1 /*-
2 * Copyright (c) 2003-2011 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27 #include "archive_platform.h"
28
29 /*
30 * Basic resizable string support, to simplify manipulating arbitrary-sized
31 * strings while minimizing heap activity.
32 *
33 * In particular, the buffer used by a string object is only grown, it
34 * never shrinks, so you can clear and reuse the same string object
35 * without incurring additional memory allocations.
36 */
37
38 #ifdef HAVE_ERRNO_H
39 #include <errno.h>
40 #endif
41 #ifdef HAVE_ICONV_H
42 #include <iconv.h>
43 #endif
44 #ifdef HAVE_LANGINFO_H
45 #include <langinfo.h>
46 #endif
47 #ifdef HAVE_LOCALCHARSET_H
48 #include <localcharset.h>
49 #endif
50 #ifdef HAVE_STDLIB_H
51 #include <stdlib.h>
52 #endif
53 #ifdef HAVE_STRING_H
54 #include <string.h>
55 #endif
56 #ifdef HAVE_WCHAR_H
57 #include <wchar.h>
58 #endif
59 #if defined(_WIN32) && !defined(__CYGWIN__)
60 #include <windows.h>
61 #include <locale.h>
62 #endif
63
64 #include "archive_endian.h"
65 #include "archive_private.h"
66 #include "archive_string.h"
67 #include "archive_string_composition.h"
68
69 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
70 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
71 #endif
72
73 #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove)
74 #define wmemmove(a,b,i) (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t))
75 #endif
76
77 #undef max
78 #define max(a, b) ((a)>(b)?(a):(b))
79
80 struct archive_string_conv {
81 struct archive_string_conv *next;
82 char *from_charset;
83 char *to_charset;
84 unsigned from_cp;
85 unsigned to_cp;
86 /* Set 1 if from_charset and to_charset are the same. */
87 int same;
88 int flag;
89 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified
90 * charset. */
91 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from
92 * specified charset. */
93 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */
94 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting
95 * MBS. */
96 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive
97 * 2.x in the wrong assumption. */
98 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C.
99 * Before UTF-8 characters are actually
100 * processed. */
101 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D.
102 * Before UTF-8 characters are actually
103 * processed.
104 * Currently this only for MAC OS X. */
105 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */
106 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */
107 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */
108 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */
109 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */
110 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */
111 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
112 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
113
114 #if HAVE_ICONV
115 iconv_t cd;
116 iconv_t cd_w;/* Use at archive_mstring on
117 * Windows. */
118 #endif
119 /* A temporary buffer for normalization. */
120 struct archive_string utftmp;
121 int (*converter[2])(struct archive_string *, const void *, size_t,
122 struct archive_string_conv *);
123 int nconverter;
124 };
125
126 #define CP_C_LOCALE 0 /* "C" locale only for this file. */
127 #define CP_UTF16LE 1200
128 #define CP_UTF16BE 1201
129
130 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
131 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
132 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
133 #define UNICODE_MAX 0x10FFFF
134 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */
135 /* Set U+FFFD(Replacement character) in UTF-8. */
136 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd};
137
138 static struct archive_string_conv *find_sconv_object(struct archive *,
139 const char *, const char *);
140 static void add_sconv_object(struct archive *, struct archive_string_conv *);
141 static struct archive_string_conv *create_sconv_object(const char *,
142 const char *, unsigned, int);
143 static void free_sconv_object(struct archive_string_conv *);
144 static struct archive_string_conv *get_sconv_object(struct archive *,
145 const char *, const char *, int);
146 static unsigned make_codepage_from_charset(const char *);
147 static unsigned get_current_codepage(void);
148 static unsigned get_current_oemcp(void);
149 static size_t mbsnbytes(const void *, size_t);
150 static size_t utf16nbytes(const void *, size_t);
151 #if defined(_WIN32) && !defined(__CYGWIN__)
152 static int archive_wstring_append_from_mbs_in_codepage(
153 struct archive_wstring *, const char *, size_t,
154 struct archive_string_conv *);
155 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
156 const wchar_t *, size_t, struct archive_string_conv *);
157 static int is_big_endian(void);
158 static int strncat_in_codepage(struct archive_string *, const void *,
159 size_t, struct archive_string_conv *);
160 static int win_strncat_from_utf16be(struct archive_string *, const void *,
161 size_t, struct archive_string_conv *);
162 static int win_strncat_from_utf16le(struct archive_string *, const void *,
163 size_t, struct archive_string_conv *);
164 static int win_strncat_to_utf16be(struct archive_string *, const void *,
165 size_t, struct archive_string_conv *);
166 static int win_strncat_to_utf16le(struct archive_string *, const void *,
167 size_t, struct archive_string_conv *);
168 #endif
169 static int best_effort_strncat_from_utf16be(struct archive_string *,
170 const void *, size_t, struct archive_string_conv *);
171 static int best_effort_strncat_from_utf16le(struct archive_string *,
172 const void *, size_t, struct archive_string_conv *);
173 static int best_effort_strncat_to_utf16be(struct archive_string *,
174 const void *, size_t, struct archive_string_conv *);
175 static int best_effort_strncat_to_utf16le(struct archive_string *,
176 const void *, size_t, struct archive_string_conv *);
177 #if defined(HAVE_ICONV)
178 static int iconv_strncat_in_locale(struct archive_string *, const void *,
179 size_t, struct archive_string_conv *);
180 #endif
181 static int best_effort_strncat_in_locale(struct archive_string *,
182 const void *, size_t, struct archive_string_conv *);
183 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
184 static int utf8_to_unicode(uint32_t *, const char *, size_t);
185 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
186 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
187 static size_t unicode_to_utf8(char *, size_t, uint32_t);
188 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
189 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
190 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
191 static int strncat_from_utf8_libarchive2(struct archive_string *,
192 const void *, size_t, struct archive_string_conv *);
193 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
194 size_t, struct archive_string_conv *);
195 static int archive_string_normalize_C(struct archive_string *, const void *,
196 size_t, struct archive_string_conv *);
197 static int archive_string_normalize_D(struct archive_string *, const void *,
198 size_t, struct archive_string_conv *);
199 static int archive_string_append_unicode(struct archive_string *,
200 const void *, size_t, struct archive_string_conv *);
201
202 static struct archive_string *
archive_string_append(struct archive_string * as,const char * p,size_t s)203 archive_string_append(struct archive_string *as, const char *p, size_t s)
204 {
205 if (archive_string_ensure(as, as->length + s + 1) == NULL)
206 return (NULL);
207 if (s)
208 memmove(as->s + as->length, p, s);
209 as->length += s;
210 as->s[as->length] = 0;
211 return (as);
212 }
213
214 static struct archive_wstring *
archive_wstring_append(struct archive_wstring * as,const wchar_t * p,size_t s)215 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
216 {
217 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
218 return (NULL);
219 if (s)
220 wmemmove(as->s + as->length, p, s);
221 as->length += s;
222 as->s[as->length] = 0;
223 return (as);
224 }
225
226 struct archive_string *
archive_array_append(struct archive_string * as,const char * p,size_t s)227 archive_array_append(struct archive_string *as, const char *p, size_t s)
228 {
229 return archive_string_append(as, p, s);
230 }
231
232 void
archive_string_concat(struct archive_string * dest,struct archive_string * src)233 archive_string_concat(struct archive_string *dest, struct archive_string *src)
234 {
235 if (archive_string_append(dest, src->s, src->length) == NULL)
236 __archive_errx(1, "Out of memory");
237 }
238
239 void
archive_wstring_concat(struct archive_wstring * dest,struct archive_wstring * src)240 archive_wstring_concat(struct archive_wstring *dest,
241 struct archive_wstring *src)
242 {
243 if (archive_wstring_append(dest, src->s, src->length) == NULL)
244 __archive_errx(1, "Out of memory");
245 }
246
247 void
archive_string_free(struct archive_string * as)248 archive_string_free(struct archive_string *as)
249 {
250 as->length = 0;
251 as->buffer_length = 0;
252 free(as->s);
253 as->s = NULL;
254 }
255
256 void
archive_wstring_free(struct archive_wstring * as)257 archive_wstring_free(struct archive_wstring *as)
258 {
259 as->length = 0;
260 as->buffer_length = 0;
261 free(as->s);
262 as->s = NULL;
263 }
264
265 struct archive_wstring *
archive_wstring_ensure(struct archive_wstring * as,size_t s)266 archive_wstring_ensure(struct archive_wstring *as, size_t s)
267 {
268 return (struct archive_wstring *)
269 archive_string_ensure((struct archive_string *)as,
270 s * sizeof(wchar_t));
271 }
272
273 /* Returns NULL on any allocation failure. */
274 struct archive_string *
archive_string_ensure(struct archive_string * as,size_t s)275 archive_string_ensure(struct archive_string *as, size_t s)
276 {
277 char *p;
278 size_t new_length;
279
280 /* If buffer is already big enough, don't reallocate. */
281 if (as->s && (s <= as->buffer_length))
282 return (as);
283
284 /*
285 * Growing the buffer at least exponentially ensures that
286 * append operations are always linear in the number of
287 * characters appended. Using a smaller growth rate for
288 * larger buffers reduces memory waste somewhat at the cost of
289 * a larger constant factor.
290 */
291 if (as->buffer_length < 32)
292 /* Start with a minimum 32-character buffer. */
293 new_length = 32;
294 else if (as->buffer_length < 8192)
295 /* Buffers under 8k are doubled for speed. */
296 new_length = as->buffer_length + as->buffer_length;
297 else {
298 /* Buffers 8k and over grow by at least 25% each time. */
299 new_length = as->buffer_length + as->buffer_length / 4;
300 /* Be safe: If size wraps, fail. */
301 if (new_length < as->buffer_length) {
302 /* On failure, wipe the string and return NULL. */
303 archive_string_free(as);
304 errno = ENOMEM;/* Make sure errno has ENOMEM. */
305 return (NULL);
306 }
307 }
308 /*
309 * The computation above is a lower limit to how much we'll
310 * grow the buffer. In any case, we have to grow it enough to
311 * hold the request.
312 */
313 if (new_length < s)
314 new_length = s;
315 /* Now we can reallocate the buffer. */
316 p = realloc(as->s, new_length);
317 if (p == NULL) {
318 /* On failure, wipe the string and return NULL. */
319 archive_string_free(as);
320 errno = ENOMEM;/* Make sure errno has ENOMEM. */
321 return (NULL);
322 }
323
324 as->s = p;
325 as->buffer_length = new_length;
326 return (as);
327 }
328
329 /*
330 * TODO: See if there's a way to avoid scanning
331 * the source string twice. Then test to see
332 * if it actually helps (remember that we're almost
333 * always called with pretty short arguments, so
334 * such an optimization might not help).
335 */
336 struct archive_string *
archive_strncat(struct archive_string * as,const void * _p,size_t n)337 archive_strncat(struct archive_string *as, const void *_p, size_t n)
338 {
339 size_t s;
340 const char *p, *pp;
341
342 p = (const char *)_p;
343
344 /* Like strlen(p), except won't examine positions beyond p[n]. */
345 s = 0;
346 pp = p;
347 while (s < n && *pp) {
348 pp++;
349 s++;
350 }
351 if ((as = archive_string_append(as, p, s)) == NULL)
352 __archive_errx(1, "Out of memory");
353 return (as);
354 }
355
356 struct archive_wstring *
archive_wstrncat(struct archive_wstring * as,const wchar_t * p,size_t n)357 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
358 {
359 size_t s;
360 const wchar_t *pp;
361
362 /* Like strlen(p), except won't examine positions beyond p[n]. */
363 s = 0;
364 pp = p;
365 while (s < n && *pp) {
366 pp++;
367 s++;
368 }
369 if ((as = archive_wstring_append(as, p, s)) == NULL)
370 __archive_errx(1, "Out of memory");
371 return (as);
372 }
373
374 struct archive_string *
archive_strcat(struct archive_string * as,const void * p)375 archive_strcat(struct archive_string *as, const void *p)
376 {
377 /* strcat is just strncat without an effective limit.
378 * Assert that we'll never get called with a source
379 * string over 16MB.
380 * TODO: Review all uses of strcat in the source
381 * and try to replace them with strncat().
382 */
383 return archive_strncat(as, p, 0x1000000);
384 }
385
386 struct archive_wstring *
archive_wstrcat(struct archive_wstring * as,const wchar_t * p)387 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
388 {
389 /* Ditto. */
390 return archive_wstrncat(as, p, 0x1000000);
391 }
392
393 struct archive_string *
archive_strappend_char(struct archive_string * as,char c)394 archive_strappend_char(struct archive_string *as, char c)
395 {
396 if ((as = archive_string_append(as, &c, 1)) == NULL)
397 __archive_errx(1, "Out of memory");
398 return (as);
399 }
400
401 struct archive_wstring *
archive_wstrappend_wchar(struct archive_wstring * as,wchar_t c)402 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
403 {
404 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
405 __archive_errx(1, "Out of memory");
406 return (as);
407 }
408
409 /*
410 * Get the "current character set" name to use with iconv.
411 * On FreeBSD, the empty character set name "" chooses
412 * the correct character encoding for the current locale,
413 * so this isn't necessary.
414 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
415 * on that system, we have to explicitly call nl_langinfo()
416 * to get the right name. Not sure about other platforms.
417 *
418 * NOTE: GNU libiconv does not recognize the character-set name
419 * which some platform nl_langinfo(CODESET) returns, so we should
420 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
421 */
422 static const char *
default_iconv_charset(const char * charset)423 default_iconv_charset(const char *charset) {
424 if (charset != NULL && charset[0] != '\0')
425 return charset;
426 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
427 /* locale_charset() is broken on Mac OS */
428 return locale_charset();
429 #elif HAVE_NL_LANGINFO
430 return nl_langinfo(CODESET);
431 #else
432 return "";
433 #endif
434 }
435
436 #if defined(_WIN32) && !defined(__CYGWIN__)
437
438 /*
439 * Convert MBS to WCS.
440 * Note: returns -1 if conversion fails.
441 */
442 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)443 archive_wstring_append_from_mbs(struct archive_wstring *dest,
444 const char *p, size_t len)
445 {
446 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
447 }
448
449 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)450 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
451 const char *s, size_t length, struct archive_string_conv *sc)
452 {
453 int count, ret = 0;
454 UINT from_cp;
455
456 if (sc != NULL)
457 from_cp = sc->from_cp;
458 else
459 from_cp = get_current_codepage();
460
461 if (from_cp == CP_C_LOCALE) {
462 /*
463 * "C" locale special processing.
464 */
465 wchar_t *ws;
466 const unsigned char *mp;
467
468 if (NULL == archive_wstring_ensure(dest,
469 dest->length + length + 1))
470 return (-1);
471
472 ws = dest->s + dest->length;
473 mp = (const unsigned char *)s;
474 count = 0;
475 while (count < (int)length && *mp) {
476 *ws++ = (wchar_t)*mp++;
477 count++;
478 }
479 } else if (sc != NULL &&
480 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
481 /*
482 * Normalize UTF-8 and UTF-16BE and convert it directly
483 * to UTF-16 as wchar_t.
484 */
485 struct archive_string u16;
486 int saved_flag = sc->flag;/* save current flag. */
487
488 if (is_big_endian())
489 sc->flag |= SCONV_TO_UTF16BE;
490 else
491 sc->flag |= SCONV_TO_UTF16LE;
492
493 if (sc->flag & SCONV_FROM_UTF16) {
494 /*
495 * UTF-16BE/LE NFD ===> UTF-16 NFC
496 * UTF-16BE/LE NFC ===> UTF-16 NFD
497 */
498 count = (int)utf16nbytes(s, length);
499 } else {
500 /*
501 * UTF-8 NFD ===> UTF-16 NFC
502 * UTF-8 NFC ===> UTF-16 NFD
503 */
504 count = (int)mbsnbytes(s, length);
505 }
506 u16.s = (char *)dest->s;
507 u16.length = dest->length << 1;;
508 u16.buffer_length = dest->buffer_length;
509 if (sc->flag & SCONV_NORMALIZATION_C)
510 ret = archive_string_normalize_C(&u16, s, count, sc);
511 else
512 ret = archive_string_normalize_D(&u16, s, count, sc);
513 dest->s = (wchar_t *)u16.s;
514 dest->length = u16.length >> 1;
515 dest->buffer_length = u16.buffer_length;
516 sc->flag = saved_flag;/* restore the saved flag. */
517 return (ret);
518 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
519 count = (int)utf16nbytes(s, length);
520 count >>= 1; /* to be WCS length */
521 /* Allocate memory for WCS. */
522 if (NULL == archive_wstring_ensure(dest,
523 dest->length + count + 1))
524 return (-1);
525 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
526 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
527 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
528 int b;
529 for (b = 0; b < count; b++) {
530 uint16_t val = archive_le16dec(u16+b);
531 archive_be16enc(u16+b, val);
532 }
533 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
534 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
535 int b;
536 for (b = 0; b < count; b++) {
537 uint16_t val = archive_be16dec(u16+b);
538 archive_le16enc(u16+b, val);
539 }
540 }
541 } else {
542 DWORD mbflag;
543 size_t buffsize;
544
545 if (sc == NULL)
546 mbflag = 0;
547 else if (sc->flag & SCONV_FROM_CHARSET) {
548 /* Do not trust the length which comes from
549 * an archive file. */
550 length = mbsnbytes(s, length);
551 mbflag = 0;
552 } else
553 mbflag = MB_PRECOMPOSED;
554
555 mbflag |= MB_ERR_INVALID_CHARS;
556
557 buffsize = dest->length + length + 1;
558 do {
559 /* Allocate memory for WCS. */
560 if (NULL == archive_wstring_ensure(dest, buffsize))
561 return (-1);
562 /* Convert MBS to WCS. */
563 count = MultiByteToWideChar(from_cp,
564 mbflag, s, (int)length, dest->s + dest->length,
565 (int)(dest->buffer_length >> 1) -1);
566 if (count == 0 &&
567 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
568 /* Expand the WCS buffer. */
569 buffsize = dest->buffer_length << 1;
570 continue;
571 }
572 if (count == 0 && length != 0)
573 ret = -1;
574 break;
575 } while (1);
576 }
577 dest->length += count;
578 dest->s[dest->length] = L'\0';
579 return (ret);
580 }
581
582 #else
583
584 /*
585 * Convert MBS to WCS.
586 * Note: returns -1 if conversion fails.
587 */
588 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)589 archive_wstring_append_from_mbs(struct archive_wstring *dest,
590 const char *p, size_t len)
591 {
592 size_t r;
593 int ret_val = 0;
594 /*
595 * No single byte will be more than one wide character,
596 * so this length estimate will always be big enough.
597 */
598 // size_t wcs_length = len;
599 size_t mbs_length = len;
600 const char *mbs = p;
601 wchar_t *wcs;
602 #if HAVE_MBRTOWC
603 mbstate_t shift_state;
604
605 memset(&shift_state, 0, sizeof(shift_state));
606 #endif
607 /*
608 * As we decided to have wcs_length == mbs_length == len
609 * we can use len here instead of wcs_length
610 */
611 if (NULL == archive_wstring_ensure(dest, dest->length + len + 1))
612 return (-1);
613 wcs = dest->s + dest->length;
614 /*
615 * We cannot use mbsrtowcs/mbstowcs here because those may convert
616 * extra MBS when strlen(p) > len and one wide character consists of
617 * multi bytes.
618 */
619 while (*mbs && mbs_length > 0) {
620 /*
621 * The buffer we allocated is always big enough.
622 * Keep this code path in a comment if we decide to choose
623 * smaller wcs_length in the future
624 */
625 /*
626 if (wcs_length == 0) {
627 dest->length = wcs - dest->s;
628 dest->s[dest->length] = L'\0';
629 wcs_length = mbs_length;
630 if (NULL == archive_wstring_ensure(dest,
631 dest->length + wcs_length + 1))
632 return (-1);
633 wcs = dest->s + dest->length;
634 }
635 */
636 #if HAVE_MBRTOWC
637 r = mbrtowc(wcs, mbs, mbs_length, &shift_state);
638 #else
639 r = mbtowc(wcs, mbs, mbs_length);
640 #endif
641 if (r == (size_t)-1 || r == (size_t)-2) {
642 ret_val = -1;
643 break;
644 }
645 if (r == 0 || r > mbs_length)
646 break;
647 wcs++;
648 // wcs_length--;
649 mbs += r;
650 mbs_length -= r;
651 }
652 dest->length = wcs - dest->s;
653 dest->s[dest->length] = L'\0';
654 return (ret_val);
655 }
656
657 #endif
658
659 #if defined(_WIN32) && !defined(__CYGWIN__)
660
661 /*
662 * WCS ==> MBS.
663 * Note: returns -1 if conversion fails.
664 *
665 * Win32 builds use WideCharToMultiByte from the Windows API.
666 * (Maybe Cygwin should too? WideCharToMultiByte will know a
667 * lot more about local character encodings than the wcrtomb()
668 * wrapper is going to know.)
669 */
670 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)671 archive_string_append_from_wcs(struct archive_string *as,
672 const wchar_t *w, size_t len)
673 {
674 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
675 }
676
677 static int
archive_string_append_from_wcs_in_codepage(struct archive_string * as,const wchar_t * ws,size_t len,struct archive_string_conv * sc)678 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
679 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
680 {
681 BOOL defchar_used, *dp;
682 int count, ret = 0;
683 UINT to_cp;
684 int wslen = (int)len;
685
686 if (sc != NULL)
687 to_cp = sc->to_cp;
688 else
689 to_cp = get_current_codepage();
690
691 if (to_cp == CP_C_LOCALE) {
692 /*
693 * "C" locale special processing.
694 */
695 const wchar_t *wp = ws;
696 char *p;
697
698 if (NULL == archive_string_ensure(as,
699 as->length + wslen +1))
700 return (-1);
701 p = as->s + as->length;
702 count = 0;
703 defchar_used = 0;
704 while (count < wslen && *wp) {
705 if (*wp > 255) {
706 *p++ = '?';
707 wp++;
708 defchar_used = 1;
709 } else
710 *p++ = (char)*wp++;
711 count++;
712 }
713 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
714 uint16_t *u16;
715
716 if (NULL ==
717 archive_string_ensure(as, as->length + len * 2 + 2))
718 return (-1);
719 u16 = (uint16_t *)(as->s + as->length);
720 count = 0;
721 defchar_used = 0;
722 if (sc->flag & SCONV_TO_UTF16BE) {
723 while (count < (int)len && *ws) {
724 archive_be16enc(u16+count, *ws);
725 ws++;
726 count++;
727 }
728 } else {
729 while (count < (int)len && *ws) {
730 archive_le16enc(u16+count, *ws);
731 ws++;
732 count++;
733 }
734 }
735 count <<= 1; /* to be byte size */
736 } else {
737 /* Make sure the MBS buffer has plenty to set. */
738 if (NULL ==
739 archive_string_ensure(as, as->length + len * 2 + 1))
740 return (-1);
741 do {
742 defchar_used = 0;
743 if (to_cp == CP_UTF8 || sc == NULL)
744 dp = NULL;
745 else
746 dp = &defchar_used;
747 count = WideCharToMultiByte(to_cp, 0, ws, wslen,
748 as->s + as->length,
749 (int)as->buffer_length - (int)as->length - 1, NULL, dp);
750 if (count == 0 &&
751 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
752 /* Expand the MBS buffer and retry. */
753 if (NULL == archive_string_ensure(as,
754 as->buffer_length + len))
755 return (-1);
756 continue;
757 }
758 if (count == 0)
759 ret = -1;
760 break;
761 } while (1);
762 }
763 as->length += count;
764 as->s[as->length] = '\0';
765 return (defchar_used?-1:ret);
766 }
767
768 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
769
770 /*
771 * Translates a wide character string into current locale character set
772 * and appends to the archive_string. Note: returns -1 if conversion
773 * fails.
774 */
775 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)776 archive_string_append_from_wcs(struct archive_string *as,
777 const wchar_t *w, size_t len)
778 {
779 /* We cannot use the standard wcstombs() here because it
780 * cannot tell us how big the output buffer should be. So
781 * I've built a loop around wcrtomb() or wctomb() that
782 * converts a character at a time and resizes the string as
783 * needed. We prefer wcrtomb() when it's available because
784 * it's thread-safe. */
785 int n, ret_val = 0;
786 char *p;
787 char *end;
788 #if HAVE_WCRTOMB
789 mbstate_t shift_state;
790
791 memset(&shift_state, 0, sizeof(shift_state));
792 #else
793 /* Clear the shift state before starting. */
794 wctomb(NULL, L'\0');
795 #endif
796 /*
797 * Allocate buffer for MBS.
798 * We need this allocation here since it is possible that
799 * as->s is still NULL.
800 */
801 if (archive_string_ensure(as, as->length + len + 1) == NULL)
802 return (-1);
803
804 p = as->s + as->length;
805 end = as->s + as->buffer_length - MB_CUR_MAX -1;
806 while (*w != L'\0' && len > 0) {
807 if (p >= end) {
808 as->length = p - as->s;
809 as->s[as->length] = '\0';
810 /* Re-allocate buffer for MBS. */
811 if (archive_string_ensure(as,
812 as->length + max(len * 2,
813 (size_t)MB_CUR_MAX) + 1) == NULL)
814 return (-1);
815 p = as->s + as->length;
816 end = as->s + as->buffer_length - MB_CUR_MAX -1;
817 }
818 #if HAVE_WCRTOMB
819 n = wcrtomb(p, *w++, &shift_state);
820 #else
821 n = wctomb(p, *w++);
822 #endif
823 if (n == -1) {
824 if (errno == EILSEQ) {
825 /* Skip an illegal wide char. */
826 *p++ = '?';
827 ret_val = -1;
828 } else {
829 ret_val = -1;
830 break;
831 }
832 } else
833 p += n;
834 len--;
835 }
836 as->length = p - as->s;
837 as->s[as->length] = '\0';
838 return (ret_val);
839 }
840
841 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
842
843 /*
844 * TODO: Test if __STDC_ISO_10646__ is defined.
845 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
846 * one character at a time. If a non-Windows platform doesn't have
847 * either of these, fall back to the built-in UTF8 conversion.
848 */
849 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)850 archive_string_append_from_wcs(struct archive_string *as,
851 const wchar_t *w, size_t len)
852 {
853 (void)as;/* UNUSED */
854 (void)w;/* UNUSED */
855 (void)len;/* UNUSED */
856 errno = ENOSYS;
857 return (-1);
858 }
859
860 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
861
862 /*
863 * Find a string conversion object by a pair of 'from' charset name
864 * and 'to' charset name from an archive object.
865 * Return NULL if not found.
866 */
867 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)868 find_sconv_object(struct archive *a, const char *fc, const char *tc)
869 {
870 struct archive_string_conv *sc;
871
872 if (a == NULL)
873 return (NULL);
874
875 for (sc = a->sconv; sc != NULL; sc = sc->next) {
876 if (strcmp(sc->from_charset, fc) == 0 &&
877 strcmp(sc->to_charset, tc) == 0)
878 break;
879 }
880 return (sc);
881 }
882
883 /*
884 * Register a string object to an archive object.
885 */
886 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)887 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
888 {
889 struct archive_string_conv **psc;
890
891 /* Add a new sconv to sconv list. */
892 psc = &(a->sconv);
893 while (*psc != NULL)
894 psc = &((*psc)->next);
895 *psc = sc;
896 }
897
898 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))899 add_converter(struct archive_string_conv *sc, int (*converter)
900 (struct archive_string *, const void *, size_t,
901 struct archive_string_conv *))
902 {
903 if (sc == NULL || sc->nconverter >= 2)
904 __archive_errx(1, "Programming error");
905 sc->converter[sc->nconverter++] = converter;
906 }
907
908 static void
setup_converter(struct archive_string_conv * sc)909 setup_converter(struct archive_string_conv *sc)
910 {
911
912 /* Reset. */
913 sc->nconverter = 0;
914
915 /*
916 * Perform special sequence for the incorrect UTF-8 filenames
917 * made by libarchive2.x.
918 */
919 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
920 add_converter(sc, strncat_from_utf8_libarchive2);
921 return;
922 }
923
924 /*
925 * Convert a string to UTF-16BE/LE.
926 */
927 if (sc->flag & SCONV_TO_UTF16) {
928 /*
929 * If the current locale is UTF-8, we can translate
930 * a UTF-8 string into a UTF-16BE string.
931 */
932 if (sc->flag & SCONV_FROM_UTF8) {
933 add_converter(sc, archive_string_append_unicode);
934 return;
935 }
936
937 #if defined(_WIN32) && !defined(__CYGWIN__)
938 if (sc->flag & SCONV_WIN_CP) {
939 if (sc->flag & SCONV_TO_UTF16BE)
940 add_converter(sc, win_strncat_to_utf16be);
941 else
942 add_converter(sc, win_strncat_to_utf16le);
943 return;
944 }
945 #endif
946
947 #if defined(HAVE_ICONV)
948 if (sc->cd != (iconv_t)-1) {
949 add_converter(sc, iconv_strncat_in_locale);
950 return;
951 }
952 #endif
953
954 if (sc->flag & SCONV_BEST_EFFORT) {
955 if (sc->flag & SCONV_TO_UTF16BE)
956 add_converter(sc,
957 best_effort_strncat_to_utf16be);
958 else
959 add_converter(sc,
960 best_effort_strncat_to_utf16le);
961 } else
962 /* Make sure we have no converter. */
963 sc->nconverter = 0;
964 return;
965 }
966
967 /*
968 * Convert a string from UTF-16BE/LE.
969 */
970 if (sc->flag & SCONV_FROM_UTF16) {
971 /*
972 * At least we should normalize a UTF-16BE string.
973 */
974 if (sc->flag & SCONV_NORMALIZATION_D)
975 add_converter(sc,archive_string_normalize_D);
976 else if (sc->flag & SCONV_NORMALIZATION_C)
977 add_converter(sc, archive_string_normalize_C);
978
979 if (sc->flag & SCONV_TO_UTF8) {
980 /*
981 * If the current locale is UTF-8, we can translate
982 * a UTF-16BE/LE string into a UTF-8 string directly.
983 */
984 if (!(sc->flag &
985 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
986 add_converter(sc,
987 archive_string_append_unicode);
988 return;
989 }
990
991 #if defined(_WIN32) && !defined(__CYGWIN__)
992 if (sc->flag & SCONV_WIN_CP) {
993 if (sc->flag & SCONV_FROM_UTF16BE)
994 add_converter(sc, win_strncat_from_utf16be);
995 else
996 add_converter(sc, win_strncat_from_utf16le);
997 return;
998 }
999 #endif
1000
1001 #if defined(HAVE_ICONV)
1002 if (sc->cd != (iconv_t)-1) {
1003 add_converter(sc, iconv_strncat_in_locale);
1004 return;
1005 }
1006 #endif
1007
1008 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1009 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1010 add_converter(sc, best_effort_strncat_from_utf16be);
1011 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1012 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1013 add_converter(sc, best_effort_strncat_from_utf16le);
1014 else
1015 /* Make sure we have no converter. */
1016 sc->nconverter = 0;
1017 return;
1018 }
1019
1020 if (sc->flag & SCONV_FROM_UTF8) {
1021 /*
1022 * At least we should normalize a UTF-8 string.
1023 */
1024 if (sc->flag & SCONV_NORMALIZATION_D)
1025 add_converter(sc,archive_string_normalize_D);
1026 else if (sc->flag & SCONV_NORMALIZATION_C)
1027 add_converter(sc, archive_string_normalize_C);
1028
1029 /*
1030 * Copy UTF-8 string with a check of CESU-8.
1031 * Apparently, iconv does not check surrogate pairs in UTF-8
1032 * when both from-charset and to-charset are UTF-8, and then
1033 * we use our UTF-8 copy code.
1034 */
1035 if (sc->flag & SCONV_TO_UTF8) {
1036 /*
1037 * If the current locale is UTF-8, we can translate
1038 * a UTF-16BE string into a UTF-8 string directly.
1039 */
1040 if (!(sc->flag &
1041 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1042 add_converter(sc, strncat_from_utf8_to_utf8);
1043 return;
1044 }
1045 }
1046
1047 #if defined(_WIN32) && !defined(__CYGWIN__)
1048 /*
1049 * On Windows we can use Windows API for a string conversion.
1050 */
1051 if (sc->flag & SCONV_WIN_CP) {
1052 add_converter(sc, strncat_in_codepage);
1053 return;
1054 }
1055 #endif
1056
1057 #if HAVE_ICONV
1058 if (sc->cd != (iconv_t)-1) {
1059 add_converter(sc, iconv_strncat_in_locale);
1060 /*
1061 * iconv generally does not support UTF-8-MAC and so
1062 * we have to the output of iconv from NFC to NFD if
1063 * need.
1064 */
1065 if ((sc->flag & SCONV_FROM_CHARSET) &&
1066 (sc->flag & SCONV_TO_UTF8)) {
1067 if (sc->flag & SCONV_NORMALIZATION_D)
1068 add_converter(sc, archive_string_normalize_D);
1069 }
1070 return;
1071 }
1072 #endif
1073
1074 /*
1075 * Try conversion in the best effort or no conversion.
1076 */
1077 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1078 add_converter(sc, best_effort_strncat_in_locale);
1079 else
1080 /* Make sure we have no converter. */
1081 sc->nconverter = 0;
1082 }
1083
1084 /*
1085 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1086 * and CP932 which are referenced in create_sconv_object().
1087 */
1088 static const char *
canonical_charset_name(const char * charset)1089 canonical_charset_name(const char *charset)
1090 {
1091 char cs[16];
1092 char *p;
1093 const char *s;
1094
1095 if (charset == NULL || charset[0] == '\0'
1096 || strlen(charset) > 15)
1097 return (charset);
1098
1099 /* Copy name to uppercase. */
1100 p = cs;
1101 s = charset;
1102 while (*s) {
1103 char c = *s++;
1104 if (c >= 'a' && c <= 'z')
1105 c -= 'a' - 'A';
1106 *p++ = c;
1107 }
1108 *p++ = '\0';
1109
1110 if (strcmp(cs, "UTF-8") == 0 ||
1111 strcmp(cs, "UTF8") == 0)
1112 return ("UTF-8");
1113 if (strcmp(cs, "UTF-16BE") == 0 ||
1114 strcmp(cs, "UTF16BE") == 0)
1115 return ("UTF-16BE");
1116 if (strcmp(cs, "UTF-16LE") == 0 ||
1117 strcmp(cs, "UTF16LE") == 0)
1118 return ("UTF-16LE");
1119 if (strcmp(cs, "CP932") == 0)
1120 return ("CP932");
1121 return (charset);
1122 }
1123
1124 /*
1125 * Create a string conversion object.
1126 */
1127 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1128 create_sconv_object(const char *fc, const char *tc,
1129 unsigned current_codepage, int flag)
1130 {
1131 struct archive_string_conv *sc;
1132
1133 sc = calloc(1, sizeof(*sc));
1134 if (sc == NULL)
1135 return (NULL);
1136 sc->next = NULL;
1137 sc->from_charset = strdup(fc);
1138 if (sc->from_charset == NULL) {
1139 free(sc);
1140 return (NULL);
1141 }
1142 sc->to_charset = strdup(tc);
1143 if (sc->to_charset == NULL) {
1144 free(sc->from_charset);
1145 free(sc);
1146 return (NULL);
1147 }
1148 archive_string_init(&sc->utftmp);
1149
1150 if (flag & SCONV_TO_CHARSET) {
1151 /*
1152 * Convert characters from the current locale charset to
1153 * a specified charset.
1154 */
1155 sc->from_cp = current_codepage;
1156 sc->to_cp = make_codepage_from_charset(tc);
1157 #if defined(_WIN32) && !defined(__CYGWIN__)
1158 if (IsValidCodePage(sc->to_cp))
1159 flag |= SCONV_WIN_CP;
1160 #endif
1161 } else if (flag & SCONV_FROM_CHARSET) {
1162 /*
1163 * Convert characters from a specified charset to
1164 * the current locale charset.
1165 */
1166 sc->to_cp = current_codepage;
1167 sc->from_cp = make_codepage_from_charset(fc);
1168 #if defined(_WIN32) && !defined(__CYGWIN__)
1169 if (IsValidCodePage(sc->from_cp))
1170 flag |= SCONV_WIN_CP;
1171 #endif
1172 }
1173
1174 /*
1175 * Check if "from charset" and "to charset" are the same.
1176 */
1177 if (strcmp(fc, tc) == 0 ||
1178 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1179 sc->same = 1;
1180 else
1181 sc->same = 0;
1182
1183 /*
1184 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1185 */
1186 if (strcmp(tc, "UTF-8") == 0)
1187 flag |= SCONV_TO_UTF8;
1188 else if (strcmp(tc, "UTF-16BE") == 0)
1189 flag |= SCONV_TO_UTF16BE;
1190 else if (strcmp(tc, "UTF-16LE") == 0)
1191 flag |= SCONV_TO_UTF16LE;
1192 if (strcmp(fc, "UTF-8") == 0)
1193 flag |= SCONV_FROM_UTF8;
1194 else if (strcmp(fc, "UTF-16BE") == 0)
1195 flag |= SCONV_FROM_UTF16BE;
1196 else if (strcmp(fc, "UTF-16LE") == 0)
1197 flag |= SCONV_FROM_UTF16LE;
1198 #if defined(_WIN32) && !defined(__CYGWIN__)
1199 if (sc->to_cp == CP_UTF8)
1200 flag |= SCONV_TO_UTF8;
1201 else if (sc->to_cp == CP_UTF16BE)
1202 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1203 else if (sc->to_cp == CP_UTF16LE)
1204 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1205 if (sc->from_cp == CP_UTF8)
1206 flag |= SCONV_FROM_UTF8;
1207 else if (sc->from_cp == CP_UTF16BE)
1208 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1209 else if (sc->from_cp == CP_UTF16LE)
1210 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1211 #endif
1212
1213 /*
1214 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1215 * handle it. So we have to translate NFD characters to NFC ones
1216 * ourselves before iconv handles. Another reason is to prevent
1217 * that the same sight of two filenames, one is NFC and other
1218 * is NFD, would be in its directory.
1219 * On Mac OS X, although its filesystem layer automatically
1220 * convert filenames to NFD, it would be useful for filename
1221 * comparing to find out the same filenames that we normalize
1222 * that to be NFD ourselves.
1223 */
1224 if ((flag & SCONV_FROM_CHARSET) &&
1225 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1226 #if defined(__APPLE__)
1227 if (flag & SCONV_TO_UTF8)
1228 flag |= SCONV_NORMALIZATION_D;
1229 else
1230 #endif
1231 flag |= SCONV_NORMALIZATION_C;
1232 }
1233 #if defined(__APPLE__)
1234 /*
1235 * In case writing an archive file, make sure that a filename
1236 * going to be passed to iconv is a Unicode NFC string since
1237 * a filename in HFS Plus filesystem is a Unicode NFD one and
1238 * iconv cannot handle it with "UTF-8" charset. It is simpler
1239 * than a use of "UTF-8-MAC" charset.
1240 */
1241 if ((flag & SCONV_TO_CHARSET) &&
1242 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1243 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1244 flag |= SCONV_NORMALIZATION_C;
1245 /*
1246 * In case reading an archive file. make sure that a filename
1247 * will be passed to users is a Unicode NFD string in order to
1248 * correctly compare the filename with other one which comes
1249 * from HFS Plus filesystem.
1250 */
1251 if ((flag & SCONV_FROM_CHARSET) &&
1252 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1253 (flag & SCONV_TO_UTF8))
1254 flag |= SCONV_NORMALIZATION_D;
1255 #endif
1256
1257 #if defined(HAVE_ICONV)
1258 sc->cd_w = (iconv_t)-1;
1259 /*
1260 * Create an iconv object.
1261 */
1262 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1263 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1264 (flag & SCONV_WIN_CP)) {
1265 /* This case we won't use iconv. */
1266 sc->cd = (iconv_t)-1;
1267 } else {
1268 sc->cd = iconv_open(tc, fc);
1269 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1270 /*
1271 * Unfortunately, all of iconv implements do support
1272 * "CP932" character-set, so we should use "SJIS"
1273 * instead if iconv_open failed.
1274 */
1275 if (strcmp(tc, "CP932") == 0)
1276 sc->cd = iconv_open("SJIS", fc);
1277 else if (strcmp(fc, "CP932") == 0)
1278 sc->cd = iconv_open(tc, "SJIS");
1279 }
1280 #if defined(_WIN32) && !defined(__CYGWIN__)
1281 /*
1282 * archive_mstring on Windows directly convert multi-bytes
1283 * into archive_wstring in order not to depend on locale
1284 * so that you can do a I18N programming. This will be
1285 * used only in archive_mstring_copy_mbs_len_l so far.
1286 */
1287 if (flag & SCONV_FROM_CHARSET) {
1288 sc->cd_w = iconv_open("UTF-8", fc);
1289 if (sc->cd_w == (iconv_t)-1 &&
1290 (sc->flag & SCONV_BEST_EFFORT)) {
1291 if (strcmp(fc, "CP932") == 0)
1292 sc->cd_w = iconv_open("UTF-8", "SJIS");
1293 }
1294 }
1295 #endif /* _WIN32 && !__CYGWIN__ */
1296 }
1297 #endif /* HAVE_ICONV */
1298
1299 sc->flag = flag;
1300
1301 /*
1302 * Set up converters.
1303 */
1304 setup_converter(sc);
1305
1306 return (sc);
1307 }
1308
1309 /*
1310 * Free a string conversion object.
1311 */
1312 static void
free_sconv_object(struct archive_string_conv * sc)1313 free_sconv_object(struct archive_string_conv *sc)
1314 {
1315 free(sc->from_charset);
1316 free(sc->to_charset);
1317 archive_string_free(&sc->utftmp);
1318 #if HAVE_ICONV
1319 if (sc->cd != (iconv_t)-1)
1320 iconv_close(sc->cd);
1321 if (sc->cd_w != (iconv_t)-1)
1322 iconv_close(sc->cd_w);
1323 #endif
1324 free(sc);
1325 }
1326
1327 #if defined(_WIN32) && !defined(__CYGWIN__)
1328 # if defined(WINAPI_FAMILY_PARTITION) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP)
1329 # define GetOEMCP() CP_OEMCP
1330 # endif
1331
1332 static unsigned
my_atoi(const char * p)1333 my_atoi(const char *p)
1334 {
1335 unsigned cp;
1336
1337 cp = 0;
1338 while (*p) {
1339 if (*p >= '0' && *p <= '9')
1340 cp = cp * 10 + (*p - '0');
1341 else
1342 return (-1);
1343 p++;
1344 }
1345 return (cp);
1346 }
1347
1348 /*
1349 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1350 * Return -1 if failed.
1351 *
1352 * Note: This translation code may be insufficient.
1353 */
1354 static struct charset {
1355 const char *name;
1356 unsigned cp;
1357 } charsets[] = {
1358 /* MUST BE SORTED! */
1359 {"ASCII", 1252},
1360 {"ASMO-708", 708},
1361 {"BIG5", 950},
1362 {"CHINESE", 936},
1363 {"CP367", 1252},
1364 {"CP819", 1252},
1365 {"CP1025", 21025},
1366 {"DOS-720", 720},
1367 {"DOS-862", 862},
1368 {"EUC-CN", 51936},
1369 {"EUC-JP", 51932},
1370 {"EUC-KR", 949},
1371 {"EUCCN", 51936},
1372 {"EUCJP", 51932},
1373 {"EUCKR", 949},
1374 {"GB18030", 54936},
1375 {"GB2312", 936},
1376 {"HEBREW", 1255},
1377 {"HZ-GB-2312", 52936},
1378 {"IBM273", 20273},
1379 {"IBM277", 20277},
1380 {"IBM278", 20278},
1381 {"IBM280", 20280},
1382 {"IBM284", 20284},
1383 {"IBM285", 20285},
1384 {"IBM290", 20290},
1385 {"IBM297", 20297},
1386 {"IBM367", 1252},
1387 {"IBM420", 20420},
1388 {"IBM423", 20423},
1389 {"IBM424", 20424},
1390 {"IBM819", 1252},
1391 {"IBM871", 20871},
1392 {"IBM880", 20880},
1393 {"IBM905", 20905},
1394 {"IBM924", 20924},
1395 {"ISO-8859-1", 28591},
1396 {"ISO-8859-13", 28603},
1397 {"ISO-8859-15", 28605},
1398 {"ISO-8859-2", 28592},
1399 {"ISO-8859-3", 28593},
1400 {"ISO-8859-4", 28594},
1401 {"ISO-8859-5", 28595},
1402 {"ISO-8859-6", 28596},
1403 {"ISO-8859-7", 28597},
1404 {"ISO-8859-8", 28598},
1405 {"ISO-8859-9", 28599},
1406 {"ISO8859-1", 28591},
1407 {"ISO8859-13", 28603},
1408 {"ISO8859-15", 28605},
1409 {"ISO8859-2", 28592},
1410 {"ISO8859-3", 28593},
1411 {"ISO8859-4", 28594},
1412 {"ISO8859-5", 28595},
1413 {"ISO8859-6", 28596},
1414 {"ISO8859-7", 28597},
1415 {"ISO8859-8", 28598},
1416 {"ISO8859-9", 28599},
1417 {"JOHAB", 1361},
1418 {"KOI8-R", 20866},
1419 {"KOI8-U", 21866},
1420 {"KS_C_5601-1987", 949},
1421 {"LATIN1", 1252},
1422 {"LATIN2", 28592},
1423 {"MACINTOSH", 10000},
1424 {"SHIFT-JIS", 932},
1425 {"SHIFT_JIS", 932},
1426 {"SJIS", 932},
1427 {"US", 1252},
1428 {"US-ASCII", 1252},
1429 {"UTF-16", 1200},
1430 {"UTF-16BE", 1201},
1431 {"UTF-16LE", 1200},
1432 {"UTF-8", CP_UTF8},
1433 {"X-EUROPA", 29001},
1434 {"X-MAC-ARABIC", 10004},
1435 {"X-MAC-CE", 10029},
1436 {"X-MAC-CHINESEIMP", 10008},
1437 {"X-MAC-CHINESETRAD", 10002},
1438 {"X-MAC-CROATIAN", 10082},
1439 {"X-MAC-CYRILLIC", 10007},
1440 {"X-MAC-GREEK", 10006},
1441 {"X-MAC-HEBREW", 10005},
1442 {"X-MAC-ICELANDIC", 10079},
1443 {"X-MAC-JAPANESE", 10001},
1444 {"X-MAC-KOREAN", 10003},
1445 {"X-MAC-ROMANIAN", 10010},
1446 {"X-MAC-THAI", 10021},
1447 {"X-MAC-TURKISH", 10081},
1448 {"X-MAC-UKRAINIAN", 10017},
1449 };
1450 static unsigned
make_codepage_from_charset(const char * charset)1451 make_codepage_from_charset(const char *charset)
1452 {
1453 char cs[16];
1454 char *p;
1455 unsigned cp;
1456 int a, b;
1457
1458 if (charset == NULL || strlen(charset) > 15)
1459 return -1;
1460
1461 /* Copy name to uppercase. */
1462 p = cs;
1463 while (*charset) {
1464 char c = *charset++;
1465 if (c >= 'a' && c <= 'z')
1466 c -= 'a' - 'A';
1467 *p++ = c;
1468 }
1469 *p++ = '\0';
1470 cp = -1;
1471
1472 /* Look it up in the table first, so that we can easily
1473 * override CP367, which we map to 1252 instead of 367. */
1474 a = 0;
1475 b = sizeof(charsets)/sizeof(charsets[0]);
1476 while (b > a) {
1477 int c = (b + a) / 2;
1478 int r = strcmp(charsets[c].name, cs);
1479 if (r < 0)
1480 a = c + 1;
1481 else if (r > 0)
1482 b = c;
1483 else
1484 return charsets[c].cp;
1485 }
1486
1487 /* If it's not in the table, try to parse it. */
1488 switch (*cs) {
1489 case 'C':
1490 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1491 cp = my_atoi(cs + 2);
1492 } else if (strcmp(cs, "CP_ACP") == 0)
1493 cp = get_current_codepage();
1494 else if (strcmp(cs, "CP_OEMCP") == 0)
1495 cp = get_current_oemcp();
1496 break;
1497 case 'I':
1498 if (cs[1] == 'B' && cs[2] == 'M' &&
1499 cs[3] >= '0' && cs[3] <= '9') {
1500 cp = my_atoi(cs + 3);
1501 }
1502 break;
1503 case 'W':
1504 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1505 cp = my_atoi(cs + 8);
1506 if (cp != 874 && (cp < 1250 || cp > 1258))
1507 cp = -1;/* This may invalid code. */
1508 }
1509 break;
1510 }
1511 return (cp);
1512 }
1513
1514 /*
1515 * Return ANSI Code Page of current locale set by setlocale().
1516 */
1517 static unsigned
get_current_codepage(void)1518 get_current_codepage(void)
1519 {
1520 char *locale, *p;
1521 unsigned cp;
1522
1523 locale = setlocale(LC_CTYPE, NULL);
1524 if (locale == NULL)
1525 return (GetACP());
1526 if (locale[0] == 'C' && locale[1] == '\0')
1527 return (CP_C_LOCALE);
1528 p = strrchr(locale, '.');
1529 if (p == NULL)
1530 return (GetACP());
1531 if ((strcmp(p+1, "utf8") == 0) || (strcmp(p+1, "UTF-8") == 0))
1532 return CP_UTF8;
1533 cp = my_atoi(p+1);
1534 if ((int)cp <= 0)
1535 return (GetACP());
1536 return (cp);
1537 }
1538
1539 /*
1540 * Translation table between Locale Name and ACP/OEMCP.
1541 */
1542 static struct {
1543 unsigned acp;
1544 unsigned ocp;
1545 const char *locale;
1546 } acp_ocp_map[] = {
1547 { 950, 950, "Chinese_Taiwan" },
1548 { 936, 936, "Chinese_People's Republic of China" },
1549 { 950, 950, "Chinese_Taiwan" },
1550 { 1250, 852, "Czech_Czech Republic" },
1551 { 1252, 850, "Danish_Denmark" },
1552 { 1252, 850, "Dutch_Netherlands" },
1553 { 1252, 850, "Dutch_Belgium" },
1554 { 1252, 437, "English_United States" },
1555 { 1252, 850, "English_Australia" },
1556 { 1252, 850, "English_Canada" },
1557 { 1252, 850, "English_New Zealand" },
1558 { 1252, 850, "English_United Kingdom" },
1559 { 1252, 437, "English_United States" },
1560 { 1252, 850, "Finnish_Finland" },
1561 { 1252, 850, "French_France" },
1562 { 1252, 850, "French_Belgium" },
1563 { 1252, 850, "French_Canada" },
1564 { 1252, 850, "French_Switzerland" },
1565 { 1252, 850, "German_Germany" },
1566 { 1252, 850, "German_Austria" },
1567 { 1252, 850, "German_Switzerland" },
1568 { 1253, 737, "Greek_Greece" },
1569 { 1250, 852, "Hungarian_Hungary" },
1570 { 1252, 850, "Icelandic_Iceland" },
1571 { 1252, 850, "Italian_Italy" },
1572 { 1252, 850, "Italian_Switzerland" },
1573 { 932, 932, "Japanese_Japan" },
1574 { 949, 949, "Korean_Korea" },
1575 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1576 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1577 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1578 { 1250, 852, "Polish_Poland" },
1579 { 1252, 850, "Portuguese_Portugal" },
1580 { 1252, 850, "Portuguese_Brazil" },
1581 { 1251, 866, "Russian_Russia" },
1582 { 1250, 852, "Slovak_Slovakia" },
1583 { 1252, 850, "Spanish_Spain" },
1584 { 1252, 850, "Spanish_Mexico" },
1585 { 1252, 850, "Spanish_Spain" },
1586 { 1252, 850, "Swedish_Sweden" },
1587 { 1254, 857, "Turkish_Turkey" },
1588 { 0, 0, NULL}
1589 };
1590
1591 /*
1592 * Return OEM Code Page of current locale set by setlocale().
1593 */
1594 static unsigned
get_current_oemcp(void)1595 get_current_oemcp(void)
1596 {
1597 int i;
1598 char *locale, *p;
1599 size_t len;
1600
1601 locale = setlocale(LC_CTYPE, NULL);
1602 if (locale == NULL)
1603 return (GetOEMCP());
1604 if (locale[0] == 'C' && locale[1] == '\0')
1605 return (CP_C_LOCALE);
1606
1607 p = strrchr(locale, '.');
1608 if (p == NULL)
1609 return (GetOEMCP());
1610 len = p - locale;
1611 for (i = 0; acp_ocp_map[i].acp; i++) {
1612 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1613 return (acp_ocp_map[i].ocp);
1614 }
1615 return (GetOEMCP());
1616 }
1617 #else
1618
1619 /*
1620 * POSIX platform does not use CodePage.
1621 */
1622
1623 static unsigned
get_current_codepage(void)1624 get_current_codepage(void)
1625 {
1626 return (-1);/* Unknown */
1627 }
1628 static unsigned
make_codepage_from_charset(const char * charset)1629 make_codepage_from_charset(const char *charset)
1630 {
1631 (void)charset; /* UNUSED */
1632 return (-1);/* Unknown */
1633 }
1634 static unsigned
get_current_oemcp(void)1635 get_current_oemcp(void)
1636 {
1637 return (-1);/* Unknown */
1638 }
1639
1640 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1641
1642 /*
1643 * Return a string conversion object.
1644 */
1645 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1646 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1647 {
1648 struct archive_string_conv *sc;
1649 unsigned current_codepage;
1650
1651 /* Check if we have made the sconv object. */
1652 sc = find_sconv_object(a, fc, tc);
1653 if (sc != NULL)
1654 return (sc);
1655
1656 if (a == NULL)
1657 current_codepage = get_current_codepage();
1658 else
1659 current_codepage = a->current_codepage;
1660
1661 sc = create_sconv_object(canonical_charset_name(fc),
1662 canonical_charset_name(tc), current_codepage, flag);
1663 if (sc == NULL) {
1664 if (a != NULL)
1665 archive_set_error(a, ENOMEM,
1666 "Could not allocate memory for "
1667 "a string conversion object");
1668 return (NULL);
1669 }
1670
1671 /*
1672 * If there is no converter for current string conversion object,
1673 * we cannot handle this conversion.
1674 */
1675 if (sc->nconverter == 0) {
1676 if (a != NULL) {
1677 #if HAVE_ICONV
1678 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1679 "iconv_open failed : Cannot handle ``%s''",
1680 (flag & SCONV_TO_CHARSET)?tc:fc);
1681 #else
1682 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1683 "A character-set conversion not fully supported "
1684 "on this platform");
1685 #endif
1686 }
1687 /* Failed; free a sconv object. */
1688 free_sconv_object(sc);
1689 return (NULL);
1690 }
1691
1692 /*
1693 * Success!
1694 */
1695 if (a != NULL)
1696 add_sconv_object(a, sc);
1697 return (sc);
1698 }
1699
1700 static const char *
get_current_charset(struct archive * a)1701 get_current_charset(struct archive *a)
1702 {
1703 const char *cur_charset;
1704
1705 if (a == NULL)
1706 cur_charset = default_iconv_charset("");
1707 else {
1708 cur_charset = default_iconv_charset(a->current_code);
1709 if (a->current_code == NULL) {
1710 a->current_code = strdup(cur_charset);
1711 a->current_codepage = get_current_codepage();
1712 a->current_oemcp = get_current_oemcp();
1713 }
1714 }
1715 return (cur_charset);
1716 }
1717
1718 /*
1719 * Make and Return a string conversion object.
1720 * Return NULL if the platform does not support the specified conversion
1721 * and best_effort is 0.
1722 * If best_effort is set, A string conversion object must be returned
1723 * unless memory allocation for the object fails, but the conversion
1724 * might fail when non-ASCII code is found.
1725 */
1726 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1727 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1728 int best_effort)
1729 {
1730 int flag = SCONV_TO_CHARSET;
1731
1732 if (best_effort)
1733 flag |= SCONV_BEST_EFFORT;
1734 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1735 }
1736
1737 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1738 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1739 int best_effort)
1740 {
1741 int flag = SCONV_FROM_CHARSET;
1742
1743 if (best_effort)
1744 flag |= SCONV_BEST_EFFORT;
1745 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1746 }
1747
1748 /*
1749 * archive_string_default_conversion_*_archive() are provided for Windows
1750 * platform because other archiver application use CP_OEMCP for
1751 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1752 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1753 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1754 * So we should make a string conversion between CP_ACP and CP_OEMCP
1755 * for compatibility.
1756 */
1757 #if defined(_WIN32) && !defined(__CYGWIN__)
1758 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1759 archive_string_default_conversion_for_read(struct archive *a)
1760 {
1761 const char *cur_charset = get_current_charset(a);
1762 char oemcp[16];
1763
1764 /* NOTE: a check of cur_charset is unneeded but we need
1765 * that get_current_charset() has been surely called at
1766 * this time whatever C compiler optimized. */
1767 if (cur_charset != NULL &&
1768 (a->current_codepage == CP_C_LOCALE ||
1769 a->current_codepage == a->current_oemcp))
1770 return (NULL);/* no conversion. */
1771
1772 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1773 /* Make sure a null termination must be set. */
1774 oemcp[sizeof(oemcp)-1] = '\0';
1775 return (get_sconv_object(a, oemcp, cur_charset,
1776 SCONV_FROM_CHARSET));
1777 }
1778
1779 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1780 archive_string_default_conversion_for_write(struct archive *a)
1781 {
1782 const char *cur_charset = get_current_charset(a);
1783 char oemcp[16];
1784
1785 /* NOTE: a check of cur_charset is unneeded but we need
1786 * that get_current_charset() has been surely called at
1787 * this time whatever C compiler optimized. */
1788 if (cur_charset != NULL &&
1789 (a->current_codepage == CP_C_LOCALE ||
1790 a->current_codepage == a->current_oemcp))
1791 return (NULL);/* no conversion. */
1792
1793 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1794 /* Make sure a null termination must be set. */
1795 oemcp[sizeof(oemcp)-1] = '\0';
1796 return (get_sconv_object(a, cur_charset, oemcp,
1797 SCONV_TO_CHARSET));
1798 }
1799 #else
1800 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1801 archive_string_default_conversion_for_read(struct archive *a)
1802 {
1803 (void)a; /* UNUSED */
1804 return (NULL);
1805 }
1806
1807 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1808 archive_string_default_conversion_for_write(struct archive *a)
1809 {
1810 (void)a; /* UNUSED */
1811 return (NULL);
1812 }
1813 #endif
1814
1815 /*
1816 * Dispose of all character conversion objects in the archive object.
1817 */
1818 void
archive_string_conversion_free(struct archive * a)1819 archive_string_conversion_free(struct archive *a)
1820 {
1821 struct archive_string_conv *sc;
1822 struct archive_string_conv *sc_next;
1823
1824 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1825 sc_next = sc->next;
1826 free_sconv_object(sc);
1827 }
1828 a->sconv = NULL;
1829 free(a->current_code);
1830 a->current_code = NULL;
1831 }
1832
1833 /*
1834 * Return a conversion charset name.
1835 */
1836 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1837 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1838 {
1839 if (sc->flag & SCONV_TO_CHARSET)
1840 return (sc->to_charset);
1841 else
1842 return (sc->from_charset);
1843 }
1844
1845 /*
1846 * Change the behavior of a string conversion.
1847 */
1848 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1849 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1850 {
1851 switch (opt) {
1852 /*
1853 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1854 * assumption that wchar_t was Unicode.
1855 * This option enables simulating the assumption in order to read
1856 * that filename correctly.
1857 */
1858 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1859 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1860 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1861 /*
1862 * Nothing to do for it since wchar_t on these platforms
1863 * is really Unicode.
1864 */
1865 (void)sc; /* UNUSED */
1866 #else
1867 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1868 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1869 /* Set up string converters. */
1870 setup_converter(sc);
1871 }
1872 #endif
1873 break;
1874 case SCONV_SET_OPT_NORMALIZATION_C:
1875 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1876 sc->flag |= SCONV_NORMALIZATION_C;
1877 sc->flag &= ~SCONV_NORMALIZATION_D;
1878 /* Set up string converters. */
1879 setup_converter(sc);
1880 }
1881 break;
1882 case SCONV_SET_OPT_NORMALIZATION_D:
1883 #if defined(HAVE_ICONV)
1884 /*
1885 * If iconv will take the string, do not change the
1886 * setting of the normalization.
1887 */
1888 if (!(sc->flag & SCONV_WIN_CP) &&
1889 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1890 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1891 break;
1892 #endif
1893 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1894 sc->flag |= SCONV_NORMALIZATION_D;
1895 sc->flag &= ~SCONV_NORMALIZATION_C;
1896 /* Set up string converters. */
1897 setup_converter(sc);
1898 }
1899 break;
1900 default:
1901 break;
1902 }
1903 }
1904
1905 /*
1906 *
1907 * Copy one archive_string to another in locale conversion.
1908 *
1909 * archive_strncat_l();
1910 * archive_strncpy_l();
1911 *
1912 */
1913
1914 static size_t
mbsnbytes(const void * _p,size_t n)1915 mbsnbytes(const void *_p, size_t n)
1916 {
1917 size_t s;
1918 const char *p, *pp;
1919
1920 if (_p == NULL)
1921 return (0);
1922 p = (const char *)_p;
1923
1924 /* Like strlen(p), except won't examine positions beyond p[n]. */
1925 s = 0;
1926 pp = p;
1927 while (s < n && *pp) {
1928 pp++;
1929 s++;
1930 }
1931 return (s);
1932 }
1933
1934 static size_t
utf16nbytes(const void * _p,size_t n)1935 utf16nbytes(const void *_p, size_t n)
1936 {
1937 size_t s;
1938 const char *p, *pp;
1939
1940 if (_p == NULL)
1941 return (0);
1942 p = (const char *)_p;
1943
1944 /* Like strlen(p), except won't examine positions beyond p[n]. */
1945 s = 0;
1946 pp = p;
1947 n >>= 1;
1948 while (s < n && (pp[0] || pp[1])) {
1949 pp += 2;
1950 s++;
1951 }
1952 return (s<<1);
1953 }
1954
1955 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1956 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1957 struct archive_string_conv *sc)
1958 {
1959 as->length = 0;
1960 return (archive_strncat_l(as, _p, n, sc));
1961 }
1962
1963 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1964 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1965 struct archive_string_conv *sc)
1966 {
1967 const void *s;
1968 size_t length = 0;
1969 int i, r = 0, r2;
1970
1971 if (_p != NULL && n > 0) {
1972 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
1973 length = utf16nbytes(_p, n);
1974 else
1975 length = mbsnbytes(_p, n);
1976 }
1977
1978 /* We must allocate memory even if there is no data for conversion
1979 * or copy. This simulates archive_string_append behavior. */
1980 if (length == 0) {
1981 int tn = 1;
1982 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1983 tn = 2;
1984 if (archive_string_ensure(as, as->length + tn) == NULL)
1985 return (-1);
1986 as->s[as->length] = 0;
1987 if (tn == 2)
1988 as->s[as->length+1] = 0;
1989 return (0);
1990 }
1991
1992 /*
1993 * If sc is NULL, we just make a copy.
1994 */
1995 if (sc == NULL) {
1996 if (archive_string_append(as, _p, length) == NULL)
1997 return (-1);/* No memory */
1998 return (0);
1999 }
2000
2001 s = _p;
2002 i = 0;
2003 if (sc->nconverter > 1) {
2004 sc->utftmp.length = 0;
2005 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2006 if (r2 != 0 && errno == ENOMEM)
2007 return (r2);
2008 if (r > r2)
2009 r = r2;
2010 s = sc->utftmp.s;
2011 length = sc->utftmp.length;
2012 ++i;
2013 }
2014 r2 = sc->converter[i](as, s, length, sc);
2015 if (r > r2)
2016 r = r2;
2017 return (r);
2018 }
2019
2020 #if HAVE_ICONV
2021
2022 /*
2023 * Return -1 if conversion fails.
2024 */
2025 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2026 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2027 size_t length, struct archive_string_conv *sc)
2028 {
2029 ICONV_CONST char *itp;
2030 size_t remaining;
2031 iconv_t cd;
2032 char *outp;
2033 size_t avail, bs;
2034 int return_value = 0; /* success */
2035 int to_size, from_size;
2036
2037 if (sc->flag & SCONV_TO_UTF16)
2038 to_size = 2;
2039 else
2040 to_size = 1;
2041 if (sc->flag & SCONV_FROM_UTF16)
2042 from_size = 2;
2043 else
2044 from_size = 1;
2045
2046 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2047 return (-1);
2048
2049 cd = sc->cd;
2050 itp = (char *)(uintptr_t)_p;
2051 remaining = length;
2052 outp = as->s + as->length;
2053 avail = as->buffer_length - as->length - to_size;
2054 while (remaining >= (size_t)from_size) {
2055 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2056
2057 if (result != (size_t)-1)
2058 break; /* Conversion completed. */
2059
2060 if (errno == EILSEQ || errno == EINVAL) {
2061 /*
2062 * If an output charset is UTF-8 or UTF-16BE/LE,
2063 * unknown character should be U+FFFD
2064 * (replacement character).
2065 */
2066 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2067 size_t rbytes;
2068 if (sc->flag & SCONV_TO_UTF8)
2069 rbytes = sizeof(utf8_replacement_char);
2070 else
2071 rbytes = 2;
2072
2073 if (avail < rbytes) {
2074 as->length = outp - as->s;
2075 bs = as->buffer_length +
2076 (remaining * to_size) + rbytes;
2077 if (NULL ==
2078 archive_string_ensure(as, bs))
2079 return (-1);
2080 outp = as->s + as->length;
2081 avail = as->buffer_length
2082 - as->length - to_size;
2083 }
2084 if (sc->flag & SCONV_TO_UTF8)
2085 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2086 else if (sc->flag & SCONV_TO_UTF16BE)
2087 archive_be16enc(outp, UNICODE_R_CHAR);
2088 else
2089 archive_le16enc(outp, UNICODE_R_CHAR);
2090 outp += rbytes;
2091 avail -= rbytes;
2092 } else {
2093 /* Skip the illegal input bytes. */
2094 *outp++ = '?';
2095 avail--;
2096 }
2097 itp += from_size;
2098 remaining -= from_size;
2099 return_value = -1; /* failure */
2100 } else {
2101 /* E2BIG no output buffer,
2102 * Increase an output buffer. */
2103 as->length = outp - as->s;
2104 bs = as->buffer_length + remaining * 2;
2105 if (NULL == archive_string_ensure(as, bs))
2106 return (-1);
2107 outp = as->s + as->length;
2108 avail = as->buffer_length - as->length - to_size;
2109 }
2110 }
2111 as->length = outp - as->s;
2112 as->s[as->length] = 0;
2113 if (to_size == 2)
2114 as->s[as->length+1] = 0;
2115 return (return_value);
2116 }
2117
2118 #endif /* HAVE_ICONV */
2119
2120
2121 #if defined(_WIN32) && !defined(__CYGWIN__)
2122
2123 /*
2124 * Translate a string from a some CodePage to an another CodePage by
2125 * Windows APIs, and copy the result. Return -1 if conversion fails.
2126 */
2127 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2128 strncat_in_codepage(struct archive_string *as,
2129 const void *_p, size_t length, struct archive_string_conv *sc)
2130 {
2131 const char *s = (const char *)_p;
2132 struct archive_wstring aws;
2133 size_t l;
2134 int r, saved_flag;
2135
2136 archive_string_init(&aws);
2137 saved_flag = sc->flag;
2138 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2139 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2140 sc->flag = saved_flag;
2141 if (r != 0) {
2142 archive_wstring_free(&aws);
2143 if (errno != ENOMEM)
2144 archive_string_append(as, s, length);
2145 return (-1);
2146 }
2147
2148 l = as->length;
2149 r = archive_string_append_from_wcs_in_codepage(
2150 as, aws.s, aws.length, sc);
2151 if (r != 0 && errno != ENOMEM && l == as->length)
2152 archive_string_append(as, s, length);
2153 archive_wstring_free(&aws);
2154 return (r);
2155 }
2156
2157 /*
2158 * Test whether MBS ==> WCS is okay.
2159 */
2160 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2161 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2162 {
2163 const char *p = (const char *)_p;
2164 unsigned codepage;
2165 DWORD mbflag = MB_ERR_INVALID_CHARS;
2166
2167 if (sc->flag & SCONV_FROM_CHARSET)
2168 codepage = sc->to_cp;
2169 else
2170 codepage = sc->from_cp;
2171
2172 if (codepage == CP_C_LOCALE)
2173 return (0);
2174 if (codepage != CP_UTF8)
2175 mbflag |= MB_PRECOMPOSED;
2176
2177 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2178 return (-1); /* Invalid */
2179 return (0); /* Okay */
2180 }
2181
2182 #else
2183
2184 /*
2185 * Test whether MBS ==> WCS is okay.
2186 */
2187 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2188 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2189 {
2190 const char *p = (const char *)_p;
2191 size_t r;
2192
2193 #if HAVE_MBRTOWC
2194 mbstate_t shift_state;
2195
2196 memset(&shift_state, 0, sizeof(shift_state));
2197 #else
2198 /* Clear the shift state before starting. */
2199 mbtowc(NULL, NULL, 0);
2200 #endif
2201 while (n) {
2202 wchar_t wc;
2203
2204 #if HAVE_MBRTOWC
2205 r = mbrtowc(&wc, p, n, &shift_state);
2206 #else
2207 r = mbtowc(&wc, p, n);
2208 #endif
2209 if (r == (size_t)-1 || r == (size_t)-2)
2210 return (-1);/* Invalid. */
2211 if (r == 0)
2212 break;
2213 p += r;
2214 n -= r;
2215 }
2216 (void)sc; /* UNUSED */
2217 return (0); /* All Okey. */
2218 }
2219
2220 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2221
2222 /*
2223 * Basically returns -1 because we cannot make a conversion of charset
2224 * without iconv but in some cases this would return 0.
2225 * Returns 0 if all copied characters are ASCII.
2226 * Returns 0 if both from-locale and to-locale are the same and those
2227 * can be WCS with no error.
2228 */
2229 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2230 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2231 size_t length, struct archive_string_conv *sc)
2232 {
2233 size_t remaining;
2234 const uint8_t *itp;
2235 int return_value = 0; /* success */
2236
2237 /*
2238 * If both from-locale and to-locale is the same, this makes a copy.
2239 * And then this checks all copied MBS can be WCS if so returns 0.
2240 */
2241 if (sc->same) {
2242 if (archive_string_append(as, _p, length) == NULL)
2243 return (-1);/* No memory */
2244 return (invalid_mbs(_p, length, sc));
2245 }
2246
2247 /*
2248 * If a character is ASCII, this just copies it. If not, this
2249 * assigns '?' character instead but in UTF-8 locale this assigns
2250 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2251 * a Replacement Character in Unicode.
2252 */
2253
2254 remaining = length;
2255 itp = (const uint8_t *)_p;
2256 while (*itp && remaining > 0) {
2257 if (*itp > 127) {
2258 // Non-ASCII: Substitute with suitable replacement
2259 if (sc->flag & SCONV_TO_UTF8) {
2260 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2261 __archive_errx(1, "Out of memory");
2262 }
2263 } else {
2264 archive_strappend_char(as, '?');
2265 }
2266 return_value = -1;
2267 } else {
2268 archive_strappend_char(as, *itp);
2269 }
2270 ++itp;
2271 }
2272 return (return_value);
2273 }
2274
2275
2276 /*
2277 * Unicode conversion functions.
2278 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2279 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2280 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2281 * - UTF-16BE <===> UTF-8.
2282 *
2283 */
2284
2285 /*
2286 * Utility to convert a single UTF-8 sequence.
2287 *
2288 * Usually return used bytes, return used byte in negative value when
2289 * a unicode character is replaced with U+FFFD.
2290 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2291 * Recommended Practice for Replacement Characters.
2292 */
2293 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2294 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2295 {
2296 static const char utf8_count[256] = {
2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2299 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2300 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2301 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2302 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2303 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2304 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2305 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2306 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2307 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2308 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2309 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2310 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2311 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2312 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2313 };
2314 int ch, i;
2315 int cnt;
2316 uint32_t wc;
2317
2318 /* Sanity check. */
2319 if (n == 0)
2320 return (0);
2321 /*
2322 * Decode 1-4 bytes depending on the value of the first byte.
2323 */
2324 ch = (unsigned char)*s;
2325 if (ch == 0)
2326 return (0); /* Standard: return 0 for end-of-string. */
2327 cnt = utf8_count[ch];
2328
2329 /* Invalid sequence or there are not plenty bytes. */
2330 if ((int)n < cnt) {
2331 cnt = (int)n;
2332 for (i = 1; i < cnt; i++) {
2333 if ((s[i] & 0xc0) != 0x80) {
2334 cnt = i;
2335 break;
2336 }
2337 }
2338 goto invalid_sequence;
2339 }
2340
2341 /* Make a Unicode code point from a single UTF-8 sequence. */
2342 switch (cnt) {
2343 case 1: /* 1 byte sequence. */
2344 *pwc = ch & 0x7f;
2345 return (cnt);
2346 case 2: /* 2 bytes sequence. */
2347 if ((s[1] & 0xc0) != 0x80) {
2348 cnt = 1;
2349 goto invalid_sequence;
2350 }
2351 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2352 return (cnt);
2353 case 3: /* 3 bytes sequence. */
2354 if ((s[1] & 0xc0) != 0x80) {
2355 cnt = 1;
2356 goto invalid_sequence;
2357 }
2358 if ((s[2] & 0xc0) != 0x80) {
2359 cnt = 2;
2360 goto invalid_sequence;
2361 }
2362 wc = ((ch & 0x0f) << 12)
2363 | ((s[1] & 0x3f) << 6)
2364 | (s[2] & 0x3f);
2365 if (wc < 0x800)
2366 goto invalid_sequence;/* Overlong sequence. */
2367 break;
2368 case 4: /* 4 bytes sequence. */
2369 if ((s[1] & 0xc0) != 0x80) {
2370 cnt = 1;
2371 goto invalid_sequence;
2372 }
2373 if ((s[2] & 0xc0) != 0x80) {
2374 cnt = 2;
2375 goto invalid_sequence;
2376 }
2377 if ((s[3] & 0xc0) != 0x80) {
2378 cnt = 3;
2379 goto invalid_sequence;
2380 }
2381 wc = ((ch & 0x07) << 18)
2382 | ((s[1] & 0x3f) << 12)
2383 | ((s[2] & 0x3f) << 6)
2384 | (s[3] & 0x3f);
2385 if (wc < 0x10000)
2386 goto invalid_sequence;/* Overlong sequence. */
2387 break;
2388 default: /* Others are all invalid sequence. */
2389 if (ch == 0xc0 || ch == 0xc1)
2390 cnt = 2;
2391 else if (ch >= 0xf5 && ch <= 0xf7)
2392 cnt = 4;
2393 else if (ch >= 0xf8 && ch <= 0xfb)
2394 cnt = 5;
2395 else if (ch == 0xfc || ch == 0xfd)
2396 cnt = 6;
2397 else
2398 cnt = 1;
2399 if ((int)n < cnt)
2400 cnt = (int)n;
2401 for (i = 1; i < cnt; i++) {
2402 if ((s[i] & 0xc0) != 0x80) {
2403 cnt = i;
2404 break;
2405 }
2406 }
2407 goto invalid_sequence;
2408 }
2409
2410 /* The code point larger than 0x10FFFF is not legal
2411 * Unicode values. */
2412 if (wc > UNICODE_MAX)
2413 goto invalid_sequence;
2414 /* Correctly gets a Unicode, returns used bytes. */
2415 *pwc = wc;
2416 return (cnt);
2417 invalid_sequence:
2418 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2419 return (cnt * -1);
2420 }
2421
2422 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2423 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2424 {
2425 int cnt;
2426
2427 cnt = _utf8_to_unicode(pwc, s, n);
2428 /* Any of Surrogate pair is not legal Unicode values. */
2429 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2430 return (-3);
2431 return (cnt);
2432 }
2433
2434 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2435 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2436 {
2437 uc -= 0xD800;
2438 uc *= 0x400;
2439 uc += uc2 - 0xDC00;
2440 uc += 0x10000;
2441 return (uc);
2442 }
2443
2444 /*
2445 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2446 * removing surrogate pairs.
2447 *
2448 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2449 *
2450 * Usually return used bytes, return used byte in negative value when
2451 * a unicode character is replaced with U+FFFD.
2452 */
2453 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2454 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2455 {
2456 uint32_t wc = 0;
2457 int cnt;
2458
2459 cnt = _utf8_to_unicode(&wc, s, n);
2460 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2461 uint32_t wc2 = 0;
2462 if (n - 3 < 3) {
2463 /* Invalid byte sequence. */
2464 goto invalid_sequence;
2465 }
2466 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2467 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2468 /* Invalid byte sequence. */
2469 goto invalid_sequence;
2470 }
2471 wc = combine_surrogate_pair(wc, wc2);
2472 cnt = 6;
2473 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2474 /* Invalid byte sequence. */
2475 goto invalid_sequence;
2476 }
2477 *pwc = wc;
2478 return (cnt);
2479 invalid_sequence:
2480 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2481 if (cnt > 0)
2482 cnt *= -1;
2483 return (cnt);
2484 }
2485
2486 /*
2487 * Convert a Unicode code point to a single UTF-8 sequence.
2488 *
2489 * NOTE:This function does not check if the Unicode is legal or not.
2490 * Please you definitely check it before calling this.
2491 */
2492 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2493 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2494 {
2495 char *_p = p;
2496
2497 /* Invalid Unicode char maps to Replacement character */
2498 if (uc > UNICODE_MAX)
2499 uc = UNICODE_R_CHAR;
2500 /* Translate code point to UTF8 */
2501 if (uc <= 0x7f) {
2502 if (remaining == 0)
2503 return (0);
2504 *p++ = (char)uc;
2505 } else if (uc <= 0x7ff) {
2506 if (remaining < 2)
2507 return (0);
2508 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2509 *p++ = 0x80 | (uc & 0x3f);
2510 } else if (uc <= 0xffff) {
2511 if (remaining < 3)
2512 return (0);
2513 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2514 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2515 *p++ = 0x80 | (uc & 0x3f);
2516 } else {
2517 if (remaining < 4)
2518 return (0);
2519 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2520 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2521 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2522 *p++ = 0x80 | (uc & 0x3f);
2523 }
2524 return (p - _p);
2525 }
2526
2527 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2528 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2529 {
2530 return (utf16_to_unicode(pwc, s, n, 1));
2531 }
2532
2533 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2534 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2535 {
2536 return (utf16_to_unicode(pwc, s, n, 0));
2537 }
2538
2539 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2540 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2541 {
2542 const char *utf16 = s;
2543 unsigned uc;
2544
2545 if (n == 0)
2546 return (0);
2547 if (n == 1) {
2548 /* set the Replacement Character instead. */
2549 *pwc = UNICODE_R_CHAR;
2550 return (-1);
2551 }
2552
2553 if (be)
2554 uc = archive_be16dec(utf16);
2555 else
2556 uc = archive_le16dec(utf16);
2557 utf16 += 2;
2558
2559 /* If this is a surrogate pair, assemble the full code point.*/
2560 if (IS_HIGH_SURROGATE_LA(uc)) {
2561 unsigned uc2;
2562
2563 if (n >= 4) {
2564 if (be)
2565 uc2 = archive_be16dec(utf16);
2566 else
2567 uc2 = archive_le16dec(utf16);
2568 } else
2569 uc2 = 0;
2570 if (IS_LOW_SURROGATE_LA(uc2)) {
2571 uc = combine_surrogate_pair(uc, uc2);
2572 utf16 += 2;
2573 } else {
2574 /* Undescribed code point should be U+FFFD
2575 * (replacement character). */
2576 *pwc = UNICODE_R_CHAR;
2577 return (-2);
2578 }
2579 }
2580
2581 /*
2582 * Surrogate pair values(0xd800 through 0xdfff) are only
2583 * used by UTF-16, so, after above calculation, the code
2584 * must not be surrogate values, and Unicode has no codes
2585 * larger than 0x10ffff. Thus, those are not legal Unicode
2586 * values.
2587 */
2588 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2589 /* Undescribed code point should be U+FFFD
2590 * (replacement character). */
2591 *pwc = UNICODE_R_CHAR;
2592 return (((int)(utf16 - s)) * -1);
2593 }
2594 *pwc = uc;
2595 return ((int)(utf16 - s));
2596 }
2597
2598 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2599 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2600 {
2601 char *utf16 = p;
2602
2603 if (uc > 0xffff) {
2604 /* We have a code point that won't fit into a
2605 * wchar_t; convert it to a surrogate pair. */
2606 if (remaining < 4)
2607 return (0);
2608 uc -= 0x10000;
2609 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2610 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2611 return (4);
2612 } else {
2613 if (remaining < 2)
2614 return (0);
2615 archive_be16enc(utf16, uc);
2616 return (2);
2617 }
2618 }
2619
2620 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2621 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2622 {
2623 char *utf16 = p;
2624
2625 if (uc > 0xffff) {
2626 /* We have a code point that won't fit into a
2627 * wchar_t; convert it to a surrogate pair. */
2628 if (remaining < 4)
2629 return (0);
2630 uc -= 0x10000;
2631 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2632 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2633 return (4);
2634 } else {
2635 if (remaining < 2)
2636 return (0);
2637 archive_le16enc(utf16, uc);
2638 return (2);
2639 }
2640 }
2641
2642 /*
2643 * Append new UTF-8 string to existing UTF-8 string.
2644 * Existing string is assumed to already be in proper form;
2645 * the new string will have invalid sequences replaced and
2646 * surrogate pairs canonicalized.
2647 */
2648 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _src,size_t len,struct archive_string_conv * sc)2649 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_src,
2650 size_t len, struct archive_string_conv *sc)
2651 {
2652 int ret = 0;
2653 const char *src = _src;
2654 (void)sc; /* UNUSED */
2655
2656 /* Pre-extend the destination */
2657 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2658 return (-1);
2659
2660 /* Invariant: src points to the first UTF8 byte that hasn't
2661 * been copied to the destination `as`. */
2662 for (;;) {
2663 int n;
2664 uint32_t uc;
2665 const char *e = src;
2666
2667 /* Skip UTF-8 sequences until we reach end-of-string or
2668 * a code point that needs conversion. */
2669 while ((n = utf8_to_unicode(&uc, e, len)) > 0) {
2670 e += n;
2671 len -= n;
2672 }
2673 /* Copy the part that doesn't need conversion */
2674 if (e > src) {
2675 if (archive_string_append(as, src, e - src) == NULL)
2676 return (-1);
2677 src = e;
2678 }
2679
2680 if (n == 0) {
2681 /* We reached end-of-string */
2682 return (ret);
2683 } else {
2684 /* Next code point needs conversion */
2685 char t[4];
2686 size_t w;
2687
2688 /* Try decoding a surrogate pair */
2689 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2690 n = cesu8_to_unicode(&uc, src, len);
2691 }
2692 /* Not a (valid) surrogate, so use a replacement char */
2693 if (n < 0) {
2694 ret = -1; /* Return -1 if we used any replacement */
2695 n *= -1;
2696 }
2697 /* Consume converted code point */
2698 src += n;
2699 len -= n;
2700 /* Convert and append new UTF-8 sequence. */
2701 w = unicode_to_utf8(t, sizeof(t), uc);
2702 if (archive_string_append(as, t, w) == NULL)
2703 return (-1);
2704 }
2705 }
2706 }
2707
2708 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2709 archive_string_append_unicode(struct archive_string *as, const void *_p,
2710 size_t len, struct archive_string_conv *sc)
2711 {
2712 const char *s;
2713 char *p, *endp;
2714 uint32_t uc;
2715 size_t w;
2716 int n, ret = 0, ts, tm;
2717 int (*parse)(uint32_t *, const char *, size_t);
2718 size_t (*unparse)(char *, size_t, uint32_t);
2719
2720 if (sc->flag & SCONV_TO_UTF16BE) {
2721 unparse = unicode_to_utf16be;
2722 ts = 2;
2723 } else if (sc->flag & SCONV_TO_UTF16LE) {
2724 unparse = unicode_to_utf16le;
2725 ts = 2;
2726 } else if (sc->flag & SCONV_TO_UTF8) {
2727 unparse = unicode_to_utf8;
2728 ts = 1;
2729 } else {
2730 /*
2731 * This case is going to be converted to another
2732 * character-set through iconv.
2733 */
2734 if (sc->flag & SCONV_FROM_UTF16BE) {
2735 unparse = unicode_to_utf16be;
2736 ts = 2;
2737 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2738 unparse = unicode_to_utf16le;
2739 ts = 2;
2740 } else {
2741 unparse = unicode_to_utf8;
2742 ts = 1;
2743 }
2744 }
2745
2746 if (sc->flag & SCONV_FROM_UTF16BE) {
2747 parse = utf16be_to_unicode;
2748 tm = 1;
2749 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2750 parse = utf16le_to_unicode;
2751 tm = 1;
2752 } else {
2753 parse = cesu8_to_unicode;
2754 tm = ts;
2755 }
2756
2757 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2758 return (-1);
2759
2760 s = (const char *)_p;
2761 p = as->s + as->length;
2762 endp = as->s + as->buffer_length - ts;
2763 while ((n = parse(&uc, s, len)) != 0) {
2764 if (n < 0) {
2765 /* Use a replaced unicode character. */
2766 n *= -1;
2767 ret = -1;
2768 }
2769 s += n;
2770 len -= n;
2771 while ((w = unparse(p, endp - p, uc)) == 0) {
2772 /* There is not enough output buffer so
2773 * we have to expand it. */
2774 as->length = p - as->s;
2775 if (archive_string_ensure(as,
2776 as->buffer_length + len * tm + ts) == NULL)
2777 return (-1);
2778 p = as->s + as->length;
2779 endp = as->s + as->buffer_length - ts;
2780 }
2781 p += w;
2782 }
2783 as->length = p - as->s;
2784 as->s[as->length] = '\0';
2785 if (ts == 2)
2786 as->s[as->length+1] = '\0';
2787 return (ret);
2788 }
2789
2790 /*
2791 * Following Constants for Hangul compositions this information comes from
2792 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2793 */
2794 #define HC_SBASE 0xAC00
2795 #define HC_LBASE 0x1100
2796 #define HC_VBASE 0x1161
2797 #define HC_TBASE 0x11A7
2798 #define HC_LCOUNT 19
2799 #define HC_VCOUNT 21
2800 #define HC_TCOUNT 28
2801 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2802 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2803
2804 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2805 get_nfc(uint32_t uc, uint32_t uc2)
2806 {
2807 int t, b;
2808
2809 t = 0;
2810 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2811 while (b >= t) {
2812 int m = (t + b) / 2;
2813 if (u_composition_table[m].cp1 < uc)
2814 t = m + 1;
2815 else if (u_composition_table[m].cp1 > uc)
2816 b = m - 1;
2817 else if (u_composition_table[m].cp2 < uc2)
2818 t = m + 1;
2819 else if (u_composition_table[m].cp2 > uc2)
2820 b = m - 1;
2821 else
2822 return (u_composition_table[m].nfc);
2823 }
2824 return (0);
2825 }
2826
2827 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2828 * Characters. */
2829
2830 /*
2831 * Update first code point.
2832 */
2833 #define UPDATE_UC(new_uc) do { \
2834 uc = new_uc; \
2835 ucptr = NULL; \
2836 } while (0)
2837
2838 /*
2839 * Replace first code point with second code point.
2840 */
2841 #define REPLACE_UC_WITH_UC2() do { \
2842 uc = uc2; \
2843 ucptr = uc2ptr; \
2844 n = n2; \
2845 } while (0)
2846
2847 #define EXPAND_BUFFER() do { \
2848 as->length = p - as->s; \
2849 if (archive_string_ensure(as, \
2850 as->buffer_length + len * tm + ts) == NULL)\
2851 return (-1); \
2852 p = as->s + as->length; \
2853 endp = as->s + as->buffer_length - ts; \
2854 } while (0)
2855
2856 #define UNPARSE(p, endp, uc) do { \
2857 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2858 EXPAND_BUFFER(); \
2859 } \
2860 p += w; \
2861 } while (0)
2862
2863 /*
2864 * Write first code point.
2865 * If the code point has not be changed from its original code,
2866 * this just copies it from its original buffer pointer.
2867 * If not, this converts it to UTF-8 byte sequence and copies it.
2868 */
2869 #define WRITE_UC() do { \
2870 if (ucptr) { \
2871 if (p + n > endp) \
2872 EXPAND_BUFFER(); \
2873 switch (n) { \
2874 case 4: \
2875 *p++ = *ucptr++; \
2876 /* FALL THROUGH */ \
2877 case 3: \
2878 *p++ = *ucptr++; \
2879 /* FALL THROUGH */ \
2880 case 2: \
2881 *p++ = *ucptr++; \
2882 /* FALL THROUGH */ \
2883 case 1: \
2884 *p++ = *ucptr; \
2885 break; \
2886 } \
2887 ucptr = NULL; \
2888 } else { \
2889 UNPARSE(p, endp, uc); \
2890 } \
2891 } while (0)
2892
2893 /*
2894 * Collect following decomposable code points.
2895 */
2896 #define COLLECT_CPS(start) do { \
2897 int _i; \
2898 for (_i = start; _i < FDC_MAX ; _i++) { \
2899 nx = parse(&ucx[_i], s, len); \
2900 if (nx <= 0) \
2901 break; \
2902 cx = CCC(ucx[_i]); \
2903 if (cl >= cx && cl != 228 && cx != 228)\
2904 break; \
2905 s += nx; \
2906 len -= nx; \
2907 cl = cx; \
2908 ccx[_i] = cx; \
2909 } \
2910 if (_i >= FDC_MAX) { \
2911 ret = -1; \
2912 ucx_size = FDC_MAX; \
2913 } else \
2914 ucx_size = _i; \
2915 } while (0)
2916
2917 /*
2918 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2919 *
2920 * TODO: Convert composition exclusions, which are never converted
2921 * from NFC,NFD,NFKC and NFKD, to Form C.
2922 */
2923 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2924 archive_string_normalize_C(struct archive_string *as, const void *_p,
2925 size_t len, struct archive_string_conv *sc)
2926 {
2927 const char *s = (const char *)_p;
2928 char *p, *endp;
2929 uint32_t uc, uc2;
2930 size_t w;
2931 int always_replace, n, n2, ret = 0, spair, ts, tm;
2932 int (*parse)(uint32_t *, const char *, size_t);
2933 size_t (*unparse)(char *, size_t, uint32_t);
2934
2935 always_replace = 1;
2936 ts = 1;/* text size. */
2937 if (sc->flag & SCONV_TO_UTF16BE) {
2938 unparse = unicode_to_utf16be;
2939 ts = 2;
2940 if (sc->flag & SCONV_FROM_UTF16BE)
2941 always_replace = 0;
2942 } else if (sc->flag & SCONV_TO_UTF16LE) {
2943 unparse = unicode_to_utf16le;
2944 ts = 2;
2945 if (sc->flag & SCONV_FROM_UTF16LE)
2946 always_replace = 0;
2947 } else if (sc->flag & SCONV_TO_UTF8) {
2948 unparse = unicode_to_utf8;
2949 if (sc->flag & SCONV_FROM_UTF8)
2950 always_replace = 0;
2951 } else {
2952 /*
2953 * This case is going to be converted to another
2954 * character-set through iconv.
2955 */
2956 always_replace = 0;
2957 if (sc->flag & SCONV_FROM_UTF16BE) {
2958 unparse = unicode_to_utf16be;
2959 ts = 2;
2960 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2961 unparse = unicode_to_utf16le;
2962 ts = 2;
2963 } else {
2964 unparse = unicode_to_utf8;
2965 }
2966 }
2967
2968 if (sc->flag & SCONV_FROM_UTF16BE) {
2969 parse = utf16be_to_unicode;
2970 tm = 1;
2971 spair = 4;/* surrogate pair size in UTF-16. */
2972 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2973 parse = utf16le_to_unicode;
2974 tm = 1;
2975 spair = 4;/* surrogate pair size in UTF-16. */
2976 } else {
2977 parse = cesu8_to_unicode;
2978 tm = ts;
2979 spair = 6;/* surrogate pair size in UTF-8. */
2980 }
2981
2982 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2983 return (-1);
2984
2985 p = as->s + as->length;
2986 endp = as->s + as->buffer_length - ts;
2987 while ((n = parse(&uc, s, len)) != 0) {
2988 const char *ucptr, *uc2ptr;
2989
2990 if (n < 0) {
2991 /* Use a replaced unicode character. */
2992 UNPARSE(p, endp, uc);
2993 s += n*-1;
2994 len -= n*-1;
2995 ret = -1;
2996 continue;
2997 } else if (n == spair || always_replace)
2998 /* uc is converted from a surrogate pair.
2999 * this should be treated as a changed code. */
3000 ucptr = NULL;
3001 else
3002 ucptr = s;
3003 s += n;
3004 len -= n;
3005
3006 /* Read second code point. */
3007 while ((n2 = parse(&uc2, s, len)) > 0) {
3008 uint32_t ucx[FDC_MAX];
3009 int ccx[FDC_MAX];
3010 int cl, cx, i, nx, ucx_size;
3011 int LIndex,SIndex;
3012 uint32_t nfc;
3013
3014 if (n2 == spair || always_replace)
3015 /* uc2 is converted from a surrogate pair.
3016 * this should be treated as a changed code. */
3017 uc2ptr = NULL;
3018 else
3019 uc2ptr = s;
3020 s += n2;
3021 len -= n2;
3022
3023 /*
3024 * If current second code point is out of decomposable
3025 * code points, finding compositions is unneeded.
3026 */
3027 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3028 WRITE_UC();
3029 REPLACE_UC_WITH_UC2();
3030 continue;
3031 }
3032
3033 /*
3034 * Try to combine current code points.
3035 */
3036 /*
3037 * We have to combine Hangul characters according to
3038 * http://uniicode.org/reports/tr15/#Hangul
3039 */
3040 if (0 <= (LIndex = uc - HC_LBASE) &&
3041 LIndex < HC_LCOUNT) {
3042 /*
3043 * Hangul Composition.
3044 * 1. Two current code points are L and V.
3045 */
3046 int VIndex = uc2 - HC_VBASE;
3047 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3048 /* Make syllable of form LV. */
3049 UPDATE_UC(HC_SBASE +
3050 (LIndex * HC_VCOUNT + VIndex) *
3051 HC_TCOUNT);
3052 } else {
3053 WRITE_UC();
3054 REPLACE_UC_WITH_UC2();
3055 }
3056 continue;
3057 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3058 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3059 /*
3060 * Hangul Composition.
3061 * 2. Two current code points are LV and T.
3062 */
3063 int TIndex = uc2 - HC_TBASE;
3064 if (0 < TIndex && TIndex < HC_TCOUNT) {
3065 /* Make syllable of form LVT. */
3066 UPDATE_UC(uc + TIndex);
3067 } else {
3068 WRITE_UC();
3069 REPLACE_UC_WITH_UC2();
3070 }
3071 continue;
3072 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3073 /* A composition to current code points
3074 * is found. */
3075 UPDATE_UC(nfc);
3076 continue;
3077 } else if ((cl = CCC(uc2)) == 0) {
3078 /* Clearly 'uc2' the second code point is not
3079 * a decomposable code. */
3080 WRITE_UC();
3081 REPLACE_UC_WITH_UC2();
3082 continue;
3083 }
3084
3085 /*
3086 * Collect following decomposable code points.
3087 */
3088 cx = 0;
3089 ucx[0] = uc2;
3090 ccx[0] = cl;
3091 COLLECT_CPS(1);
3092
3093 /*
3094 * Find a composed code in the collected code points.
3095 */
3096 i = 1;
3097 while (i < ucx_size) {
3098 int j;
3099
3100 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3101 i++;
3102 continue;
3103 }
3104
3105 /*
3106 * nfc is composed of uc and ucx[i].
3107 */
3108 UPDATE_UC(nfc);
3109
3110 /*
3111 * Remove ucx[i] by shifting
3112 * following code points.
3113 */
3114 for (j = i; j+1 < ucx_size; j++) {
3115 ucx[j] = ucx[j+1];
3116 ccx[j] = ccx[j+1];
3117 }
3118 ucx_size --;
3119
3120 /*
3121 * Collect following code points blocked
3122 * by ucx[i] the removed code point.
3123 */
3124 if (ucx_size > 0 && i == ucx_size &&
3125 nx > 0 && cx == cl) {
3126 cl = ccx[ucx_size-1];
3127 COLLECT_CPS(ucx_size);
3128 }
3129 /*
3130 * Restart finding a composed code with
3131 * the updated uc from the top of the
3132 * collected code points.
3133 */
3134 i = 0;
3135 }
3136
3137 /*
3138 * Apparently the current code points are not
3139 * decomposed characters or already composed.
3140 */
3141 WRITE_UC();
3142 for (i = 0; i < ucx_size; i++)
3143 UNPARSE(p, endp, ucx[i]);
3144
3145 /*
3146 * Flush out remaining canonical combining characters.
3147 */
3148 if (nx > 0 && cx == cl && len > 0) {
3149 while ((nx = parse(&ucx[0], s, len))
3150 > 0) {
3151 cx = CCC(ucx[0]);
3152 if (cl > cx)
3153 break;
3154 s += nx;
3155 len -= nx;
3156 cl = cx;
3157 UNPARSE(p, endp, ucx[0]);
3158 }
3159 }
3160 break;
3161 }
3162 if (n2 < 0) {
3163 WRITE_UC();
3164 /* Use a replaced unicode character. */
3165 UNPARSE(p, endp, uc2);
3166 s += n2*-1;
3167 len -= n2*-1;
3168 ret = -1;
3169 continue;
3170 } else if (n2 == 0) {
3171 WRITE_UC();
3172 break;
3173 }
3174 }
3175 as->length = p - as->s;
3176 as->s[as->length] = '\0';
3177 if (ts == 2)
3178 as->s[as->length+1] = '\0';
3179 return (ret);
3180 }
3181
3182 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3183 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3184 {
3185 int t, b;
3186
3187 /*
3188 * These are not converted to NFD on Mac OS.
3189 */
3190 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3191 (uc >= 0xF900 && uc <= 0xFAFF) ||
3192 (uc >= 0x2F800 && uc <= 0x2FAFF))
3193 return (0);
3194 /*
3195 * Those code points are not converted to NFD on Mac OS.
3196 * I do not know the reason because it is undocumented.
3197 * NFC NFD
3198 * 1109A ==> 11099 110BA
3199 * 1109C ==> 1109B 110BA
3200 * 110AB ==> 110A5 110BA
3201 */
3202 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3203 return (0);
3204
3205 t = 0;
3206 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3207 while (b >= t) {
3208 int m = (t + b) / 2;
3209 if (u_decomposition_table[m].nfc < uc)
3210 t = m + 1;
3211 else if (u_decomposition_table[m].nfc > uc)
3212 b = m - 1;
3213 else {
3214 *cp1 = u_decomposition_table[m].cp1;
3215 *cp2 = u_decomposition_table[m].cp2;
3216 return (1);
3217 }
3218 }
3219 return (0);
3220 }
3221
3222 #define REPLACE_UC_WITH(cp) do { \
3223 uc = cp; \
3224 ucptr = NULL; \
3225 } while (0)
3226
3227 /*
3228 * Normalize UTF-8 characters to Form D and copy the result.
3229 */
3230 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3231 archive_string_normalize_D(struct archive_string *as, const void *_p,
3232 size_t len, struct archive_string_conv *sc)
3233 {
3234 const char *s = (const char *)_p;
3235 char *p, *endp;
3236 uint32_t uc, uc2;
3237 size_t w;
3238 int always_replace, n, n2, ret = 0, spair, ts, tm;
3239 int (*parse)(uint32_t *, const char *, size_t);
3240 size_t (*unparse)(char *, size_t, uint32_t);
3241
3242 always_replace = 1;
3243 ts = 1;/* text size. */
3244 if (sc->flag & SCONV_TO_UTF16BE) {
3245 unparse = unicode_to_utf16be;
3246 ts = 2;
3247 if (sc->flag & SCONV_FROM_UTF16BE)
3248 always_replace = 0;
3249 } else if (sc->flag & SCONV_TO_UTF16LE) {
3250 unparse = unicode_to_utf16le;
3251 ts = 2;
3252 if (sc->flag & SCONV_FROM_UTF16LE)
3253 always_replace = 0;
3254 } else if (sc->flag & SCONV_TO_UTF8) {
3255 unparse = unicode_to_utf8;
3256 if (sc->flag & SCONV_FROM_UTF8)
3257 always_replace = 0;
3258 } else {
3259 /*
3260 * This case is going to be converted to another
3261 * character-set through iconv.
3262 */
3263 always_replace = 0;
3264 if (sc->flag & SCONV_FROM_UTF16BE) {
3265 unparse = unicode_to_utf16be;
3266 ts = 2;
3267 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3268 unparse = unicode_to_utf16le;
3269 ts = 2;
3270 } else {
3271 unparse = unicode_to_utf8;
3272 }
3273 }
3274
3275 if (sc->flag & SCONV_FROM_UTF16BE) {
3276 parse = utf16be_to_unicode;
3277 tm = 1;
3278 spair = 4;/* surrogate pair size in UTF-16. */
3279 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3280 parse = utf16le_to_unicode;
3281 tm = 1;
3282 spair = 4;/* surrogate pair size in UTF-16. */
3283 } else {
3284 parse = cesu8_to_unicode;
3285 tm = ts;
3286 spair = 6;/* surrogate pair size in UTF-8. */
3287 }
3288
3289 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3290 return (-1);
3291
3292 p = as->s + as->length;
3293 endp = as->s + as->buffer_length - ts;
3294 while ((n = parse(&uc, s, len)) != 0) {
3295 const char *ucptr;
3296 uint32_t cp1, cp2;
3297 int SIndex;
3298 struct {
3299 uint32_t uc;
3300 int ccc;
3301 } fdc[FDC_MAX];
3302 int fdi, fdj;
3303 int ccc;
3304
3305 check_first_code:
3306 if (n < 0) {
3307 /* Use a replaced unicode character. */
3308 UNPARSE(p, endp, uc);
3309 s += n*-1;
3310 len -= n*-1;
3311 ret = -1;
3312 continue;
3313 } else if (n == spair || always_replace)
3314 /* uc is converted from a surrogate pair.
3315 * this should be treated as a changed code. */
3316 ucptr = NULL;
3317 else
3318 ucptr = s;
3319 s += n;
3320 len -= n;
3321
3322 /* Hangul Decomposition. */
3323 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3324 int L = HC_LBASE + SIndex / HC_NCOUNT;
3325 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3326 int T = HC_TBASE + SIndex % HC_TCOUNT;
3327
3328 REPLACE_UC_WITH(L);
3329 WRITE_UC();
3330 REPLACE_UC_WITH(V);
3331 WRITE_UC();
3332 if (T != HC_TBASE) {
3333 REPLACE_UC_WITH(T);
3334 WRITE_UC();
3335 }
3336 continue;
3337 }
3338 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3339 WRITE_UC();
3340 continue;
3341 }
3342
3343 fdi = 0;
3344 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3345 int k;
3346
3347 for (k = fdi; k > 0; k--)
3348 fdc[k] = fdc[k-1];
3349 fdc[0].ccc = CCC(cp2);
3350 fdc[0].uc = cp2;
3351 fdi++;
3352 REPLACE_UC_WITH(cp1);
3353 }
3354
3355 /* Read following code points. */
3356 while ((n2 = parse(&uc2, s, len)) > 0 &&
3357 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3358 int j, k;
3359
3360 s += n2;
3361 len -= n2;
3362 for (j = 0; j < fdi; j++) {
3363 if (fdc[j].ccc > ccc)
3364 break;
3365 }
3366 if (j < fdi) {
3367 for (k = fdi; k > j; k--)
3368 fdc[k] = fdc[k-1];
3369 fdc[j].ccc = ccc;
3370 fdc[j].uc = uc2;
3371 } else {
3372 fdc[fdi].ccc = ccc;
3373 fdc[fdi].uc = uc2;
3374 }
3375 fdi++;
3376 }
3377
3378 WRITE_UC();
3379 for (fdj = 0; fdj < fdi; fdj++) {
3380 REPLACE_UC_WITH(fdc[fdj].uc);
3381 WRITE_UC();
3382 }
3383
3384 if (n2 == 0)
3385 break;
3386 REPLACE_UC_WITH(uc2);
3387 n = n2;
3388 goto check_first_code;
3389 }
3390 as->length = p - as->s;
3391 as->s[as->length] = '\0';
3392 if (ts == 2)
3393 as->s[as->length+1] = '\0';
3394 return (ret);
3395 }
3396
3397 /*
3398 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3399 * that WCS is Unicode. It is true for several platforms but some are false.
3400 * And then people who did not use UTF-8 locale on the non Unicode WCS
3401 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3402 * now cannot get right filename from libarchive 3.x and later since we
3403 * fixed the wrong assumption and it is incompatible to older its versions.
3404 * So we provide special option, "compat-2x.x", for resolving it.
3405 * That option enable the string conversion of libarchive 2.x.
3406 *
3407 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3408 * locale character set and appends to the archive_string.
3409 * Note: returns -1 if conversion fails.
3410 */
3411 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3412 strncat_from_utf8_libarchive2(struct archive_string *as,
3413 const void *_p, size_t len, struct archive_string_conv *sc)
3414 {
3415 const char *s;
3416 int n;
3417 char *p;
3418 char *end;
3419 uint32_t unicode;
3420 #if HAVE_WCRTOMB
3421 mbstate_t shift_state;
3422
3423 memset(&shift_state, 0, sizeof(shift_state));
3424 #else
3425 /* Clear the shift state before starting. */
3426 wctomb(NULL, L'\0');
3427 #endif
3428 (void)sc; /* UNUSED */
3429 /*
3430 * Allocate buffer for MBS.
3431 * We need this allocation here since it is possible that
3432 * as->s is still NULL.
3433 */
3434 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3435 return (-1);
3436
3437 s = (const char *)_p;
3438 p = as->s + as->length;
3439 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3440 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3441 wchar_t wc;
3442
3443 if (p >= end) {
3444 as->length = p - as->s;
3445 /* Re-allocate buffer for MBS. */
3446 if (archive_string_ensure(as,
3447 as->length + max(len * 2,
3448 (size_t)MB_CUR_MAX) + 1) == NULL)
3449 return (-1);
3450 p = as->s + as->length;
3451 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3452 }
3453
3454 /*
3455 * As libarchive 2.x, translates the UTF-8 characters into
3456 * wide-characters in the assumption that WCS is Unicode.
3457 */
3458 if (n < 0) {
3459 n *= -1;
3460 wc = L'?';
3461 } else
3462 wc = (wchar_t)unicode;
3463
3464 s += n;
3465 len -= n;
3466 /*
3467 * Translates the wide-character into the current locale MBS.
3468 */
3469 #if HAVE_WCRTOMB
3470 n = (int)wcrtomb(p, wc, &shift_state);
3471 #else
3472 n = (int)wctomb(p, wc);
3473 #endif
3474 if (n == -1)
3475 return (-1);
3476 p += n;
3477 }
3478 as->length = p - as->s;
3479 as->s[as->length] = '\0';
3480 return (0);
3481 }
3482
3483
3484 /*
3485 * Conversion functions between current locale dependent MBS and UTF-16BE.
3486 * strncat_from_utf16be() : UTF-16BE --> MBS
3487 * strncat_to_utf16be() : MBS --> UTF16BE
3488 */
3489
3490 #if defined(_WIN32) && !defined(__CYGWIN__)
3491
3492 /*
3493 * Convert a UTF-16BE/LE string to current locale and copy the result.
3494 * Return -1 if conversion fails.
3495 */
3496 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3497 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3498 struct archive_string_conv *sc, int be)
3499 {
3500 struct archive_string tmp;
3501 const char *u16;
3502 int ll;
3503 BOOL defchar;
3504 char *mbs;
3505 size_t mbs_size, b;
3506 int ret = 0;
3507
3508 bytes &= ~1;
3509 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3510 return (-1);
3511
3512 mbs = as->s + as->length;
3513 mbs_size = as->buffer_length - as->length -1;
3514
3515 if (sc->to_cp == CP_C_LOCALE) {
3516 /*
3517 * "C" locale special process.
3518 */
3519 u16 = _p;
3520 ll = 0;
3521 for (b = 0; b < bytes; b += 2) {
3522 uint16_t val;
3523 if (be)
3524 val = archive_be16dec(u16+b);
3525 else
3526 val = archive_le16dec(u16+b);
3527 if (val > 255) {
3528 *mbs++ = '?';
3529 ret = -1;
3530 } else
3531 *mbs++ = (char)(val&0xff);
3532 ll++;
3533 }
3534 as->length += ll;
3535 as->s[as->length] = '\0';
3536 return (ret);
3537 }
3538
3539 archive_string_init(&tmp);
3540 if (be) {
3541 if (is_big_endian()) {
3542 u16 = _p;
3543 } else {
3544 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3545 return (-1);
3546 memcpy(tmp.s, _p, bytes);
3547 for (b = 0; b < bytes; b += 2) {
3548 uint16_t val = archive_be16dec(tmp.s+b);
3549 archive_le16enc(tmp.s+b, val);
3550 }
3551 u16 = tmp.s;
3552 }
3553 } else {
3554 if (!is_big_endian()) {
3555 u16 = _p;
3556 } else {
3557 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3558 return (-1);
3559 memcpy(tmp.s, _p, bytes);
3560 for (b = 0; b < bytes; b += 2) {
3561 uint16_t val = archive_le16dec(tmp.s+b);
3562 archive_be16enc(tmp.s+b, val);
3563 }
3564 u16 = tmp.s;
3565 }
3566 }
3567
3568 do {
3569 defchar = 0;
3570 ll = WideCharToMultiByte(sc->to_cp, 0,
3571 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3572 NULL, &defchar);
3573 /* Exit loop if we succeeded */
3574 if (ll != 0 ||
3575 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3576 break;
3577 }
3578 /* Else expand buffer and loop to try again. */
3579 ll = WideCharToMultiByte(sc->to_cp, 0,
3580 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3581 if (archive_string_ensure(as, ll +1) == NULL)
3582 return (-1);
3583 mbs = as->s + as->length;
3584 mbs_size = as->buffer_length - as->length -1;
3585 } while (1);
3586 archive_string_free(&tmp);
3587 as->length += ll;
3588 as->s[as->length] = '\0';
3589 if (ll == 0 || defchar)
3590 ret = -1;
3591 return (ret);
3592 }
3593
3594 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3595 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3596 size_t bytes, struct archive_string_conv *sc)
3597 {
3598 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3599 }
3600
3601 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3602 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3603 size_t bytes, struct archive_string_conv *sc)
3604 {
3605 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3606 }
3607
3608 static int
is_big_endian(void)3609 is_big_endian(void)
3610 {
3611 uint16_t d = 1;
3612
3613 return (archive_be16dec(&d) == 1);
3614 }
3615
3616 /*
3617 * Convert a current locale string to UTF-16BE/LE and copy the result.
3618 * Return -1 if conversion fails.
3619 */
3620 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3621 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3622 size_t length, struct archive_string_conv *sc, int bigendian)
3623 {
3624 const char *s = (const char *)_p;
3625 char *u16;
3626 size_t count, avail;
3627
3628 if (archive_string_ensure(as16,
3629 as16->length + (length + 1) * 2) == NULL)
3630 return (-1);
3631
3632 u16 = as16->s + as16->length;
3633 avail = as16->buffer_length - 2;
3634 if (sc->from_cp == CP_C_LOCALE) {
3635 /*
3636 * "C" locale special process.
3637 */
3638 count = 0;
3639 while (count < length && *s) {
3640 if (bigendian)
3641 archive_be16enc(u16, *s);
3642 else
3643 archive_le16enc(u16, *s);
3644 u16 += 2;
3645 s++;
3646 count++;
3647 }
3648 as16->length += count << 1;
3649 as16->s[as16->length] = 0;
3650 as16->s[as16->length+1] = 0;
3651 return (0);
3652 }
3653 do {
3654 count = MultiByteToWideChar(sc->from_cp,
3655 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3656 /* Exit loop if we succeeded */
3657 if (count != 0 ||
3658 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3659 break;
3660 }
3661 /* Expand buffer and try again */
3662 count = MultiByteToWideChar(sc->from_cp,
3663 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3664 if (archive_string_ensure(as16, (count +1) * 2)
3665 == NULL)
3666 return (-1);
3667 u16 = as16->s + as16->length;
3668 avail = as16->buffer_length - 2;
3669 } while (1);
3670 as16->length += count * 2;
3671 as16->s[as16->length] = 0;
3672 as16->s[as16->length+1] = 0;
3673 if (count == 0)
3674 return (-1);
3675
3676 if (is_big_endian()) {
3677 if (!bigendian) {
3678 while (count > 0) {
3679 uint16_t v = archive_be16dec(u16);
3680 archive_le16enc(u16, v);
3681 u16 += 2;
3682 count--;
3683 }
3684 }
3685 } else {
3686 if (bigendian) {
3687 while (count > 0) {
3688 uint16_t v = archive_le16dec(u16);
3689 archive_be16enc(u16, v);
3690 u16 += 2;
3691 count--;
3692 }
3693 }
3694 }
3695 return (0);
3696 }
3697
3698 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3699 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3700 size_t length, struct archive_string_conv *sc)
3701 {
3702 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3703 }
3704
3705 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3706 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3707 size_t length, struct archive_string_conv *sc)
3708 {
3709 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3710 }
3711
3712 #endif /* _WIN32 && !__CYGWIN__ */
3713
3714 /*
3715 * Do the best effort for conversions.
3716 * We cannot handle UTF-16BE character-set without such iconv,
3717 * but there is a chance if a string consists just ASCII code or
3718 * a current locale is UTF-8.
3719 */
3720
3721 /*
3722 * Convert a UTF-16BE string to current locale and copy the result.
3723 * Return -1 if conversion fails.
3724 */
3725 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3726 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3727 size_t bytes, struct archive_string_conv *sc, int be)
3728 {
3729 const char *utf16 = (const char *)_p;
3730 char *mbs;
3731 uint32_t uc;
3732 int n, ret;
3733
3734 (void)sc; /* UNUSED */
3735 /*
3736 * Other case, we should do the best effort.
3737 * If all character are ASCII(<0x7f), we can convert it.
3738 * if not , we set a alternative character and return -1.
3739 */
3740 ret = 0;
3741 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3742 return (-1);
3743 mbs = as->s + as->length;
3744
3745 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3746 if (n < 0) {
3747 n *= -1;
3748 ret = -1;
3749 }
3750 bytes -= n;
3751 utf16 += n;
3752
3753 if (uc > 127) {
3754 /* We cannot handle it. */
3755 *mbs++ = '?';
3756 ret = -1;
3757 } else
3758 *mbs++ = (char)uc;
3759 }
3760 as->length = mbs - as->s;
3761 as->s[as->length] = '\0';
3762 return (ret);
3763 }
3764
3765 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3766 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3767 size_t bytes, struct archive_string_conv *sc)
3768 {
3769 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3770 }
3771
3772 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3773 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3774 size_t bytes, struct archive_string_conv *sc)
3775 {
3776 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3777 }
3778
3779 /*
3780 * Convert a current locale string to UTF-16BE/LE and copy the result.
3781 * Return -1 if conversion fails.
3782 */
3783 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3784 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3785 size_t length, struct archive_string_conv *sc, int bigendian)
3786 {
3787 const char *s = (const char *)_p;
3788 char *utf16;
3789 size_t remaining;
3790 int ret;
3791
3792 (void)sc; /* UNUSED */
3793 /*
3794 * Other case, we should do the best effort.
3795 * If all character are ASCII(<0x7f), we can convert it.
3796 * if not , we set a alternative character and return -1.
3797 */
3798 ret = 0;
3799 remaining = length;
3800
3801 if (archive_string_ensure(as16,
3802 as16->length + (length + 1) * 2) == NULL)
3803 return (-1);
3804
3805 utf16 = as16->s + as16->length;
3806 while (remaining--) {
3807 unsigned c = *s++;
3808 if (c > 127) {
3809 /* We cannot handle it. */
3810 c = UNICODE_R_CHAR;
3811 ret = -1;
3812 }
3813 if (bigendian)
3814 archive_be16enc(utf16, c);
3815 else
3816 archive_le16enc(utf16, c);
3817 utf16 += 2;
3818 }
3819 as16->length = utf16 - as16->s;
3820 as16->s[as16->length] = 0;
3821 as16->s[as16->length+1] = 0;
3822 return (ret);
3823 }
3824
3825 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3826 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3827 size_t length, struct archive_string_conv *sc)
3828 {
3829 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3830 }
3831
3832 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3833 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3834 size_t length, struct archive_string_conv *sc)
3835 {
3836 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3837 }
3838
3839
3840 /*
3841 * Multistring operations.
3842 */
3843
3844 void
archive_mstring_clean(struct archive_mstring * aes)3845 archive_mstring_clean(struct archive_mstring *aes)
3846 {
3847 archive_wstring_free(&(aes->aes_wcs));
3848 archive_string_free(&(aes->aes_mbs));
3849 archive_string_free(&(aes->aes_utf8));
3850 archive_string_free(&(aes->aes_mbs_in_locale));
3851 aes->aes_set = 0;
3852 }
3853
3854 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3855 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3856 {
3857 dest->aes_set = src->aes_set;
3858 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3859 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3860 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3861 }
3862
3863 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3864 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3865 const char **p)
3866 {
3867 struct archive_string_conv *sc;
3868 int r;
3869
3870 /* If we already have a UTF8 form, return that immediately. */
3871 if (aes->aes_set & AES_SET_UTF8) {
3872 *p = aes->aes_utf8.s;
3873 return (0);
3874 }
3875
3876 *p = NULL;
3877 #if defined(_WIN32) && !defined(__CYGWIN__)
3878 /*
3879 * On Windows, first try converting from WCS because (1) there's no
3880 * guarantee that the conversion to MBS will succeed, e.g. when using
3881 * CP_ACP, and (2) that's more efficient than converting to MBS, just to
3882 * convert back to WCS again before finally converting to UTF-8
3883 */
3884 if ((aes->aes_set & AES_SET_WCS) != 0) {
3885 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3886 if (sc == NULL)
3887 return (-1);/* Couldn't allocate memory for sc. */
3888 archive_string_empty(&(aes->aes_utf8));
3889 r = archive_string_append_from_wcs_in_codepage(&(aes->aes_utf8),
3890 aes->aes_wcs.s, aes->aes_wcs.length, sc);
3891 if (a == NULL)
3892 free_sconv_object(sc);
3893 if (r == 0) {
3894 aes->aes_set |= AES_SET_UTF8;
3895 *p = aes->aes_utf8.s;
3896 return (0);/* success. */
3897 } else
3898 return (-1);/* failure. */
3899 }
3900 #endif
3901 /* Try converting WCS to MBS first if MBS does not exist yet. */
3902 if ((aes->aes_set & AES_SET_MBS) == 0) {
3903 const char *pm; /* unused */
3904 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3905 }
3906 if (aes->aes_set & AES_SET_MBS) {
3907 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3908 if (sc == NULL)
3909 return (-1);/* Couldn't allocate memory for sc. */
3910 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3911 aes->aes_mbs.length, sc);
3912 if (a == NULL)
3913 free_sconv_object(sc);
3914 if (r == 0) {
3915 aes->aes_set |= AES_SET_UTF8;
3916 *p = aes->aes_utf8.s;
3917 return (0);/* success. */
3918 } else
3919 return (-1);/* failure. */
3920 }
3921 return (0);/* success. */
3922 }
3923
3924 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3925 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3926 const char **p)
3927 {
3928 struct archive_string_conv *sc;
3929 int r, ret = 0;
3930
3931 /* If we already have an MBS form, return that immediately. */
3932 if (aes->aes_set & AES_SET_MBS) {
3933 *p = aes->aes_mbs.s;
3934 return (ret);
3935 }
3936
3937 *p = NULL;
3938 /* If there's a WCS form, try converting with the native locale. */
3939 if (aes->aes_set & AES_SET_WCS) {
3940 archive_string_empty(&(aes->aes_mbs));
3941 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3942 aes->aes_wcs.s, aes->aes_wcs.length);
3943 *p = aes->aes_mbs.s;
3944 if (r == 0) {
3945 aes->aes_set |= AES_SET_MBS;
3946 return (ret);
3947 } else
3948 ret = -1;
3949 }
3950
3951 /* If there's a UTF-8 form, try converting with the native locale. */
3952 if (aes->aes_set & AES_SET_UTF8) {
3953 archive_string_empty(&(aes->aes_mbs));
3954 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
3955 if (sc == NULL)
3956 return (-1);/* Couldn't allocate memory for sc. */
3957 r = archive_strncpy_l(&(aes->aes_mbs),
3958 aes->aes_utf8.s, aes->aes_utf8.length, sc);
3959 if (a == NULL)
3960 free_sconv_object(sc);
3961 *p = aes->aes_mbs.s;
3962 if (r == 0) {
3963 aes->aes_set |= AES_SET_MBS;
3964 ret = 0;/* success; overwrite previous error. */
3965 } else
3966 ret = -1;/* failure. */
3967 }
3968 return (ret);
3969 }
3970
3971 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)3972 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3973 const wchar_t **wp)
3974 {
3975 int r, ret = 0;
3976
3977 (void)a;/* UNUSED */
3978 /* Return WCS form if we already have it. */
3979 if (aes->aes_set & AES_SET_WCS) {
3980 *wp = aes->aes_wcs.s;
3981 return (ret);
3982 }
3983
3984 *wp = NULL;
3985 #if defined(_WIN32) && !defined(__CYGWIN__)
3986 /*
3987 * On Windows, prefer converting from UTF-8 directly to WCS because:
3988 * (1) there's no guarantee that the string can be represented in MBS (e.g.
3989 * with CP_ACP), and (2) in order to convert from UTF-8 to MBS, we're going
3990 * to need to convert from UTF-8 to WCS anyway and its wasteful to throw
3991 * away that intermediate result
3992 */
3993 if (aes->aes_set & AES_SET_UTF8) {
3994 struct archive_string_conv *sc;
3995
3996 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
3997 if (sc != NULL) {
3998 archive_wstring_empty((&aes->aes_wcs));
3999 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4000 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4001 if (a == NULL)
4002 free_sconv_object(sc);
4003 if (r == 0) {
4004 aes->aes_set |= AES_SET_WCS;
4005 *wp = aes->aes_wcs.s;
4006 return (0);
4007 }
4008 }
4009 }
4010 #endif
4011 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
4012 if ((aes->aes_set & AES_SET_MBS) == 0) {
4013 const char *p; /* unused */
4014 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
4015 }
4016 /* Try converting MBS to WCS using native locale. */
4017 if (aes->aes_set & AES_SET_MBS) {
4018 archive_wstring_empty(&(aes->aes_wcs));
4019 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
4020 aes->aes_mbs.s, aes->aes_mbs.length);
4021 if (r == 0) {
4022 aes->aes_set |= AES_SET_WCS;
4023 *wp = aes->aes_wcs.s;
4024 } else
4025 ret = -1;/* failure. */
4026 }
4027 return (ret);
4028 }
4029
4030 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)4031 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
4032 const char **p, size_t *length, struct archive_string_conv *sc)
4033 {
4034 int ret = 0;
4035 #if defined(_WIN32) && !defined(__CYGWIN__)
4036 int r;
4037
4038 /*
4039 * Internationalization programming on Windows must use Wide
4040 * characters because Windows platform cannot make locale UTF-8.
4041 */
4042 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4043 archive_string_empty(&(aes->aes_mbs_in_locale));
4044 r = archive_string_append_from_wcs_in_codepage(
4045 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4046 aes->aes_wcs.length, sc);
4047 if (r == 0) {
4048 *p = aes->aes_mbs_in_locale.s;
4049 if (length != NULL)
4050 *length = aes->aes_mbs_in_locale.length;
4051 return (0);
4052 } else if (errno == ENOMEM)
4053 return (-1);
4054 else
4055 ret = -1;
4056 }
4057 #endif
4058
4059 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4060 * with the native locale to be used for translating it to specified
4061 * character-set. */
4062 if ((aes->aes_set & AES_SET_MBS) == 0) {
4063 const char *pm; /* unused */
4064 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
4065 }
4066 /* If we already have an MBS form, use it to be translated to
4067 * specified character-set. */
4068 if (aes->aes_set & AES_SET_MBS) {
4069 if (sc == NULL) {
4070 /* Conversion is unneeded. */
4071 *p = aes->aes_mbs.s;
4072 if (length != NULL)
4073 *length = aes->aes_mbs.length;
4074 return (0);
4075 }
4076 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4077 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4078 *p = aes->aes_mbs_in_locale.s;
4079 if (length != NULL)
4080 *length = aes->aes_mbs_in_locale.length;
4081 } else {
4082 *p = NULL;
4083 if (length != NULL)
4084 *length = 0;
4085 }
4086 return (ret);
4087 }
4088
4089 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4090 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4091 {
4092 if (mbs == NULL) {
4093 aes->aes_set = 0;
4094 return (0);
4095 }
4096 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4097 }
4098
4099 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4100 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4101 size_t len)
4102 {
4103 if (mbs == NULL) {
4104 aes->aes_set = 0;
4105 return (0);
4106 }
4107 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4108 archive_strncpy(&(aes->aes_mbs), mbs, len);
4109 archive_string_empty(&(aes->aes_utf8));
4110 archive_wstring_empty(&(aes->aes_wcs));
4111 return (0);
4112 }
4113
4114 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4115 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4116 {
4117 return archive_mstring_copy_wcs_len(aes, wcs,
4118 wcs == NULL ? 0 : wcslen(wcs));
4119 }
4120
4121 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4122 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4123 {
4124 if (utf8 == NULL) {
4125 aes->aes_set = 0;
4126 return (0);
4127 }
4128 aes->aes_set = AES_SET_UTF8;
4129 archive_string_empty(&(aes->aes_mbs));
4130 archive_string_empty(&(aes->aes_wcs));
4131 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4132 return (int)strlen(utf8);
4133 }
4134
4135 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4136 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4137 size_t len)
4138 {
4139 if (wcs == NULL) {
4140 aes->aes_set = 0;
4141 return (0);
4142 }
4143 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4144 archive_string_empty(&(aes->aes_mbs));
4145 archive_string_empty(&(aes->aes_utf8));
4146 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4147 return (0);
4148 }
4149
4150 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4151 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4152 const char *mbs, size_t len, struct archive_string_conv *sc)
4153 {
4154 int r;
4155
4156 if (mbs == NULL) {
4157 aes->aes_set = 0;
4158 return (0);
4159 }
4160 archive_string_empty(&(aes->aes_mbs));
4161 archive_wstring_empty(&(aes->aes_wcs));
4162 archive_string_empty(&(aes->aes_utf8));
4163 #if defined(_WIN32) && !defined(__CYGWIN__)
4164 /*
4165 * Internationalization programming on Windows must use Wide
4166 * characters because Windows platform cannot make locale UTF-8.
4167 */
4168 if (sc == NULL) {
4169 if (archive_string_append(&(aes->aes_mbs),
4170 mbs, mbsnbytes(mbs, len)) == NULL) {
4171 aes->aes_set = 0;
4172 r = -1;
4173 } else {
4174 aes->aes_set = AES_SET_MBS;
4175 r = 0;
4176 }
4177 #if defined(HAVE_ICONV)
4178 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4179 /*
4180 * This case happens only when MultiByteToWideChar() cannot
4181 * handle sc->from_cp, and we have to iconv in order to
4182 * translate character-set to wchar_t,UTF-16.
4183 */
4184 iconv_t cd = sc->cd;
4185 unsigned from_cp;
4186 int flag;
4187
4188 /*
4189 * Translate multi-bytes from some character-set to UTF-8.
4190 */
4191 sc->cd = sc->cd_w;
4192 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4193 sc->cd = cd;
4194 if (r != 0) {
4195 aes->aes_set = 0;
4196 return (r);
4197 }
4198 aes->aes_set = AES_SET_UTF8;
4199
4200 /*
4201 * Append the UTF-8 string into wstring.
4202 */
4203 flag = sc->flag;
4204 sc->flag &= ~(SCONV_NORMALIZATION_C
4205 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4206 from_cp = sc->from_cp;
4207 sc->from_cp = CP_UTF8;
4208 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4209 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4210 sc->flag = flag;
4211 sc->from_cp = from_cp;
4212 if (r == 0)
4213 aes->aes_set |= AES_SET_WCS;
4214 #endif
4215 } else {
4216 r = archive_wstring_append_from_mbs_in_codepage(
4217 &(aes->aes_wcs), mbs, len, sc);
4218 if (r == 0)
4219 aes->aes_set = AES_SET_WCS;
4220 else
4221 aes->aes_set = 0;
4222 }
4223 #else
4224 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4225 if (r == 0)
4226 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4227 else
4228 aes->aes_set = 0;
4229 #endif
4230 return (r);
4231 }
4232
4233 /*
4234 * The 'update' form tries to proactively update all forms of
4235 * this string (WCS and MBS) and returns an error if any of
4236 * them fail. This is used by the 'pax' handler, for instance,
4237 * to detect and report character-conversion failures early while
4238 * still allowing clients to get potentially useful values from
4239 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4240 * strive to give the user something useful, so you can get hopefully
4241 * usable values even if some of the character conversions are failing.)
4242 */
4243 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4244 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4245 const char *utf8)
4246 {
4247 struct archive_string_conv *sc;
4248 int r;
4249
4250 if (utf8 == NULL) {
4251 aes->aes_set = 0;
4252 return (0); /* Succeeded in clearing everything. */
4253 }
4254
4255 /* Save the UTF8 string. */
4256 archive_strcpy(&(aes->aes_utf8), utf8);
4257
4258 /* Empty the mbs and wcs strings. */
4259 archive_string_empty(&(aes->aes_mbs));
4260 archive_wstring_empty(&(aes->aes_wcs));
4261
4262 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4263
4264 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4265 if (sc == NULL)
4266 return (-1);/* Couldn't allocate memory for sc. */
4267
4268 #if defined(_WIN32) && !defined(__CYGWIN__)
4269 /* On Windows, there's no good way to convert from UTF8 -> MBS directly, so
4270 * prefer to first convert to WCS as (1) it's wasteful to throw away the
4271 * intermediate result, and (2) WCS will still be set even if we fail to
4272 * convert to MBS (e.g. with ACP that can't represent the characters) */
4273 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4274 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4275
4276 if (a == NULL)
4277 free_sconv_object(sc);
4278 if (r != 0)
4279 return (-1); /* This will guarantee we can't convert to MBS */
4280 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS; /* Both UTF8 and WCS set. */
4281
4282 /* Try converting WCS to MBS, return false on failure. */
4283 if (archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s,
4284 aes->aes_wcs.length))
4285 return (-1);
4286 #else
4287 /* Try converting UTF-8 to MBS, return false on failure. */
4288 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4289
4290 if (a == NULL)
4291 free_sconv_object(sc);
4292 if (r != 0)
4293 return (-1);
4294 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4295
4296 /* Try converting MBS to WCS, return false on failure. */
4297 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4298 aes->aes_mbs.length))
4299 return (-1);
4300 #endif
4301
4302 /* All conversions succeeded. */
4303 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4304
4305 return (0);
4306 }
4307