fdc[0].ccc = CCC(cp2);
3426 fdc[0].uc = cp2;
3427 fdi++;
3428 REPLACE_UC_WITH(cp1);
3429 }
3430
3431 /* Read following code points. */
3432 while ((n2 = parse(&uc2, s, len)) > 0 &&
3433 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3434 int j, k;
3435
3436 s += n2;
3437 len -= n2;
3438 for (j = 0; j < fdi; j++) {
3439 if (fdc[j].ccc > ccc)
3440 break;
3441 }
3442 if (j < fdi) {
3443 for (k = fdi; k > j; k--)
3444 fdc[k] = fdc[k-1];
3445 fdc[j].ccc = ccc;
3446 fdc[j].uc = uc2;
3447 } else {
3448 fdc[fdi].ccc = ccc;
3449 fdc[fdi].uc = uc2;
3450 }
3451 fdi++;
3452 }
3453
3454 WRITE_UC();
3455 for (fdj = 0; fdj < fdi; fdj++) {
3456 REPLACE_UC_WITH(fdc[fdj].uc);
3457 WRITE_UC();
3458 }
3459
3460 if (n2 == 0)
3461 break;
3462 REPLACE_UC_WITH(uc2);
3463 n = n2;
3464 goto check_first_code;
3465 }
3466 as->length = p - as->s;
3467 as->s[as->length] = '\0';
3468 if (ts == 2)
3469 as->s[as->length+1] = '\0';
3470 return (ret);
3471 }
3472
3473 /*
3474 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3475 * that WCS is Unicode. It is true for several platforms but some are false.
3476 * And then people who did not use UTF-8 locale on the non Unicode WCS
3477 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3478 * now cannot get right filename from libarchive 3.x and later since we
3479 * fixed the wrong assumption and it is incompatible to older its versions.
3480 * So we provide special option, "compat-2x.x", for resolving it.
3481 * That option enable the string conversion of libarchive 2.x.
3482 *
3483 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3484 * locale character set and appends to the archive_string.
3485 * Note: returns -1 if conversion fails.
3486 */
3487 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3488 strncat_from_utf8_libarchive2(struct archive_string *as,
3489 const void *_p, size_t len, struct archive_string_conv *sc)
3490 {
3491 const char *s;
3492 int n;
3493 char *p;
3494 char *end;
3495 uint32_t unicode;
3496 #if HAVE_WCRTOMB
3497 mbstate_t shift_state;
3498
3499 memset(&shift_state, 0, sizeof(shift_state));
3500 #else
3501 /* Clear the shift state before starting. */
3502 wctomb(NULL, L'\0');
3503 #endif
3504 (void)sc; /* UNUSED */
3505 /*
3506 * Allocate buffer for MBS.
3507 * We need this allocation here since it is possible that
3508 * as->s is still NULL.
3509 */
3510 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3511 return (-1);
3512
3513 s = (const char *)_p;
3514 p = as->s + as->length;
3515 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3516 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3517 wchar_t wc;
3518
3519 if (p >= end) {
3520 as->length = p - as->s;
3521 /* Re-allocate buffer for MBS. */
3522 if (archive_string_ensure(as,
3523 as->length + max(len * 2,
3524 (size_t)MB_CUR_MAX) + 1) == NULL)
3525 return (-1);
3526 p = as->s + as->length;
3527 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3528 }
3529
3530 /*
3531 * As libarchive 2.x, translates the UTF-8 characters into
3532 * wide-characters in the assumption that WCS is Unicode.
3533 */
3534 if (n < 0) {
3535 n *= -1;
3536 wc = L'?';
3537 } else
3538 wc = (wchar_t)unicode;
3539
3540 s += n;
3541 len -= n;
3542 /*
3543 * Translates the wide-character into the current locale MBS.
3544 */
3545 #if HAVE_WCRTOMB
3546 n = (int)wcrtomb(p, wc, &shift_state);
3547 #else
3548 n = (int)wctomb(p, wc);
3549 #endif
3550 if (n == -1)
3551 return (-1);
3552 p += n;
3553 }
3554 as->length = p - as->s;
3555 as->s[as->length] = '\0';
3556 return (0);
3557 }
3558
3559
3560 /*
3561 * Conversion functions between current locale dependent MBS and UTF-16BE.
3562 * strncat_from_utf16be() : UTF-16BE --> MBS
3563 * strncat_to_utf16be() : MBS --> UTF16BE
3564 */
3565
3566 #if defined(_WIN32) && !defined(__CYGWIN__)
3567
3568 /*
3569 * Convert a UTF-16BE/LE string to current locale and copy the result.
3570 * Return -1 if conversion fails.
3571 */
3572 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3573 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3574 struct archive_string_conv *sc, int be)
3575 {
3576 struct archive_string tmp;
3577 const char *u16;
3578 BOOL defchar;
3579 char *mbs;
3580 size_t mbs_size, b, ll;
3581 int ret = 0;
3582
3583 bytes &= ~1;
3584 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3585 return (-1);
3586
3587 mbs = as->s + as->length;
3588 mbs_size = as->buffer_length - as->length -1;
3589
3590 if (sc->to_cp == CP_C_LOCALE) {
3591 /*
3592 * "C" locale special processing.
3593 */
3594 u16 = _p;
3595 ll = 0;
3596 for (b = 0; b < bytes; b += 2) {
3597 uint16_t val;
3598 if (be)
3599 val = archive_be16dec(u16+b);
3600 else
3601 val = archive_le16dec(u16+b);
3602 if (val > 255) {
3603 *mbs++ = '?';
3604 ret = -1;
3605 } else
3606 *mbs++ = (char)(val&0xff);
3607 ll++;
3608 }
3609 as->length += ll;
3610 as->s[as->length] = '\0';
3611 return (ret);
3612 }
3613
3614 archive_string_init(&tmp);
3615 if (be) {
3616 if (IS_BIG_ENDIAN) {
3617 u16 = _p;
3618 } else {
3619 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3620 return (-1);
3621 memcpy(tmp.s, _p, bytes);
3622 for (b = 0; b < bytes; b += 2) {
3623 uint16_t val = archive_be16dec(tmp.s+b);
3624 archive_le16enc(tmp.s+b, val);
3625 }
3626 u16 = tmp.s;
3627 }
3628 } else {
3629 if (!IS_BIG_ENDIAN) {
3630 u16 = _p;
3631 } else {
3632 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3633 return (-1);
3634 memcpy(tmp.s, _p, bytes);
3635 for (b = 0; b < bytes; b += 2) {
3636 uint16_t val = archive_le16dec(tmp.s+b);
3637 archive_be16enc(tmp.s+b, val);
3638 }
3639 u16 = tmp.s;
3640 }
3641 }
3642
3643 do {
3644 int r;
3645 defchar = 0;
3646 /* WideCharToMultiByte is limited to int. */
3647 if (bytes > (size_t)INT_MAX || mbs_size > (size_t)INT_MAX)
3648 return (-1);
3649 r = WideCharToMultiByte(sc->to_cp, 0,
3650 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3651 NULL, &defchar);
3652 /* Exit loop if we succeeded */
3653 if (r != 0 ||
3654 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3655 ll = (size_t)r;
3656 break;
3657 }
3658 /* Else expand buffer and loop to try again. */
3659 r = WideCharToMultiByte(sc->to_cp, 0,
3660 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3661 ll = (size_t)r;
3662 if (archive_string_ensure(as, ll +1) == NULL)
3663 return (-1);
3664 mbs = as->s + as->length;
3665 mbs_size = as->buffer_length - as->length -1;
3666 } while (1);
3667 archive_string_free(&tmp);
3668 as->length += ll;
3669 as->s[as->length] = '\0';
3670 if (ll == 0 || defchar)
3671 ret = -1;
3672 return (ret);
3673 }
3674
3675 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3676 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3677 size_t bytes, struct archive_string_conv *sc)
3678 {
3679 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3680 }
3681
3682 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3683 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3684 size_t bytes, struct archive_string_conv *sc)
3685 {
3686 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3687 }
3688
3689 /*
3690 * Convert a current locale string to UTF-16BE/LE and copy the result.
3691 * Return -1 if conversion fails.
3692 */
3693 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3694 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3695 size_t length, struct archive_string_conv *sc, int bigendian)
3696 {
3697 const char *s = (const char *)_p;
3698 char *u16;
3699 size_t count, avail;
3700
3701 if (archive_string_ensure(as16,
3702 as16->length + (length + 1) * 2) == NULL)
3703 return (-1);
3704
3705 u16 = as16->s + as16->length;
3706 avail = as16->buffer_length - 2;
3707 if (sc->from_cp == CP_C_LOCALE) {
3708 /*
3709 * "C" locale special processing.
3710 */
3711 count = 0;
3712 while (count < length && *s) {
3713 if (bigendian)
3714 archive_be16enc(u16, *s);
3715 else
3716 archive_le16enc(u16, *s);
3717 u16 += 2;
3718 s++;
3719 count++;
3720 }
3721 as16->length += count << 1;
3722 as16->s[as16->length] = 0;
3723 as16->s[as16->length+1] = 0;
3724 return (0);
3725 }
3726 do {
3727 int r;
3728 if (length > (size_t)INT_MAX || (avail >> 1) > (size_t)INT_MAX)
3729 return (-1);
3730 r = MultiByteToWideChar(sc->from_cp,
3731 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3732 /* Exit loop if we succeeded */
3733 if (r != 0 ||
3734 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3735 count = (size_t)r;
3736 break;
3737 }
3738 /* Expand buffer and try again */
3739 r = MultiByteToWideChar(sc->from_cp,
3740 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3741 count = (size_t)r;
3742 if (archive_string_ensure(as16, (count +1) * 2)
3743 == NULL)
3744 return (-1);
3745 u16 = as16->s + as16->length;
3746 avail = as16->buffer_length - 2;
3747 } while (1);
3748 as16->length += count * 2;
3749 as16->s[as16->length] = 0;
3750 as16->s[as16->length+1] = 0;
3751 if (count == 0)
3752 return (-1);
3753
3754 if (IS_BIG_ENDIAN) {
3755 if (!bigendian) {
3756 while (count > 0) {
3757 uint16_t v = archive_be16dec(u16);
3758 archive_le16enc(u16, v);
3759 u16 += 2;
3760 count--;
3761 }
3762 }
3763 } else {
3764 if (bigendian) {
3765 while (count > 0) {
3766 uint16_t v = archive_le16dec(u16);
3767 archive_be16enc(u16, v);
3768 u16 += 2;
3769 count--;
3770 }
3771 }
3772 }
3773 return (0);
3774 }
3775
3776 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3777 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3778 size_t length, struct archive_string_conv *sc)
3779 {
3780 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3781 }
3782
3783 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3784 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3785 size_t length, struct archive_string_conv *sc)
3786 {
3787 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3788 }
3789
3790 #endif /* _WIN32 && !__CYGWIN__ */
3791
3792 /*
3793 * Do the best effort for conversions.
3794 * We cannot handle UTF-16BE character-set without such iconv,
3795 * but there is a chance if a string consists just ASCII code or
3796 * a current locale is UTF-8.
3797 */
3798
3799 /*
3800 * Convert a UTF-16BE string to current locale and copy the result.
3801 * Return -1 if conversion fails.
3802 */
3803 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3804 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3805 size_t bytes, struct archive_string_conv *sc, int be)
3806 {
3807 const char *utf16 = (const char *)_p;
3808 char *mbs;
3809 uint32_t uc;
3810 int n, ret;
3811
3812 (void)sc; /* UNUSED */
3813 /*
3814 * Other case, we should do the best effort.
3815 * If all character are ASCII(<0x7f), we can convert it.
3816 * if not , we set a alternative character and return -1.
3817 */
3818 ret = 0;
3819 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3820 return (-1);
3821 mbs = as->s + as->length;
3822
3823 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3824 if (n < 0) {
3825 n *= -1;
3826 ret = -1;
3827 }
3828 bytes -= n;
3829 utf16 += n;
3830
3831 if (uc > 127) {
3832 /* We cannot handle it. */
3833 *mbs++ = '?';
3834 ret = -1;
3835 } else
3836 *mbs++ = (char)uc;
3837 }
3838 as->length = mbs - as->s;
3839 as->s[as->length] = '\0';
3840 return (ret);
3841 }
3842
3843 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3844 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3845 size_t bytes, struct archive_string_conv *sc)
3846 {
3847 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3848 }
3849
3850 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3851 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3852 size_t bytes, struct archive_string_conv *sc)
3853 {
3854 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3855 }
3856
3857 /*
3858 * Convert a current locale string to UTF-16BE/LE and copy the result.
3859 * Return -1 if conversion fails.
3860 */
3861 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3862 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3863 size_t length, struct archive_string_conv *sc, int bigendian)
3864 {
3865 const char *s = (const char *)_p;
3866 char *utf16;
3867 size_t remaining;
3868 int ret;
3869
3870 (void)sc; /* UNUSED */
3871 /*
3872 * Other case, we should do the best effort.
3873 * If all character are ASCII(<0x7f), we can convert it.
3874 * if not , we set a alternative character and return -1.
3875 */
3876 ret = 0;
3877 remaining = length;
3878
3879 if (archive_string_ensure(as16,
3880 as16->length + (length + 1) * 2) == NULL)
3881 return (-1);
3882
3883 utf16 = as16->s + as16->length;
3884 while (remaining--) {
3885 unsigned c = *s++;
3886 if (c > 127) {
3887 /* We cannot handle it. */
3888 c = UNICODE_R_CHAR;
3889 ret = -1;
3890 }
3891 if (bigendian)
3892 archive_be16enc(utf16, (uint16_t)c);
3893 else
3894 archive_le16enc(utf16, (uint16_t)c);
3895 utf16 += 2;
3896 }
3897 as16->length = utf16 - as16->s;
3898 as16->s[as16->length] = 0;
3899 as16->s[as16->length+1] = 0;
3900 return (ret);
3901 }
3902
3903 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3904 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3905 size_t length, struct archive_string_conv *sc)
3906 {
3907 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3908 }
3909
3910 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3911 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3912 size_t length, struct archive_string_conv *sc)
3913 {
3914 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3915 }
3916
3917
3918 /*
3919 * Multistring operations.
3920 */
3921
3922 void
archive_mstring_clean(struct archive_mstring * aes)3923 archive_mstring_clean(struct archive_mstring *aes)
3924 {
3925 archive_wstring_free(&(aes->aes_wcs));
3926 archive_string_free(&(aes->aes_mbs));
3927 archive_string_free(&(aes->aes_utf8));
3928 archive_string_free(&(aes->aes_mbs_in_locale));
3929 aes->aes_set = 0;
3930 }
3931
3932 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3933 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3934 {
3935 dest->aes_set = src->aes_set;
3936 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3937 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3938 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3939 }
3940
3941 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3942 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3943 const char **p)
3944 {
3945 struct archive_string_conv *sc;
3946 int r;
3947
3948 /* If we already have a UTF8 form, return that immediately. */
3949 if (aes->aes_set & AES_SET_UTF8) {
3950 *p = aes->aes_utf8.s;
3951 return (0);
3952 }
3953
3954 *p = NULL;
3955 #if defined(_WIN32) && !defined(__CYGWIN__)
3956 /*
3957 * On Windows, first try converting from WCS because (1) there's no
3958 * guarantee that the conversion to MBS will succeed, e.g. when using
3959 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3960 * convert back to WCS again before finally converting to UTF-8
3961 */
3962 if ((aes->aes_set & AES_SET_WCS) != 0) {
3963 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3964 if (sc == NULL)
3965 return (-1);/* Couldn't allocate memory for sc. */
3966 archive_string_empty(&(aes->aes_utf8));
3967 r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3968 aes->aes_wcs.s, aes->aes_wcs.length, sc);
3969 if (a == NULL)
3970 free_sconv_object(sc);
3971 if (r == 0) {
3972 aes->aes_set |= AES_SET_UTF8;
3973 *p = aes->aes_utf8.s;
3974 return (0);/* success. */
3975 } else
3976 return (-1);/* failure. */
3977 }
3978 #endif
3979 /* Try converting WCS to MBS first if MBS does not exist yet. */
3980 if ((aes->aes_set & AES_SET_MBS) == 0) {
3981 const char *pm; /* unused */
3982 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3983 }
3984 if (aes->aes_set & AES_SET_MBS) {
3985 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3986 if (sc == NULL)
3987 return (-1);/* Couldn't allocate memory for sc. */
3988 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3989 aes->aes_mbs.length, sc);
3990 if (a == NULL)
3991 free_sconv_object(sc);
3992 if (r == 0) {
3993 aes->aes_set |= AES_SET_UTF8;
3994 *p = aes->aes_utf8.s;
3995 return (0);/* success. */
3996 } else
3997 return (-1);/* failure. */
3998 }
3999 return (0);/* success. */
4000 }
4001
4002 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)4003 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
4004 const char **p)
4005 {
4006 struct archive_string_conv *sc;
4007 int r, ret = 0;
4008
4009 /* If we already have an MBS form, return that immediately. */
4010 if (aes->aes_set & AES_SET_MBS) {
4011 *p = aes->aes_mbs.s;
4012 return (ret);
4013 }
4014
4015 *p = NULL;
4016 /* If there's a WCS form, try converting with the native locale. */
4017 if (aes->aes_set & AES_SET_WCS) {
4018 archive_string_empty(&(aes->aes_mbs));
4019 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4020 aes->aes_wcs.s, aes->aes_wcs.length);
4021 *p = aes->aes_mbs.s;
4022 if (r == 0) {
4023 aes->aes_set |= AES_SET_MBS;
4024 return (ret);
4025 } else
4026 ret = -1;
4027 }
4028
4029 /* If there's a UTF-8 form, try converting with the native locale. */
4030 if (aes->aes_set & AES_SET_UTF8) {
4031 archive_string_empty(&(aes->aes_mbs));
4032 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4033 if (sc == NULL)
4034 return (-1);/* Couldn't allocate memory for sc. */
4035 r = archive_strncpy_l(&(aes->aes_mbs),
4036 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4037 if (a == NULL)
4038 free_sconv_object(sc);
4039 *p = aes->aes_mbs.s;
4040 if (r == 0) {
4041 aes->aes_set |= AES_SET_MBS;
4042 ret = 0;/* success; overwrite previous error. */
4043 } else
4044 ret = -1;/* failure. */
4045 }
4046 return (ret);
4047 }
4048
4049 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)4050 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
4051 const wchar_t **wp)
4052 {
4053 int r, ret = 0;
4054
4055 (void)a;/* UNUSED */
4056 /* Return WCS form if we already have it. */
4057 if (aes->aes_set & AES_SET_WCS) {
4058 *wp = aes->aes_wcs.s;
4059 return (ret);
4060 }
4061
4062 *wp = NULL;
4063 #if defined(_WIN32) && !defined(__CYGWIN__)
4064 /*
4065 * On Windows, prefer converting from UTF-8 directly to WCS because:
4066 * (1) there's no guarantee that the string can be represented in MBS (e.g.
4067 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
4068 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
4069 * away that intermediate result
4070 */
4071 if (aes->aes_set & AES_SET_UTF8) {
4072 struct archive_string_conv *sc;
4073
4074 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4075 if (sc != NULL) {
4076 archive_wstring_empty((&aes->aes_wcs));
4077 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4078 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4079 if (a == NULL)
4080 free_sconv_object(sc);
4081 if (r == 0) {
4082 aes->aes_set |= AES_SET_WCS;
4083 *wp = aes->aes_wcs.s;
4084 return (0);
4085 }
4086 }
4087 }
4088 #endif
4089 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
4090 if ((aes->aes_set & AES_SET_MBS) == 0) {
4091 const char *p; /* unused */
4092 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4093 }
4094 /* Try converting MBS to WCS using native locale. */
4095 if (aes->aes_set & AES_SET_MBS) {
4096 archive_wstring_empty(&(aes->aes_wcs));
4097 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4098 aes->aes_mbs.s, aes->aes_mbs.length);
4099 if (r == 0) {
4100 aes->aes_set |= AES_SET_WCS;
4101 *wp = aes->aes_wcs.s;
4102 } else
4103 ret = -1;/* failure. */
4104 }
4105 return (ret);
4106 }
4107
4108 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)4109 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4110 const char **p, size_t *length, struct archive_string_conv *sc)
4111 {
4112 int ret = 0;
4113 #if defined(_WIN32) && !defined(__CYGWIN__)
4114 int r;
4115
4116 /*
4117 * Internationalization programming on Windows must use Wide
4118 * characters because Windows platform cannot make locale UTF-8.
4119 */
4120 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4121 archive_string_empty(&(aes->aes_mbs_in_locale));
4122 r = archive_string_append_from_wcs_in_codepage(
4123 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4124 aes->aes_wcs.length, sc);
4125 if (r == 0) {
4126 *p = aes->aes_mbs_in_locale.s;
4127 if (length != NULL)
4128 *length = aes->aes_mbs_in_locale.length;
4129 return (0);
4130 } else if (errno == ENOMEM)
4131 return (-1);
4132 else
4133 ret = -1;
4134 }
4135 #endif
4136
4137 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4138 * with the native locale to be used for translating it to specified
4139 * character-set. */
4140 if ((aes->aes_set & AES_SET_MBS) == 0) {
4141 const char *pm; /* unused */
4142 if (archive_mstring_get_mbs(a, aes, &pm) != 0) {
4143 /* We have another form, but failed to convert it to
4144 * the native locale. Transitively, we've failed to
4145 * convert it to the specified character set. */
4146 ret = -1;
4147 }
4148 }
4149 /* If we already have an MBS form, use it to be translated to
4150 * specified character-set. */
4151 if (aes->aes_set & AES_SET_MBS) {
4152 if (sc == NULL) {
4153 /* Conversion is unneeded. */
4154 *p = aes->aes_mbs.s;
4155 if (length != NULL)
4156 *length = aes->aes_mbs.length;
4157 return (0);
4158 }
4159 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4160 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4161 *p = aes->aes_mbs_in_locale.s;
4162 if (length != NULL)
4163 *length = aes->aes_mbs_in_locale.length;
4164 } else {
4165 /* Either we have no string in any form,
4166 * or conversion failed and set 'ret != 0'. */
4167 *p = NULL;
4168 if (length != NULL)
4169 *length = 0;
4170 }
4171 return (ret);
4172 }
4173
4174 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4175 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4176 {
4177 if (mbs == NULL) {
4178 aes->aes_set = 0;
4179 return (0);
4180 }
4181 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4182 }
4183
4184 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4185 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4186 size_t len)
4187 {
4188 if (mbs == NULL) {
4189 aes->aes_set = 0;
4190 return (0);
4191 }
4192 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4193 archive_strncpy(&(aes->aes_mbs), mbs, len);
4194 archive_string_empty(&(aes->aes_utf8));
4195 archive_wstring_empty(&(aes->aes_wcs));
4196 return (0);
4197 }
4198
4199 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4200 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4201 {
4202 return archive_mstring_copy_wcs_len(aes, wcs,
4203 wcs == NULL ? 0 : wcslen(wcs));
4204 }
4205
4206 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4207 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4208 {
4209 if (utf8 == NULL) {
4210 aes->aes_set = 0;
4211 return (0);
4212 }
4213 aes->aes_set = AES_SET_UTF8;
4214 archive_string_empty(&(aes->aes_mbs));
4215 archive_string_empty(&(aes->aes_wcs));
4216 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4217 return (int)strlen(utf8);
4218 }
4219
4220 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4221 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4222 size_t len)
4223 {
4224 if (wcs == NULL) {
4225 aes->aes_set = 0;
4226 return (0);
4227 }
4228 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4229 archive_string_empty(&(aes->aes_mbs));
4230 archive_string_empty(&(aes->aes_utf8));
4231 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4232 return (0);
4233 }
4234
4235 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4236 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4237 const char *mbs, size_t len, struct archive_string_conv *sc)
4238 {
4239 int r;
4240
4241 if (mbs == NULL) {
4242 aes->aes_set = 0;
4243 return (0);
4244 }
4245 archive_string_empty(&(aes->aes_mbs));
4246 archive_wstring_empty(&(aes->aes_wcs));
4247 archive_string_empty(&(aes->aes_utf8));
4248 #if defined(_WIN32) && !defined(__CYGWIN__)
4249 /*
4250 * Internationalization programming on Windows must use Wide
4251 * characters because Windows platform cannot make locale UTF-8.
4252 */
4253 if (sc == NULL) {
4254 if (archive_string_append(&(aes->aes_mbs),
4255 mbs, mbsnbytes(mbs, len)) == NULL) {
4256 aes->aes_set = 0;
4257 r = -1;
4258 } else {
4259 aes->aes_set = AES_SET_MBS;
4260 r = 0;
4261 }
4262 #if defined(HAVE_ICONV)
4263 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4264 /*
4265 * This case happens only when MultiByteToWideChar() cannot
4266 * handle sc->from_cp, and we have to iconv in order to
4267 * translate character-set to wchar_t,UTF-16.
4268 */
4269 iconv_t cd = sc->cd;
4270 unsigned from_cp;
4271 int flag;
4272
4273 /*
4274 * Translate multi-bytes from some character-set to UTF-8.
4275 */
4276 sc->cd = sc->cd_w;
4277 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4278 sc->cd = cd;
4279 if (r != 0) {
4280 aes->aes_set = 0;
4281 return (r);
4282 }
4283 aes->aes_set = AES_SET_UTF8;
4284
4285 /*
4286 * Append the UTF-8 string into wstring.
4287 */
4288 flag = sc->flag;
4289 sc->flag &= ~(SCONV_NORMALIZATION_C
4290 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4291 from_cp = sc->from_cp;
4292 sc->from_cp = CP_UTF8;
4293 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4294 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4295 sc->flag = flag;
4296 sc->from_cp = from_cp;
4297 if (r == 0)
4298 aes->aes_set |= AES_SET_WCS;
4299 #endif
4300 } else {
4301 r = archive_wstring_append_from_mbs_in_codepage(
4302 &(aes->aes_wcs), mbs, len, sc);
4303 if (r == 0)
4304 aes->aes_set = AES_SET_WCS;
4305 else
4306 aes->aes_set = 0;
4307 }
4308 #else
4309 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4310 if (r == 0)
4311 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4312 else
4313 aes->aes_set = 0;
4314 #endif
4315 return (r);
4316 }
4317
4318 /*
4319 * The 'update' form tries to proactively update all forms of
4320 * this string (WCS and MBS) and returns an error if any of
4321 * them fail. This is used by the 'pax' handler, for instance,
4322 * to detect and report character-conversion failures early while
4323 * still allowing clients to get potentially useful values from
4324 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4325 * strive to give the user something useful, so you can get hopefully
4326 * usable values even if some of the character conversions are failing.)
4327 */
4328 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4329 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4330 const char *utf8)
4331 {
4332 struct archive_string_conv *sc;
4333 int r;
4334
4335 if (utf8 == NULL) {
4336 aes->aes_set = 0;
4337 return (0); /* Succeeded in clearing everything. */
4338 }
4339
4340 /* Save the UTF8 string. */
4341 archive_strcpy(&(aes->aes_utf8), utf8);
4342
4343 /* Empty the mbs and wcs strings. */
4344 archive_string_empty(&(aes->aes_mbs));
4345 archive_wstring_empty(&(aes->aes_wcs));
4346
4347 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4348
4349 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4350 if (sc == NULL)
4351 return (-1);/* Couldn't allocate memory for sc. */
4352
4353 #if defined(_WIN32) && !defined(__CYGWIN__)
4354 /* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4355 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4356 * intermediate result, and (2) WCS will still be set even if we fail to
4357 * convert to MBS (e.g. with ACP that can't represent the characters) */
4358 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4359 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4360
4361 if (a == NULL)
4362 free_sconv_object(sc);
4363 if (r != 0)
4364 return (-1); /* This will guarantee we can't convert to MBS */
4365 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4366
4367 /* Try converting WCS to MBS, return false on failure. */
4368 if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4369 aes->aes_wcs.length))
4370 return (-1);
4371 #else
4372 /* Try converting UTF-8 to MBS, return false on failure. */
4373 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4374
4375 if (a == NULL)
4376 free_sconv_object(sc);
4377 if (r != 0)
4378 return (-1);
4379 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4380
4381 /* Try converting MBS to WCS, return false on failure. */
4382 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4383 aes->aes_mbs.length))
4384 return (-1);
4385 #endif
4386
4387 /* All conversions succeeded. */
4388 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4389
4390 return (0);
4391 }
4392