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