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 __FBSDID("$FreeBSD: head/lib/libarchive/archive_string.c 201095 2009-12-28 02:33:22Z kientzle $");
29
30 /*
31 * Basic resizable string support, to simplify manipulating arbitrary-sized
32 * strings while minimizing heap activity.
33 *
34 * In particular, the buffer used by a string object is only grown, it
35 * never shrinks, so you can clear and reuse the same string object
36 * without incurring additional memory allocations.
37 */
38
39 #ifdef HAVE_ERRNO_H
40 #include <errno.h>
41 #endif
42 #ifdef HAVE_ICONV_H
43 #include <iconv.h>
44 #endif
45 #ifdef HAVE_LANGINFO_H
46 #include <langinfo.h>
47 #endif
48 #ifdef HAVE_LOCALCHARSET_H
49 #include <localcharset.h>
50 #endif
51 #ifdef HAVE_STDLIB_H
52 #include <stdlib.h>
53 #endif
54 #ifdef HAVE_STRING_H
55 #include <string.h>
56 #endif
57 #ifdef HAVE_WCHAR_H
58 #include <wchar.h>
59 #endif
60 #if defined(_WIN32) && !defined(__CYGWIN__)
61 #include <windows.h>
62 #include <locale.h>
63 #endif
64
65 #include "archive_endian.h"
66 #include "archive_private.h"
67 #include "archive_string.h"
68 #include "archive_string_composition.h"
69
70 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
71 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
72 #endif
73
74 #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove)
75 #define wmemmove(a,b,i) (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t))
76 #endif
77
78 struct archive_string_conv {
79 struct archive_string_conv *next;
80 char *from_charset;
81 char *to_charset;
82 unsigned from_cp;
83 unsigned to_cp;
84 /* Set 1 if from_charset and to_charset are the same. */
85 int same;
86 int flag;
87 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified
88 * charset. */
89 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from
90 * specified charset. */
91 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */
92 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting
93 * MBS. */
94 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive
95 * 2.x in the wrong assumption. */
96 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C.
97 * Before UTF-8 characters are actually
98 * processed. */
99 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D.
100 * Before UTF-8 characters are actually
101 * processed.
102 * Currently this only for MAC OS X. */
103 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */
104 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */
105 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */
106 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */
107 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */
108 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */
109 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
110 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
111
112 #if HAVE_ICONV
113 iconv_t cd;
114 iconv_t cd_w;/* Use at archive_mstring on
115 * Windows. */
116 #endif
117 /* A temporary buffer for normalization. */
118 struct archive_string utftmp;
119 int (*converter[2])(struct archive_string *, const void *, size_t,
120 struct archive_string_conv *);
121 int nconverter;
122 };
123
124 #define CP_C_LOCALE 0 /* "C" locale only for this file. */
125 #define CP_UTF16LE 1200
126 #define CP_UTF16BE 1201
127
128 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
129 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
130 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
131 #define UNICODE_MAX 0x10FFFF
132 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */
133 /* Set U+FFFD(Replacement character) in UTF-8. */
134 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd};
135
136 static struct archive_string_conv *find_sconv_object(struct archive *,
137 const char *, const char *);
138 static void add_sconv_object(struct archive *, struct archive_string_conv *);
139 static struct archive_string_conv *create_sconv_object(const char *,
140 const char *, unsigned, int);
141 static void free_sconv_object(struct archive_string_conv *);
142 static struct archive_string_conv *get_sconv_object(struct archive *,
143 const char *, const char *, int);
144 static unsigned make_codepage_from_charset(const char *);
145 static unsigned get_current_codepage(void);
146 static unsigned get_current_oemcp(void);
147 static size_t mbsnbytes(const void *, size_t);
148 static size_t utf16nbytes(const void *, size_t);
149 #if defined(_WIN32) && !defined(__CYGWIN__)
150 static int archive_wstring_append_from_mbs_in_codepage(
151 struct archive_wstring *, const char *, size_t,
152 struct archive_string_conv *);
153 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
154 const wchar_t *, size_t, struct archive_string_conv *);
155 static int is_big_endian(void);
156 static int strncat_in_codepage(struct archive_string *, const void *,
157 size_t, struct archive_string_conv *);
158 static int win_strncat_from_utf16be(struct archive_string *, const void *,
159 size_t, struct archive_string_conv *);
160 static int win_strncat_from_utf16le(struct archive_string *, const void *,
161 size_t, struct archive_string_conv *);
162 static int win_strncat_to_utf16be(struct archive_string *, const void *,
163 size_t, struct archive_string_conv *);
164 static int win_strncat_to_utf16le(struct archive_string *, const void *,
165 size_t, struct archive_string_conv *);
166 #endif
167 static int best_effort_strncat_from_utf16be(struct archive_string *,
168 const void *, size_t, struct archive_string_conv *);
169 static int best_effort_strncat_from_utf16le(struct archive_string *,
170 const void *, size_t, struct archive_string_conv *);
171 static int best_effort_strncat_to_utf16be(struct archive_string *,
172 const void *, size_t, struct archive_string_conv *);
173 static int best_effort_strncat_to_utf16le(struct archive_string *,
174 const void *, size_t, struct archive_string_conv *);
175 #if defined(HAVE_ICONV)
176 static int iconv_strncat_in_locale(struct archive_string *, const void *,
177 size_t, struct archive_string_conv *);
178 #endif
179 static int best_effort_strncat_in_locale(struct archive_string *,
180 const void *, size_t, struct archive_string_conv *);
181 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
182 static int utf8_to_unicode(uint32_t *, const char *, size_t);
183 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
184 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
185 static size_t unicode_to_utf8(char *, size_t, uint32_t);
186 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
187 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
188 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
189 static int strncat_from_utf8_libarchive2(struct archive_string *,
190 const void *, size_t, struct archive_string_conv *);
191 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
192 size_t, struct archive_string_conv *);
193 static int archive_string_normalize_C(struct archive_string *, const void *,
194 size_t, struct archive_string_conv *);
195 static int archive_string_normalize_D(struct archive_string *, const void *,
196 size_t, struct archive_string_conv *);
197 static int archive_string_append_unicode(struct archive_string *,
198 const void *, size_t, struct archive_string_conv *);
199
200 static struct archive_string *
archive_string_append(struct archive_string * as,const char * p,size_t s)201 archive_string_append(struct archive_string *as, const char *p, size_t s)
202 {
203 if (archive_string_ensure(as, as->length + s + 1) == NULL)
204 return (NULL);
205 if (s)
206 memmove(as->s + as->length, p, s);
207 as->length += s;
208 as->s[as->length] = 0;
209 return (as);
210 }
211
212 static struct archive_wstring *
archive_wstring_append(struct archive_wstring * as,const wchar_t * p,size_t s)213 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
214 {
215 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
216 return (NULL);
217 if (s)
218 wmemmove(as->s + as->length, p, s);
219 as->length += s;
220 as->s[as->length] = 0;
221 return (as);
222 }
223
224 struct archive_string *
archive_array_append(struct archive_string * as,const char * p,size_t s)225 archive_array_append(struct archive_string *as, const char *p, size_t s)
226 {
227 return archive_string_append(as, p, s);
228 }
229
230 void
archive_string_concat(struct archive_string * dest,struct archive_string * src)231 archive_string_concat(struct archive_string *dest, struct archive_string *src)
232 {
233 if (archive_string_append(dest, src->s, src->length) == NULL)
234 __archive_errx(1, "Out of memory");
235 }
236
237 void
archive_wstring_concat(struct archive_wstring * dest,struct archive_wstring * src)238 archive_wstring_concat(struct archive_wstring *dest,
239 struct archive_wstring *src)
240 {
241 if (archive_wstring_append(dest, src->s, src->length) == NULL)
242 __archive_errx(1, "Out of memory");
243 }
244
245 void
archive_string_free(struct archive_string * as)246 archive_string_free(struct archive_string *as)
247 {
248 as->length = 0;
249 as->buffer_length = 0;
250 free(as->s);
251 as->s = NULL;
252 }
253
254 void
archive_wstring_free(struct archive_wstring * as)255 archive_wstring_free(struct archive_wstring *as)
256 {
257 as->length = 0;
258 as->buffer_length = 0;
259 free(as->s);
260 as->s = NULL;
261 }
262
263 struct archive_wstring *
archive_wstring_ensure(struct archive_wstring * as,size_t s)264 archive_wstring_ensure(struct archive_wstring *as, size_t s)
265 {
266 return (struct archive_wstring *)
267 archive_string_ensure((struct archive_string *)as,
268 s * sizeof(wchar_t));
269 }
270
271 /* Returns NULL on any allocation failure. */
272 struct archive_string *
archive_string_ensure(struct archive_string * as,size_t s)273 archive_string_ensure(struct archive_string *as, size_t s)
274 {
275 char *p;
276 size_t new_length;
277
278 /* If buffer is already big enough, don't reallocate. */
279 if (as->s && (s <= as->buffer_length))
280 return (as);
281
282 /*
283 * Growing the buffer at least exponentially ensures that
284 * append operations are always linear in the number of
285 * characters appended. Using a smaller growth rate for
286 * larger buffers reduces memory waste somewhat at the cost of
287 * a larger constant factor.
288 */
289 if (as->buffer_length < 32)
290 /* Start with a minimum 32-character buffer. */
291 new_length = 32;
292 else if (as->buffer_length < 8192)
293 /* Buffers under 8k are doubled for speed. */
294 new_length = as->buffer_length + as->buffer_length;
295 else {
296 /* Buffers 8k and over grow by at least 25% each time. */
297 new_length = as->buffer_length + as->buffer_length / 4;
298 /* Be safe: If size wraps, fail. */
299 if (new_length < as->buffer_length) {
300 /* On failure, wipe the string and return NULL. */
301 archive_string_free(as);
302 errno = ENOMEM;/* Make sure errno has ENOMEM. */
303 return (NULL);
304 }
305 }
306 /*
307 * The computation above is a lower limit to how much we'll
308 * grow the buffer. In any case, we have to grow it enough to
309 * hold the request.
310 */
311 if (new_length < s)
312 new_length = s;
313 /* Now we can reallocate the buffer. */
314 p = (char *)realloc(as->s, new_length);
315 if (p == NULL) {
316 /* On failure, wipe the string and return NULL. */
317 archive_string_free(as);
318 errno = ENOMEM;/* Make sure errno has ENOMEM. */
319 return (NULL);
320 }
321
322 as->s = p;
323 as->buffer_length = new_length;
324 return (as);
325 }
326
327 /*
328 * TODO: See if there's a way to avoid scanning
329 * the source string twice. Then test to see
330 * if it actually helps (remember that we're almost
331 * always called with pretty short arguments, so
332 * such an optimization might not help).
333 */
334 struct archive_string *
archive_strncat(struct archive_string * as,const void * _p,size_t n)335 archive_strncat(struct archive_string *as, const void *_p, size_t n)
336 {
337 size_t s;
338 const char *p, *pp;
339
340 p = (const char *)_p;
341
342 /* Like strlen(p), except won't examine positions beyond p[n]. */
343 s = 0;
344 pp = p;
345 while (s < n && *pp) {
346 pp++;
347 s++;
348 }
349 if ((as = archive_string_append(as, p, s)) == NULL)
350 __archive_errx(1, "Out of memory");
351 return (as);
352 }
353
354 struct archive_wstring *
archive_wstrncat(struct archive_wstring * as,const wchar_t * p,size_t n)355 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
356 {
357 size_t s;
358 const wchar_t *pp;
359
360 /* Like strlen(p), except won't examine positions beyond p[n]. */
361 s = 0;
362 pp = p;
363 while (s < n && *pp) {
364 pp++;
365 s++;
366 }
367 if ((as = archive_wstring_append(as, p, s)) == NULL)
368 __archive_errx(1, "Out of memory");
369 return (as);
370 }
371
372 struct archive_string *
archive_strcat(struct archive_string * as,const void * p)373 archive_strcat(struct archive_string *as, const void *p)
374 {
375 /* strcat is just strncat without an effective limit.
376 * Assert that we'll never get called with a source
377 * string over 16MB.
378 * TODO: Review all uses of strcat in the source
379 * and try to replace them with strncat().
380 */
381 return archive_strncat(as, p, 0x1000000);
382 }
383
384 struct archive_wstring *
archive_wstrcat(struct archive_wstring * as,const wchar_t * p)385 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
386 {
387 /* Ditto. */
388 return archive_wstrncat(as, p, 0x1000000);
389 }
390
391 struct archive_string *
archive_strappend_char(struct archive_string * as,char c)392 archive_strappend_char(struct archive_string *as, char c)
393 {
394 if ((as = archive_string_append(as, &c, 1)) == NULL)
395 __archive_errx(1, "Out of memory");
396 return (as);
397 }
398
399 struct archive_wstring *
archive_wstrappend_wchar(struct archive_wstring * as,wchar_t c)400 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
401 {
402 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
403 __archive_errx(1, "Out of memory");
404 return (as);
405 }
406
407 /*
408 * Get the "current character set" name to use with iconv.
409 * On FreeBSD, the empty character set name "" chooses
410 * the correct character encoding for the current locale,
411 * so this isn't necessary.
412 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
413 * on that system, we have to explicitly call nl_langinfo()
414 * to get the right name. Not sure about other platforms.
415 *
416 * NOTE: GNU libiconv does not recognize the character-set name
417 * which some platform nl_langinfo(CODESET) returns, so we should
418 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
419 */
420 static const char *
default_iconv_charset(const char * charset)421 default_iconv_charset(const char *charset) {
422 if (charset != NULL && charset[0] != '\0')
423 return charset;
424 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
425 /* locale_charset() is broken on Mac OS */
426 return locale_charset();
427 #elif HAVE_NL_LANGINFO
428 return nl_langinfo(CODESET);
429 #else
430 return "";
431 #endif
432 }
433
434 #if defined(_WIN32) && !defined(__CYGWIN__)
435
436 /*
437 * Convert MBS to WCS.
438 * Note: returns -1 if conversion fails.
439 */
440 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)441 archive_wstring_append_from_mbs(struct archive_wstring *dest,
442 const char *p, size_t len)
443 {
444 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
445 }
446
447 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)448 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
449 const char *s, size_t length, struct archive_string_conv *sc)
450 {
451 int count, ret = 0;
452 UINT from_cp;
453
454 if (sc != NULL)
455 from_cp = sc->from_cp;
456 else
457 from_cp = get_current_codepage();
458
459 if (from_cp == CP_C_LOCALE) {
460 /*
461 * "C" locale special process.
462 */
463 wchar_t *ws;
464 const unsigned char *mp;
465
466 if (NULL == archive_wstring_ensure(dest,
467 dest->length + length + 1))
468 return (-1);
469
470 ws = dest->s + dest->length;
471 mp = (const unsigned char *)s;
472 count = 0;
473 while (count < (int)length && *mp) {
474 *ws++ = (wchar_t)*mp++;
475 count++;
476 }
477 } else if (sc != NULL &&
478 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
479 /*
480 * Normalize UTF-8 and UTF-16BE and convert it directly
481 * to UTF-16 as wchar_t.
482 */
483 struct archive_string u16;
484 int saved_flag = sc->flag;/* save current flag. */
485
486 if (is_big_endian())
487 sc->flag |= SCONV_TO_UTF16BE;
488 else
489 sc->flag |= SCONV_TO_UTF16LE;
490
491 if (sc->flag & SCONV_FROM_UTF16) {
492 /*
493 * UTF-16BE/LE NFD ===> UTF-16 NFC
494 * UTF-16BE/LE NFC ===> UTF-16 NFD
495 */
496 count = (int)utf16nbytes(s, length);
497 } else {
498 /*
499 * UTF-8 NFD ===> UTF-16 NFC
500 * UTF-8 NFC ===> UTF-16 NFD
501 */
502 count = (int)mbsnbytes(s, length);
503 }
504 u16.s = (char *)dest->s;
505 u16.length = dest->length << 1;;
506 u16.buffer_length = dest->buffer_length;
507 if (sc->flag & SCONV_NORMALIZATION_C)
508 ret = archive_string_normalize_C(&u16, s, count, sc);
509 else
510 ret = archive_string_normalize_D(&u16, s, count, sc);
511 dest->s = (wchar_t *)u16.s;
512 dest->length = u16.length >> 1;
513 dest->buffer_length = u16.buffer_length;
514 sc->flag = saved_flag;/* restore the saved flag. */
515 return (ret);
516 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
517 count = (int)utf16nbytes(s, length);
518 count >>= 1; /* to be WCS length */
519 /* Allocate memory for WCS. */
520 if (NULL == archive_wstring_ensure(dest,
521 dest->length + count + 1))
522 return (-1);
523 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
524 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
525 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
526 int b;
527 for (b = 0; b < count; b++) {
528 uint16_t val = archive_le16dec(u16+b);
529 archive_be16enc(u16+b, val);
530 }
531 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
532 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
533 int b;
534 for (b = 0; b < count; b++) {
535 uint16_t val = archive_be16dec(u16+b);
536 archive_le16enc(u16+b, val);
537 }
538 }
539 } else {
540 DWORD mbflag;
541 size_t buffsize;
542
543 if (sc == NULL)
544 mbflag = 0;
545 else if (sc->flag & SCONV_FROM_CHARSET) {
546 /* Do not trust the length which comes from
547 * an archive file. */
548 length = mbsnbytes(s, length);
549 mbflag = 0;
550 } else
551 mbflag = MB_PRECOMPOSED;
552
553 buffsize = dest->length + length + 1;
554 do {
555 /* Allocate memory for WCS. */
556 if (NULL == archive_wstring_ensure(dest, buffsize))
557 return (-1);
558 /* Convert MBS to WCS. */
559 count = MultiByteToWideChar(from_cp,
560 mbflag, s, (int)length, dest->s + dest->length,
561 (int)(dest->buffer_length >> 1) -1);
562 if (count == 0 &&
563 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
564 /* Expand the WCS buffer. */
565 buffsize = dest->buffer_length << 1;
566 continue;
567 }
568 if (count == 0 && length != 0)
569 ret = -1;
570 break;
571 } while (1);
572 }
573 dest->length += count;
574 dest->s[dest->length] = L'\0';
575 return (ret);
576 }
577
578 #else
579
580 /*
581 * Convert MBS to WCS.
582 * Note: returns -1 if conversion fails.
583 */
584 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)585 archive_wstring_append_from_mbs(struct archive_wstring *dest,
586 const char *p, size_t len)
587 {
588 size_t r;
589 int ret_val = 0;
590 /*
591 * No single byte will be more than one wide character,
592 * so this length estimate will always be big enough.
593 */
594 size_t wcs_length = len;
595 size_t mbs_length = len;
596 const char *mbs = p;
597 wchar_t *wcs;
598 #if HAVE_MBRTOWC
599 mbstate_t shift_state;
600
601 memset(&shift_state, 0, sizeof(shift_state));
602 #endif
603 if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1))
604 return (-1);
605 wcs = dest->s + dest->length;
606 /*
607 * We cannot use mbsrtowcs/mbstowcs here because those may convert
608 * extra MBS when strlen(p) > len and one wide character consists of
609 * multi bytes.
610 */
611 while (*mbs && mbs_length > 0) {
612 if (wcs_length == 0) {
613 dest->length = wcs - dest->s;
614 dest->s[dest->length] = L'\0';
615 wcs_length = mbs_length;
616 if (NULL == archive_wstring_ensure(dest,
617 dest->length + wcs_length + 1))
618 return (-1);
619 wcs = dest->s + dest->length;
620 }
621 #if HAVE_MBRTOWC
622 r = mbrtowc(wcs, mbs, wcs_length, &shift_state);
623 #else
624 r = mbtowc(wcs, mbs, wcs_length);
625 #endif
626 if (r == (size_t)-1 || r == (size_t)-2) {
627 ret_val = -1;
628 if (errno == EILSEQ) {
629 ++mbs;
630 --mbs_length;
631 continue;
632 } else
633 break;
634 }
635 if (r == 0 || r > mbs_length)
636 break;
637 wcs++;
638 wcs_length--;
639 mbs += r;
640 mbs_length -= r;
641 }
642 dest->length = wcs - dest->s;
643 dest->s[dest->length] = L'\0';
644 return (ret_val);
645 }
646
647 #endif
648
649 #if defined(_WIN32) && !defined(__CYGWIN__)
650
651 /*
652 * WCS ==> MBS.
653 * Note: returns -1 if conversion fails.
654 *
655 * Win32 builds use WideCharToMultiByte from the Windows API.
656 * (Maybe Cygwin should too? WideCharToMultiByte will know a
657 * lot more about local character encodings than the wcrtomb()
658 * wrapper is going to know.)
659 */
660 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)661 archive_string_append_from_wcs(struct archive_string *as,
662 const wchar_t *w, size_t len)
663 {
664 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
665 }
666
667 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)668 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
669 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
670 {
671 BOOL defchar_used, *dp;
672 int count, ret = 0;
673 UINT to_cp;
674 int wslen = (int)len;
675
676 if (sc != NULL)
677 to_cp = sc->to_cp;
678 else
679 to_cp = get_current_codepage();
680
681 if (to_cp == CP_C_LOCALE) {
682 /*
683 * "C" locale special process.
684 */
685 const wchar_t *wp = ws;
686 char *p;
687
688 if (NULL == archive_string_ensure(as,
689 as->length + wslen +1))
690 return (-1);
691 p = as->s + as->length;
692 count = 0;
693 defchar_used = 0;
694 while (count < wslen && *wp) {
695 if (*wp > 255) {
696 *p++ = '?';
697 wp++;
698 defchar_used = 1;
699 } else
700 *p++ = (char)*wp++;
701 count++;
702 }
703 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
704 uint16_t *u16;
705
706 if (NULL ==
707 archive_string_ensure(as, as->length + len * 2 + 2))
708 return (-1);
709 u16 = (uint16_t *)(as->s + as->length);
710 count = 0;
711 defchar_used = 0;
712 if (sc->flag & SCONV_TO_UTF16BE) {
713 while (count < (int)len && *ws) {
714 archive_be16enc(u16+count, *ws);
715 ws++;
716 count++;
717 }
718 } else {
719 while (count < (int)len && *ws) {
720 archive_le16enc(u16+count, *ws);
721 ws++;
722 count++;
723 }
724 }
725 count <<= 1; /* to be byte size */
726 } else {
727 /* Make sure the MBS buffer has plenty to set. */
728 if (NULL ==
729 archive_string_ensure(as, as->length + len * 2 + 1))
730 return (-1);
731 do {
732 defchar_used = 0;
733 if (to_cp == CP_UTF8 || sc == NULL)
734 dp = NULL;
735 else
736 dp = &defchar_used;
737 count = WideCharToMultiByte(to_cp, 0, ws, wslen,
738 as->s + as->length, (int)as->buffer_length-1, NULL, dp);
739 if (count == 0 &&
740 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
741 /* Expand the MBS buffer and retry. */
742 if (NULL == archive_string_ensure(as,
743 as->buffer_length + len))
744 return (-1);
745 continue;
746 }
747 if (count == 0)
748 ret = -1;
749 break;
750 } while (1);
751 }
752 as->length += count;
753 as->s[as->length] = '\0';
754 return (defchar_used?-1:ret);
755 }
756
757 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
758
759 /*
760 * Translates a wide character string into current locale character set
761 * and appends to the archive_string. Note: returns -1 if conversion
762 * fails.
763 */
764 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)765 archive_string_append_from_wcs(struct archive_string *as,
766 const wchar_t *w, size_t len)
767 {
768 /* We cannot use the standard wcstombs() here because it
769 * cannot tell us how big the output buffer should be. So
770 * I've built a loop around wcrtomb() or wctomb() that
771 * converts a character at a time and resizes the string as
772 * needed. We prefer wcrtomb() when it's available because
773 * it's thread-safe. */
774 int n, ret_val = 0;
775 char *p;
776 char *end;
777 #if HAVE_WCRTOMB
778 mbstate_t shift_state;
779
780 memset(&shift_state, 0, sizeof(shift_state));
781 #else
782 /* Clear the shift state before starting. */
783 wctomb(NULL, L'\0');
784 #endif
785 /*
786 * Allocate buffer for MBS.
787 * We need this allocation here since it is possible that
788 * as->s is still NULL.
789 */
790 if (archive_string_ensure(as, as->length + len + 1) == NULL)
791 return (-1);
792
793 p = as->s + as->length;
794 end = as->s + as->buffer_length - MB_CUR_MAX -1;
795 while (*w != L'\0' && len > 0) {
796 if (p >= end) {
797 as->length = p - as->s;
798 as->s[as->length] = '\0';
799 /* Re-allocate buffer for MBS. */
800 if (archive_string_ensure(as,
801 as->length + len * 2 + 1) == NULL)
802 return (-1);
803 p = as->s + as->length;
804 end = as->s + as->buffer_length - MB_CUR_MAX -1;
805 }
806 #if HAVE_WCRTOMB
807 n = wcrtomb(p, *w++, &shift_state);
808 #else
809 n = wctomb(p, *w++);
810 #endif
811 if (n == -1) {
812 if (errno == EILSEQ) {
813 /* Skip an illegal wide char. */
814 *p++ = '?';
815 ret_val = -1;
816 } else {
817 ret_val = -1;
818 break;
819 }
820 } else
821 p += n;
822 len--;
823 }
824 as->length = p - as->s;
825 as->s[as->length] = '\0';
826 return (ret_val);
827 }
828
829 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
830
831 /*
832 * TODO: Test if __STDC_ISO_10646__ is defined.
833 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
834 * one character at a time. If a non-Windows platform doesn't have
835 * either of these, fall back to the built-in UTF8 conversion.
836 */
837 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)838 archive_string_append_from_wcs(struct archive_string *as,
839 const wchar_t *w, size_t len)
840 {
841 (void)as;/* UNUSED */
842 (void)w;/* UNUSED */
843 (void)len;/* UNUSED */
844 errno = ENOSYS;
845 return (-1);
846 }
847
848 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
849
850 /*
851 * Find a string conversion object by a pair of 'from' charset name
852 * and 'to' charset name from an archive object.
853 * Return NULL if not found.
854 */
855 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)856 find_sconv_object(struct archive *a, const char *fc, const char *tc)
857 {
858 struct archive_string_conv *sc;
859
860 if (a == NULL)
861 return (NULL);
862
863 for (sc = a->sconv; sc != NULL; sc = sc->next) {
864 if (strcmp(sc->from_charset, fc) == 0 &&
865 strcmp(sc->to_charset, tc) == 0)
866 break;
867 }
868 return (sc);
869 }
870
871 /*
872 * Register a string object to an archive object.
873 */
874 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)875 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
876 {
877 struct archive_string_conv **psc;
878
879 /* Add a new sconv to sconv list. */
880 psc = &(a->sconv);
881 while (*psc != NULL)
882 psc = &((*psc)->next);
883 *psc = sc;
884 }
885
886 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))887 add_converter(struct archive_string_conv *sc, int (*converter)
888 (struct archive_string *, const void *, size_t,
889 struct archive_string_conv *))
890 {
891 if (sc == NULL || sc->nconverter >= 2)
892 __archive_errx(1, "Programing error");
893 sc->converter[sc->nconverter++] = converter;
894 }
895
896 static void
setup_converter(struct archive_string_conv * sc)897 setup_converter(struct archive_string_conv *sc)
898 {
899
900 /* Reset. */
901 sc->nconverter = 0;
902
903 /*
904 * Perform special sequence for the incorrect UTF-8 filenames
905 * made by libarchive2.x.
906 */
907 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
908 add_converter(sc, strncat_from_utf8_libarchive2);
909 return;
910 }
911
912 /*
913 * Convert a string to UTF-16BE/LE.
914 */
915 if (sc->flag & SCONV_TO_UTF16) {
916 /*
917 * If the current locale is UTF-8, we can translate
918 * a UTF-8 string into a UTF-16BE string.
919 */
920 if (sc->flag & SCONV_FROM_UTF8) {
921 add_converter(sc, archive_string_append_unicode);
922 return;
923 }
924
925 #if defined(_WIN32) && !defined(__CYGWIN__)
926 if (sc->flag & SCONV_WIN_CP) {
927 if (sc->flag & SCONV_TO_UTF16BE)
928 add_converter(sc, win_strncat_to_utf16be);
929 else
930 add_converter(sc, win_strncat_to_utf16le);
931 return;
932 }
933 #endif
934
935 #if defined(HAVE_ICONV)
936 if (sc->cd != (iconv_t)-1) {
937 add_converter(sc, iconv_strncat_in_locale);
938 return;
939 }
940 #endif
941
942 if (sc->flag & SCONV_BEST_EFFORT) {
943 if (sc->flag & SCONV_TO_UTF16BE)
944 add_converter(sc,
945 best_effort_strncat_to_utf16be);
946 else
947 add_converter(sc,
948 best_effort_strncat_to_utf16le);
949 } else
950 /* Make sure we have no converter. */
951 sc->nconverter = 0;
952 return;
953 }
954
955 /*
956 * Convert a string from UTF-16BE/LE.
957 */
958 if (sc->flag & SCONV_FROM_UTF16) {
959 /*
960 * At least we should normalize a UTF-16BE string.
961 */
962 if (sc->flag & SCONV_NORMALIZATION_D)
963 add_converter(sc,archive_string_normalize_D);
964 else if (sc->flag & SCONV_NORMALIZATION_C)
965 add_converter(sc, archive_string_normalize_C);
966
967 if (sc->flag & SCONV_TO_UTF8) {
968 /*
969 * If the current locale is UTF-8, we can translate
970 * a UTF-16BE/LE string into a UTF-8 string directly.
971 */
972 if (!(sc->flag &
973 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
974 add_converter(sc,
975 archive_string_append_unicode);
976 return;
977 }
978
979 #if defined(_WIN32) && !defined(__CYGWIN__)
980 if (sc->flag & SCONV_WIN_CP) {
981 if (sc->flag & SCONV_FROM_UTF16BE)
982 add_converter(sc, win_strncat_from_utf16be);
983 else
984 add_converter(sc, win_strncat_from_utf16le);
985 return;
986 }
987 #endif
988
989 #if defined(HAVE_ICONV)
990 if (sc->cd != (iconv_t)-1) {
991 add_converter(sc, iconv_strncat_in_locale);
992 return;
993 }
994 #endif
995
996 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
997 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
998 add_converter(sc, best_effort_strncat_from_utf16be);
999 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1000 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1001 add_converter(sc, best_effort_strncat_from_utf16le);
1002 else
1003 /* Make sure we have no converter. */
1004 sc->nconverter = 0;
1005 return;
1006 }
1007
1008 if (sc->flag & SCONV_FROM_UTF8) {
1009 /*
1010 * At least we should normalize a UTF-8 string.
1011 */
1012 if (sc->flag & SCONV_NORMALIZATION_D)
1013 add_converter(sc,archive_string_normalize_D);
1014 else if (sc->flag & SCONV_NORMALIZATION_C)
1015 add_converter(sc, archive_string_normalize_C);
1016
1017 /*
1018 * Copy UTF-8 string with a check of CESU-8.
1019 * Apparently, iconv does not check surrogate pairs in UTF-8
1020 * when both from-charset and to-charset are UTF-8, and then
1021 * we use our UTF-8 copy code.
1022 */
1023 if (sc->flag & SCONV_TO_UTF8) {
1024 /*
1025 * If the current locale is UTF-8, we can translate
1026 * a UTF-16BE string into a UTF-8 string directly.
1027 */
1028 if (!(sc->flag &
1029 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1030 add_converter(sc, strncat_from_utf8_to_utf8);
1031 return;
1032 }
1033 }
1034
1035 #if defined(_WIN32) && !defined(__CYGWIN__)
1036 /*
1037 * On Windows we can use Windows API for a string conversion.
1038 */
1039 if (sc->flag & SCONV_WIN_CP) {
1040 add_converter(sc, strncat_in_codepage);
1041 return;
1042 }
1043 #endif
1044
1045 #if HAVE_ICONV
1046 if (sc->cd != (iconv_t)-1) {
1047 add_converter(sc, iconv_strncat_in_locale);
1048 /*
1049 * iconv generally does not support UTF-8-MAC and so
1050 * we have to the output of iconv from NFC to NFD if
1051 * need.
1052 */
1053 if ((sc->flag & SCONV_FROM_CHARSET) &&
1054 (sc->flag & SCONV_TO_UTF8)) {
1055 if (sc->flag & SCONV_NORMALIZATION_D)
1056 add_converter(sc, archive_string_normalize_D);
1057 }
1058 return;
1059 }
1060 #endif
1061
1062 /*
1063 * Try conversion in the best effort or no conversion.
1064 */
1065 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1066 add_converter(sc, best_effort_strncat_in_locale);
1067 else
1068 /* Make sure we have no converter. */
1069 sc->nconverter = 0;
1070 }
1071
1072 /*
1073 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1074 * and CP932 which are referenced in create_sconv_object().
1075 */
1076 static const char *
canonical_charset_name(const char * charset)1077 canonical_charset_name(const char *charset)
1078 {
1079 char cs[16];
1080 char *p;
1081 const char *s;
1082
1083 if (charset == NULL || charset[0] == '\0'
1084 || strlen(charset) > 15)
1085 return (charset);
1086
1087 /* Copy name to uppercase. */
1088 p = cs;
1089 s = charset;
1090 while (*s) {
1091 char c = *s++;
1092 if (c >= 'a' && c <= 'z')
1093 c -= 'a' - 'A';
1094 *p++ = c;
1095 }
1096 *p++ = '\0';
1097
1098 if (strcmp(cs, "UTF-8") == 0 ||
1099 strcmp(cs, "UTF8") == 0)
1100 return ("UTF-8");
1101 if (strcmp(cs, "UTF-16BE") == 0 ||
1102 strcmp(cs, "UTF16BE") == 0)
1103 return ("UTF-16BE");
1104 if (strcmp(cs, "UTF-16LE") == 0 ||
1105 strcmp(cs, "UTF16LE") == 0)
1106 return ("UTF-16LE");
1107 if (strcmp(cs, "CP932") == 0)
1108 return ("CP932");
1109 return (charset);
1110 }
1111
1112 /*
1113 * Create a string conversion object.
1114 */
1115 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1116 create_sconv_object(const char *fc, const char *tc,
1117 unsigned current_codepage, int flag)
1118 {
1119 struct archive_string_conv *sc;
1120
1121 sc = calloc(1, sizeof(*sc));
1122 if (sc == NULL)
1123 return (NULL);
1124 sc->next = NULL;
1125 sc->from_charset = strdup(fc);
1126 if (sc->from_charset == NULL) {
1127 free(sc);
1128 return (NULL);
1129 }
1130 sc->to_charset = strdup(tc);
1131 if (sc->to_charset == NULL) {
1132 free(sc->from_charset);
1133 free(sc);
1134 return (NULL);
1135 }
1136 archive_string_init(&sc->utftmp);
1137
1138 if (flag & SCONV_TO_CHARSET) {
1139 /*
1140 * Convert characters from the current locale charset to
1141 * a specified charset.
1142 */
1143 sc->from_cp = current_codepage;
1144 sc->to_cp = make_codepage_from_charset(tc);
1145 #if defined(_WIN32) && !defined(__CYGWIN__)
1146 if (IsValidCodePage(sc->to_cp))
1147 flag |= SCONV_WIN_CP;
1148 #endif
1149 } else if (flag & SCONV_FROM_CHARSET) {
1150 /*
1151 * Convert characters from a specified charset to
1152 * the current locale charset.
1153 */
1154 sc->to_cp = current_codepage;
1155 sc->from_cp = make_codepage_from_charset(fc);
1156 #if defined(_WIN32) && !defined(__CYGWIN__)
1157 if (IsValidCodePage(sc->from_cp))
1158 flag |= SCONV_WIN_CP;
1159 #endif
1160 }
1161
1162 /*
1163 * Check if "from charset" and "to charset" are the same.
1164 */
1165 if (strcmp(fc, tc) == 0 ||
1166 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1167 sc->same = 1;
1168 else
1169 sc->same = 0;
1170
1171 /*
1172 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1173 */
1174 if (strcmp(tc, "UTF-8") == 0)
1175 flag |= SCONV_TO_UTF8;
1176 else if (strcmp(tc, "UTF-16BE") == 0)
1177 flag |= SCONV_TO_UTF16BE;
1178 else if (strcmp(tc, "UTF-16LE") == 0)
1179 flag |= SCONV_TO_UTF16LE;
1180 if (strcmp(fc, "UTF-8") == 0)
1181 flag |= SCONV_FROM_UTF8;
1182 else if (strcmp(fc, "UTF-16BE") == 0)
1183 flag |= SCONV_FROM_UTF16BE;
1184 else if (strcmp(fc, "UTF-16LE") == 0)
1185 flag |= SCONV_FROM_UTF16LE;
1186 #if defined(_WIN32) && !defined(__CYGWIN__)
1187 if (sc->to_cp == CP_UTF8)
1188 flag |= SCONV_TO_UTF8;
1189 else if (sc->to_cp == CP_UTF16BE)
1190 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1191 else if (sc->to_cp == CP_UTF16LE)
1192 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1193 if (sc->from_cp == CP_UTF8)
1194 flag |= SCONV_FROM_UTF8;
1195 else if (sc->from_cp == CP_UTF16BE)
1196 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1197 else if (sc->from_cp == CP_UTF16LE)
1198 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1199 #endif
1200
1201 /*
1202 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1203 * handle it. So we have to translate NFD characters to NFC ones
1204 * ourselves before iconv handles. Another reason is to prevent
1205 * that the same sight of two filenames, one is NFC and other
1206 * is NFD, would be in its directory.
1207 * On Mac OS X, although its filesystem layer automatically
1208 * convert filenames to NFD, it would be useful for filename
1209 * comparing to find out the same filenames that we normalize
1210 * that to be NFD ourselves.
1211 */
1212 if ((flag & SCONV_FROM_CHARSET) &&
1213 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1214 #if defined(__APPLE__)
1215 if (flag & SCONV_TO_UTF8)
1216 flag |= SCONV_NORMALIZATION_D;
1217 else
1218 #endif
1219 flag |= SCONV_NORMALIZATION_C;
1220 }
1221 #if defined(__APPLE__)
1222 /*
1223 * In case writing an archive file, make sure that a filename
1224 * going to be passed to iconv is a Unicode NFC string since
1225 * a filename in HFS Plus filesystem is a Unicode NFD one and
1226 * iconv cannot handle it with "UTF-8" charset. It is simpler
1227 * than a use of "UTF-8-MAC" charset.
1228 */
1229 if ((flag & SCONV_TO_CHARSET) &&
1230 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1231 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1232 flag |= SCONV_NORMALIZATION_C;
1233 /*
1234 * In case reading an archive file. make sure that a filename
1235 * will be passed to users is a Unicode NFD string in order to
1236 * correctly compare the filename with other one which comes
1237 * from HFS Plus filesystem.
1238 */
1239 if ((flag & SCONV_FROM_CHARSET) &&
1240 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1241 (flag & SCONV_TO_UTF8))
1242 flag |= SCONV_NORMALIZATION_D;
1243 #endif
1244
1245 #if defined(HAVE_ICONV)
1246 sc->cd_w = (iconv_t)-1;
1247 /*
1248 * Create an iconv object.
1249 */
1250 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1251 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1252 (flag & SCONV_WIN_CP)) {
1253 /* This case we won't use iconv. */
1254 sc->cd = (iconv_t)-1;
1255 } else {
1256 sc->cd = iconv_open(tc, fc);
1257 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1258 /*
1259 * Unfortunately, all of iconv implements do support
1260 * "CP932" character-set, so we should use "SJIS"
1261 * instead if iconv_open failed.
1262 */
1263 if (strcmp(tc, "CP932") == 0)
1264 sc->cd = iconv_open("SJIS", fc);
1265 else if (strcmp(fc, "CP932") == 0)
1266 sc->cd = iconv_open(tc, "SJIS");
1267 }
1268 #if defined(_WIN32) && !defined(__CYGWIN__)
1269 /*
1270 * archive_mstring on Windows directly convert multi-bytes
1271 * into archive_wstring in order not to depend on locale
1272 * so that you can do a I18N programming. This will be
1273 * used only in archive_mstring_copy_mbs_len_l so far.
1274 */
1275 if (flag & SCONV_FROM_CHARSET) {
1276 sc->cd_w = iconv_open("UTF-8", fc);
1277 if (sc->cd_w == (iconv_t)-1 &&
1278 (sc->flag & SCONV_BEST_EFFORT)) {
1279 if (strcmp(fc, "CP932") == 0)
1280 sc->cd_w = iconv_open("UTF-8", "SJIS");
1281 }
1282 }
1283 #endif /* _WIN32 && !__CYGWIN__ */
1284 }
1285 #endif /* HAVE_ICONV */
1286
1287 sc->flag = flag;
1288
1289 /*
1290 * Set up converters.
1291 */
1292 setup_converter(sc);
1293
1294 return (sc);
1295 }
1296
1297 /*
1298 * Free a string conversion object.
1299 */
1300 static void
free_sconv_object(struct archive_string_conv * sc)1301 free_sconv_object(struct archive_string_conv *sc)
1302 {
1303 free(sc->from_charset);
1304 free(sc->to_charset);
1305 archive_string_free(&sc->utftmp);
1306 #if HAVE_ICONV
1307 if (sc->cd != (iconv_t)-1)
1308 iconv_close(sc->cd);
1309 if (sc->cd_w != (iconv_t)-1)
1310 iconv_close(sc->cd_w);
1311 #endif
1312 free(sc);
1313 }
1314
1315 #if defined(_WIN32) && !defined(__CYGWIN__)
1316 static unsigned
my_atoi(const char * p)1317 my_atoi(const char *p)
1318 {
1319 unsigned cp;
1320
1321 cp = 0;
1322 while (*p) {
1323 if (*p >= '0' && *p <= '9')
1324 cp = cp * 10 + (*p - '0');
1325 else
1326 return (-1);
1327 p++;
1328 }
1329 return (cp);
1330 }
1331
1332 /*
1333 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1334 * Return -1 if failed.
1335 *
1336 * Note: This translation code may be insufficient.
1337 */
1338 static struct charset {
1339 const char *name;
1340 unsigned cp;
1341 } charsets[] = {
1342 /* MUST BE SORTED! */
1343 {"ASCII", 1252},
1344 {"ASMO-708", 708},
1345 {"BIG5", 950},
1346 {"CHINESE", 936},
1347 {"CP367", 1252},
1348 {"CP819", 1252},
1349 {"CP1025", 21025},
1350 {"DOS-720", 720},
1351 {"DOS-862", 862},
1352 {"EUC-CN", 51936},
1353 {"EUC-JP", 51932},
1354 {"EUC-KR", 949},
1355 {"EUCCN", 51936},
1356 {"EUCJP", 51932},
1357 {"EUCKR", 949},
1358 {"GB18030", 54936},
1359 {"GB2312", 936},
1360 {"HEBREW", 1255},
1361 {"HZ-GB-2312", 52936},
1362 {"IBM273", 20273},
1363 {"IBM277", 20277},
1364 {"IBM278", 20278},
1365 {"IBM280", 20280},
1366 {"IBM284", 20284},
1367 {"IBM285", 20285},
1368 {"IBM290", 20290},
1369 {"IBM297", 20297},
1370 {"IBM367", 1252},
1371 {"IBM420", 20420},
1372 {"IBM423", 20423},
1373 {"IBM424", 20424},
1374 {"IBM819", 1252},
1375 {"IBM871", 20871},
1376 {"IBM880", 20880},
1377 {"IBM905", 20905},
1378 {"IBM924", 20924},
1379 {"ISO-8859-1", 28591},
1380 {"ISO-8859-13", 28603},
1381 {"ISO-8859-15", 28605},
1382 {"ISO-8859-2", 28592},
1383 {"ISO-8859-3", 28593},
1384 {"ISO-8859-4", 28594},
1385 {"ISO-8859-5", 28595},
1386 {"ISO-8859-6", 28596},
1387 {"ISO-8859-7", 28597},
1388 {"ISO-8859-8", 28598},
1389 {"ISO-8859-9", 28599},
1390 {"ISO8859-1", 28591},
1391 {"ISO8859-13", 28603},
1392 {"ISO8859-15", 28605},
1393 {"ISO8859-2", 28592},
1394 {"ISO8859-3", 28593},
1395 {"ISO8859-4", 28594},
1396 {"ISO8859-5", 28595},
1397 {"ISO8859-6", 28596},
1398 {"ISO8859-7", 28597},
1399 {"ISO8859-8", 28598},
1400 {"ISO8859-9", 28599},
1401 {"JOHAB", 1361},
1402 {"KOI8-R", 20866},
1403 {"KOI8-U", 21866},
1404 {"KS_C_5601-1987", 949},
1405 {"LATIN1", 1252},
1406 {"LATIN2", 28592},
1407 {"MACINTOSH", 10000},
1408 {"SHIFT-JIS", 932},
1409 {"SHIFT_JIS", 932},
1410 {"SJIS", 932},
1411 {"US", 1252},
1412 {"US-ASCII", 1252},
1413 {"UTF-16", 1200},
1414 {"UTF-16BE", 1201},
1415 {"UTF-16LE", 1200},
1416 {"UTF-8", CP_UTF8},
1417 {"X-EUROPA", 29001},
1418 {"X-MAC-ARABIC", 10004},
1419 {"X-MAC-CE", 10029},
1420 {"X-MAC-CHINESEIMP", 10008},
1421 {"X-MAC-CHINESETRAD", 10002},
1422 {"X-MAC-CROATIAN", 10082},
1423 {"X-MAC-CYRILLIC", 10007},
1424 {"X-MAC-GREEK", 10006},
1425 {"X-MAC-HEBREW", 10005},
1426 {"X-MAC-ICELANDIC", 10079},
1427 {"X-MAC-JAPANESE", 10001},
1428 {"X-MAC-KOREAN", 10003},
1429 {"X-MAC-ROMANIAN", 10010},
1430 {"X-MAC-THAI", 10021},
1431 {"X-MAC-TURKISH", 10081},
1432 {"X-MAC-UKRAINIAN", 10017},
1433 };
1434 static unsigned
make_codepage_from_charset(const char * charset)1435 make_codepage_from_charset(const char *charset)
1436 {
1437 char cs[16];
1438 char *p;
1439 unsigned cp;
1440 int a, b;
1441
1442 if (charset == NULL || strlen(charset) > 15)
1443 return -1;
1444
1445 /* Copy name to uppercase. */
1446 p = cs;
1447 while (*charset) {
1448 char c = *charset++;
1449 if (c >= 'a' && c <= 'z')
1450 c -= 'a' - 'A';
1451 *p++ = c;
1452 }
1453 *p++ = '\0';
1454 cp = -1;
1455
1456 /* Look it up in the table first, so that we can easily
1457 * override CP367, which we map to 1252 instead of 367. */
1458 a = 0;
1459 b = sizeof(charsets)/sizeof(charsets[0]);
1460 while (b > a) {
1461 int c = (b + a) / 2;
1462 int r = strcmp(charsets[c].name, cs);
1463 if (r < 0)
1464 a = c + 1;
1465 else if (r > 0)
1466 b = c;
1467 else
1468 return charsets[c].cp;
1469 }
1470
1471 /* If it's not in the table, try to parse it. */
1472 switch (*cs) {
1473 case 'C':
1474 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1475 cp = my_atoi(cs + 2);
1476 } else if (strcmp(cs, "CP_ACP") == 0)
1477 cp = get_current_codepage();
1478 else if (strcmp(cs, "CP_OEMCP") == 0)
1479 cp = get_current_oemcp();
1480 break;
1481 case 'I':
1482 if (cs[1] == 'B' && cs[2] == 'M' &&
1483 cs[3] >= '0' && cs[3] <= '9') {
1484 cp = my_atoi(cs + 3);
1485 }
1486 break;
1487 case 'W':
1488 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1489 cp = my_atoi(cs + 8);
1490 if (cp != 874 && (cp < 1250 || cp > 1258))
1491 cp = -1;/* This may invalid code. */
1492 }
1493 break;
1494 }
1495 return (cp);
1496 }
1497
1498 /*
1499 * Return ANSI Code Page of current locale set by setlocale().
1500 */
1501 static unsigned
get_current_codepage(void)1502 get_current_codepage(void)
1503 {
1504 char *locale, *p;
1505 unsigned cp;
1506
1507 locale = setlocale(LC_CTYPE, NULL);
1508 if (locale == NULL)
1509 return (GetACP());
1510 if (locale[0] == 'C' && locale[1] == '\0')
1511 return (CP_C_LOCALE);
1512 p = strrchr(locale, '.');
1513 if (p == NULL)
1514 return (GetACP());
1515 cp = my_atoi(p+1);
1516 if (cp <= 0)
1517 return (GetACP());
1518 return (cp);
1519 }
1520
1521 /*
1522 * Translation table between Locale Name and ACP/OEMCP.
1523 */
1524 static struct {
1525 unsigned acp;
1526 unsigned ocp;
1527 const char *locale;
1528 } acp_ocp_map[] = {
1529 { 950, 950, "Chinese_Taiwan" },
1530 { 936, 936, "Chinese_People's Republic of China" },
1531 { 950, 950, "Chinese_Taiwan" },
1532 { 1250, 852, "Czech_Czech Republic" },
1533 { 1252, 850, "Danish_Denmark" },
1534 { 1252, 850, "Dutch_Netherlands" },
1535 { 1252, 850, "Dutch_Belgium" },
1536 { 1252, 437, "English_United States" },
1537 { 1252, 850, "English_Australia" },
1538 { 1252, 850, "English_Canada" },
1539 { 1252, 850, "English_New Zealand" },
1540 { 1252, 850, "English_United Kingdom" },
1541 { 1252, 437, "English_United States" },
1542 { 1252, 850, "Finnish_Finland" },
1543 { 1252, 850, "French_France" },
1544 { 1252, 850, "French_Belgium" },
1545 { 1252, 850, "French_Canada" },
1546 { 1252, 850, "French_Switzerland" },
1547 { 1252, 850, "German_Germany" },
1548 { 1252, 850, "German_Austria" },
1549 { 1252, 850, "German_Switzerland" },
1550 { 1253, 737, "Greek_Greece" },
1551 { 1250, 852, "Hungarian_Hungary" },
1552 { 1252, 850, "Icelandic_Iceland" },
1553 { 1252, 850, "Italian_Italy" },
1554 { 1252, 850, "Italian_Switzerland" },
1555 { 932, 932, "Japanese_Japan" },
1556 { 949, 949, "Korean_Korea" },
1557 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1558 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1559 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1560 { 1250, 852, "Polish_Poland" },
1561 { 1252, 850, "Portuguese_Portugal" },
1562 { 1252, 850, "Portuguese_Brazil" },
1563 { 1251, 866, "Russian_Russia" },
1564 { 1250, 852, "Slovak_Slovakia" },
1565 { 1252, 850, "Spanish_Spain" },
1566 { 1252, 850, "Spanish_Mexico" },
1567 { 1252, 850, "Spanish_Spain" },
1568 { 1252, 850, "Swedish_Sweden" },
1569 { 1254, 857, "Turkish_Turkey" },
1570 { 0, 0, NULL}
1571 };
1572
1573 /*
1574 * Return OEM Code Page of current locale set by setlocale().
1575 */
1576 static unsigned
get_current_oemcp(void)1577 get_current_oemcp(void)
1578 {
1579 int i;
1580 char *locale, *p;
1581 size_t len;
1582
1583 locale = setlocale(LC_CTYPE, NULL);
1584 if (locale == NULL)
1585 return (GetOEMCP());
1586 if (locale[0] == 'C' && locale[1] == '\0')
1587 return (CP_C_LOCALE);
1588
1589 p = strrchr(locale, '.');
1590 if (p == NULL)
1591 return (GetOEMCP());
1592 len = p - locale;
1593 for (i = 0; acp_ocp_map[i].acp; i++) {
1594 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1595 return (acp_ocp_map[i].ocp);
1596 }
1597 return (GetOEMCP());
1598 }
1599 #else
1600
1601 /*
1602 * POSIX platform does not use CodePage.
1603 */
1604
1605 static unsigned
get_current_codepage(void)1606 get_current_codepage(void)
1607 {
1608 return (-1);/* Unknown */
1609 }
1610 static unsigned
make_codepage_from_charset(const char * charset)1611 make_codepage_from_charset(const char *charset)
1612 {
1613 (void)charset; /* UNUSED */
1614 return (-1);/* Unknown */
1615 }
1616 static unsigned
get_current_oemcp(void)1617 get_current_oemcp(void)
1618 {
1619 return (-1);/* Unknown */
1620 }
1621
1622 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1623
1624 /*
1625 * Return a string conversion object.
1626 */
1627 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1628 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1629 {
1630 struct archive_string_conv *sc;
1631 unsigned current_codepage;
1632
1633 /* Check if we have made the sconv object. */
1634 sc = find_sconv_object(a, fc, tc);
1635 if (sc != NULL)
1636 return (sc);
1637
1638 if (a == NULL)
1639 current_codepage = get_current_codepage();
1640 else
1641 current_codepage = a->current_codepage;
1642
1643 sc = create_sconv_object(canonical_charset_name(fc),
1644 canonical_charset_name(tc), current_codepage, flag);
1645 if (sc == NULL) {
1646 if (a != NULL)
1647 archive_set_error(a, ENOMEM,
1648 "Could not allocate memory for "
1649 "a string conversion object");
1650 return (NULL);
1651 }
1652
1653 /*
1654 * If there is no converter for current string conversion object,
1655 * we cannot handle this conversion.
1656 */
1657 if (sc->nconverter == 0) {
1658 if (a != NULL) {
1659 #if HAVE_ICONV
1660 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1661 "iconv_open failed : Cannot handle ``%s''",
1662 (flag & SCONV_TO_CHARSET)?tc:fc);
1663 #else
1664 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1665 "A character-set conversion not fully supported "
1666 "on this platform");
1667 #endif
1668 }
1669 /* Failed; free a sconv object. */
1670 free_sconv_object(sc);
1671 return (NULL);
1672 }
1673
1674 /*
1675 * Success!
1676 */
1677 if (a != NULL)
1678 add_sconv_object(a, sc);
1679 return (sc);
1680 }
1681
1682 static const char *
get_current_charset(struct archive * a)1683 get_current_charset(struct archive *a)
1684 {
1685 const char *cur_charset;
1686
1687 if (a == NULL)
1688 cur_charset = default_iconv_charset("");
1689 else {
1690 cur_charset = default_iconv_charset(a->current_code);
1691 if (a->current_code == NULL) {
1692 a->current_code = strdup(cur_charset);
1693 a->current_codepage = get_current_codepage();
1694 a->current_oemcp = get_current_oemcp();
1695 }
1696 }
1697 return (cur_charset);
1698 }
1699
1700 /*
1701 * Make and Return a string conversion object.
1702 * Return NULL if the platform does not support the specified conversion
1703 * and best_effort is 0.
1704 * If best_effort is set, A string conversion object must be returned
1705 * unless memory allocation for the object fails, but the conversion
1706 * might fail when non-ASCII code is found.
1707 */
1708 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1709 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1710 int best_effort)
1711 {
1712 int flag = SCONV_TO_CHARSET;
1713
1714 if (best_effort)
1715 flag |= SCONV_BEST_EFFORT;
1716 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1717 }
1718
1719 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1720 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1721 int best_effort)
1722 {
1723 int flag = SCONV_FROM_CHARSET;
1724
1725 if (best_effort)
1726 flag |= SCONV_BEST_EFFORT;
1727 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1728 }
1729
1730 /*
1731 * archive_string_default_conversion_*_archive() are provided for Windows
1732 * platform because other archiver application use CP_OEMCP for
1733 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1734 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1735 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1736 * So we should make a string conversion between CP_ACP and CP_OEMCP
1737 * for compatibility.
1738 */
1739 #if defined(_WIN32) && !defined(__CYGWIN__)
1740 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1741 archive_string_default_conversion_for_read(struct archive *a)
1742 {
1743 const char *cur_charset = get_current_charset(a);
1744 char oemcp[16];
1745
1746 /* NOTE: a check of cur_charset is unneeded but we need
1747 * that get_current_charset() has been surely called at
1748 * this time whatever C compiler optimized. */
1749 if (cur_charset != NULL &&
1750 (a->current_codepage == CP_C_LOCALE ||
1751 a->current_codepage == a->current_oemcp))
1752 return (NULL);/* no conversion. */
1753
1754 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1755 /* Make sure a null termination must be set. */
1756 oemcp[sizeof(oemcp)-1] = '\0';
1757 return (get_sconv_object(a, oemcp, cur_charset,
1758 SCONV_FROM_CHARSET));
1759 }
1760
1761 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1762 archive_string_default_conversion_for_write(struct archive *a)
1763 {
1764 const char *cur_charset = get_current_charset(a);
1765 char oemcp[16];
1766
1767 /* NOTE: a check of cur_charset is unneeded but we need
1768 * that get_current_charset() has been surely called at
1769 * this time whatever C compiler optimized. */
1770 if (cur_charset != NULL &&
1771 (a->current_codepage == CP_C_LOCALE ||
1772 a->current_codepage == a->current_oemcp))
1773 return (NULL);/* no conversion. */
1774
1775 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1776 /* Make sure a null termination must be set. */
1777 oemcp[sizeof(oemcp)-1] = '\0';
1778 return (get_sconv_object(a, cur_charset, oemcp,
1779 SCONV_TO_CHARSET));
1780 }
1781 #else
1782 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1783 archive_string_default_conversion_for_read(struct archive *a)
1784 {
1785 (void)a; /* UNUSED */
1786 return (NULL);
1787 }
1788
1789 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1790 archive_string_default_conversion_for_write(struct archive *a)
1791 {
1792 (void)a; /* UNUSED */
1793 return (NULL);
1794 }
1795 #endif
1796
1797 /*
1798 * Dispose of all character conversion objects in the archive object.
1799 */
1800 void
archive_string_conversion_free(struct archive * a)1801 archive_string_conversion_free(struct archive *a)
1802 {
1803 struct archive_string_conv *sc;
1804 struct archive_string_conv *sc_next;
1805
1806 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1807 sc_next = sc->next;
1808 free_sconv_object(sc);
1809 }
1810 a->sconv = NULL;
1811 free(a->current_code);
1812 a->current_code = NULL;
1813 }
1814
1815 /*
1816 * Return a conversion charset name.
1817 */
1818 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1819 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1820 {
1821 if (sc->flag & SCONV_TO_CHARSET)
1822 return (sc->to_charset);
1823 else
1824 return (sc->from_charset);
1825 }
1826
1827 /*
1828 * Change the behavior of a string conversion.
1829 */
1830 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1831 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1832 {
1833 switch (opt) {
1834 /*
1835 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1836 * assumption that wchar_t was Unicode.
1837 * This option enables simulating the assumption in order to read
1838 * that filename correctly.
1839 */
1840 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1841 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1842 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1843 /*
1844 * Nothing to do for it since wchar_t on these platforms
1845 * is really Unicode.
1846 */
1847 (void)sc; /* UNUSED */
1848 #else
1849 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1850 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1851 /* Set up string converters. */
1852 setup_converter(sc);
1853 }
1854 #endif
1855 break;
1856 case SCONV_SET_OPT_NORMALIZATION_C:
1857 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1858 sc->flag |= SCONV_NORMALIZATION_C;
1859 sc->flag &= ~SCONV_NORMALIZATION_D;
1860 /* Set up string converters. */
1861 setup_converter(sc);
1862 }
1863 break;
1864 case SCONV_SET_OPT_NORMALIZATION_D:
1865 #if defined(HAVE_ICONV)
1866 /*
1867 * If iconv will take the string, do not change the
1868 * setting of the normalization.
1869 */
1870 if (!(sc->flag & SCONV_WIN_CP) &&
1871 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1872 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1873 break;
1874 #endif
1875 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1876 sc->flag |= SCONV_NORMALIZATION_D;
1877 sc->flag &= ~SCONV_NORMALIZATION_C;
1878 /* Set up string converters. */
1879 setup_converter(sc);
1880 }
1881 break;
1882 default:
1883 break;
1884 }
1885 }
1886
1887 /*
1888 *
1889 * Copy one archive_string to another in locale conversion.
1890 *
1891 * archive_strncat_l();
1892 * archive_strncpy_l();
1893 *
1894 */
1895
1896 static size_t
mbsnbytes(const void * _p,size_t n)1897 mbsnbytes(const void *_p, size_t n)
1898 {
1899 size_t s;
1900 const char *p, *pp;
1901
1902 if (_p == NULL)
1903 return (0);
1904 p = (const char *)_p;
1905
1906 /* Like strlen(p), except won't examine positions beyond p[n]. */
1907 s = 0;
1908 pp = p;
1909 while (s < n && *pp) {
1910 pp++;
1911 s++;
1912 }
1913 return (s);
1914 }
1915
1916 static size_t
utf16nbytes(const void * _p,size_t n)1917 utf16nbytes(const void *_p, size_t n)
1918 {
1919 size_t s;
1920 const char *p, *pp;
1921
1922 if (_p == NULL)
1923 return (0);
1924 p = (const char *)_p;
1925
1926 /* Like strlen(p), except won't examine positions beyond p[n]. */
1927 s = 0;
1928 pp = p;
1929 n >>= 1;
1930 while (s < n && (pp[0] || pp[1])) {
1931 pp += 2;
1932 s++;
1933 }
1934 return (s<<1);
1935 }
1936
1937 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1938 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1939 struct archive_string_conv *sc)
1940 {
1941 as->length = 0;
1942 return (archive_strncat_l(as, _p, n, sc));
1943 }
1944
1945 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1946 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1947 struct archive_string_conv *sc)
1948 {
1949 const void *s;
1950 size_t length = 0;
1951 int i, r = 0, r2;
1952
1953 if (_p != NULL && n > 0) {
1954 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
1955 length = utf16nbytes(_p, n);
1956 else
1957 length = mbsnbytes(_p, n);
1958 }
1959
1960 /* We must allocate memory even if there is no data for conversion
1961 * or copy. This simulates archive_string_append behavior. */
1962 if (length == 0) {
1963 int tn = 1;
1964 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1965 tn = 2;
1966 if (archive_string_ensure(as, as->length + tn) == NULL)
1967 return (-1);
1968 as->s[as->length] = 0;
1969 if (tn == 2)
1970 as->s[as->length+1] = 0;
1971 return (0);
1972 }
1973
1974 /*
1975 * If sc is NULL, we just make a copy.
1976 */
1977 if (sc == NULL) {
1978 if (archive_string_append(as, _p, length) == NULL)
1979 return (-1);/* No memory */
1980 return (0);
1981 }
1982
1983 s = _p;
1984 i = 0;
1985 if (sc->nconverter > 1) {
1986 sc->utftmp.length = 0;
1987 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
1988 if (r2 != 0 && errno == ENOMEM)
1989 return (r2);
1990 if (r > r2)
1991 r = r2;
1992 s = sc->utftmp.s;
1993 length = sc->utftmp.length;
1994 ++i;
1995 }
1996 r2 = sc->converter[i](as, s, length, sc);
1997 if (r > r2)
1998 r = r2;
1999 return (r);
2000 }
2001
2002 #if HAVE_ICONV
2003
2004 /*
2005 * Return -1 if conversion fails.
2006 */
2007 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2008 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2009 size_t length, struct archive_string_conv *sc)
2010 {
2011 ICONV_CONST char *itp;
2012 size_t remaining;
2013 iconv_t cd;
2014 char *outp;
2015 size_t avail, bs;
2016 int return_value = 0; /* success */
2017 int to_size, from_size;
2018
2019 if (sc->flag & SCONV_TO_UTF16)
2020 to_size = 2;
2021 else
2022 to_size = 1;
2023 if (sc->flag & SCONV_FROM_UTF16)
2024 from_size = 2;
2025 else
2026 from_size = 1;
2027
2028 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2029 return (-1);
2030
2031 cd = sc->cd;
2032 itp = (char *)(uintptr_t)_p;
2033 remaining = length;
2034 outp = as->s + as->length;
2035 avail = as->buffer_length - as->length - to_size;
2036 while (remaining >= (size_t)from_size) {
2037 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2038
2039 if (result != (size_t)-1)
2040 break; /* Conversion completed. */
2041
2042 if (errno == EILSEQ || errno == EINVAL) {
2043 /*
2044 * If an output charset is UTF-8 or UTF-16BE/LE,
2045 * unknown character should be U+FFFD
2046 * (replacement character).
2047 */
2048 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2049 size_t rbytes;
2050 if (sc->flag & SCONV_TO_UTF8)
2051 rbytes = sizeof(utf8_replacement_char);
2052 else
2053 rbytes = 2;
2054
2055 if (avail < rbytes) {
2056 as->length = outp - as->s;
2057 bs = as->buffer_length +
2058 (remaining * to_size) + rbytes;
2059 if (NULL ==
2060 archive_string_ensure(as, bs))
2061 return (-1);
2062 outp = as->s + as->length;
2063 avail = as->buffer_length
2064 - as->length - to_size;
2065 }
2066 if (sc->flag & SCONV_TO_UTF8)
2067 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2068 else if (sc->flag & SCONV_TO_UTF16BE)
2069 archive_be16enc(outp, UNICODE_R_CHAR);
2070 else
2071 archive_le16enc(outp, UNICODE_R_CHAR);
2072 outp += rbytes;
2073 avail -= rbytes;
2074 } else {
2075 /* Skip the illegal input bytes. */
2076 *outp++ = '?';
2077 avail--;
2078 }
2079 itp += from_size;
2080 remaining -= from_size;
2081 return_value = -1; /* failure */
2082 } else {
2083 /* E2BIG no output buffer,
2084 * Increase an output buffer. */
2085 as->length = outp - as->s;
2086 bs = as->buffer_length + remaining * 2;
2087 if (NULL == archive_string_ensure(as, bs))
2088 return (-1);
2089 outp = as->s + as->length;
2090 avail = as->buffer_length - as->length - to_size;
2091 }
2092 }
2093 as->length = outp - as->s;
2094 as->s[as->length] = 0;
2095 if (to_size == 2)
2096 as->s[as->length+1] = 0;
2097 return (return_value);
2098 }
2099
2100 #endif /* HAVE_ICONV */
2101
2102
2103 #if defined(_WIN32) && !defined(__CYGWIN__)
2104
2105 /*
2106 * Translate a string from a some CodePage to an another CodePage by
2107 * Windows APIs, and copy the result. Return -1 if conversion fails.
2108 */
2109 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2110 strncat_in_codepage(struct archive_string *as,
2111 const void *_p, size_t length, struct archive_string_conv *sc)
2112 {
2113 const char *s = (const char *)_p;
2114 struct archive_wstring aws;
2115 size_t l;
2116 int r, saved_flag;
2117
2118 archive_string_init(&aws);
2119 saved_flag = sc->flag;
2120 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2121 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2122 sc->flag = saved_flag;
2123 if (r != 0) {
2124 archive_wstring_free(&aws);
2125 if (errno != ENOMEM)
2126 archive_string_append(as, s, length);
2127 return (-1);
2128 }
2129
2130 l = as->length;
2131 r = archive_string_append_from_wcs_in_codepage(
2132 as, aws.s, aws.length, sc);
2133 if (r != 0 && errno != ENOMEM && l == as->length)
2134 archive_string_append(as, s, length);
2135 archive_wstring_free(&aws);
2136 return (r);
2137 }
2138
2139 /*
2140 * Test whether MBS ==> WCS is okay.
2141 */
2142 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2143 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2144 {
2145 const char *p = (const char *)_p;
2146 unsigned codepage;
2147 DWORD mbflag = MB_ERR_INVALID_CHARS;
2148
2149 if (sc->flag & SCONV_FROM_CHARSET)
2150 codepage = sc->to_cp;
2151 else
2152 codepage = sc->from_cp;
2153
2154 if (codepage == CP_C_LOCALE)
2155 return (0);
2156 if (codepage != CP_UTF8)
2157 mbflag |= MB_PRECOMPOSED;
2158
2159 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2160 return (-1); /* Invalid */
2161 return (0); /* Okay */
2162 }
2163
2164 #else
2165
2166 /*
2167 * Test whether MBS ==> WCS is okay.
2168 */
2169 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2170 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2171 {
2172 const char *p = (const char *)_p;
2173 size_t r;
2174
2175 #if HAVE_MBRTOWC
2176 mbstate_t shift_state;
2177
2178 memset(&shift_state, 0, sizeof(shift_state));
2179 #else
2180 /* Clear the shift state before starting. */
2181 mbtowc(NULL, NULL, 0);
2182 #endif
2183 while (n) {
2184 wchar_t wc;
2185
2186 #if HAVE_MBRTOWC
2187 r = mbrtowc(&wc, p, n, &shift_state);
2188 #else
2189 r = mbtowc(&wc, p, n);
2190 #endif
2191 if (r == (size_t)-1 || r == (size_t)-2)
2192 return (-1);/* Invalid. */
2193 if (r == 0)
2194 break;
2195 p += r;
2196 n -= r;
2197 }
2198 (void)sc; /* UNUSED */
2199 return (0); /* All Okey. */
2200 }
2201
2202 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2203
2204 /*
2205 * Basically returns -1 because we cannot make a conversion of charset
2206 * without iconv but in some cases this would return 0.
2207 * Returns 0 if all copied characters are ASCII.
2208 * Returns 0 if both from-locale and to-locale are the same and those
2209 * can be WCS with no error.
2210 */
2211 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2212 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2213 size_t length, struct archive_string_conv *sc)
2214 {
2215 size_t remaining;
2216 const uint8_t *itp;
2217 int return_value = 0; /* success */
2218
2219 /*
2220 * If both from-locale and to-locale is the same, this makes a copy.
2221 * And then this checks all copied MBS can be WCS if so returns 0.
2222 */
2223 if (sc->same) {
2224 if (archive_string_append(as, _p, length) == NULL)
2225 return (-1);/* No memory */
2226 return (invalid_mbs(_p, length, sc));
2227 }
2228
2229 /*
2230 * If a character is ASCII, this just copies it. If not, this
2231 * assigns '?' character instead but in UTF-8 locale this assigns
2232 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2233 * a Replacement Character in Unicode.
2234 */
2235
2236 remaining = length;
2237 itp = (const uint8_t *)_p;
2238 while (*itp && remaining > 0) {
2239 if (*itp > 127) {
2240 // Non-ASCII: Substitute with suitable replacement
2241 if (sc->flag & SCONV_TO_UTF8) {
2242 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2243 __archive_errx(1, "Out of memory");
2244 }
2245 } else {
2246 archive_strappend_char(as, '?');
2247 }
2248 return_value = -1;
2249 } else {
2250 archive_strappend_char(as, *itp);
2251 }
2252 ++itp;
2253 }
2254 return (return_value);
2255 }
2256
2257
2258 /*
2259 * Unicode conversion functions.
2260 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2261 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2262 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2263 * - UTF-16BE <===> UTF-8.
2264 *
2265 */
2266
2267 /*
2268 * Utility to convert a single UTF-8 sequence.
2269 *
2270 * Usually return used bytes, return used byte in negative value when
2271 * a unicode character is replaced with U+FFFD.
2272 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2273 * Recommended Practice for Replacement Characters.
2274 */
2275 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2276 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2277 {
2278 static const char utf8_count[256] = {
2279 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2280 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2281 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2282 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2283 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2284 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2285 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2286 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2287 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2288 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2289 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2290 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2291 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2292 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2293 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2294 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2295 };
2296 int ch, i;
2297 int cnt;
2298 uint32_t wc;
2299
2300 /* Sanity check. */
2301 if (n == 0)
2302 return (0);
2303 /*
2304 * Decode 1-4 bytes depending on the value of the first byte.
2305 */
2306 ch = (unsigned char)*s;
2307 if (ch == 0)
2308 return (0); /* Standard: return 0 for end-of-string. */
2309 cnt = utf8_count[ch];
2310
2311 /* Invalid sequence or there are not plenty bytes. */
2312 if ((int)n < cnt) {
2313 cnt = (int)n;
2314 for (i = 1; i < cnt; i++) {
2315 if ((s[i] & 0xc0) != 0x80) {
2316 cnt = i;
2317 break;
2318 }
2319 }
2320 goto invalid_sequence;
2321 }
2322
2323 /* Make a Unicode code point from a single UTF-8 sequence. */
2324 switch (cnt) {
2325 case 1: /* 1 byte sequence. */
2326 *pwc = ch & 0x7f;
2327 return (cnt);
2328 case 2: /* 2 bytes sequence. */
2329 if ((s[1] & 0xc0) != 0x80) {
2330 cnt = 1;
2331 goto invalid_sequence;
2332 }
2333 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2334 return (cnt);
2335 case 3: /* 3 bytes sequence. */
2336 if ((s[1] & 0xc0) != 0x80) {
2337 cnt = 1;
2338 goto invalid_sequence;
2339 }
2340 if ((s[2] & 0xc0) != 0x80) {
2341 cnt = 2;
2342 goto invalid_sequence;
2343 }
2344 wc = ((ch & 0x0f) << 12)
2345 | ((s[1] & 0x3f) << 6)
2346 | (s[2] & 0x3f);
2347 if (wc < 0x800)
2348 goto invalid_sequence;/* Overlong sequence. */
2349 break;
2350 case 4: /* 4 bytes sequence. */
2351 if ((s[1] & 0xc0) != 0x80) {
2352 cnt = 1;
2353 goto invalid_sequence;
2354 }
2355 if ((s[2] & 0xc0) != 0x80) {
2356 cnt = 2;
2357 goto invalid_sequence;
2358 }
2359 if ((s[3] & 0xc0) != 0x80) {
2360 cnt = 3;
2361 goto invalid_sequence;
2362 }
2363 wc = ((ch & 0x07) << 18)
2364 | ((s[1] & 0x3f) << 12)
2365 | ((s[2] & 0x3f) << 6)
2366 | (s[3] & 0x3f);
2367 if (wc < 0x10000)
2368 goto invalid_sequence;/* Overlong sequence. */
2369 break;
2370 default: /* Others are all invalid sequence. */
2371 if (ch == 0xc0 || ch == 0xc1)
2372 cnt = 2;
2373 else if (ch >= 0xf5 && ch <= 0xf7)
2374 cnt = 4;
2375 else if (ch >= 0xf8 && ch <= 0xfb)
2376 cnt = 5;
2377 else if (ch == 0xfc || ch == 0xfd)
2378 cnt = 6;
2379 else
2380 cnt = 1;
2381 if ((int)n < cnt)
2382 cnt = (int)n;
2383 for (i = 1; i < cnt; i++) {
2384 if ((s[i] & 0xc0) != 0x80) {
2385 cnt = i;
2386 break;
2387 }
2388 }
2389 goto invalid_sequence;
2390 }
2391
2392 /* The code point larger than 0x10FFFF is not legal
2393 * Unicode values. */
2394 if (wc > UNICODE_MAX)
2395 goto invalid_sequence;
2396 /* Correctly gets a Unicode, returns used bytes. */
2397 *pwc = wc;
2398 return (cnt);
2399 invalid_sequence:
2400 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2401 return (cnt * -1);
2402 }
2403
2404 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2405 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2406 {
2407 int cnt;
2408
2409 cnt = _utf8_to_unicode(pwc, s, n);
2410 /* Any of Surrogate pair is not legal Unicode values. */
2411 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2412 return (-3);
2413 return (cnt);
2414 }
2415
2416 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2417 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2418 {
2419 uc -= 0xD800;
2420 uc *= 0x400;
2421 uc += uc2 - 0xDC00;
2422 uc += 0x10000;
2423 return (uc);
2424 }
2425
2426 /*
2427 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2428 * removing surrogate pairs.
2429 *
2430 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2431 *
2432 * Usually return used bytes, return used byte in negative value when
2433 * a unicode character is replaced with U+FFFD.
2434 */
2435 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2436 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2437 {
2438 uint32_t wc = 0;
2439 int cnt;
2440
2441 cnt = _utf8_to_unicode(&wc, s, n);
2442 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2443 uint32_t wc2 = 0;
2444 if (n - 3 < 3) {
2445 /* Invalid byte sequence. */
2446 goto invalid_sequence;
2447 }
2448 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2449 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2450 /* Invalid byte sequence. */
2451 goto invalid_sequence;
2452 }
2453 wc = combine_surrogate_pair(wc, wc2);
2454 cnt = 6;
2455 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2456 /* Invalid byte sequence. */
2457 goto invalid_sequence;
2458 }
2459 *pwc = wc;
2460 return (cnt);
2461 invalid_sequence:
2462 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2463 if (cnt > 0)
2464 cnt *= -1;
2465 return (cnt);
2466 }
2467
2468 /*
2469 * Convert a Unicode code point to a single UTF-8 sequence.
2470 *
2471 * NOTE:This function does not check if the Unicode is legal or not.
2472 * Please you definitely check it before calling this.
2473 */
2474 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2475 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2476 {
2477 char *_p = p;
2478
2479 /* Invalid Unicode char maps to Replacement character */
2480 if (uc > UNICODE_MAX)
2481 uc = UNICODE_R_CHAR;
2482 /* Translate code point to UTF8 */
2483 if (uc <= 0x7f) {
2484 if (remaining == 0)
2485 return (0);
2486 *p++ = (char)uc;
2487 } else if (uc <= 0x7ff) {
2488 if (remaining < 2)
2489 return (0);
2490 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2491 *p++ = 0x80 | (uc & 0x3f);
2492 } else if (uc <= 0xffff) {
2493 if (remaining < 3)
2494 return (0);
2495 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2496 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2497 *p++ = 0x80 | (uc & 0x3f);
2498 } else {
2499 if (remaining < 4)
2500 return (0);
2501 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2502 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2503 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2504 *p++ = 0x80 | (uc & 0x3f);
2505 }
2506 return (p - _p);
2507 }
2508
2509 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2510 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2511 {
2512 return (utf16_to_unicode(pwc, s, n, 1));
2513 }
2514
2515 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2516 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2517 {
2518 return (utf16_to_unicode(pwc, s, n, 0));
2519 }
2520
2521 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2522 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2523 {
2524 const char *utf16 = s;
2525 unsigned uc;
2526
2527 if (n == 0)
2528 return (0);
2529 if (n == 1) {
2530 /* set the Replacement Character instead. */
2531 *pwc = UNICODE_R_CHAR;
2532 return (-1);
2533 }
2534
2535 if (be)
2536 uc = archive_be16dec(utf16);
2537 else
2538 uc = archive_le16dec(utf16);
2539 utf16 += 2;
2540
2541 /* If this is a surrogate pair, assemble the full code point.*/
2542 if (IS_HIGH_SURROGATE_LA(uc)) {
2543 unsigned uc2;
2544
2545 if (n >= 4) {
2546 if (be)
2547 uc2 = archive_be16dec(utf16);
2548 else
2549 uc2 = archive_le16dec(utf16);
2550 } else
2551 uc2 = 0;
2552 if (IS_LOW_SURROGATE_LA(uc2)) {
2553 uc = combine_surrogate_pair(uc, uc2);
2554 utf16 += 2;
2555 } else {
2556 /* Undescribed code point should be U+FFFD
2557 * (replacement character). */
2558 *pwc = UNICODE_R_CHAR;
2559 return (-2);
2560 }
2561 }
2562
2563 /*
2564 * Surrogate pair values(0xd800 through 0xdfff) are only
2565 * used by UTF-16, so, after above calculation, the code
2566 * must not be surrogate values, and Unicode has no codes
2567 * larger than 0x10ffff. Thus, those are not legal Unicode
2568 * values.
2569 */
2570 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2571 /* Undescribed code point should be U+FFFD
2572 * (replacement character). */
2573 *pwc = UNICODE_R_CHAR;
2574 return (((int)(utf16 - s)) * -1);
2575 }
2576 *pwc = uc;
2577 return ((int)(utf16 - s));
2578 }
2579
2580 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2581 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2582 {
2583 char *utf16 = p;
2584
2585 if (uc > 0xffff) {
2586 /* We have a code point that won't fit into a
2587 * wchar_t; convert it to a surrogate pair. */
2588 if (remaining < 4)
2589 return (0);
2590 uc -= 0x10000;
2591 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2592 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2593 return (4);
2594 } else {
2595 if (remaining < 2)
2596 return (0);
2597 archive_be16enc(utf16, uc);
2598 return (2);
2599 }
2600 }
2601
2602 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2603 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2604 {
2605 char *utf16 = p;
2606
2607 if (uc > 0xffff) {
2608 /* We have a code point that won't fit into a
2609 * wchar_t; convert it to a surrogate pair. */
2610 if (remaining < 4)
2611 return (0);
2612 uc -= 0x10000;
2613 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2614 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2615 return (4);
2616 } else {
2617 if (remaining < 2)
2618 return (0);
2619 archive_le16enc(utf16, uc);
2620 return (2);
2621 }
2622 }
2623
2624 /*
2625 * Copy UTF-8 string in checking surrogate pair.
2626 * If any surrogate pair are found, it would be canonicalized.
2627 */
2628 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2629 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p,
2630 size_t len, struct archive_string_conv *sc)
2631 {
2632 const char *s;
2633 char *p, *endp;
2634 int n, ret = 0;
2635
2636 (void)sc; /* UNUSED */
2637
2638 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2639 return (-1);
2640
2641 s = (const char *)_p;
2642 p = as->s + as->length;
2643 endp = as->s + as->buffer_length -1;
2644 do {
2645 uint32_t uc;
2646 const char *ss = s;
2647 size_t w;
2648
2649 /*
2650 * Forward byte sequence until a conversion of that is needed.
2651 */
2652 while ((n = utf8_to_unicode(&uc, s, len)) > 0) {
2653 s += n;
2654 len -= n;
2655 }
2656 if (ss < s) {
2657 if (p + (s - ss) > endp) {
2658 as->length = p - as->s;
2659 if (archive_string_ensure(as,
2660 as->buffer_length + len + 1) == NULL)
2661 return (-1);
2662 p = as->s + as->length;
2663 endp = as->s + as->buffer_length -1;
2664 }
2665
2666 memcpy(p, ss, s - ss);
2667 p += s - ss;
2668 }
2669
2670 /*
2671 * If n is negative, current byte sequence needs a replacement.
2672 */
2673 if (n < 0) {
2674 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2675 /* Current byte sequence may be CESU-8. */
2676 n = cesu8_to_unicode(&uc, s, len);
2677 }
2678 if (n < 0) {
2679 ret = -1;
2680 n *= -1;/* Use a replaced unicode character. */
2681 }
2682
2683 /* Rebuild UTF-8 byte sequence. */
2684 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) {
2685 as->length = p - as->s;
2686 if (archive_string_ensure(as,
2687 as->buffer_length + len + 1) == NULL)
2688 return (-1);
2689 p = as->s + as->length;
2690 endp = as->s + as->buffer_length -1;
2691 }
2692 p += w;
2693 s += n;
2694 len -= n;
2695 }
2696 } while (n > 0);
2697 as->length = p - as->s;
2698 as->s[as->length] = '\0';
2699 return (ret);
2700 }
2701
2702 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2703 archive_string_append_unicode(struct archive_string *as, const void *_p,
2704 size_t len, struct archive_string_conv *sc)
2705 {
2706 const char *s;
2707 char *p, *endp;
2708 uint32_t uc;
2709 size_t w;
2710 int n, ret = 0, ts, tm;
2711 int (*parse)(uint32_t *, const char *, size_t);
2712 size_t (*unparse)(char *, size_t, uint32_t);
2713
2714 if (sc->flag & SCONV_TO_UTF16BE) {
2715 unparse = unicode_to_utf16be;
2716 ts = 2;
2717 } else if (sc->flag & SCONV_TO_UTF16LE) {
2718 unparse = unicode_to_utf16le;
2719 ts = 2;
2720 } else if (sc->flag & SCONV_TO_UTF8) {
2721 unparse = unicode_to_utf8;
2722 ts = 1;
2723 } else {
2724 /*
2725 * This case is going to be converted to another
2726 * character-set through iconv.
2727 */
2728 if (sc->flag & SCONV_FROM_UTF16BE) {
2729 unparse = unicode_to_utf16be;
2730 ts = 2;
2731 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2732 unparse = unicode_to_utf16le;
2733 ts = 2;
2734 } else {
2735 unparse = unicode_to_utf8;
2736 ts = 1;
2737 }
2738 }
2739
2740 if (sc->flag & SCONV_FROM_UTF16BE) {
2741 parse = utf16be_to_unicode;
2742 tm = 1;
2743 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2744 parse = utf16le_to_unicode;
2745 tm = 1;
2746 } else {
2747 parse = cesu8_to_unicode;
2748 tm = ts;
2749 }
2750
2751 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2752 return (-1);
2753
2754 s = (const char *)_p;
2755 p = as->s + as->length;
2756 endp = as->s + as->buffer_length - ts;
2757 while ((n = parse(&uc, s, len)) != 0) {
2758 if (n < 0) {
2759 /* Use a replaced unicode character. */
2760 n *= -1;
2761 ret = -1;
2762 }
2763 s += n;
2764 len -= n;
2765 while ((w = unparse(p, endp - p, uc)) == 0) {
2766 /* There is not enough output buffer so
2767 * we have to expand it. */
2768 as->length = p - as->s;
2769 if (archive_string_ensure(as,
2770 as->buffer_length + len * tm + ts) == NULL)
2771 return (-1);
2772 p = as->s + as->length;
2773 endp = as->s + as->buffer_length - ts;
2774 }
2775 p += w;
2776 }
2777 as->length = p - as->s;
2778 as->s[as->length] = '\0';
2779 if (ts == 2)
2780 as->s[as->length+1] = '\0';
2781 return (ret);
2782 }
2783
2784 /*
2785 * Following Constants for Hangul compositions this information comes from
2786 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2787 */
2788 #define HC_SBASE 0xAC00
2789 #define HC_LBASE 0x1100
2790 #define HC_VBASE 0x1161
2791 #define HC_TBASE 0x11A7
2792 #define HC_LCOUNT 19
2793 #define HC_VCOUNT 21
2794 #define HC_TCOUNT 28
2795 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2796 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2797
2798 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2799 get_nfc(uint32_t uc, uint32_t uc2)
2800 {
2801 int t, b;
2802
2803 t = 0;
2804 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2805 while (b >= t) {
2806 int m = (t + b) / 2;
2807 if (u_composition_table[m].cp1 < uc)
2808 t = m + 1;
2809 else if (u_composition_table[m].cp1 > uc)
2810 b = m - 1;
2811 else if (u_composition_table[m].cp2 < uc2)
2812 t = m + 1;
2813 else if (u_composition_table[m].cp2 > uc2)
2814 b = m - 1;
2815 else
2816 return (u_composition_table[m].nfc);
2817 }
2818 return (0);
2819 }
2820
2821 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2822 * Characters. */
2823
2824 /*
2825 * Update first code point.
2826 */
2827 #define UPDATE_UC(new_uc) do { \
2828 uc = new_uc; \
2829 ucptr = NULL; \
2830 } while (0)
2831
2832 /*
2833 * Replace first code point with second code point.
2834 */
2835 #define REPLACE_UC_WITH_UC2() do { \
2836 uc = uc2; \
2837 ucptr = uc2ptr; \
2838 n = n2; \
2839 } while (0)
2840
2841 #define EXPAND_BUFFER() do { \
2842 as->length = p - as->s; \
2843 if (archive_string_ensure(as, \
2844 as->buffer_length + len * tm + ts) == NULL)\
2845 return (-1); \
2846 p = as->s + as->length; \
2847 endp = as->s + as->buffer_length - ts; \
2848 } while (0)
2849
2850 #define UNPARSE(p, endp, uc) do { \
2851 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2852 EXPAND_BUFFER(); \
2853 } \
2854 p += w; \
2855 } while (0)
2856
2857 /*
2858 * Write first code point.
2859 * If the code point has not be changed from its original code,
2860 * this just copies it from its original buffer pointer.
2861 * If not, this converts it to UTF-8 byte sequence and copies it.
2862 */
2863 #define WRITE_UC() do { \
2864 if (ucptr) { \
2865 if (p + n > endp) \
2866 EXPAND_BUFFER(); \
2867 switch (n) { \
2868 case 4: \
2869 *p++ = *ucptr++; \
2870 /* FALL THROUGH */ \
2871 case 3: \
2872 *p++ = *ucptr++; \
2873 /* FALL THROUGH */ \
2874 case 2: \
2875 *p++ = *ucptr++; \
2876 /* FALL THROUGH */ \
2877 case 1: \
2878 *p++ = *ucptr; \
2879 break; \
2880 } \
2881 ucptr = NULL; \
2882 } else { \
2883 UNPARSE(p, endp, uc); \
2884 } \
2885 } while (0)
2886
2887 /*
2888 * Collect following decomposable code points.
2889 */
2890 #define COLLECT_CPS(start) do { \
2891 int _i; \
2892 for (_i = start; _i < FDC_MAX ; _i++) { \
2893 nx = parse(&ucx[_i], s, len); \
2894 if (nx <= 0) \
2895 break; \
2896 cx = CCC(ucx[_i]); \
2897 if (cl >= cx && cl != 228 && cx != 228)\
2898 break; \
2899 s += nx; \
2900 len -= nx; \
2901 cl = cx; \
2902 ccx[_i] = cx; \
2903 } \
2904 if (_i >= FDC_MAX) { \
2905 ret = -1; \
2906 ucx_size = FDC_MAX; \
2907 } else \
2908 ucx_size = _i; \
2909 } while (0)
2910
2911 /*
2912 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2913 *
2914 * TODO: Convert composition exclusions, which are never converted
2915 * from NFC,NFD,NFKC and NFKD, to Form C.
2916 */
2917 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2918 archive_string_normalize_C(struct archive_string *as, const void *_p,
2919 size_t len, struct archive_string_conv *sc)
2920 {
2921 const char *s = (const char *)_p;
2922 char *p, *endp;
2923 uint32_t uc, uc2;
2924 size_t w;
2925 int always_replace, n, n2, ret = 0, spair, ts, tm;
2926 int (*parse)(uint32_t *, const char *, size_t);
2927 size_t (*unparse)(char *, size_t, uint32_t);
2928
2929 always_replace = 1;
2930 ts = 1;/* text size. */
2931 if (sc->flag & SCONV_TO_UTF16BE) {
2932 unparse = unicode_to_utf16be;
2933 ts = 2;
2934 if (sc->flag & SCONV_FROM_UTF16BE)
2935 always_replace = 0;
2936 } else if (sc->flag & SCONV_TO_UTF16LE) {
2937 unparse = unicode_to_utf16le;
2938 ts = 2;
2939 if (sc->flag & SCONV_FROM_UTF16LE)
2940 always_replace = 0;
2941 } else if (sc->flag & SCONV_TO_UTF8) {
2942 unparse = unicode_to_utf8;
2943 if (sc->flag & SCONV_FROM_UTF8)
2944 always_replace = 0;
2945 } else {
2946 /*
2947 * This case is going to be converted to another
2948 * character-set through iconv.
2949 */
2950 always_replace = 0;
2951 if (sc->flag & SCONV_FROM_UTF16BE) {
2952 unparse = unicode_to_utf16be;
2953 ts = 2;
2954 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2955 unparse = unicode_to_utf16le;
2956 ts = 2;
2957 } else {
2958 unparse = unicode_to_utf8;
2959 }
2960 }
2961
2962 if (sc->flag & SCONV_FROM_UTF16BE) {
2963 parse = utf16be_to_unicode;
2964 tm = 1;
2965 spair = 4;/* surrogate pair size in UTF-16. */
2966 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2967 parse = utf16le_to_unicode;
2968 tm = 1;
2969 spair = 4;/* surrogate pair size in UTF-16. */
2970 } else {
2971 parse = cesu8_to_unicode;
2972 tm = ts;
2973 spair = 6;/* surrogate pair size in UTF-8. */
2974 }
2975
2976 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2977 return (-1);
2978
2979 p = as->s + as->length;
2980 endp = as->s + as->buffer_length - ts;
2981 while ((n = parse(&uc, s, len)) != 0) {
2982 const char *ucptr, *uc2ptr;
2983
2984 if (n < 0) {
2985 /* Use a replaced unicode character. */
2986 UNPARSE(p, endp, uc);
2987 s += n*-1;
2988 len -= n*-1;
2989 ret = -1;
2990 continue;
2991 } else if (n == spair || always_replace)
2992 /* uc is converted from a surrogate pair.
2993 * this should be treated as a changed code. */
2994 ucptr = NULL;
2995 else
2996 ucptr = s;
2997 s += n;
2998 len -= n;
2999
3000 /* Read second code point. */
3001 while ((n2 = parse(&uc2, s, len)) > 0) {
3002 uint32_t ucx[FDC_MAX];
3003 int ccx[FDC_MAX];
3004 int cl, cx, i, nx, ucx_size;
3005 int LIndex,SIndex;
3006 uint32_t nfc;
3007
3008 if (n2 == spair || always_replace)
3009 /* uc2 is converted from a surrogate pair.
3010 * this should be treated as a changed code. */
3011 uc2ptr = NULL;
3012 else
3013 uc2ptr = s;
3014 s += n2;
3015 len -= n2;
3016
3017 /*
3018 * If current second code point is out of decomposable
3019 * code points, finding compositions is unneeded.
3020 */
3021 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3022 WRITE_UC();
3023 REPLACE_UC_WITH_UC2();
3024 continue;
3025 }
3026
3027 /*
3028 * Try to combine current code points.
3029 */
3030 /*
3031 * We have to combine Hangul characters according to
3032 * http://uniicode.org/reports/tr15/#Hangul
3033 */
3034 if (0 <= (LIndex = uc - HC_LBASE) &&
3035 LIndex < HC_LCOUNT) {
3036 /*
3037 * Hangul Composition.
3038 * 1. Two current code points are L and V.
3039 */
3040 int VIndex = uc2 - HC_VBASE;
3041 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3042 /* Make syllable of form LV. */
3043 UPDATE_UC(HC_SBASE +
3044 (LIndex * HC_VCOUNT + VIndex) *
3045 HC_TCOUNT);
3046 } else {
3047 WRITE_UC();
3048 REPLACE_UC_WITH_UC2();
3049 }
3050 continue;
3051 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3052 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3053 /*
3054 * Hangul Composition.
3055 * 2. Two current code points are LV and T.
3056 */
3057 int TIndex = uc2 - HC_TBASE;
3058 if (0 < TIndex && TIndex < HC_TCOUNT) {
3059 /* Make syllable of form LVT. */
3060 UPDATE_UC(uc + TIndex);
3061 } else {
3062 WRITE_UC();
3063 REPLACE_UC_WITH_UC2();
3064 }
3065 continue;
3066 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3067 /* A composition to current code points
3068 * is found. */
3069 UPDATE_UC(nfc);
3070 continue;
3071 } else if ((cl = CCC(uc2)) == 0) {
3072 /* Clearly 'uc2' the second code point is not
3073 * a decomposable code. */
3074 WRITE_UC();
3075 REPLACE_UC_WITH_UC2();
3076 continue;
3077 }
3078
3079 /*
3080 * Collect following decomposable code points.
3081 */
3082 cx = 0;
3083 ucx[0] = uc2;
3084 ccx[0] = cl;
3085 COLLECT_CPS(1);
3086
3087 /*
3088 * Find a composed code in the collected code points.
3089 */
3090 i = 1;
3091 while (i < ucx_size) {
3092 int j;
3093
3094 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3095 i++;
3096 continue;
3097 }
3098
3099 /*
3100 * nfc is composed of uc and ucx[i].
3101 */
3102 UPDATE_UC(nfc);
3103
3104 /*
3105 * Remove ucx[i] by shifting
3106 * following code points.
3107 */
3108 for (j = i; j+1 < ucx_size; j++) {
3109 ucx[j] = ucx[j+1];
3110 ccx[j] = ccx[j+1];
3111 }
3112 ucx_size --;
3113
3114 /*
3115 * Collect following code points blocked
3116 * by ucx[i] the removed code point.
3117 */
3118 if (ucx_size > 0 && i == ucx_size &&
3119 nx > 0 && cx == cl) {
3120 cl = ccx[ucx_size-1];
3121 COLLECT_CPS(ucx_size);
3122 }
3123 /*
3124 * Restart finding a composed code with
3125 * the updated uc from the top of the
3126 * collected code points.
3127 */
3128 i = 0;
3129 }
3130
3131 /*
3132 * Apparently the current code points are not
3133 * decomposed characters or already composed.
3134 */
3135 WRITE_UC();
3136 for (i = 0; i < ucx_size; i++)
3137 UNPARSE(p, endp, ucx[i]);
3138
3139 /*
3140 * Flush out remaining canonical combining characters.
3141 */
3142 if (nx > 0 && cx == cl && len > 0) {
3143 while ((nx = parse(&ucx[0], s, len))
3144 > 0) {
3145 cx = CCC(ucx[0]);
3146 if (cl > cx)
3147 break;
3148 s += nx;
3149 len -= nx;
3150 cl = cx;
3151 UNPARSE(p, endp, ucx[0]);
3152 }
3153 }
3154 break;
3155 }
3156 if (n2 < 0) {
3157 WRITE_UC();
3158 /* Use a replaced unicode character. */
3159 UNPARSE(p, endp, uc2);
3160 s += n2*-1;
3161 len -= n2*-1;
3162 ret = -1;
3163 continue;
3164 } else if (n2 == 0) {
3165 WRITE_UC();
3166 break;
3167 }
3168 }
3169 as->length = p - as->s;
3170 as->s[as->length] = '\0';
3171 if (ts == 2)
3172 as->s[as->length+1] = '\0';
3173 return (ret);
3174 }
3175
3176 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3177 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3178 {
3179 int t, b;
3180
3181 /*
3182 * These are not converted to NFD on Mac OS.
3183 */
3184 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3185 (uc >= 0xF900 && uc <= 0xFAFF) ||
3186 (uc >= 0x2F800 && uc <= 0x2FAFF))
3187 return (0);
3188 /*
3189 * Those code points are not converted to NFD on Mac OS.
3190 * I do not know the reason because it is undocumented.
3191 * NFC NFD
3192 * 1109A ==> 11099 110BA
3193 * 1109C ==> 1109B 110BA
3194 * 110AB ==> 110A5 110BA
3195 */
3196 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3197 return (0);
3198
3199 t = 0;
3200 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3201 while (b >= t) {
3202 int m = (t + b) / 2;
3203 if (u_decomposition_table[m].nfc < uc)
3204 t = m + 1;
3205 else if (u_decomposition_table[m].nfc > uc)
3206 b = m - 1;
3207 else {
3208 *cp1 = u_decomposition_table[m].cp1;
3209 *cp2 = u_decomposition_table[m].cp2;
3210 return (1);
3211 }
3212 }
3213 return (0);
3214 }
3215
3216 #define REPLACE_UC_WITH(cp) do { \
3217 uc = cp; \
3218 ucptr = NULL; \
3219 } while (0)
3220
3221 /*
3222 * Normalize UTF-8 characters to Form D and copy the result.
3223 */
3224 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3225 archive_string_normalize_D(struct archive_string *as, const void *_p,
3226 size_t len, struct archive_string_conv *sc)
3227 {
3228 const char *s = (const char *)_p;
3229 char *p, *endp;
3230 uint32_t uc, uc2;
3231 size_t w;
3232 int always_replace, n, n2, ret = 0, spair, ts, tm;
3233 int (*parse)(uint32_t *, const char *, size_t);
3234 size_t (*unparse)(char *, size_t, uint32_t);
3235
3236 always_replace = 1;
3237 ts = 1;/* text size. */
3238 if (sc->flag & SCONV_TO_UTF16BE) {
3239 unparse = unicode_to_utf16be;
3240 ts = 2;
3241 if (sc->flag & SCONV_FROM_UTF16BE)
3242 always_replace = 0;
3243 } else if (sc->flag & SCONV_TO_UTF16LE) {
3244 unparse = unicode_to_utf16le;
3245 ts = 2;
3246 if (sc->flag & SCONV_FROM_UTF16LE)
3247 always_replace = 0;
3248 } else if (sc->flag & SCONV_TO_UTF8) {
3249 unparse = unicode_to_utf8;
3250 if (sc->flag & SCONV_FROM_UTF8)
3251 always_replace = 0;
3252 } else {
3253 /*
3254 * This case is going to be converted to another
3255 * character-set through iconv.
3256 */
3257 always_replace = 0;
3258 if (sc->flag & SCONV_FROM_UTF16BE) {
3259 unparse = unicode_to_utf16be;
3260 ts = 2;
3261 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3262 unparse = unicode_to_utf16le;
3263 ts = 2;
3264 } else {
3265 unparse = unicode_to_utf8;
3266 }
3267 }
3268
3269 if (sc->flag & SCONV_FROM_UTF16BE) {
3270 parse = utf16be_to_unicode;
3271 tm = 1;
3272 spair = 4;/* surrogate pair size in UTF-16. */
3273 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3274 parse = utf16le_to_unicode;
3275 tm = 1;
3276 spair = 4;/* surrogate pair size in UTF-16. */
3277 } else {
3278 parse = cesu8_to_unicode;
3279 tm = ts;
3280 spair = 6;/* surrogate pair size in UTF-8. */
3281 }
3282
3283 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3284 return (-1);
3285
3286 p = as->s + as->length;
3287 endp = as->s + as->buffer_length - ts;
3288 while ((n = parse(&uc, s, len)) != 0) {
3289 const char *ucptr;
3290 uint32_t cp1, cp2;
3291 int SIndex;
3292 struct {
3293 uint32_t uc;
3294 int ccc;
3295 } fdc[FDC_MAX];
3296 int fdi, fdj;
3297 int ccc;
3298
3299 check_first_code:
3300 if (n < 0) {
3301 /* Use a replaced unicode character. */
3302 UNPARSE(p, endp, uc);
3303 s += n*-1;
3304 len -= n*-1;
3305 ret = -1;
3306 continue;
3307 } else if (n == spair || always_replace)
3308 /* uc is converted from a surrogate pair.
3309 * this should be treated as a changed code. */
3310 ucptr = NULL;
3311 else
3312 ucptr = s;
3313 s += n;
3314 len -= n;
3315
3316 /* Hangul Decomposition. */
3317 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3318 int L = HC_LBASE + SIndex / HC_NCOUNT;
3319 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3320 int T = HC_TBASE + SIndex % HC_TCOUNT;
3321
3322 REPLACE_UC_WITH(L);
3323 WRITE_UC();
3324 REPLACE_UC_WITH(V);
3325 WRITE_UC();
3326 if (T != HC_TBASE) {
3327 REPLACE_UC_WITH(T);
3328 WRITE_UC();
3329 }
3330 continue;
3331 }
3332 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3333 WRITE_UC();
3334 continue;
3335 }
3336
3337 fdi = 0;
3338 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3339 int k;
3340
3341 for (k = fdi; k > 0; k--)
3342 fdc[k] = fdc[k-1];
3343 fdc[0].ccc = CCC(cp2);
3344 fdc[0].uc = cp2;
3345 fdi++;
3346 REPLACE_UC_WITH(cp1);
3347 }
3348
3349 /* Read following code points. */
3350 while ((n2 = parse(&uc2, s, len)) > 0 &&
3351 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3352 int j, k;
3353
3354 s += n2;
3355 len -= n2;
3356 for (j = 0; j < fdi; j++) {
3357 if (fdc[j].ccc > ccc)
3358 break;
3359 }
3360 if (j < fdi) {
3361 for (k = fdi; k > j; k--)
3362 fdc[k] = fdc[k-1];
3363 fdc[j].ccc = ccc;
3364 fdc[j].uc = uc2;
3365 } else {
3366 fdc[fdi].ccc = ccc;
3367 fdc[fdi].uc = uc2;
3368 }
3369 fdi++;
3370 }
3371
3372 WRITE_UC();
3373 for (fdj = 0; fdj < fdi; fdj++) {
3374 REPLACE_UC_WITH(fdc[fdj].uc);
3375 WRITE_UC();
3376 }
3377
3378 if (n2 == 0)
3379 break;
3380 REPLACE_UC_WITH(uc2);
3381 n = n2;
3382 goto check_first_code;
3383 }
3384 as->length = p - as->s;
3385 as->s[as->length] = '\0';
3386 if (ts == 2)
3387 as->s[as->length+1] = '\0';
3388 return (ret);
3389 }
3390
3391 /*
3392 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3393 * that WCS is Unicode. It is true for several platforms but some are false.
3394 * And then people who did not use UTF-8 locale on the non Unicode WCS
3395 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3396 * now cannot get right filename from libarchive 3.x and later since we
3397 * fixed the wrong assumption and it is incompatible to older its versions.
3398 * So we provide special option, "compat-2x.x", for resolving it.
3399 * That option enable the string conversion of libarchive 2.x.
3400 *
3401 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3402 * locale character set and appends to the archive_string.
3403 * Note: returns -1 if conversion fails.
3404 */
3405 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3406 strncat_from_utf8_libarchive2(struct archive_string *as,
3407 const void *_p, size_t len, struct archive_string_conv *sc)
3408 {
3409 const char *s;
3410 int n;
3411 char *p;
3412 char *end;
3413 uint32_t unicode;
3414 #if HAVE_WCRTOMB
3415 mbstate_t shift_state;
3416
3417 memset(&shift_state, 0, sizeof(shift_state));
3418 #else
3419 /* Clear the shift state before starting. */
3420 wctomb(NULL, L'\0');
3421 #endif
3422 (void)sc; /* UNUSED */
3423 /*
3424 * Allocate buffer for MBS.
3425 * We need this allocation here since it is possible that
3426 * as->s is still NULL.
3427 */
3428 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3429 return (-1);
3430
3431 s = (const char *)_p;
3432 p = as->s + as->length;
3433 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3434 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3435 wchar_t wc;
3436
3437 if (p >= end) {
3438 as->length = p - as->s;
3439 /* Re-allocate buffer for MBS. */
3440 if (archive_string_ensure(as,
3441 as->length + len * 2 + 1) == NULL)
3442 return (-1);
3443 p = as->s + as->length;
3444 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3445 }
3446
3447 /*
3448 * As libarchive 2.x, translates the UTF-8 characters into
3449 * wide-characters in the assumption that WCS is Unicode.
3450 */
3451 if (n < 0) {
3452 n *= -1;
3453 wc = L'?';
3454 } else
3455 wc = (wchar_t)unicode;
3456
3457 s += n;
3458 len -= n;
3459 /*
3460 * Translates the wide-character into the current locale MBS.
3461 */
3462 #if HAVE_WCRTOMB
3463 n = (int)wcrtomb(p, wc, &shift_state);
3464 #else
3465 n = (int)wctomb(p, wc);
3466 #endif
3467 if (n == -1)
3468 return (-1);
3469 p += n;
3470 }
3471 as->length = p - as->s;
3472 as->s[as->length] = '\0';
3473 return (0);
3474 }
3475
3476
3477 /*
3478 * Conversion functions between current locale dependent MBS and UTF-16BE.
3479 * strncat_from_utf16be() : UTF-16BE --> MBS
3480 * strncat_to_utf16be() : MBS --> UTF16BE
3481 */
3482
3483 #if defined(_WIN32) && !defined(__CYGWIN__)
3484
3485 /*
3486 * Convert a UTF-16BE/LE string to current locale and copy the result.
3487 * Return -1 if conversion fails.
3488 */
3489 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3490 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3491 struct archive_string_conv *sc, int be)
3492 {
3493 struct archive_string tmp;
3494 const char *u16;
3495 int ll;
3496 BOOL defchar;
3497 char *mbs;
3498 size_t mbs_size, b;
3499 int ret = 0;
3500
3501 bytes &= ~1;
3502 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3503 return (-1);
3504
3505 mbs = as->s + as->length;
3506 mbs_size = as->buffer_length - as->length -1;
3507
3508 if (sc->to_cp == CP_C_LOCALE) {
3509 /*
3510 * "C" locale special process.
3511 */
3512 u16 = _p;
3513 ll = 0;
3514 for (b = 0; b < bytes; b += 2) {
3515 uint16_t val;
3516 if (be)
3517 val = archive_be16dec(u16+b);
3518 else
3519 val = archive_le16dec(u16+b);
3520 if (val > 255) {
3521 *mbs++ = '?';
3522 ret = -1;
3523 } else
3524 *mbs++ = (char)(val&0xff);
3525 ll++;
3526 }
3527 as->length += ll;
3528 as->s[as->length] = '\0';
3529 return (ret);
3530 }
3531
3532 archive_string_init(&tmp);
3533 if (be) {
3534 if (is_big_endian()) {
3535 u16 = _p;
3536 } else {
3537 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3538 return (-1);
3539 memcpy(tmp.s, _p, bytes);
3540 for (b = 0; b < bytes; b += 2) {
3541 uint16_t val = archive_be16dec(tmp.s+b);
3542 archive_le16enc(tmp.s+b, val);
3543 }
3544 u16 = tmp.s;
3545 }
3546 } else {
3547 if (!is_big_endian()) {
3548 u16 = _p;
3549 } else {
3550 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3551 return (-1);
3552 memcpy(tmp.s, _p, bytes);
3553 for (b = 0; b < bytes; b += 2) {
3554 uint16_t val = archive_le16dec(tmp.s+b);
3555 archive_be16enc(tmp.s+b, val);
3556 }
3557 u16 = tmp.s;
3558 }
3559 }
3560
3561 do {
3562 defchar = 0;
3563 ll = WideCharToMultiByte(sc->to_cp, 0,
3564 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3565 NULL, &defchar);
3566 /* Exit loop if we succeeded */
3567 if (ll != 0 ||
3568 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3569 break;
3570 }
3571 /* Else expand buffer and loop to try again. */
3572 ll = WideCharToMultiByte(sc->to_cp, 0,
3573 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3574 if (archive_string_ensure(as, ll +1) == NULL)
3575 return (-1);
3576 mbs = as->s + as->length;
3577 mbs_size = as->buffer_length - as->length -1;
3578 } while (1);
3579 archive_string_free(&tmp);
3580 as->length += ll;
3581 as->s[as->length] = '\0';
3582 if (ll == 0 || defchar)
3583 ret = -1;
3584 return (ret);
3585 }
3586
3587 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3588 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3589 size_t bytes, struct archive_string_conv *sc)
3590 {
3591 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3592 }
3593
3594 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3595 win_strncat_from_utf16le(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, 0));
3599 }
3600
3601 static int
is_big_endian(void)3602 is_big_endian(void)
3603 {
3604 uint16_t d = 1;
3605
3606 return (archive_be16dec(&d) == 1);
3607 }
3608
3609 /*
3610 * Convert a current locale string to UTF-16BE/LE and copy the result.
3611 * Return -1 if conversion fails.
3612 */
3613 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3614 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3615 size_t length, struct archive_string_conv *sc, int bigendian)
3616 {
3617 const char *s = (const char *)_p;
3618 char *u16;
3619 size_t count, avail;
3620
3621 if (archive_string_ensure(as16,
3622 as16->length + (length + 1) * 2) == NULL)
3623 return (-1);
3624
3625 u16 = as16->s + as16->length;
3626 avail = as16->buffer_length - 2;
3627 if (sc->from_cp == CP_C_LOCALE) {
3628 /*
3629 * "C" locale special process.
3630 */
3631 count = 0;
3632 while (count < length && *s) {
3633 if (bigendian)
3634 archive_be16enc(u16, *s);
3635 else
3636 archive_le16enc(u16, *s);
3637 u16 += 2;
3638 s++;
3639 count++;
3640 }
3641 as16->length += count << 1;
3642 as16->s[as16->length] = 0;
3643 as16->s[as16->length+1] = 0;
3644 return (0);
3645 }
3646 do {
3647 count = MultiByteToWideChar(sc->from_cp,
3648 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3649 /* Exit loop if we succeeded */
3650 if (count != 0 ||
3651 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3652 break;
3653 }
3654 /* Expand buffer and try again */
3655 count = MultiByteToWideChar(sc->from_cp,
3656 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3657 if (archive_string_ensure(as16, (count +1) * 2)
3658 == NULL)
3659 return (-1);
3660 u16 = as16->s + as16->length;
3661 avail = as16->buffer_length - 2;
3662 } while (1);
3663 as16->length += count * 2;
3664 as16->s[as16->length] = 0;
3665 as16->s[as16->length+1] = 0;
3666 if (count == 0)
3667 return (-1);
3668
3669 if (is_big_endian()) {
3670 if (!bigendian) {
3671 while (count > 0) {
3672 uint16_t v = archive_be16dec(u16);
3673 archive_le16enc(u16, v);
3674 u16 += 2;
3675 count--;
3676 }
3677 }
3678 } else {
3679 if (bigendian) {
3680 while (count > 0) {
3681 uint16_t v = archive_le16dec(u16);
3682 archive_be16enc(u16, v);
3683 u16 += 2;
3684 count--;
3685 }
3686 }
3687 }
3688 return (0);
3689 }
3690
3691 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3692 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3693 size_t length, struct archive_string_conv *sc)
3694 {
3695 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3696 }
3697
3698 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3699 win_strncat_to_utf16le(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, 0));
3703 }
3704
3705 #endif /* _WIN32 && !__CYGWIN__ */
3706
3707 /*
3708 * Do the best effort for conversions.
3709 * We cannot handle UTF-16BE character-set without such iconv,
3710 * but there is a chance if a string consists just ASCII code or
3711 * a current locale is UTF-8.
3712 */
3713
3714 /*
3715 * Convert a UTF-16BE string to current locale and copy the result.
3716 * Return -1 if conversion fails.
3717 */
3718 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3719 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3720 size_t bytes, struct archive_string_conv *sc, int be)
3721 {
3722 const char *utf16 = (const char *)_p;
3723 char *mbs;
3724 uint32_t uc;
3725 int n, ret;
3726
3727 (void)sc; /* UNUSED */
3728 /*
3729 * Other case, we should do the best effort.
3730 * If all character are ASCII(<0x7f), we can convert it.
3731 * if not , we set a alternative character and return -1.
3732 */
3733 ret = 0;
3734 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3735 return (-1);
3736 mbs = as->s + as->length;
3737
3738 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3739 if (n < 0) {
3740 n *= -1;
3741 ret = -1;
3742 }
3743 bytes -= n;
3744 utf16 += n;
3745
3746 if (uc > 127) {
3747 /* We cannot handle it. */
3748 *mbs++ = '?';
3749 ret = -1;
3750 } else
3751 *mbs++ = (char)uc;
3752 }
3753 as->length = mbs - as->s;
3754 as->s[as->length] = '\0';
3755 return (ret);
3756 }
3757
3758 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3759 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3760 size_t bytes, struct archive_string_conv *sc)
3761 {
3762 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3763 }
3764
3765 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3766 best_effort_strncat_from_utf16le(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, 0));
3770 }
3771
3772 /*
3773 * Convert a current locale string to UTF-16BE/LE and copy the result.
3774 * Return -1 if conversion fails.
3775 */
3776 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3777 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3778 size_t length, struct archive_string_conv *sc, int bigendian)
3779 {
3780 const char *s = (const char *)_p;
3781 char *utf16;
3782 size_t remaining;
3783 int ret;
3784
3785 (void)sc; /* UNUSED */
3786 /*
3787 * Other case, we should do the best effort.
3788 * If all character are ASCII(<0x7f), we can convert it.
3789 * if not , we set a alternative character and return -1.
3790 */
3791 ret = 0;
3792 remaining = length;
3793
3794 if (archive_string_ensure(as16,
3795 as16->length + (length + 1) * 2) == NULL)
3796 return (-1);
3797
3798 utf16 = as16->s + as16->length;
3799 while (remaining--) {
3800 unsigned c = *s++;
3801 if (c > 127) {
3802 /* We cannot handle it. */
3803 c = UNICODE_R_CHAR;
3804 ret = -1;
3805 }
3806 if (bigendian)
3807 archive_be16enc(utf16, c);
3808 else
3809 archive_le16enc(utf16, c);
3810 utf16 += 2;
3811 }
3812 as16->length = utf16 - as16->s;
3813 as16->s[as16->length] = 0;
3814 as16->s[as16->length+1] = 0;
3815 return (ret);
3816 }
3817
3818 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3819 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3820 size_t length, struct archive_string_conv *sc)
3821 {
3822 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3823 }
3824
3825 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3826 best_effort_strncat_to_utf16le(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, 0));
3830 }
3831
3832
3833 /*
3834 * Multistring operations.
3835 */
3836
3837 void
archive_mstring_clean(struct archive_mstring * aes)3838 archive_mstring_clean(struct archive_mstring *aes)
3839 {
3840 archive_wstring_free(&(aes->aes_wcs));
3841 archive_string_free(&(aes->aes_mbs));
3842 archive_string_free(&(aes->aes_utf8));
3843 archive_string_free(&(aes->aes_mbs_in_locale));
3844 aes->aes_set = 0;
3845 }
3846
3847 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3848 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3849 {
3850 dest->aes_set = src->aes_set;
3851 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3852 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3853 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3854 }
3855
3856 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3857 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3858 const char **p)
3859 {
3860 struct archive_string_conv *sc;
3861 int r;
3862
3863 /* If we already have a UTF8 form, return that immediately. */
3864 if (aes->aes_set & AES_SET_UTF8) {
3865 *p = aes->aes_utf8.s;
3866 return (0);
3867 }
3868
3869 *p = NULL;
3870 if (aes->aes_set & AES_SET_MBS) {
3871 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3872 if (sc == NULL)
3873 return (-1);/* Couldn't allocate memory for sc. */
3874 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3875 aes->aes_mbs.length, sc);
3876 if (a == NULL)
3877 free_sconv_object(sc);
3878 if (r == 0) {
3879 aes->aes_set |= AES_SET_UTF8;
3880 *p = aes->aes_utf8.s;
3881 return (0);/* success. */
3882 } else
3883 return (-1);/* failure. */
3884 }
3885 return (0);/* success. */
3886 }
3887
3888 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3889 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3890 const char **p)
3891 {
3892 int r, ret = 0;
3893
3894 (void)a; /* UNUSED */
3895 /* If we already have an MBS form, return that immediately. */
3896 if (aes->aes_set & AES_SET_MBS) {
3897 *p = aes->aes_mbs.s;
3898 return (ret);
3899 }
3900
3901 *p = NULL;
3902 /* If there's a WCS form, try converting with the native locale. */
3903 if (aes->aes_set & AES_SET_WCS) {
3904 archive_string_empty(&(aes->aes_mbs));
3905 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3906 aes->aes_wcs.s, aes->aes_wcs.length);
3907 *p = aes->aes_mbs.s;
3908 if (r == 0) {
3909 aes->aes_set |= AES_SET_MBS;
3910 return (ret);
3911 } else
3912 ret = -1;
3913 }
3914
3915 /*
3916 * Only a UTF-8 form cannot avail because its conversion already
3917 * failed at archive_mstring_update_utf8().
3918 */
3919 return (ret);
3920 }
3921
3922 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)3923 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3924 const wchar_t **wp)
3925 {
3926 int r, ret = 0;
3927
3928 (void)a;/* UNUSED */
3929 /* Return WCS form if we already have it. */
3930 if (aes->aes_set & AES_SET_WCS) {
3931 *wp = aes->aes_wcs.s;
3932 return (ret);
3933 }
3934
3935 *wp = NULL;
3936 /* Try converting MBS to WCS using native locale. */
3937 if (aes->aes_set & AES_SET_MBS) {
3938 archive_wstring_empty(&(aes->aes_wcs));
3939 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
3940 aes->aes_mbs.s, aes->aes_mbs.length);
3941 if (r == 0) {
3942 aes->aes_set |= AES_SET_WCS;
3943 *wp = aes->aes_wcs.s;
3944 } else
3945 ret = -1;/* failure. */
3946 }
3947 return (ret);
3948 }
3949
3950 int
archive_mstring_get_mbs_l(struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)3951 archive_mstring_get_mbs_l(struct archive_mstring *aes,
3952 const char **p, size_t *length, struct archive_string_conv *sc)
3953 {
3954 int r, ret = 0;
3955
3956 #if defined(_WIN32) && !defined(__CYGWIN__)
3957 /*
3958 * Internationalization programming on Windows must use Wide
3959 * characters because Windows platform cannot make locale UTF-8.
3960 */
3961 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
3962 archive_string_empty(&(aes->aes_mbs_in_locale));
3963 r = archive_string_append_from_wcs_in_codepage(
3964 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
3965 aes->aes_wcs.length, sc);
3966 if (r == 0) {
3967 *p = aes->aes_mbs_in_locale.s;
3968 if (length != NULL)
3969 *length = aes->aes_mbs_in_locale.length;
3970 return (0);
3971 } else if (errno == ENOMEM)
3972 return (-1);
3973 else
3974 ret = -1;
3975 }
3976 #endif
3977
3978 /* If there is not an MBS form but is a WCS form, try converting
3979 * with the native locale to be used for translating it to specified
3980 * character-set. */
3981 if ((aes->aes_set & AES_SET_MBS) == 0 &&
3982 (aes->aes_set & AES_SET_WCS) != 0) {
3983 archive_string_empty(&(aes->aes_mbs));
3984 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3985 aes->aes_wcs.s, aes->aes_wcs.length);
3986 if (r == 0)
3987 aes->aes_set |= AES_SET_MBS;
3988 else if (errno == ENOMEM)
3989 return (-1);
3990 else
3991 ret = -1;
3992 }
3993 /* If we already have an MBS form, use it to be translated to
3994 * specified character-set. */
3995 if (aes->aes_set & AES_SET_MBS) {
3996 if (sc == NULL) {
3997 /* Conversion is unneeded. */
3998 *p = aes->aes_mbs.s;
3999 if (length != NULL)
4000 *length = aes->aes_mbs.length;
4001 return (0);
4002 }
4003 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4004 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4005 *p = aes->aes_mbs_in_locale.s;
4006 if (length != NULL)
4007 *length = aes->aes_mbs_in_locale.length;
4008 } else {
4009 *p = NULL;
4010 if (length != NULL)
4011 *length = 0;
4012 }
4013 return (ret);
4014 }
4015
4016 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4017 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4018 {
4019 if (mbs == NULL) {
4020 aes->aes_set = 0;
4021 return (0);
4022 }
4023 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4024 }
4025
4026 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4027 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4028 size_t len)
4029 {
4030 if (mbs == NULL) {
4031 aes->aes_set = 0;
4032 return (0);
4033 }
4034 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4035 archive_strncpy(&(aes->aes_mbs), mbs, len);
4036 archive_string_empty(&(aes->aes_utf8));
4037 archive_wstring_empty(&(aes->aes_wcs));
4038 return (0);
4039 }
4040
4041 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4042 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4043 {
4044 return archive_mstring_copy_wcs_len(aes, wcs,
4045 wcs == NULL ? 0 : wcslen(wcs));
4046 }
4047
4048 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4049 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4050 {
4051 if (utf8 == NULL) {
4052 aes->aes_set = 0;
4053 }
4054 aes->aes_set = AES_SET_UTF8;
4055 archive_string_empty(&(aes->aes_mbs));
4056 archive_string_empty(&(aes->aes_wcs));
4057 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4058 return (int)strlen(utf8);
4059 }
4060
4061 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4062 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4063 size_t len)
4064 {
4065 if (wcs == NULL) {
4066 aes->aes_set = 0;
4067 }
4068 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4069 archive_string_empty(&(aes->aes_mbs));
4070 archive_string_empty(&(aes->aes_utf8));
4071 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4072 return (0);
4073 }
4074
4075 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4076 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4077 const char *mbs, size_t len, struct archive_string_conv *sc)
4078 {
4079 int r;
4080
4081 if (mbs == NULL) {
4082 aes->aes_set = 0;
4083 return (0);
4084 }
4085 archive_string_empty(&(aes->aes_mbs));
4086 archive_wstring_empty(&(aes->aes_wcs));
4087 archive_string_empty(&(aes->aes_utf8));
4088 #if defined(_WIN32) && !defined(__CYGWIN__)
4089 /*
4090 * Internationalization programming on Windows must use Wide
4091 * characters because Windows platform cannot make locale UTF-8.
4092 */
4093 if (sc == NULL) {
4094 if (archive_string_append(&(aes->aes_mbs),
4095 mbs, mbsnbytes(mbs, len)) == NULL) {
4096 aes->aes_set = 0;
4097 r = -1;
4098 } else {
4099 aes->aes_set = AES_SET_MBS;
4100 r = 0;
4101 }
4102 #if defined(HAVE_ICONV)
4103 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4104 /*
4105 * This case happens only when MultiByteToWideChar() cannot
4106 * handle sc->from_cp, and we have to iconv in order to
4107 * translate character-set to wchar_t,UTF-16.
4108 */
4109 iconv_t cd = sc->cd;
4110 unsigned from_cp;
4111 int flag;
4112
4113 /*
4114 * Translate multi-bytes from some character-set to UTF-8.
4115 */
4116 sc->cd = sc->cd_w;
4117 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4118 sc->cd = cd;
4119 if (r != 0) {
4120 aes->aes_set = 0;
4121 return (r);
4122 }
4123 aes->aes_set = AES_SET_UTF8;
4124
4125 /*
4126 * Append the UTF-8 string into wstring.
4127 */
4128 flag = sc->flag;
4129 sc->flag &= ~(SCONV_NORMALIZATION_C
4130 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4131 from_cp = sc->from_cp;
4132 sc->from_cp = CP_UTF8;
4133 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4134 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4135 sc->flag = flag;
4136 sc->from_cp = from_cp;
4137 if (r == 0)
4138 aes->aes_set |= AES_SET_WCS;
4139 #endif
4140 } else {
4141 r = archive_wstring_append_from_mbs_in_codepage(
4142 &(aes->aes_wcs), mbs, len, sc);
4143 if (r == 0)
4144 aes->aes_set = AES_SET_WCS;
4145 else
4146 aes->aes_set = 0;
4147 }
4148 #else
4149 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4150 if (r == 0)
4151 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4152 else
4153 aes->aes_set = 0;
4154 #endif
4155 return (r);
4156 }
4157
4158 /*
4159 * The 'update' form tries to proactively update all forms of
4160 * this string (WCS and MBS) and returns an error if any of
4161 * them fail. This is used by the 'pax' handler, for instance,
4162 * to detect and report character-conversion failures early while
4163 * still allowing clients to get potentially useful values from
4164 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4165 * strive to give the user something useful, so you can get hopefully
4166 * usable values even if some of the character conversions are failing.)
4167 */
4168 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4169 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4170 const char *utf8)
4171 {
4172 struct archive_string_conv *sc;
4173 int r;
4174
4175 if (utf8 == NULL) {
4176 aes->aes_set = 0;
4177 return (0); /* Succeeded in clearing everything. */
4178 }
4179
4180 /* Save the UTF8 string. */
4181 archive_strcpy(&(aes->aes_utf8), utf8);
4182
4183 /* Empty the mbs and wcs strings. */
4184 archive_string_empty(&(aes->aes_mbs));
4185 archive_wstring_empty(&(aes->aes_wcs));
4186
4187 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4188
4189 /* Try converting UTF-8 to MBS, return false on failure. */
4190 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4191 if (sc == NULL)
4192 return (-1);/* Couldn't allocate memory for sc. */
4193 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4194 if (a == NULL)
4195 free_sconv_object(sc);
4196 if (r != 0)
4197 return (-1);
4198 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4199
4200 /* Try converting MBS to WCS, return false on failure. */
4201 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4202 aes->aes_mbs.length))
4203 return (-1);
4204 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4205
4206 /* All conversions succeeded. */
4207 return (0);
4208 }
4209