1 /*-
2 * Copyright (c) 2003-2011 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27 #include "archive_platform.h"
28 __FBSDID("$FreeBSD: head/lib/libarchive/archive_string.c 201095 2009-12-28 02:33:22Z kientzle $");
29
30 /*
31 * Basic resizable string support, to simplify manipulating arbitrary-sized
32 * strings while minimizing heap activity.
33 *
34 * In particular, the buffer used by a string object is only grown, it
35 * never shrinks, so you can clear and reuse the same string object
36 * without incurring additional memory allocations.
37 */
38
39 #ifdef HAVE_ERRNO_H
40 #include <errno.h>
41 #endif
42 #ifdef HAVE_ICONV_H
43 #include <iconv.h>
44 #endif
45 #ifdef HAVE_LANGINFO_H
46 #include <langinfo.h>
47 #endif
48 #ifdef HAVE_LOCALCHARSET_H
49 #include <localcharset.h>
50 #endif
51 #ifdef HAVE_STDLIB_H
52 #include <stdlib.h>
53 #endif
54 #ifdef HAVE_STRING_H
55 #include <string.h>
56 #endif
57 #ifdef HAVE_WCHAR_H
58 #include <wchar.h>
59 #endif
60 #if defined(_WIN32) && !defined(__CYGWIN__)
61 #include <windows.h>
62 #include <locale.h>
63 #endif
64
65 #include "archive_endian.h"
66 #include "archive_private.h"
67 #include "archive_string.h"
68 #include "archive_string_composition.h"
69
70 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
71 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
72 #endif
73
74 #if !defined(HAVE_WMEMMOVE) && !defined(wmemmove)
75 #define wmemmove(a,b,i) (wchar_t *)memmove((a), (b), (i) * sizeof(wchar_t))
76 #endif
77
78 #undef max
79 #define max(a, b) ((a)>(b)?(a):(b))
80
81 struct archive_string_conv {
82 struct archive_string_conv *next;
83 char *from_charset;
84 char *to_charset;
85 unsigned from_cp;
86 unsigned to_cp;
87 /* Set 1 if from_charset and to_charset are the same. */
88 int same;
89 int flag;
90 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified
91 * charset. */
92 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from
93 * specified charset. */
94 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */
95 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting
96 * MBS. */
97 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive
98 * 2.x in the wrong assumption. */
99 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C.
100 * Before UTF-8 characters are actually
101 * processed. */
102 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D.
103 * Before UTF-8 characters are actually
104 * processed.
105 * Currently this only for MAC OS X. */
106 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */
107 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */
108 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */
109 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */
110 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */
111 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */
112 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
113 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
114
115 #if HAVE_ICONV
116 iconv_t cd;
117 iconv_t cd_w;/* Use at archive_mstring on
118 * Windows. */
119 #endif
120 /* A temporary buffer for normalization. */
121 struct archive_string utftmp;
122 int (*converter[2])(struct archive_string *, const void *, size_t,
123 struct archive_string_conv *);
124 int nconverter;
125 };
126
127 #define CP_C_LOCALE 0 /* "C" locale only for this file. */
128 #define CP_UTF16LE 1200
129 #define CP_UTF16BE 1201
130
131 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
132 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
133 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
134 #define UNICODE_MAX 0x10FFFF
135 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */
136 /* Set U+FFFD(Replacement character) in UTF-8. */
137 static const char utf8_replacement_char[] = {0xef, 0xbf, 0xbd};
138
139 static struct archive_string_conv *find_sconv_object(struct archive *,
140 const char *, const char *);
141 static void add_sconv_object(struct archive *, struct archive_string_conv *);
142 static struct archive_string_conv *create_sconv_object(const char *,
143 const char *, unsigned, int);
144 static void free_sconv_object(struct archive_string_conv *);
145 static struct archive_string_conv *get_sconv_object(struct archive *,
146 const char *, const char *, int);
147 static unsigned make_codepage_from_charset(const char *);
148 static unsigned get_current_codepage(void);
149 static unsigned get_current_oemcp(void);
150 static size_t mbsnbytes(const void *, size_t);
151 static size_t utf16nbytes(const void *, size_t);
152 #if defined(_WIN32) && !defined(__CYGWIN__)
153 static int archive_wstring_append_from_mbs_in_codepage(
154 struct archive_wstring *, const char *, size_t,
155 struct archive_string_conv *);
156 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
157 const wchar_t *, size_t, struct archive_string_conv *);
158 static int is_big_endian(void);
159 static int strncat_in_codepage(struct archive_string *, const void *,
160 size_t, struct archive_string_conv *);
161 static int win_strncat_from_utf16be(struct archive_string *, const void *,
162 size_t, struct archive_string_conv *);
163 static int win_strncat_from_utf16le(struct archive_string *, const void *,
164 size_t, struct archive_string_conv *);
165 static int win_strncat_to_utf16be(struct archive_string *, const void *,
166 size_t, struct archive_string_conv *);
167 static int win_strncat_to_utf16le(struct archive_string *, const void *,
168 size_t, struct archive_string_conv *);
169 #endif
170 static int best_effort_strncat_from_utf16be(struct archive_string *,
171 const void *, size_t, struct archive_string_conv *);
172 static int best_effort_strncat_from_utf16le(struct archive_string *,
173 const void *, size_t, struct archive_string_conv *);
174 static int best_effort_strncat_to_utf16be(struct archive_string *,
175 const void *, size_t, struct archive_string_conv *);
176 static int best_effort_strncat_to_utf16le(struct archive_string *,
177 const void *, size_t, struct archive_string_conv *);
178 #if defined(HAVE_ICONV)
179 static int iconv_strncat_in_locale(struct archive_string *, const void *,
180 size_t, struct archive_string_conv *);
181 #endif
182 static int best_effort_strncat_in_locale(struct archive_string *,
183 const void *, size_t, struct archive_string_conv *);
184 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
185 static int utf8_to_unicode(uint32_t *, const char *, size_t);
186 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
187 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
188 static size_t unicode_to_utf8(char *, size_t, uint32_t);
189 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
190 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
191 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
192 static int strncat_from_utf8_libarchive2(struct archive_string *,
193 const void *, size_t, struct archive_string_conv *);
194 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
195 size_t, struct archive_string_conv *);
196 static int archive_string_normalize_C(struct archive_string *, const void *,
197 size_t, struct archive_string_conv *);
198 static int archive_string_normalize_D(struct archive_string *, const void *,
199 size_t, struct archive_string_conv *);
200 static int archive_string_append_unicode(struct archive_string *,
201 const void *, size_t, struct archive_string_conv *);
202
203 static struct archive_string *
archive_string_append(struct archive_string * as,const char * p,size_t s)204 archive_string_append(struct archive_string *as, const char *p, size_t s)
205 {
206 if (archive_string_ensure(as, as->length + s + 1) == NULL)
207 return (NULL);
208 if (s)
209 memmove(as->s + as->length, p, s);
210 as->length += s;
211 as->s[as->length] = 0;
212 return (as);
213 }
214
215 static struct archive_wstring *
archive_wstring_append(struct archive_wstring * as,const wchar_t * p,size_t s)216 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
217 {
218 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
219 return (NULL);
220 if (s)
221 wmemmove(as->s + as->length, p, s);
222 as->length += s;
223 as->s[as->length] = 0;
224 return (as);
225 }
226
227 struct archive_string *
archive_array_append(struct archive_string * as,const char * p,size_t s)228 archive_array_append(struct archive_string *as, const char *p, size_t s)
229 {
230 return archive_string_append(as, p, s);
231 }
232
233 void
archive_string_concat(struct archive_string * dest,struct archive_string * src)234 archive_string_concat(struct archive_string *dest, struct archive_string *src)
235 {
236 if (archive_string_append(dest, src->s, src->length) == NULL)
237 __archive_errx(1, "Out of memory");
238 }
239
240 void
archive_wstring_concat(struct archive_wstring * dest,struct archive_wstring * src)241 archive_wstring_concat(struct archive_wstring *dest,
242 struct archive_wstring *src)
243 {
244 if (archive_wstring_append(dest, src->s, src->length) == NULL)
245 __archive_errx(1, "Out of memory");
246 }
247
248 void
archive_string_free(struct archive_string * as)249 archive_string_free(struct archive_string *as)
250 {
251 as->length = 0;
252 as->buffer_length = 0;
253 free(as->s);
254 as->s = NULL;
255 }
256
257 void
archive_wstring_free(struct archive_wstring * as)258 archive_wstring_free(struct archive_wstring *as)
259 {
260 as->length = 0;
261 as->buffer_length = 0;
262 free(as->s);
263 as->s = NULL;
264 }
265
266 struct archive_wstring *
archive_wstring_ensure(struct archive_wstring * as,size_t s)267 archive_wstring_ensure(struct archive_wstring *as, size_t s)
268 {
269 return (struct archive_wstring *)
270 archive_string_ensure((struct archive_string *)as,
271 s * sizeof(wchar_t));
272 }
273
274 /* Returns NULL on any allocation failure. */
275 struct archive_string *
archive_string_ensure(struct archive_string * as,size_t s)276 archive_string_ensure(struct archive_string *as, size_t s)
277 {
278 char *p;
279 size_t new_length;
280
281 /* If buffer is already big enough, don't reallocate. */
282 if (as->s && (s <= as->buffer_length))
283 return (as);
284
285 /*
286 * Growing the buffer at least exponentially ensures that
287 * append operations are always linear in the number of
288 * characters appended. Using a smaller growth rate for
289 * larger buffers reduces memory waste somewhat at the cost of
290 * a larger constant factor.
291 */
292 if (as->buffer_length < 32)
293 /* Start with a minimum 32-character buffer. */
294 new_length = 32;
295 else if (as->buffer_length < 8192)
296 /* Buffers under 8k are doubled for speed. */
297 new_length = as->buffer_length + as->buffer_length;
298 else {
299 /* Buffers 8k and over grow by at least 25% each time. */
300 new_length = as->buffer_length + as->buffer_length / 4;
301 /* Be safe: If size wraps, fail. */
302 if (new_length < as->buffer_length) {
303 /* On failure, wipe the string and return NULL. */
304 archive_string_free(as);
305 errno = ENOMEM;/* Make sure errno has ENOMEM. */
306 return (NULL);
307 }
308 }
309 /*
310 * The computation above is a lower limit to how much we'll
311 * grow the buffer. In any case, we have to grow it enough to
312 * hold the request.
313 */
314 if (new_length < s)
315 new_length = s;
316 /* Now we can reallocate the buffer. */
317 p = (char *)realloc(as->s, new_length);
318 if (p == NULL) {
319 /* On failure, wipe the string and return NULL. */
320 archive_string_free(as);
321 errno = ENOMEM;/* Make sure errno has ENOMEM. */
322 return (NULL);
323 }
324
325 as->s = p;
326 as->buffer_length = new_length;
327 return (as);
328 }
329
330 /*
331 * TODO: See if there's a way to avoid scanning
332 * the source string twice. Then test to see
333 * if it actually helps (remember that we're almost
334 * always called with pretty short arguments, so
335 * such an optimization might not help).
336 */
337 struct archive_string *
archive_strncat(struct archive_string * as,const void * _p,size_t n)338 archive_strncat(struct archive_string *as, const void *_p, size_t n)
339 {
340 size_t s;
341 const char *p, *pp;
342
343 p = (const char *)_p;
344
345 /* Like strlen(p), except won't examine positions beyond p[n]. */
346 s = 0;
347 pp = p;
348 while (s < n && *pp) {
349 pp++;
350 s++;
351 }
352 if ((as = archive_string_append(as, p, s)) == NULL)
353 __archive_errx(1, "Out of memory");
354 return (as);
355 }
356
357 struct archive_wstring *
archive_wstrncat(struct archive_wstring * as,const wchar_t * p,size_t n)358 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
359 {
360 size_t s;
361 const wchar_t *pp;
362
363 /* Like strlen(p), except won't examine positions beyond p[n]. */
364 s = 0;
365 pp = p;
366 while (s < n && *pp) {
367 pp++;
368 s++;
369 }
370 if ((as = archive_wstring_append(as, p, s)) == NULL)
371 __archive_errx(1, "Out of memory");
372 return (as);
373 }
374
375 struct archive_string *
archive_strcat(struct archive_string * as,const void * p)376 archive_strcat(struct archive_string *as, const void *p)
377 {
378 /* strcat is just strncat without an effective limit.
379 * Assert that we'll never get called with a source
380 * string over 16MB.
381 * TODO: Review all uses of strcat in the source
382 * and try to replace them with strncat().
383 */
384 return archive_strncat(as, p, 0x1000000);
385 }
386
387 struct archive_wstring *
archive_wstrcat(struct archive_wstring * as,const wchar_t * p)388 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
389 {
390 /* Ditto. */
391 return archive_wstrncat(as, p, 0x1000000);
392 }
393
394 struct archive_string *
archive_strappend_char(struct archive_string * as,char c)395 archive_strappend_char(struct archive_string *as, char c)
396 {
397 if ((as = archive_string_append(as, &c, 1)) == NULL)
398 __archive_errx(1, "Out of memory");
399 return (as);
400 }
401
402 struct archive_wstring *
archive_wstrappend_wchar(struct archive_wstring * as,wchar_t c)403 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
404 {
405 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
406 __archive_errx(1, "Out of memory");
407 return (as);
408 }
409
410 /*
411 * Get the "current character set" name to use with iconv.
412 * On FreeBSD, the empty character set name "" chooses
413 * the correct character encoding for the current locale,
414 * so this isn't necessary.
415 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
416 * on that system, we have to explicitly call nl_langinfo()
417 * to get the right name. Not sure about other platforms.
418 *
419 * NOTE: GNU libiconv does not recognize the character-set name
420 * which some platform nl_langinfo(CODESET) returns, so we should
421 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
422 */
423 static const char *
default_iconv_charset(const char * charset)424 default_iconv_charset(const char *charset) {
425 if (charset != NULL && charset[0] != '\0')
426 return charset;
427 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
428 /* locale_charset() is broken on Mac OS */
429 return locale_charset();
430 #elif HAVE_NL_LANGINFO
431 return nl_langinfo(CODESET);
432 #else
433 return "";
434 #endif
435 }
436
437 #if defined(_WIN32) && !defined(__CYGWIN__)
438
439 /*
440 * Convert MBS to WCS.
441 * Note: returns -1 if conversion fails.
442 */
443 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)444 archive_wstring_append_from_mbs(struct archive_wstring *dest,
445 const char *p, size_t len)
446 {
447 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
448 }
449
450 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)451 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
452 const char *s, size_t length, struct archive_string_conv *sc)
453 {
454 int count, ret = 0;
455 UINT from_cp;
456
457 if (sc != NULL)
458 from_cp = sc->from_cp;
459 else
460 from_cp = get_current_codepage();
461
462 if (from_cp == CP_C_LOCALE) {
463 /*
464 * "C" locale special processing.
465 */
466 wchar_t *ws;
467 const unsigned char *mp;
468
469 if (NULL == archive_wstring_ensure(dest,
470 dest->length + length + 1))
471 return (-1);
472
473 ws = dest->s + dest->length;
474 mp = (const unsigned char *)s;
475 count = 0;
476 while (count < (int)length && *mp) {
477 *ws++ = (wchar_t)*mp++;
478 count++;
479 }
480 } else if (sc != NULL &&
481 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
482 /*
483 * Normalize UTF-8 and UTF-16BE and convert it directly
484 * to UTF-16 as wchar_t.
485 */
486 struct archive_string u16;
487 int saved_flag = sc->flag;/* save current flag. */
488
489 if (is_big_endian())
490 sc->flag |= SCONV_TO_UTF16BE;
491 else
492 sc->flag |= SCONV_TO_UTF16LE;
493
494 if (sc->flag & SCONV_FROM_UTF16) {
495 /*
496 * UTF-16BE/LE NFD ===> UTF-16 NFC
497 * UTF-16BE/LE NFC ===> UTF-16 NFD
498 */
499 count = (int)utf16nbytes(s, length);
500 } else {
501 /*
502 * UTF-8 NFD ===> UTF-16 NFC
503 * UTF-8 NFC ===> UTF-16 NFD
504 */
505 count = (int)mbsnbytes(s, length);
506 }
507 u16.s = (char *)dest->s;
508 u16.length = dest->length << 1;;
509 u16.buffer_length = dest->buffer_length;
510 if (sc->flag & SCONV_NORMALIZATION_C)
511 ret = archive_string_normalize_C(&u16, s, count, sc);
512 else
513 ret = archive_string_normalize_D(&u16, s, count, sc);
514 dest->s = (wchar_t *)u16.s;
515 dest->length = u16.length >> 1;
516 dest->buffer_length = u16.buffer_length;
517 sc->flag = saved_flag;/* restore the saved flag. */
518 return (ret);
519 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
520 count = (int)utf16nbytes(s, length);
521 count >>= 1; /* to be WCS length */
522 /* Allocate memory for WCS. */
523 if (NULL == archive_wstring_ensure(dest,
524 dest->length + count + 1))
525 return (-1);
526 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
527 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
528 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
529 int b;
530 for (b = 0; b < count; b++) {
531 uint16_t val = archive_le16dec(u16+b);
532 archive_be16enc(u16+b, val);
533 }
534 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
535 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
536 int b;
537 for (b = 0; b < count; b++) {
538 uint16_t val = archive_be16dec(u16+b);
539 archive_le16enc(u16+b, val);
540 }
541 }
542 } else {
543 DWORD mbflag;
544 size_t buffsize;
545
546 if (sc == NULL)
547 mbflag = 0;
548 else if (sc->flag & SCONV_FROM_CHARSET) {
549 /* Do not trust the length which comes from
550 * an archive file. */
551 length = mbsnbytes(s, length);
552 mbflag = 0;
553 } else
554 mbflag = MB_PRECOMPOSED;
555
556 buffsize = dest->length + length + 1;
557 do {
558 /* Allocate memory for WCS. */
559 if (NULL == archive_wstring_ensure(dest, buffsize))
560 return (-1);
561 /* Convert MBS to WCS. */
562 count = MultiByteToWideChar(from_cp,
563 mbflag, s, (int)length, dest->s + dest->length,
564 (int)(dest->buffer_length >> 1) -1);
565 if (count == 0 &&
566 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
567 /* Expand the WCS buffer. */
568 buffsize = dest->buffer_length << 1;
569 continue;
570 }
571 if (count == 0 && length != 0)
572 ret = -1;
573 break;
574 } while (1);
575 }
576 dest->length += count;
577 dest->s[dest->length] = L'\0';
578 return (ret);
579 }
580
581 #else
582
583 /*
584 * Convert MBS to WCS.
585 * Note: returns -1 if conversion fails.
586 */
587 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)588 archive_wstring_append_from_mbs(struct archive_wstring *dest,
589 const char *p, size_t len)
590 {
591 size_t r;
592 int ret_val = 0;
593 /*
594 * No single byte will be more than one wide character,
595 * so this length estimate will always be big enough.
596 */
597 // size_t wcs_length = len;
598 size_t mbs_length = len;
599 const char *mbs = p;
600 wchar_t *wcs;
601 #if HAVE_MBRTOWC
602 mbstate_t shift_state;
603
604 memset(&shift_state, 0, sizeof(shift_state));
605 #endif
606 /*
607 * As we decided to have wcs_length == mbs_length == len
608 * we can use len here instead of wcs_length
609 */
610 if (NULL == archive_wstring_ensure(dest, dest->length + len + 1))
611 return (-1);
612 wcs = dest->s + dest->length;
613 /*
614 * We cannot use mbsrtowcs/mbstowcs here because those may convert
615 * extra MBS when strlen(p) > len and one wide character consists of
616 * multi bytes.
617 */
618 while (*mbs && mbs_length > 0) {
619 /*
620 * The buffer we allocated is always big enough.
621 * Keep this code path in a comment if we decide to choose
622 * smaller wcs_length in the future
623 */
624 /*
625 if (wcs_length == 0) {
626 dest->length = wcs - dest->s;
627 dest->s[dest->length] = L'\0';
628 wcs_length = mbs_length;
629 if (NULL == archive_wstring_ensure(dest,
630 dest->length + wcs_length + 1))
631 return (-1);
632 wcs = dest->s + dest->length;
633 }
634 */
635 #if HAVE_MBRTOWC
636 r = mbrtowc(wcs, mbs, mbs_length, &shift_state);
637 #else
638 r = mbtowc(wcs, mbs, mbs_length);
639 #endif
640 if (r == (size_t)-1 || r == (size_t)-2) {
641 ret_val = -1;
642 break;
643 }
644 if (r == 0 || r > mbs_length)
645 break;
646 wcs++;
647 // wcs_length--;
648 mbs += r;
649 mbs_length -= r;
650 }
651 dest->length = wcs - dest->s;
652 dest->s[dest->length] = L'\0';
653 return (ret_val);
654 }
655
656 #endif
657
658 #if defined(_WIN32) && !defined(__CYGWIN__)
659
660 /*
661 * WCS ==> MBS.
662 * Note: returns -1 if conversion fails.
663 *
664 * Win32 builds use WideCharToMultiByte from the Windows API.
665 * (Maybe Cygwin should too? WideCharToMultiByte will know a
666 * lot more about local character encodings than the wcrtomb()
667 * wrapper is going to know.)
668 */
669 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)670 archive_string_append_from_wcs(struct archive_string *as,
671 const wchar_t *w, size_t len)
672 {
673 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
674 }
675
676 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)677 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
678 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
679 {
680 BOOL defchar_used, *dp;
681 int count, ret = 0;
682 UINT to_cp;
683 int wslen = (int)len;
684
685 if (sc != NULL)
686 to_cp = sc->to_cp;
687 else
688 to_cp = get_current_codepage();
689
690 if (to_cp == CP_C_LOCALE) {
691 /*
692 * "C" locale special processing.
693 */
694 const wchar_t *wp = ws;
695 char *p;
696
697 if (NULL == archive_string_ensure(as,
698 as->length + wslen +1))
699 return (-1);
700 p = as->s + as->length;
701 count = 0;
702 defchar_used = 0;
703 while (count < wslen && *wp) {
704 if (*wp > 255) {
705 *p++ = '?';
706 wp++;
707 defchar_used = 1;
708 } else
709 *p++ = (char)*wp++;
710 count++;
711 }
712 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
713 uint16_t *u16;
714
715 if (NULL ==
716 archive_string_ensure(as, as->length + len * 2 + 2))
717 return (-1);
718 u16 = (uint16_t *)(as->s + as->length);
719 count = 0;
720 defchar_used = 0;
721 if (sc->flag & SCONV_TO_UTF16BE) {
722 while (count < (int)len && *ws) {
723 archive_be16enc(u16+count, *ws);
724 ws++;
725 count++;
726 }
727 } else {
728 while (count < (int)len && *ws) {
729 archive_le16enc(u16+count, *ws);
730 ws++;
731 count++;
732 }
733 }
734 count <<= 1; /* to be byte size */
735 } else {
736 /* Make sure the MBS buffer has plenty to set. */
737 if (NULL ==
738 archive_string_ensure(as, as->length + len * 2 + 1))
739 return (-1);
740 do {
741 defchar_used = 0;
742 if (to_cp == CP_UTF8 || sc == NULL)
743 dp = NULL;
744 else
745 dp = &defchar_used;
746 count = WideCharToMultiByte(to_cp, 0, ws, wslen,
747 as->s + as->length,
748 (int)as->buffer_length - (int)as->length - 1, NULL, dp);
749 if (count == 0 &&
750 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
751 /* Expand the MBS buffer and retry. */
752 if (NULL == archive_string_ensure(as,
753 as->buffer_length + len))
754 return (-1);
755 continue;
756 }
757 if (count == 0)
758 ret = -1;
759 break;
760 } while (1);
761 }
762 as->length += count;
763 as->s[as->length] = '\0';
764 return (defchar_used?-1:ret);
765 }
766
767 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
768
769 /*
770 * Translates a wide character string into current locale character set
771 * and appends to the archive_string. Note: returns -1 if conversion
772 * fails.
773 */
774 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)775 archive_string_append_from_wcs(struct archive_string *as,
776 const wchar_t *w, size_t len)
777 {
778 /* We cannot use the standard wcstombs() here because it
779 * cannot tell us how big the output buffer should be. So
780 * I've built a loop around wcrtomb() or wctomb() that
781 * converts a character at a time and resizes the string as
782 * needed. We prefer wcrtomb() when it's available because
783 * it's thread-safe. */
784 int n, ret_val = 0;
785 char *p;
786 char *end;
787 #if HAVE_WCRTOMB
788 mbstate_t shift_state;
789
790 memset(&shift_state, 0, sizeof(shift_state));
791 #else
792 /* Clear the shift state before starting. */
793 wctomb(NULL, L'\0');
794 #endif
795 /*
796 * Allocate buffer for MBS.
797 * We need this allocation here since it is possible that
798 * as->s is still NULL.
799 */
800 if (archive_string_ensure(as, as->length + len + 1) == NULL)
801 return (-1);
802
803 p = as->s + as->length;
804 end = as->s + as->buffer_length - MB_CUR_MAX -1;
805 while (*w != L'\0' && len > 0) {
806 if (p >= end) {
807 as->length = p - as->s;
808 as->s[as->length] = '\0';
809 /* Re-allocate buffer for MBS. */
810 if (archive_string_ensure(as,
811 as->length + max(len * 2,
812 (size_t)MB_CUR_MAX) + 1) == NULL)
813 return (-1);
814 p = as->s + as->length;
815 end = as->s + as->buffer_length - MB_CUR_MAX -1;
816 }
817 #if HAVE_WCRTOMB
818 n = wcrtomb(p, *w++, &shift_state);
819 #else
820 n = wctomb(p, *w++);
821 #endif
822 if (n == -1) {
823 if (errno == EILSEQ) {
824 /* Skip an illegal wide char. */
825 *p++ = '?';
826 ret_val = -1;
827 } else {
828 ret_val = -1;
829 break;
830 }
831 } else
832 p += n;
833 len--;
834 }
835 as->length = p - as->s;
836 as->s[as->length] = '\0';
837 return (ret_val);
838 }
839
840 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
841
842 /*
843 * TODO: Test if __STDC_ISO_10646__ is defined.
844 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
845 * one character at a time. If a non-Windows platform doesn't have
846 * either of these, fall back to the built-in UTF8 conversion.
847 */
848 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)849 archive_string_append_from_wcs(struct archive_string *as,
850 const wchar_t *w, size_t len)
851 {
852 (void)as;/* UNUSED */
853 (void)w;/* UNUSED */
854 (void)len;/* UNUSED */
855 errno = ENOSYS;
856 return (-1);
857 }
858
859 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
860
861 /*
862 * Find a string conversion object by a pair of 'from' charset name
863 * and 'to' charset name from an archive object.
864 * Return NULL if not found.
865 */
866 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)867 find_sconv_object(struct archive *a, const char *fc, const char *tc)
868 {
869 struct archive_string_conv *sc;
870
871 if (a == NULL)
872 return (NULL);
873
874 for (sc = a->sconv; sc != NULL; sc = sc->next) {
875 if (strcmp(sc->from_charset, fc) == 0 &&
876 strcmp(sc->to_charset, tc) == 0)
877 break;
878 }
879 return (sc);
880 }
881
882 /*
883 * Register a string object to an archive object.
884 */
885 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)886 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
887 {
888 struct archive_string_conv **psc;
889
890 /* Add a new sconv to sconv list. */
891 psc = &(a->sconv);
892 while (*psc != NULL)
893 psc = &((*psc)->next);
894 *psc = sc;
895 }
896
897 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))898 add_converter(struct archive_string_conv *sc, int (*converter)
899 (struct archive_string *, const void *, size_t,
900 struct archive_string_conv *))
901 {
902 if (sc == NULL || sc->nconverter >= 2)
903 __archive_errx(1, "Programming error");
904 sc->converter[sc->nconverter++] = converter;
905 }
906
907 static void
setup_converter(struct archive_string_conv * sc)908 setup_converter(struct archive_string_conv *sc)
909 {
910
911 /* Reset. */
912 sc->nconverter = 0;
913
914 /*
915 * Perform special sequence for the incorrect UTF-8 filenames
916 * made by libarchive2.x.
917 */
918 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
919 add_converter(sc, strncat_from_utf8_libarchive2);
920 return;
921 }
922
923 /*
924 * Convert a string to UTF-16BE/LE.
925 */
926 if (sc->flag & SCONV_TO_UTF16) {
927 /*
928 * If the current locale is UTF-8, we can translate
929 * a UTF-8 string into a UTF-16BE string.
930 */
931 if (sc->flag & SCONV_FROM_UTF8) {
932 add_converter(sc, archive_string_append_unicode);
933 return;
934 }
935
936 #if defined(_WIN32) && !defined(__CYGWIN__)
937 if (sc->flag & SCONV_WIN_CP) {
938 if (sc->flag & SCONV_TO_UTF16BE)
939 add_converter(sc, win_strncat_to_utf16be);
940 else
941 add_converter(sc, win_strncat_to_utf16le);
942 return;
943 }
944 #endif
945
946 #if defined(HAVE_ICONV)
947 if (sc->cd != (iconv_t)-1) {
948 add_converter(sc, iconv_strncat_in_locale);
949 return;
950 }
951 #endif
952
953 if (sc->flag & SCONV_BEST_EFFORT) {
954 if (sc->flag & SCONV_TO_UTF16BE)
955 add_converter(sc,
956 best_effort_strncat_to_utf16be);
957 else
958 add_converter(sc,
959 best_effort_strncat_to_utf16le);
960 } else
961 /* Make sure we have no converter. */
962 sc->nconverter = 0;
963 return;
964 }
965
966 /*
967 * Convert a string from UTF-16BE/LE.
968 */
969 if (sc->flag & SCONV_FROM_UTF16) {
970 /*
971 * At least we should normalize a UTF-16BE string.
972 */
973 if (sc->flag & SCONV_NORMALIZATION_D)
974 add_converter(sc,archive_string_normalize_D);
975 else if (sc->flag & SCONV_NORMALIZATION_C)
976 add_converter(sc, archive_string_normalize_C);
977
978 if (sc->flag & SCONV_TO_UTF8) {
979 /*
980 * If the current locale is UTF-8, we can translate
981 * a UTF-16BE/LE string into a UTF-8 string directly.
982 */
983 if (!(sc->flag &
984 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
985 add_converter(sc,
986 archive_string_append_unicode);
987 return;
988 }
989
990 #if defined(_WIN32) && !defined(__CYGWIN__)
991 if (sc->flag & SCONV_WIN_CP) {
992 if (sc->flag & SCONV_FROM_UTF16BE)
993 add_converter(sc, win_strncat_from_utf16be);
994 else
995 add_converter(sc, win_strncat_from_utf16le);
996 return;
997 }
998 #endif
999
1000 #if defined(HAVE_ICONV)
1001 if (sc->cd != (iconv_t)-1) {
1002 add_converter(sc, iconv_strncat_in_locale);
1003 return;
1004 }
1005 #endif
1006
1007 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1008 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
1009 add_converter(sc, best_effort_strncat_from_utf16be);
1010 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1011 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
1012 add_converter(sc, best_effort_strncat_from_utf16le);
1013 else
1014 /* Make sure we have no converter. */
1015 sc->nconverter = 0;
1016 return;
1017 }
1018
1019 if (sc->flag & SCONV_FROM_UTF8) {
1020 /*
1021 * At least we should normalize a UTF-8 string.
1022 */
1023 if (sc->flag & SCONV_NORMALIZATION_D)
1024 add_converter(sc,archive_string_normalize_D);
1025 else if (sc->flag & SCONV_NORMALIZATION_C)
1026 add_converter(sc, archive_string_normalize_C);
1027
1028 /*
1029 * Copy UTF-8 string with a check of CESU-8.
1030 * Apparently, iconv does not check surrogate pairs in UTF-8
1031 * when both from-charset and to-charset are UTF-8, and then
1032 * we use our UTF-8 copy code.
1033 */
1034 if (sc->flag & SCONV_TO_UTF8) {
1035 /*
1036 * If the current locale is UTF-8, we can translate
1037 * a UTF-16BE string into a UTF-8 string directly.
1038 */
1039 if (!(sc->flag &
1040 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1041 add_converter(sc, strncat_from_utf8_to_utf8);
1042 return;
1043 }
1044 }
1045
1046 #if defined(_WIN32) && !defined(__CYGWIN__)
1047 /*
1048 * On Windows we can use Windows API for a string conversion.
1049 */
1050 if (sc->flag & SCONV_WIN_CP) {
1051 add_converter(sc, strncat_in_codepage);
1052 return;
1053 }
1054 #endif
1055
1056 #if HAVE_ICONV
1057 if (sc->cd != (iconv_t)-1) {
1058 add_converter(sc, iconv_strncat_in_locale);
1059 /*
1060 * iconv generally does not support UTF-8-MAC and so
1061 * we have to the output of iconv from NFC to NFD if
1062 * need.
1063 */
1064 if ((sc->flag & SCONV_FROM_CHARSET) &&
1065 (sc->flag & SCONV_TO_UTF8)) {
1066 if (sc->flag & SCONV_NORMALIZATION_D)
1067 add_converter(sc, archive_string_normalize_D);
1068 }
1069 return;
1070 }
1071 #endif
1072
1073 /*
1074 * Try conversion in the best effort or no conversion.
1075 */
1076 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1077 add_converter(sc, best_effort_strncat_in_locale);
1078 else
1079 /* Make sure we have no converter. */
1080 sc->nconverter = 0;
1081 }
1082
1083 /*
1084 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1085 * and CP932 which are referenced in create_sconv_object().
1086 */
1087 static const char *
canonical_charset_name(const char * charset)1088 canonical_charset_name(const char *charset)
1089 {
1090 char cs[16];
1091 char *p;
1092 const char *s;
1093
1094 if (charset == NULL || charset[0] == '\0'
1095 || strlen(charset) > 15)
1096 return (charset);
1097
1098 /* Copy name to uppercase. */
1099 p = cs;
1100 s = charset;
1101 while (*s) {
1102 char c = *s++;
1103 if (c >= 'a' && c <= 'z')
1104 c -= 'a' - 'A';
1105 *p++ = c;
1106 }
1107 *p++ = '\0';
1108
1109 if (strcmp(cs, "UTF-8") == 0 ||
1110 strcmp(cs, "UTF8") == 0)
1111 return ("UTF-8");
1112 if (strcmp(cs, "UTF-16BE") == 0 ||
1113 strcmp(cs, "UTF16BE") == 0)
1114 return ("UTF-16BE");
1115 if (strcmp(cs, "UTF-16LE") == 0 ||
1116 strcmp(cs, "UTF16LE") == 0)
1117 return ("UTF-16LE");
1118 if (strcmp(cs, "CP932") == 0)
1119 return ("CP932");
1120 return (charset);
1121 }
1122
1123 /*
1124 * Create a string conversion object.
1125 */
1126 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1127 create_sconv_object(const char *fc, const char *tc,
1128 unsigned current_codepage, int flag)
1129 {
1130 struct archive_string_conv *sc;
1131
1132 sc = calloc(1, sizeof(*sc));
1133 if (sc == NULL)
1134 return (NULL);
1135 sc->next = NULL;
1136 sc->from_charset = strdup(fc);
1137 if (sc->from_charset == NULL) {
1138 free(sc);
1139 return (NULL);
1140 }
1141 sc->to_charset = strdup(tc);
1142 if (sc->to_charset == NULL) {
1143 free(sc->from_charset);
1144 free(sc);
1145 return (NULL);
1146 }
1147 archive_string_init(&sc->utftmp);
1148
1149 if (flag & SCONV_TO_CHARSET) {
1150 /*
1151 * Convert characters from the current locale charset to
1152 * a specified charset.
1153 */
1154 sc->from_cp = current_codepage;
1155 sc->to_cp = make_codepage_from_charset(tc);
1156 #if defined(_WIN32) && !defined(__CYGWIN__)
1157 if (IsValidCodePage(sc->to_cp))
1158 flag |= SCONV_WIN_CP;
1159 #endif
1160 } else if (flag & SCONV_FROM_CHARSET) {
1161 /*
1162 * Convert characters from a specified charset to
1163 * the current locale charset.
1164 */
1165 sc->to_cp = current_codepage;
1166 sc->from_cp = make_codepage_from_charset(fc);
1167 #if defined(_WIN32) && !defined(__CYGWIN__)
1168 if (IsValidCodePage(sc->from_cp))
1169 flag |= SCONV_WIN_CP;
1170 #endif
1171 }
1172
1173 /*
1174 * Check if "from charset" and "to charset" are the same.
1175 */
1176 if (strcmp(fc, tc) == 0 ||
1177 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1178 sc->same = 1;
1179 else
1180 sc->same = 0;
1181
1182 /*
1183 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1184 */
1185 if (strcmp(tc, "UTF-8") == 0)
1186 flag |= SCONV_TO_UTF8;
1187 else if (strcmp(tc, "UTF-16BE") == 0)
1188 flag |= SCONV_TO_UTF16BE;
1189 else if (strcmp(tc, "UTF-16LE") == 0)
1190 flag |= SCONV_TO_UTF16LE;
1191 if (strcmp(fc, "UTF-8") == 0)
1192 flag |= SCONV_FROM_UTF8;
1193 else if (strcmp(fc, "UTF-16BE") == 0)
1194 flag |= SCONV_FROM_UTF16BE;
1195 else if (strcmp(fc, "UTF-16LE") == 0)
1196 flag |= SCONV_FROM_UTF16LE;
1197 #if defined(_WIN32) && !defined(__CYGWIN__)
1198 if (sc->to_cp == CP_UTF8)
1199 flag |= SCONV_TO_UTF8;
1200 else if (sc->to_cp == CP_UTF16BE)
1201 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1202 else if (sc->to_cp == CP_UTF16LE)
1203 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1204 if (sc->from_cp == CP_UTF8)
1205 flag |= SCONV_FROM_UTF8;
1206 else if (sc->from_cp == CP_UTF16BE)
1207 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1208 else if (sc->from_cp == CP_UTF16LE)
1209 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1210 #endif
1211
1212 /*
1213 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1214 * handle it. So we have to translate NFD characters to NFC ones
1215 * ourselves before iconv handles. Another reason is to prevent
1216 * that the same sight of two filenames, one is NFC and other
1217 * is NFD, would be in its directory.
1218 * On Mac OS X, although its filesystem layer automatically
1219 * convert filenames to NFD, it would be useful for filename
1220 * comparing to find out the same filenames that we normalize
1221 * that to be NFD ourselves.
1222 */
1223 if ((flag & SCONV_FROM_CHARSET) &&
1224 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1225 #if defined(__APPLE__)
1226 if (flag & SCONV_TO_UTF8)
1227 flag |= SCONV_NORMALIZATION_D;
1228 else
1229 #endif
1230 flag |= SCONV_NORMALIZATION_C;
1231 }
1232 #if defined(__APPLE__)
1233 /*
1234 * In case writing an archive file, make sure that a filename
1235 * going to be passed to iconv is a Unicode NFC string since
1236 * a filename in HFS Plus filesystem is a Unicode NFD one and
1237 * iconv cannot handle it with "UTF-8" charset. It is simpler
1238 * than a use of "UTF-8-MAC" charset.
1239 */
1240 if ((flag & SCONV_TO_CHARSET) &&
1241 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1242 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1243 flag |= SCONV_NORMALIZATION_C;
1244 /*
1245 * In case reading an archive file. make sure that a filename
1246 * will be passed to users is a Unicode NFD string in order to
1247 * correctly compare the filename with other one which comes
1248 * from HFS Plus filesystem.
1249 */
1250 if ((flag & SCONV_FROM_CHARSET) &&
1251 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1252 (flag & SCONV_TO_UTF8))
1253 flag |= SCONV_NORMALIZATION_D;
1254 #endif
1255
1256 #if defined(HAVE_ICONV)
1257 sc->cd_w = (iconv_t)-1;
1258 /*
1259 * Create an iconv object.
1260 */
1261 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1262 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1263 (flag & SCONV_WIN_CP)) {
1264 /* This case we won't use iconv. */
1265 sc->cd = (iconv_t)-1;
1266 } else {
1267 sc->cd = iconv_open(tc, fc);
1268 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1269 /*
1270 * Unfortunately, all of iconv implements do support
1271 * "CP932" character-set, so we should use "SJIS"
1272 * instead if iconv_open failed.
1273 */
1274 if (strcmp(tc, "CP932") == 0)
1275 sc->cd = iconv_open("SJIS", fc);
1276 else if (strcmp(fc, "CP932") == 0)
1277 sc->cd = iconv_open(tc, "SJIS");
1278 }
1279 #if defined(_WIN32) && !defined(__CYGWIN__)
1280 /*
1281 * archive_mstring on Windows directly convert multi-bytes
1282 * into archive_wstring in order not to depend on locale
1283 * so that you can do a I18N programming. This will be
1284 * used only in archive_mstring_copy_mbs_len_l so far.
1285 */
1286 if (flag & SCONV_FROM_CHARSET) {
1287 sc->cd_w = iconv_open("UTF-8", fc);
1288 if (sc->cd_w == (iconv_t)-1 &&
1289 (sc->flag & SCONV_BEST_EFFORT)) {
1290 if (strcmp(fc, "CP932") == 0)
1291 sc->cd_w = iconv_open("UTF-8", "SJIS");
1292 }
1293 }
1294 #endif /* _WIN32 && !__CYGWIN__ */
1295 }
1296 #endif /* HAVE_ICONV */
1297
1298 sc->flag = flag;
1299
1300 /*
1301 * Set up converters.
1302 */
1303 setup_converter(sc);
1304
1305 return (sc);
1306 }
1307
1308 /*
1309 * Free a string conversion object.
1310 */
1311 static void
free_sconv_object(struct archive_string_conv * sc)1312 free_sconv_object(struct archive_string_conv *sc)
1313 {
1314 free(sc->from_charset);
1315 free(sc->to_charset);
1316 archive_string_free(&sc->utftmp);
1317 #if HAVE_ICONV
1318 if (sc->cd != (iconv_t)-1)
1319 iconv_close(sc->cd);
1320 if (sc->cd_w != (iconv_t)-1)
1321 iconv_close(sc->cd_w);
1322 #endif
1323 free(sc);
1324 }
1325
1326 #if defined(_WIN32) && !defined(__CYGWIN__)
1327 static unsigned
my_atoi(const char * p)1328 my_atoi(const char *p)
1329 {
1330 unsigned cp;
1331
1332 cp = 0;
1333 while (*p) {
1334 if (*p >= '0' && *p <= '9')
1335 cp = cp * 10 + (*p - '0');
1336 else
1337 return (-1);
1338 p++;
1339 }
1340 return (cp);
1341 }
1342
1343 /*
1344 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1345 * Return -1 if failed.
1346 *
1347 * Note: This translation code may be insufficient.
1348 */
1349 static struct charset {
1350 const char *name;
1351 unsigned cp;
1352 } charsets[] = {
1353 /* MUST BE SORTED! */
1354 {"ASCII", 1252},
1355 {"ASMO-708", 708},
1356 {"BIG5", 950},
1357 {"CHINESE", 936},
1358 {"CP367", 1252},
1359 {"CP819", 1252},
1360 {"CP1025", 21025},
1361 {"DOS-720", 720},
1362 {"DOS-862", 862},
1363 {"EUC-CN", 51936},
1364 {"EUC-JP", 51932},
1365 {"EUC-KR", 949},
1366 {"EUCCN", 51936},
1367 {"EUCJP", 51932},
1368 {"EUCKR", 949},
1369 {"GB18030", 54936},
1370 {"GB2312", 936},
1371 {"HEBREW", 1255},
1372 {"HZ-GB-2312", 52936},
1373 {"IBM273", 20273},
1374 {"IBM277", 20277},
1375 {"IBM278", 20278},
1376 {"IBM280", 20280},
1377 {"IBM284", 20284},
1378 {"IBM285", 20285},
1379 {"IBM290", 20290},
1380 {"IBM297", 20297},
1381 {"IBM367", 1252},
1382 {"IBM420", 20420},
1383 {"IBM423", 20423},
1384 {"IBM424", 20424},
1385 {"IBM819", 1252},
1386 {"IBM871", 20871},
1387 {"IBM880", 20880},
1388 {"IBM905", 20905},
1389 {"IBM924", 20924},
1390 {"ISO-8859-1", 28591},
1391 {"ISO-8859-13", 28603},
1392 {"ISO-8859-15", 28605},
1393 {"ISO-8859-2", 28592},
1394 {"ISO-8859-3", 28593},
1395 {"ISO-8859-4", 28594},
1396 {"ISO-8859-5", 28595},
1397 {"ISO-8859-6", 28596},
1398 {"ISO-8859-7", 28597},
1399 {"ISO-8859-8", 28598},
1400 {"ISO-8859-9", 28599},
1401 {"ISO8859-1", 28591},
1402 {"ISO8859-13", 28603},
1403 {"ISO8859-15", 28605},
1404 {"ISO8859-2", 28592},
1405 {"ISO8859-3", 28593},
1406 {"ISO8859-4", 28594},
1407 {"ISO8859-5", 28595},
1408 {"ISO8859-6", 28596},
1409 {"ISO8859-7", 28597},
1410 {"ISO8859-8", 28598},
1411 {"ISO8859-9", 28599},
1412 {"JOHAB", 1361},
1413 {"KOI8-R", 20866},
1414 {"KOI8-U", 21866},
1415 {"KS_C_5601-1987", 949},
1416 {"LATIN1", 1252},
1417 {"LATIN2", 28592},
1418 {"MACINTOSH", 10000},
1419 {"SHIFT-JIS", 932},
1420 {"SHIFT_JIS", 932},
1421 {"SJIS", 932},
1422 {"US", 1252},
1423 {"US-ASCII", 1252},
1424 {"UTF-16", 1200},
1425 {"UTF-16BE", 1201},
1426 {"UTF-16LE", 1200},
1427 {"UTF-8", CP_UTF8},
1428 {"X-EUROPA", 29001},
1429 {"X-MAC-ARABIC", 10004},
1430 {"X-MAC-CE", 10029},
1431 {"X-MAC-CHINESEIMP", 10008},
1432 {"X-MAC-CHINESETRAD", 10002},
1433 {"X-MAC-CROATIAN", 10082},
1434 {"X-MAC-CYRILLIC", 10007},
1435 {"X-MAC-GREEK", 10006},
1436 {"X-MAC-HEBREW", 10005},
1437 {"X-MAC-ICELANDIC", 10079},
1438 {"X-MAC-JAPANESE", 10001},
1439 {"X-MAC-KOREAN", 10003},
1440 {"X-MAC-ROMANIAN", 10010},
1441 {"X-MAC-THAI", 10021},
1442 {"X-MAC-TURKISH", 10081},
1443 {"X-MAC-UKRAINIAN", 10017},
1444 };
1445 static unsigned
make_codepage_from_charset(const char * charset)1446 make_codepage_from_charset(const char *charset)
1447 {
1448 char cs[16];
1449 char *p;
1450 unsigned cp;
1451 int a, b;
1452
1453 if (charset == NULL || strlen(charset) > 15)
1454 return -1;
1455
1456 /* Copy name to uppercase. */
1457 p = cs;
1458 while (*charset) {
1459 char c = *charset++;
1460 if (c >= 'a' && c <= 'z')
1461 c -= 'a' - 'A';
1462 *p++ = c;
1463 }
1464 *p++ = '\0';
1465 cp = -1;
1466
1467 /* Look it up in the table first, so that we can easily
1468 * override CP367, which we map to 1252 instead of 367. */
1469 a = 0;
1470 b = sizeof(charsets)/sizeof(charsets[0]);
1471 while (b > a) {
1472 int c = (b + a) / 2;
1473 int r = strcmp(charsets[c].name, cs);
1474 if (r < 0)
1475 a = c + 1;
1476 else if (r > 0)
1477 b = c;
1478 else
1479 return charsets[c].cp;
1480 }
1481
1482 /* If it's not in the table, try to parse it. */
1483 switch (*cs) {
1484 case 'C':
1485 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1486 cp = my_atoi(cs + 2);
1487 } else if (strcmp(cs, "CP_ACP") == 0)
1488 cp = get_current_codepage();
1489 else if (strcmp(cs, "CP_OEMCP") == 0)
1490 cp = get_current_oemcp();
1491 break;
1492 case 'I':
1493 if (cs[1] == 'B' && cs[2] == 'M' &&
1494 cs[3] >= '0' && cs[3] <= '9') {
1495 cp = my_atoi(cs + 3);
1496 }
1497 break;
1498 case 'W':
1499 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1500 cp = my_atoi(cs + 8);
1501 if (cp != 874 && (cp < 1250 || cp > 1258))
1502 cp = -1;/* This may invalid code. */
1503 }
1504 break;
1505 }
1506 return (cp);
1507 }
1508
1509 /*
1510 * Return ANSI Code Page of current locale set by setlocale().
1511 */
1512 static unsigned
get_current_codepage(void)1513 get_current_codepage(void)
1514 {
1515 char *locale, *p;
1516 unsigned cp;
1517
1518 locale = setlocale(LC_CTYPE, NULL);
1519 if (locale == NULL)
1520 return (GetACP());
1521 if (locale[0] == 'C' && locale[1] == '\0')
1522 return (CP_C_LOCALE);
1523 p = strrchr(locale, '.');
1524 if (p == NULL)
1525 return (GetACP());
1526 if (strcmp(p+1, "utf8") == 0)
1527 return CP_UTF8;
1528 cp = my_atoi(p+1);
1529 if ((int)cp <= 0)
1530 return (GetACP());
1531 return (cp);
1532 }
1533
1534 /*
1535 * Translation table between Locale Name and ACP/OEMCP.
1536 */
1537 static struct {
1538 unsigned acp;
1539 unsigned ocp;
1540 const char *locale;
1541 } acp_ocp_map[] = {
1542 { 950, 950, "Chinese_Taiwan" },
1543 { 936, 936, "Chinese_People's Republic of China" },
1544 { 950, 950, "Chinese_Taiwan" },
1545 { 1250, 852, "Czech_Czech Republic" },
1546 { 1252, 850, "Danish_Denmark" },
1547 { 1252, 850, "Dutch_Netherlands" },
1548 { 1252, 850, "Dutch_Belgium" },
1549 { 1252, 437, "English_United States" },
1550 { 1252, 850, "English_Australia" },
1551 { 1252, 850, "English_Canada" },
1552 { 1252, 850, "English_New Zealand" },
1553 { 1252, 850, "English_United Kingdom" },
1554 { 1252, 437, "English_United States" },
1555 { 1252, 850, "Finnish_Finland" },
1556 { 1252, 850, "French_France" },
1557 { 1252, 850, "French_Belgium" },
1558 { 1252, 850, "French_Canada" },
1559 { 1252, 850, "French_Switzerland" },
1560 { 1252, 850, "German_Germany" },
1561 { 1252, 850, "German_Austria" },
1562 { 1252, 850, "German_Switzerland" },
1563 { 1253, 737, "Greek_Greece" },
1564 { 1250, 852, "Hungarian_Hungary" },
1565 { 1252, 850, "Icelandic_Iceland" },
1566 { 1252, 850, "Italian_Italy" },
1567 { 1252, 850, "Italian_Switzerland" },
1568 { 932, 932, "Japanese_Japan" },
1569 { 949, 949, "Korean_Korea" },
1570 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1571 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1572 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1573 { 1250, 852, "Polish_Poland" },
1574 { 1252, 850, "Portuguese_Portugal" },
1575 { 1252, 850, "Portuguese_Brazil" },
1576 { 1251, 866, "Russian_Russia" },
1577 { 1250, 852, "Slovak_Slovakia" },
1578 { 1252, 850, "Spanish_Spain" },
1579 { 1252, 850, "Spanish_Mexico" },
1580 { 1252, 850, "Spanish_Spain" },
1581 { 1252, 850, "Swedish_Sweden" },
1582 { 1254, 857, "Turkish_Turkey" },
1583 { 0, 0, NULL}
1584 };
1585
1586 /*
1587 * Return OEM Code Page of current locale set by setlocale().
1588 */
1589 static unsigned
get_current_oemcp(void)1590 get_current_oemcp(void)
1591 {
1592 int i;
1593 char *locale, *p;
1594 size_t len;
1595
1596 locale = setlocale(LC_CTYPE, NULL);
1597 if (locale == NULL)
1598 return (GetOEMCP());
1599 if (locale[0] == 'C' && locale[1] == '\0')
1600 return (CP_C_LOCALE);
1601
1602 p = strrchr(locale, '.');
1603 if (p == NULL)
1604 return (GetOEMCP());
1605 len = p - locale;
1606 for (i = 0; acp_ocp_map[i].acp; i++) {
1607 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1608 return (acp_ocp_map[i].ocp);
1609 }
1610 return (GetOEMCP());
1611 }
1612 #else
1613
1614 /*
1615 * POSIX platform does not use CodePage.
1616 */
1617
1618 static unsigned
get_current_codepage(void)1619 get_current_codepage(void)
1620 {
1621 return (-1);/* Unknown */
1622 }
1623 static unsigned
make_codepage_from_charset(const char * charset)1624 make_codepage_from_charset(const char *charset)
1625 {
1626 (void)charset; /* UNUSED */
1627 return (-1);/* Unknown */
1628 }
1629 static unsigned
get_current_oemcp(void)1630 get_current_oemcp(void)
1631 {
1632 return (-1);/* Unknown */
1633 }
1634
1635 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1636
1637 /*
1638 * Return a string conversion object.
1639 */
1640 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1641 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1642 {
1643 struct archive_string_conv *sc;
1644 unsigned current_codepage;
1645
1646 /* Check if we have made the sconv object. */
1647 sc = find_sconv_object(a, fc, tc);
1648 if (sc != NULL)
1649 return (sc);
1650
1651 if (a == NULL)
1652 current_codepage = get_current_codepage();
1653 else
1654 current_codepage = a->current_codepage;
1655
1656 sc = create_sconv_object(canonical_charset_name(fc),
1657 canonical_charset_name(tc), current_codepage, flag);
1658 if (sc == NULL) {
1659 if (a != NULL)
1660 archive_set_error(a, ENOMEM,
1661 "Could not allocate memory for "
1662 "a string conversion object");
1663 return (NULL);
1664 }
1665
1666 /*
1667 * If there is no converter for current string conversion object,
1668 * we cannot handle this conversion.
1669 */
1670 if (sc->nconverter == 0) {
1671 if (a != NULL) {
1672 #if HAVE_ICONV
1673 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1674 "iconv_open failed : Cannot handle ``%s''",
1675 (flag & SCONV_TO_CHARSET)?tc:fc);
1676 #else
1677 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1678 "A character-set conversion not fully supported "
1679 "on this platform");
1680 #endif
1681 }
1682 /* Failed; free a sconv object. */
1683 free_sconv_object(sc);
1684 return (NULL);
1685 }
1686
1687 /*
1688 * Success!
1689 */
1690 if (a != NULL)
1691 add_sconv_object(a, sc);
1692 return (sc);
1693 }
1694
1695 static const char *
get_current_charset(struct archive * a)1696 get_current_charset(struct archive *a)
1697 {
1698 const char *cur_charset;
1699
1700 if (a == NULL)
1701 cur_charset = default_iconv_charset("");
1702 else {
1703 cur_charset = default_iconv_charset(a->current_code);
1704 if (a->current_code == NULL) {
1705 a->current_code = strdup(cur_charset);
1706 a->current_codepage = get_current_codepage();
1707 a->current_oemcp = get_current_oemcp();
1708 }
1709 }
1710 return (cur_charset);
1711 }
1712
1713 /*
1714 * Make and Return a string conversion object.
1715 * Return NULL if the platform does not support the specified conversion
1716 * and best_effort is 0.
1717 * If best_effort is set, A string conversion object must be returned
1718 * unless memory allocation for the object fails, but the conversion
1719 * might fail when non-ASCII code is found.
1720 */
1721 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1722 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1723 int best_effort)
1724 {
1725 int flag = SCONV_TO_CHARSET;
1726
1727 if (best_effort)
1728 flag |= SCONV_BEST_EFFORT;
1729 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1730 }
1731
1732 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1733 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1734 int best_effort)
1735 {
1736 int flag = SCONV_FROM_CHARSET;
1737
1738 if (best_effort)
1739 flag |= SCONV_BEST_EFFORT;
1740 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1741 }
1742
1743 /*
1744 * archive_string_default_conversion_*_archive() are provided for Windows
1745 * platform because other archiver application use CP_OEMCP for
1746 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1747 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1748 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1749 * So we should make a string conversion between CP_ACP and CP_OEMCP
1750 * for compatibility.
1751 */
1752 #if defined(_WIN32) && !defined(__CYGWIN__)
1753 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1754 archive_string_default_conversion_for_read(struct archive *a)
1755 {
1756 const char *cur_charset = get_current_charset(a);
1757 char oemcp[16];
1758
1759 /* NOTE: a check of cur_charset is unneeded but we need
1760 * that get_current_charset() has been surely called at
1761 * this time whatever C compiler optimized. */
1762 if (cur_charset != NULL &&
1763 (a->current_codepage == CP_C_LOCALE ||
1764 a->current_codepage == a->current_oemcp))
1765 return (NULL);/* no conversion. */
1766
1767 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1768 /* Make sure a null termination must be set. */
1769 oemcp[sizeof(oemcp)-1] = '\0';
1770 return (get_sconv_object(a, oemcp, cur_charset,
1771 SCONV_FROM_CHARSET));
1772 }
1773
1774 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1775 archive_string_default_conversion_for_write(struct archive *a)
1776 {
1777 const char *cur_charset = get_current_charset(a);
1778 char oemcp[16];
1779
1780 /* NOTE: a check of cur_charset is unneeded but we need
1781 * that get_current_charset() has been surely called at
1782 * this time whatever C compiler optimized. */
1783 if (cur_charset != NULL &&
1784 (a->current_codepage == CP_C_LOCALE ||
1785 a->current_codepage == a->current_oemcp))
1786 return (NULL);/* no conversion. */
1787
1788 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1789 /* Make sure a null termination must be set. */
1790 oemcp[sizeof(oemcp)-1] = '\0';
1791 return (get_sconv_object(a, cur_charset, oemcp,
1792 SCONV_TO_CHARSET));
1793 }
1794 #else
1795 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1796 archive_string_default_conversion_for_read(struct archive *a)
1797 {
1798 (void)a; /* UNUSED */
1799 return (NULL);
1800 }
1801
1802 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1803 archive_string_default_conversion_for_write(struct archive *a)
1804 {
1805 (void)a; /* UNUSED */
1806 return (NULL);
1807 }
1808 #endif
1809
1810 /*
1811 * Dispose of all character conversion objects in the archive object.
1812 */
1813 void
archive_string_conversion_free(struct archive * a)1814 archive_string_conversion_free(struct archive *a)
1815 {
1816 struct archive_string_conv *sc;
1817 struct archive_string_conv *sc_next;
1818
1819 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1820 sc_next = sc->next;
1821 free_sconv_object(sc);
1822 }
1823 a->sconv = NULL;
1824 free(a->current_code);
1825 a->current_code = NULL;
1826 }
1827
1828 /*
1829 * Return a conversion charset name.
1830 */
1831 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1832 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1833 {
1834 if (sc->flag & SCONV_TO_CHARSET)
1835 return (sc->to_charset);
1836 else
1837 return (sc->from_charset);
1838 }
1839
1840 /*
1841 * Change the behavior of a string conversion.
1842 */
1843 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1844 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1845 {
1846 switch (opt) {
1847 /*
1848 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1849 * assumption that wchar_t was Unicode.
1850 * This option enables simulating the assumption in order to read
1851 * that filename correctly.
1852 */
1853 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1854 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1855 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1856 /*
1857 * Nothing to do for it since wchar_t on these platforms
1858 * is really Unicode.
1859 */
1860 (void)sc; /* UNUSED */
1861 #else
1862 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1863 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1864 /* Set up string converters. */
1865 setup_converter(sc);
1866 }
1867 #endif
1868 break;
1869 case SCONV_SET_OPT_NORMALIZATION_C:
1870 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1871 sc->flag |= SCONV_NORMALIZATION_C;
1872 sc->flag &= ~SCONV_NORMALIZATION_D;
1873 /* Set up string converters. */
1874 setup_converter(sc);
1875 }
1876 break;
1877 case SCONV_SET_OPT_NORMALIZATION_D:
1878 #if defined(HAVE_ICONV)
1879 /*
1880 * If iconv will take the string, do not change the
1881 * setting of the normalization.
1882 */
1883 if (!(sc->flag & SCONV_WIN_CP) &&
1884 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1885 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1886 break;
1887 #endif
1888 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1889 sc->flag |= SCONV_NORMALIZATION_D;
1890 sc->flag &= ~SCONV_NORMALIZATION_C;
1891 /* Set up string converters. */
1892 setup_converter(sc);
1893 }
1894 break;
1895 default:
1896 break;
1897 }
1898 }
1899
1900 /*
1901 *
1902 * Copy one archive_string to another in locale conversion.
1903 *
1904 * archive_strncat_l();
1905 * archive_strncpy_l();
1906 *
1907 */
1908
1909 static size_t
mbsnbytes(const void * _p,size_t n)1910 mbsnbytes(const void *_p, size_t n)
1911 {
1912 size_t s;
1913 const char *p, *pp;
1914
1915 if (_p == NULL)
1916 return (0);
1917 p = (const char *)_p;
1918
1919 /* Like strlen(p), except won't examine positions beyond p[n]. */
1920 s = 0;
1921 pp = p;
1922 while (s < n && *pp) {
1923 pp++;
1924 s++;
1925 }
1926 return (s);
1927 }
1928
1929 static size_t
utf16nbytes(const void * _p,size_t n)1930 utf16nbytes(const void *_p, size_t n)
1931 {
1932 size_t s;
1933 const char *p, *pp;
1934
1935 if (_p == NULL)
1936 return (0);
1937 p = (const char *)_p;
1938
1939 /* Like strlen(p), except won't examine positions beyond p[n]. */
1940 s = 0;
1941 pp = p;
1942 n >>= 1;
1943 while (s < n && (pp[0] || pp[1])) {
1944 pp += 2;
1945 s++;
1946 }
1947 return (s<<1);
1948 }
1949
1950 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1951 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1952 struct archive_string_conv *sc)
1953 {
1954 as->length = 0;
1955 return (archive_strncat_l(as, _p, n, sc));
1956 }
1957
1958 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1959 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1960 struct archive_string_conv *sc)
1961 {
1962 const void *s;
1963 size_t length = 0;
1964 int i, r = 0, r2;
1965
1966 if (_p != NULL && n > 0) {
1967 if (sc != NULL && (sc->flag & SCONV_FROM_UTF16))
1968 length = utf16nbytes(_p, n);
1969 else
1970 length = mbsnbytes(_p, n);
1971 }
1972
1973 /* We must allocate memory even if there is no data for conversion
1974 * or copy. This simulates archive_string_append behavior. */
1975 if (length == 0) {
1976 int tn = 1;
1977 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1978 tn = 2;
1979 if (archive_string_ensure(as, as->length + tn) == NULL)
1980 return (-1);
1981 as->s[as->length] = 0;
1982 if (tn == 2)
1983 as->s[as->length+1] = 0;
1984 return (0);
1985 }
1986
1987 /*
1988 * If sc is NULL, we just make a copy.
1989 */
1990 if (sc == NULL) {
1991 if (archive_string_append(as, _p, length) == NULL)
1992 return (-1);/* No memory */
1993 return (0);
1994 }
1995
1996 s = _p;
1997 i = 0;
1998 if (sc->nconverter > 1) {
1999 sc->utftmp.length = 0;
2000 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
2001 if (r2 != 0 && errno == ENOMEM)
2002 return (r2);
2003 if (r > r2)
2004 r = r2;
2005 s = sc->utftmp.s;
2006 length = sc->utftmp.length;
2007 ++i;
2008 }
2009 r2 = sc->converter[i](as, s, length, sc);
2010 if (r > r2)
2011 r = r2;
2012 return (r);
2013 }
2014
2015 #if HAVE_ICONV
2016
2017 /*
2018 * Return -1 if conversion fails.
2019 */
2020 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2021 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
2022 size_t length, struct archive_string_conv *sc)
2023 {
2024 ICONV_CONST char *itp;
2025 size_t remaining;
2026 iconv_t cd;
2027 char *outp;
2028 size_t avail, bs;
2029 int return_value = 0; /* success */
2030 int to_size, from_size;
2031
2032 if (sc->flag & SCONV_TO_UTF16)
2033 to_size = 2;
2034 else
2035 to_size = 1;
2036 if (sc->flag & SCONV_FROM_UTF16)
2037 from_size = 2;
2038 else
2039 from_size = 1;
2040
2041 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2042 return (-1);
2043
2044 cd = sc->cd;
2045 itp = (char *)(uintptr_t)_p;
2046 remaining = length;
2047 outp = as->s + as->length;
2048 avail = as->buffer_length - as->length - to_size;
2049 while (remaining >= (size_t)from_size) {
2050 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2051
2052 if (result != (size_t)-1)
2053 break; /* Conversion completed. */
2054
2055 if (errno == EILSEQ || errno == EINVAL) {
2056 /*
2057 * If an output charset is UTF-8 or UTF-16BE/LE,
2058 * unknown character should be U+FFFD
2059 * (replacement character).
2060 */
2061 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2062 size_t rbytes;
2063 if (sc->flag & SCONV_TO_UTF8)
2064 rbytes = sizeof(utf8_replacement_char);
2065 else
2066 rbytes = 2;
2067
2068 if (avail < rbytes) {
2069 as->length = outp - as->s;
2070 bs = as->buffer_length +
2071 (remaining * to_size) + rbytes;
2072 if (NULL ==
2073 archive_string_ensure(as, bs))
2074 return (-1);
2075 outp = as->s + as->length;
2076 avail = as->buffer_length
2077 - as->length - to_size;
2078 }
2079 if (sc->flag & SCONV_TO_UTF8)
2080 memcpy(outp, utf8_replacement_char, sizeof(utf8_replacement_char));
2081 else if (sc->flag & SCONV_TO_UTF16BE)
2082 archive_be16enc(outp, UNICODE_R_CHAR);
2083 else
2084 archive_le16enc(outp, UNICODE_R_CHAR);
2085 outp += rbytes;
2086 avail -= rbytes;
2087 } else {
2088 /* Skip the illegal input bytes. */
2089 *outp++ = '?';
2090 avail--;
2091 }
2092 itp += from_size;
2093 remaining -= from_size;
2094 return_value = -1; /* failure */
2095 } else {
2096 /* E2BIG no output buffer,
2097 * Increase an output buffer. */
2098 as->length = outp - as->s;
2099 bs = as->buffer_length + remaining * 2;
2100 if (NULL == archive_string_ensure(as, bs))
2101 return (-1);
2102 outp = as->s + as->length;
2103 avail = as->buffer_length - as->length - to_size;
2104 }
2105 }
2106 as->length = outp - as->s;
2107 as->s[as->length] = 0;
2108 if (to_size == 2)
2109 as->s[as->length+1] = 0;
2110 return (return_value);
2111 }
2112
2113 #endif /* HAVE_ICONV */
2114
2115
2116 #if defined(_WIN32) && !defined(__CYGWIN__)
2117
2118 /*
2119 * Translate a string from a some CodePage to an another CodePage by
2120 * Windows APIs, and copy the result. Return -1 if conversion fails.
2121 */
2122 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2123 strncat_in_codepage(struct archive_string *as,
2124 const void *_p, size_t length, struct archive_string_conv *sc)
2125 {
2126 const char *s = (const char *)_p;
2127 struct archive_wstring aws;
2128 size_t l;
2129 int r, saved_flag;
2130
2131 archive_string_init(&aws);
2132 saved_flag = sc->flag;
2133 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2134 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2135 sc->flag = saved_flag;
2136 if (r != 0) {
2137 archive_wstring_free(&aws);
2138 if (errno != ENOMEM)
2139 archive_string_append(as, s, length);
2140 return (-1);
2141 }
2142
2143 l = as->length;
2144 r = archive_string_append_from_wcs_in_codepage(
2145 as, aws.s, aws.length, sc);
2146 if (r != 0 && errno != ENOMEM && l == as->length)
2147 archive_string_append(as, s, length);
2148 archive_wstring_free(&aws);
2149 return (r);
2150 }
2151
2152 /*
2153 * Test whether MBS ==> WCS is okay.
2154 */
2155 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2156 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2157 {
2158 const char *p = (const char *)_p;
2159 unsigned codepage;
2160 DWORD mbflag = MB_ERR_INVALID_CHARS;
2161
2162 if (sc->flag & SCONV_FROM_CHARSET)
2163 codepage = sc->to_cp;
2164 else
2165 codepage = sc->from_cp;
2166
2167 if (codepage == CP_C_LOCALE)
2168 return (0);
2169 if (codepage != CP_UTF8)
2170 mbflag |= MB_PRECOMPOSED;
2171
2172 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2173 return (-1); /* Invalid */
2174 return (0); /* Okay */
2175 }
2176
2177 #else
2178
2179 /*
2180 * Test whether MBS ==> WCS is okay.
2181 */
2182 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2183 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2184 {
2185 const char *p = (const char *)_p;
2186 size_t r;
2187
2188 #if HAVE_MBRTOWC
2189 mbstate_t shift_state;
2190
2191 memset(&shift_state, 0, sizeof(shift_state));
2192 #else
2193 /* Clear the shift state before starting. */
2194 mbtowc(NULL, NULL, 0);
2195 #endif
2196 while (n) {
2197 wchar_t wc;
2198
2199 #if HAVE_MBRTOWC
2200 r = mbrtowc(&wc, p, n, &shift_state);
2201 #else
2202 r = mbtowc(&wc, p, n);
2203 #endif
2204 if (r == (size_t)-1 || r == (size_t)-2)
2205 return (-1);/* Invalid. */
2206 if (r == 0)
2207 break;
2208 p += r;
2209 n -= r;
2210 }
2211 (void)sc; /* UNUSED */
2212 return (0); /* All Okey. */
2213 }
2214
2215 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2216
2217 /*
2218 * Basically returns -1 because we cannot make a conversion of charset
2219 * without iconv but in some cases this would return 0.
2220 * Returns 0 if all copied characters are ASCII.
2221 * Returns 0 if both from-locale and to-locale are the same and those
2222 * can be WCS with no error.
2223 */
2224 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2225 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2226 size_t length, struct archive_string_conv *sc)
2227 {
2228 size_t remaining;
2229 const uint8_t *itp;
2230 int return_value = 0; /* success */
2231
2232 /*
2233 * If both from-locale and to-locale is the same, this makes a copy.
2234 * And then this checks all copied MBS can be WCS if so returns 0.
2235 */
2236 if (sc->same) {
2237 if (archive_string_append(as, _p, length) == NULL)
2238 return (-1);/* No memory */
2239 return (invalid_mbs(_p, length, sc));
2240 }
2241
2242 /*
2243 * If a character is ASCII, this just copies it. If not, this
2244 * assigns '?' character instead but in UTF-8 locale this assigns
2245 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2246 * a Replacement Character in Unicode.
2247 */
2248
2249 remaining = length;
2250 itp = (const uint8_t *)_p;
2251 while (*itp && remaining > 0) {
2252 if (*itp > 127) {
2253 // Non-ASCII: Substitute with suitable replacement
2254 if (sc->flag & SCONV_TO_UTF8) {
2255 if (archive_string_append(as, utf8_replacement_char, sizeof(utf8_replacement_char)) == NULL) {
2256 __archive_errx(1, "Out of memory");
2257 }
2258 } else {
2259 archive_strappend_char(as, '?');
2260 }
2261 return_value = -1;
2262 } else {
2263 archive_strappend_char(as, *itp);
2264 }
2265 ++itp;
2266 }
2267 return (return_value);
2268 }
2269
2270
2271 /*
2272 * Unicode conversion functions.
2273 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2274 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2275 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2276 * - UTF-16BE <===> UTF-8.
2277 *
2278 */
2279
2280 /*
2281 * Utility to convert a single UTF-8 sequence.
2282 *
2283 * Usually return used bytes, return used byte in negative value when
2284 * a unicode character is replaced with U+FFFD.
2285 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2286 * Recommended Practice for Replacement Characters.
2287 */
2288 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2289 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2290 {
2291 static const char utf8_count[256] = {
2292 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2293 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2294 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2295 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2296 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2299 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2300 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2301 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2302 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2303 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2304 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2305 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2306 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2307 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2308 };
2309 int ch, i;
2310 int cnt;
2311 uint32_t wc;
2312
2313 /* Sanity check. */
2314 if (n == 0)
2315 return (0);
2316 /*
2317 * Decode 1-4 bytes depending on the value of the first byte.
2318 */
2319 ch = (unsigned char)*s;
2320 if (ch == 0)
2321 return (0); /* Standard: return 0 for end-of-string. */
2322 cnt = utf8_count[ch];
2323
2324 /* Invalid sequence or there are not plenty bytes. */
2325 if ((int)n < cnt) {
2326 cnt = (int)n;
2327 for (i = 1; i < cnt; i++) {
2328 if ((s[i] & 0xc0) != 0x80) {
2329 cnt = i;
2330 break;
2331 }
2332 }
2333 goto invalid_sequence;
2334 }
2335
2336 /* Make a Unicode code point from a single UTF-8 sequence. */
2337 switch (cnt) {
2338 case 1: /* 1 byte sequence. */
2339 *pwc = ch & 0x7f;
2340 return (cnt);
2341 case 2: /* 2 bytes sequence. */
2342 if ((s[1] & 0xc0) != 0x80) {
2343 cnt = 1;
2344 goto invalid_sequence;
2345 }
2346 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2347 return (cnt);
2348 case 3: /* 3 bytes sequence. */
2349 if ((s[1] & 0xc0) != 0x80) {
2350 cnt = 1;
2351 goto invalid_sequence;
2352 }
2353 if ((s[2] & 0xc0) != 0x80) {
2354 cnt = 2;
2355 goto invalid_sequence;
2356 }
2357 wc = ((ch & 0x0f) << 12)
2358 | ((s[1] & 0x3f) << 6)
2359 | (s[2] & 0x3f);
2360 if (wc < 0x800)
2361 goto invalid_sequence;/* Overlong sequence. */
2362 break;
2363 case 4: /* 4 bytes sequence. */
2364 if ((s[1] & 0xc0) != 0x80) {
2365 cnt = 1;
2366 goto invalid_sequence;
2367 }
2368 if ((s[2] & 0xc0) != 0x80) {
2369 cnt = 2;
2370 goto invalid_sequence;
2371 }
2372 if ((s[3] & 0xc0) != 0x80) {
2373 cnt = 3;
2374 goto invalid_sequence;
2375 }
2376 wc = ((ch & 0x07) << 18)
2377 | ((s[1] & 0x3f) << 12)
2378 | ((s[2] & 0x3f) << 6)
2379 | (s[3] & 0x3f);
2380 if (wc < 0x10000)
2381 goto invalid_sequence;/* Overlong sequence. */
2382 break;
2383 default: /* Others are all invalid sequence. */
2384 if (ch == 0xc0 || ch == 0xc1)
2385 cnt = 2;
2386 else if (ch >= 0xf5 && ch <= 0xf7)
2387 cnt = 4;
2388 else if (ch >= 0xf8 && ch <= 0xfb)
2389 cnt = 5;
2390 else if (ch == 0xfc || ch == 0xfd)
2391 cnt = 6;
2392 else
2393 cnt = 1;
2394 if ((int)n < cnt)
2395 cnt = (int)n;
2396 for (i = 1; i < cnt; i++) {
2397 if ((s[i] & 0xc0) != 0x80) {
2398 cnt = i;
2399 break;
2400 }
2401 }
2402 goto invalid_sequence;
2403 }
2404
2405 /* The code point larger than 0x10FFFF is not legal
2406 * Unicode values. */
2407 if (wc > UNICODE_MAX)
2408 goto invalid_sequence;
2409 /* Correctly gets a Unicode, returns used bytes. */
2410 *pwc = wc;
2411 return (cnt);
2412 invalid_sequence:
2413 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2414 return (cnt * -1);
2415 }
2416
2417 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2418 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2419 {
2420 int cnt;
2421
2422 cnt = _utf8_to_unicode(pwc, s, n);
2423 /* Any of Surrogate pair is not legal Unicode values. */
2424 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2425 return (-3);
2426 return (cnt);
2427 }
2428
2429 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2430 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2431 {
2432 uc -= 0xD800;
2433 uc *= 0x400;
2434 uc += uc2 - 0xDC00;
2435 uc += 0x10000;
2436 return (uc);
2437 }
2438
2439 /*
2440 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2441 * removing surrogate pairs.
2442 *
2443 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2444 *
2445 * Usually return used bytes, return used byte in negative value when
2446 * a unicode character is replaced with U+FFFD.
2447 */
2448 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2449 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2450 {
2451 uint32_t wc = 0;
2452 int cnt;
2453
2454 cnt = _utf8_to_unicode(&wc, s, n);
2455 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2456 uint32_t wc2 = 0;
2457 if (n - 3 < 3) {
2458 /* Invalid byte sequence. */
2459 goto invalid_sequence;
2460 }
2461 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2462 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2463 /* Invalid byte sequence. */
2464 goto invalid_sequence;
2465 }
2466 wc = combine_surrogate_pair(wc, wc2);
2467 cnt = 6;
2468 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2469 /* Invalid byte sequence. */
2470 goto invalid_sequence;
2471 }
2472 *pwc = wc;
2473 return (cnt);
2474 invalid_sequence:
2475 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2476 if (cnt > 0)
2477 cnt *= -1;
2478 return (cnt);
2479 }
2480
2481 /*
2482 * Convert a Unicode code point to a single UTF-8 sequence.
2483 *
2484 * NOTE:This function does not check if the Unicode is legal or not.
2485 * Please you definitely check it before calling this.
2486 */
2487 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2488 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2489 {
2490 char *_p = p;
2491
2492 /* Invalid Unicode char maps to Replacement character */
2493 if (uc > UNICODE_MAX)
2494 uc = UNICODE_R_CHAR;
2495 /* Translate code point to UTF8 */
2496 if (uc <= 0x7f) {
2497 if (remaining == 0)
2498 return (0);
2499 *p++ = (char)uc;
2500 } else if (uc <= 0x7ff) {
2501 if (remaining < 2)
2502 return (0);
2503 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2504 *p++ = 0x80 | (uc & 0x3f);
2505 } else if (uc <= 0xffff) {
2506 if (remaining < 3)
2507 return (0);
2508 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2509 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2510 *p++ = 0x80 | (uc & 0x3f);
2511 } else {
2512 if (remaining < 4)
2513 return (0);
2514 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2515 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2516 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2517 *p++ = 0x80 | (uc & 0x3f);
2518 }
2519 return (p - _p);
2520 }
2521
2522 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2523 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2524 {
2525 return (utf16_to_unicode(pwc, s, n, 1));
2526 }
2527
2528 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2529 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2530 {
2531 return (utf16_to_unicode(pwc, s, n, 0));
2532 }
2533
2534 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2535 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2536 {
2537 const char *utf16 = s;
2538 unsigned uc;
2539
2540 if (n == 0)
2541 return (0);
2542 if (n == 1) {
2543 /* set the Replacement Character instead. */
2544 *pwc = UNICODE_R_CHAR;
2545 return (-1);
2546 }
2547
2548 if (be)
2549 uc = archive_be16dec(utf16);
2550 else
2551 uc = archive_le16dec(utf16);
2552 utf16 += 2;
2553
2554 /* If this is a surrogate pair, assemble the full code point.*/
2555 if (IS_HIGH_SURROGATE_LA(uc)) {
2556 unsigned uc2;
2557
2558 if (n >= 4) {
2559 if (be)
2560 uc2 = archive_be16dec(utf16);
2561 else
2562 uc2 = archive_le16dec(utf16);
2563 } else
2564 uc2 = 0;
2565 if (IS_LOW_SURROGATE_LA(uc2)) {
2566 uc = combine_surrogate_pair(uc, uc2);
2567 utf16 += 2;
2568 } else {
2569 /* Undescribed code point should be U+FFFD
2570 * (replacement character). */
2571 *pwc = UNICODE_R_CHAR;
2572 return (-2);
2573 }
2574 }
2575
2576 /*
2577 * Surrogate pair values(0xd800 through 0xdfff) are only
2578 * used by UTF-16, so, after above calculation, the code
2579 * must not be surrogate values, and Unicode has no codes
2580 * larger than 0x10ffff. Thus, those are not legal Unicode
2581 * values.
2582 */
2583 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2584 /* Undescribed code point should be U+FFFD
2585 * (replacement character). */
2586 *pwc = UNICODE_R_CHAR;
2587 return (((int)(utf16 - s)) * -1);
2588 }
2589 *pwc = uc;
2590 return ((int)(utf16 - s));
2591 }
2592
2593 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2594 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2595 {
2596 char *utf16 = p;
2597
2598 if (uc > 0xffff) {
2599 /* We have a code point that won't fit into a
2600 * wchar_t; convert it to a surrogate pair. */
2601 if (remaining < 4)
2602 return (0);
2603 uc -= 0x10000;
2604 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2605 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2606 return (4);
2607 } else {
2608 if (remaining < 2)
2609 return (0);
2610 archive_be16enc(utf16, uc);
2611 return (2);
2612 }
2613 }
2614
2615 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2616 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2617 {
2618 char *utf16 = p;
2619
2620 if (uc > 0xffff) {
2621 /* We have a code point that won't fit into a
2622 * wchar_t; convert it to a surrogate pair. */
2623 if (remaining < 4)
2624 return (0);
2625 uc -= 0x10000;
2626 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2627 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2628 return (4);
2629 } else {
2630 if (remaining < 2)
2631 return (0);
2632 archive_le16enc(utf16, uc);
2633 return (2);
2634 }
2635 }
2636
2637 /*
2638 * Copy UTF-8 string in checking surrogate pair.
2639 * If any surrogate pair are found, it would be canonicalized.
2640 */
2641 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2642 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p,
2643 size_t len, struct archive_string_conv *sc)
2644 {
2645 const char *s;
2646 char *p, *endp;
2647 int n, ret = 0;
2648
2649 (void)sc; /* UNUSED */
2650
2651 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2652 return (-1);
2653
2654 s = (const char *)_p;
2655 p = as->s + as->length;
2656 endp = as->s + as->buffer_length -1;
2657 do {
2658 uint32_t uc;
2659 const char *ss = s;
2660 size_t w;
2661
2662 /*
2663 * Forward byte sequence until a conversion of that is needed.
2664 */
2665 while ((n = utf8_to_unicode(&uc, s, len)) > 0) {
2666 s += n;
2667 len -= n;
2668 }
2669 if (ss < s) {
2670 if (p + (s - ss) > endp) {
2671 as->length = p - as->s;
2672 if (archive_string_ensure(as,
2673 as->buffer_length + len + 1) == NULL)
2674 return (-1);
2675 p = as->s + as->length;
2676 endp = as->s + as->buffer_length -1;
2677 }
2678
2679 memcpy(p, ss, s - ss);
2680 p += s - ss;
2681 }
2682
2683 /*
2684 * If n is negative, current byte sequence needs a replacement.
2685 */
2686 if (n < 0) {
2687 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2688 /* Current byte sequence may be CESU-8. */
2689 n = cesu8_to_unicode(&uc, s, len);
2690 }
2691 if (n < 0) {
2692 ret = -1;
2693 n *= -1;/* Use a replaced unicode character. */
2694 }
2695
2696 /* Rebuild UTF-8 byte sequence. */
2697 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) {
2698 as->length = p - as->s;
2699 if (archive_string_ensure(as,
2700 as->buffer_length + len + 1) == NULL)
2701 return (-1);
2702 p = as->s + as->length;
2703 endp = as->s + as->buffer_length -1;
2704 }
2705 p += w;
2706 s += n;
2707 len -= n;
2708 }
2709 } while (n > 0);
2710 as->length = p - as->s;
2711 as->s[as->length] = '\0';
2712 return (ret);
2713 }
2714
2715 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2716 archive_string_append_unicode(struct archive_string *as, const void *_p,
2717 size_t len, struct archive_string_conv *sc)
2718 {
2719 const char *s;
2720 char *p, *endp;
2721 uint32_t uc;
2722 size_t w;
2723 int n, ret = 0, ts, tm;
2724 int (*parse)(uint32_t *, const char *, size_t);
2725 size_t (*unparse)(char *, size_t, uint32_t);
2726
2727 if (sc->flag & SCONV_TO_UTF16BE) {
2728 unparse = unicode_to_utf16be;
2729 ts = 2;
2730 } else if (sc->flag & SCONV_TO_UTF16LE) {
2731 unparse = unicode_to_utf16le;
2732 ts = 2;
2733 } else if (sc->flag & SCONV_TO_UTF8) {
2734 unparse = unicode_to_utf8;
2735 ts = 1;
2736 } else {
2737 /*
2738 * This case is going to be converted to another
2739 * character-set through iconv.
2740 */
2741 if (sc->flag & SCONV_FROM_UTF16BE) {
2742 unparse = unicode_to_utf16be;
2743 ts = 2;
2744 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2745 unparse = unicode_to_utf16le;
2746 ts = 2;
2747 } else {
2748 unparse = unicode_to_utf8;
2749 ts = 1;
2750 }
2751 }
2752
2753 if (sc->flag & SCONV_FROM_UTF16BE) {
2754 parse = utf16be_to_unicode;
2755 tm = 1;
2756 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2757 parse = utf16le_to_unicode;
2758 tm = 1;
2759 } else {
2760 parse = cesu8_to_unicode;
2761 tm = ts;
2762 }
2763
2764 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2765 return (-1);
2766
2767 s = (const char *)_p;
2768 p = as->s + as->length;
2769 endp = as->s + as->buffer_length - ts;
2770 while ((n = parse(&uc, s, len)) != 0) {
2771 if (n < 0) {
2772 /* Use a replaced unicode character. */
2773 n *= -1;
2774 ret = -1;
2775 }
2776 s += n;
2777 len -= n;
2778 while ((w = unparse(p, endp - p, uc)) == 0) {
2779 /* There is not enough output buffer so
2780 * we have to expand it. */
2781 as->length = p - as->s;
2782 if (archive_string_ensure(as,
2783 as->buffer_length + len * tm + ts) == NULL)
2784 return (-1);
2785 p = as->s + as->length;
2786 endp = as->s + as->buffer_length - ts;
2787 }
2788 p += w;
2789 }
2790 as->length = p - as->s;
2791 as->s[as->length] = '\0';
2792 if (ts == 2)
2793 as->s[as->length+1] = '\0';
2794 return (ret);
2795 }
2796
2797 /*
2798 * Following Constants for Hangul compositions this information comes from
2799 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2800 */
2801 #define HC_SBASE 0xAC00
2802 #define HC_LBASE 0x1100
2803 #define HC_VBASE 0x1161
2804 #define HC_TBASE 0x11A7
2805 #define HC_LCOUNT 19
2806 #define HC_VCOUNT 21
2807 #define HC_TCOUNT 28
2808 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2809 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2810
2811 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2812 get_nfc(uint32_t uc, uint32_t uc2)
2813 {
2814 int t, b;
2815
2816 t = 0;
2817 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2818 while (b >= t) {
2819 int m = (t + b) / 2;
2820 if (u_composition_table[m].cp1 < uc)
2821 t = m + 1;
2822 else if (u_composition_table[m].cp1 > uc)
2823 b = m - 1;
2824 else if (u_composition_table[m].cp2 < uc2)
2825 t = m + 1;
2826 else if (u_composition_table[m].cp2 > uc2)
2827 b = m - 1;
2828 else
2829 return (u_composition_table[m].nfc);
2830 }
2831 return (0);
2832 }
2833
2834 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2835 * Characters. */
2836
2837 /*
2838 * Update first code point.
2839 */
2840 #define UPDATE_UC(new_uc) do { \
2841 uc = new_uc; \
2842 ucptr = NULL; \
2843 } while (0)
2844
2845 /*
2846 * Replace first code point with second code point.
2847 */
2848 #define REPLACE_UC_WITH_UC2() do { \
2849 uc = uc2; \
2850 ucptr = uc2ptr; \
2851 n = n2; \
2852 } while (0)
2853
2854 #define EXPAND_BUFFER() do { \
2855 as->length = p - as->s; \
2856 if (archive_string_ensure(as, \
2857 as->buffer_length + len * tm + ts) == NULL)\
2858 return (-1); \
2859 p = as->s + as->length; \
2860 endp = as->s + as->buffer_length - ts; \
2861 } while (0)
2862
2863 #define UNPARSE(p, endp, uc) do { \
2864 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2865 EXPAND_BUFFER(); \
2866 } \
2867 p += w; \
2868 } while (0)
2869
2870 /*
2871 * Write first code point.
2872 * If the code point has not be changed from its original code,
2873 * this just copies it from its original buffer pointer.
2874 * If not, this converts it to UTF-8 byte sequence and copies it.
2875 */
2876 #define WRITE_UC() do { \
2877 if (ucptr) { \
2878 if (p + n > endp) \
2879 EXPAND_BUFFER(); \
2880 switch (n) { \
2881 case 4: \
2882 *p++ = *ucptr++; \
2883 /* FALL THROUGH */ \
2884 case 3: \
2885 *p++ = *ucptr++; \
2886 /* FALL THROUGH */ \
2887 case 2: \
2888 *p++ = *ucptr++; \
2889 /* FALL THROUGH */ \
2890 case 1: \
2891 *p++ = *ucptr; \
2892 break; \
2893 } \
2894 ucptr = NULL; \
2895 } else { \
2896 UNPARSE(p, endp, uc); \
2897 } \
2898 } while (0)
2899
2900 /*
2901 * Collect following decomposable code points.
2902 */
2903 #define COLLECT_CPS(start) do { \
2904 int _i; \
2905 for (_i = start; _i < FDC_MAX ; _i++) { \
2906 nx = parse(&ucx[_i], s, len); \
2907 if (nx <= 0) \
2908 break; \
2909 cx = CCC(ucx[_i]); \
2910 if (cl >= cx && cl != 228 && cx != 228)\
2911 break; \
2912 s += nx; \
2913 len -= nx; \
2914 cl = cx; \
2915 ccx[_i] = cx; \
2916 } \
2917 if (_i >= FDC_MAX) { \
2918 ret = -1; \
2919 ucx_size = FDC_MAX; \
2920 } else \
2921 ucx_size = _i; \
2922 } while (0)
2923
2924 /*
2925 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2926 *
2927 * TODO: Convert composition exclusions, which are never converted
2928 * from NFC,NFD,NFKC and NFKD, to Form C.
2929 */
2930 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2931 archive_string_normalize_C(struct archive_string *as, const void *_p,
2932 size_t len, struct archive_string_conv *sc)
2933 {
2934 const char *s = (const char *)_p;
2935 char *p, *endp;
2936 uint32_t uc, uc2;
2937 size_t w;
2938 int always_replace, n, n2, ret = 0, spair, ts, tm;
2939 int (*parse)(uint32_t *, const char *, size_t);
2940 size_t (*unparse)(char *, size_t, uint32_t);
2941
2942 always_replace = 1;
2943 ts = 1;/* text size. */
2944 if (sc->flag & SCONV_TO_UTF16BE) {
2945 unparse = unicode_to_utf16be;
2946 ts = 2;
2947 if (sc->flag & SCONV_FROM_UTF16BE)
2948 always_replace = 0;
2949 } else if (sc->flag & SCONV_TO_UTF16LE) {
2950 unparse = unicode_to_utf16le;
2951 ts = 2;
2952 if (sc->flag & SCONV_FROM_UTF16LE)
2953 always_replace = 0;
2954 } else if (sc->flag & SCONV_TO_UTF8) {
2955 unparse = unicode_to_utf8;
2956 if (sc->flag & SCONV_FROM_UTF8)
2957 always_replace = 0;
2958 } else {
2959 /*
2960 * This case is going to be converted to another
2961 * character-set through iconv.
2962 */
2963 always_replace = 0;
2964 if (sc->flag & SCONV_FROM_UTF16BE) {
2965 unparse = unicode_to_utf16be;
2966 ts = 2;
2967 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2968 unparse = unicode_to_utf16le;
2969 ts = 2;
2970 } else {
2971 unparse = unicode_to_utf8;
2972 }
2973 }
2974
2975 if (sc->flag & SCONV_FROM_UTF16BE) {
2976 parse = utf16be_to_unicode;
2977 tm = 1;
2978 spair = 4;/* surrogate pair size in UTF-16. */
2979 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2980 parse = utf16le_to_unicode;
2981 tm = 1;
2982 spair = 4;/* surrogate pair size in UTF-16. */
2983 } else {
2984 parse = cesu8_to_unicode;
2985 tm = ts;
2986 spair = 6;/* surrogate pair size in UTF-8. */
2987 }
2988
2989 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2990 return (-1);
2991
2992 p = as->s + as->length;
2993 endp = as->s + as->buffer_length - ts;
2994 while ((n = parse(&uc, s, len)) != 0) {
2995 const char *ucptr, *uc2ptr;
2996
2997 if (n < 0) {
2998 /* Use a replaced unicode character. */
2999 UNPARSE(p, endp, uc);
3000 s += n*-1;
3001 len -= n*-1;
3002 ret = -1;
3003 continue;
3004 } else if (n == spair || always_replace)
3005 /* uc is converted from a surrogate pair.
3006 * this should be treated as a changed code. */
3007 ucptr = NULL;
3008 else
3009 ucptr = s;
3010 s += n;
3011 len -= n;
3012
3013 /* Read second code point. */
3014 while ((n2 = parse(&uc2, s, len)) > 0) {
3015 uint32_t ucx[FDC_MAX];
3016 int ccx[FDC_MAX];
3017 int cl, cx, i, nx, ucx_size;
3018 int LIndex,SIndex;
3019 uint32_t nfc;
3020
3021 if (n2 == spair || always_replace)
3022 /* uc2 is converted from a surrogate pair.
3023 * this should be treated as a changed code. */
3024 uc2ptr = NULL;
3025 else
3026 uc2ptr = s;
3027 s += n2;
3028 len -= n2;
3029
3030 /*
3031 * If current second code point is out of decomposable
3032 * code points, finding compositions is unneeded.
3033 */
3034 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3035 WRITE_UC();
3036 REPLACE_UC_WITH_UC2();
3037 continue;
3038 }
3039
3040 /*
3041 * Try to combine current code points.
3042 */
3043 /*
3044 * We have to combine Hangul characters according to
3045 * http://uniicode.org/reports/tr15/#Hangul
3046 */
3047 if (0 <= (LIndex = uc - HC_LBASE) &&
3048 LIndex < HC_LCOUNT) {
3049 /*
3050 * Hangul Composition.
3051 * 1. Two current code points are L and V.
3052 */
3053 int VIndex = uc2 - HC_VBASE;
3054 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3055 /* Make syllable of form LV. */
3056 UPDATE_UC(HC_SBASE +
3057 (LIndex * HC_VCOUNT + VIndex) *
3058 HC_TCOUNT);
3059 } else {
3060 WRITE_UC();
3061 REPLACE_UC_WITH_UC2();
3062 }
3063 continue;
3064 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3065 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3066 /*
3067 * Hangul Composition.
3068 * 2. Two current code points are LV and T.
3069 */
3070 int TIndex = uc2 - HC_TBASE;
3071 if (0 < TIndex && TIndex < HC_TCOUNT) {
3072 /* Make syllable of form LVT. */
3073 UPDATE_UC(uc + TIndex);
3074 } else {
3075 WRITE_UC();
3076 REPLACE_UC_WITH_UC2();
3077 }
3078 continue;
3079 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3080 /* A composition to current code points
3081 * is found. */
3082 UPDATE_UC(nfc);
3083 continue;
3084 } else if ((cl = CCC(uc2)) == 0) {
3085 /* Clearly 'uc2' the second code point is not
3086 * a decomposable code. */
3087 WRITE_UC();
3088 REPLACE_UC_WITH_UC2();
3089 continue;
3090 }
3091
3092 /*
3093 * Collect following decomposable code points.
3094 */
3095 cx = 0;
3096 ucx[0] = uc2;
3097 ccx[0] = cl;
3098 COLLECT_CPS(1);
3099
3100 /*
3101 * Find a composed code in the collected code points.
3102 */
3103 i = 1;
3104 while (i < ucx_size) {
3105 int j;
3106
3107 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3108 i++;
3109 continue;
3110 }
3111
3112 /*
3113 * nfc is composed of uc and ucx[i].
3114 */
3115 UPDATE_UC(nfc);
3116
3117 /*
3118 * Remove ucx[i] by shifting
3119 * following code points.
3120 */
3121 for (j = i; j+1 < ucx_size; j++) {
3122 ucx[j] = ucx[j+1];
3123 ccx[j] = ccx[j+1];
3124 }
3125 ucx_size --;
3126
3127 /*
3128 * Collect following code points blocked
3129 * by ucx[i] the removed code point.
3130 */
3131 if (ucx_size > 0 && i == ucx_size &&
3132 nx > 0 && cx == cl) {
3133 cl = ccx[ucx_size-1];
3134 COLLECT_CPS(ucx_size);
3135 }
3136 /*
3137 * Restart finding a composed code with
3138 * the updated uc from the top of the
3139 * collected code points.
3140 */
3141 i = 0;
3142 }
3143
3144 /*
3145 * Apparently the current code points are not
3146 * decomposed characters or already composed.
3147 */
3148 WRITE_UC();
3149 for (i = 0; i < ucx_size; i++)
3150 UNPARSE(p, endp, ucx[i]);
3151
3152 /*
3153 * Flush out remaining canonical combining characters.
3154 */
3155 if (nx > 0 && cx == cl && len > 0) {
3156 while ((nx = parse(&ucx[0], s, len))
3157 > 0) {
3158 cx = CCC(ucx[0]);
3159 if (cl > cx)
3160 break;
3161 s += nx;
3162 len -= nx;
3163 cl = cx;
3164 UNPARSE(p, endp, ucx[0]);
3165 }
3166 }
3167 break;
3168 }
3169 if (n2 < 0) {
3170 WRITE_UC();
3171 /* Use a replaced unicode character. */
3172 UNPARSE(p, endp, uc2);
3173 s += n2*-1;
3174 len -= n2*-1;
3175 ret = -1;
3176 continue;
3177 } else if (n2 == 0) {
3178 WRITE_UC();
3179 break;
3180 }
3181 }
3182 as->length = p - as->s;
3183 as->s[as->length] = '\0';
3184 if (ts == 2)
3185 as->s[as->length+1] = '\0';
3186 return (ret);
3187 }
3188
3189 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3190 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3191 {
3192 int t, b;
3193
3194 /*
3195 * These are not converted to NFD on Mac OS.
3196 */
3197 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3198 (uc >= 0xF900 && uc <= 0xFAFF) ||
3199 (uc >= 0x2F800 && uc <= 0x2FAFF))
3200 return (0);
3201 /*
3202 * Those code points are not converted to NFD on Mac OS.
3203 * I do not know the reason because it is undocumented.
3204 * NFC NFD
3205 * 1109A ==> 11099 110BA
3206 * 1109C ==> 1109B 110BA
3207 * 110AB ==> 110A5 110BA
3208 */
3209 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3210 return (0);
3211
3212 t = 0;
3213 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3214 while (b >= t) {
3215 int m = (t + b) / 2;
3216 if (u_decomposition_table[m].nfc < uc)
3217 t = m + 1;
3218 else if (u_decomposition_table[m].nfc > uc)
3219 b = m - 1;
3220 else {
3221 *cp1 = u_decomposition_table[m].cp1;
3222 *cp2 = u_decomposition_table[m].cp2;
3223 return (1);
3224 }
3225 }
3226 return (0);
3227 }
3228
3229 #define REPLACE_UC_WITH(cp) do { \
3230 uc = cp; \
3231 ucptr = NULL; \
3232 } while (0)
3233
3234 /*
3235 * Normalize UTF-8 characters to Form D and copy the result.
3236 */
3237 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3238 archive_string_normalize_D(struct archive_string *as, const void *_p,
3239 size_t len, struct archive_string_conv *sc)
3240 {
3241 const char *s = (const char *)_p;
3242 char *p, *endp;
3243 uint32_t uc, uc2;
3244 size_t w;
3245 int always_replace, n, n2, ret = 0, spair, ts, tm;
3246 int (*parse)(uint32_t *, const char *, size_t);
3247 size_t (*unparse)(char *, size_t, uint32_t);
3248
3249 always_replace = 1;
3250 ts = 1;/* text size. */
3251 if (sc->flag & SCONV_TO_UTF16BE) {
3252 unparse = unicode_to_utf16be;
3253 ts = 2;
3254 if (sc->flag & SCONV_FROM_UTF16BE)
3255 always_replace = 0;
3256 } else if (sc->flag & SCONV_TO_UTF16LE) {
3257 unparse = unicode_to_utf16le;
3258 ts = 2;
3259 if (sc->flag & SCONV_FROM_UTF16LE)
3260 always_replace = 0;
3261 } else if (sc->flag & SCONV_TO_UTF8) {
3262 unparse = unicode_to_utf8;
3263 if (sc->flag & SCONV_FROM_UTF8)
3264 always_replace = 0;
3265 } else {
3266 /*
3267 * This case is going to be converted to another
3268 * character-set through iconv.
3269 */
3270 always_replace = 0;
3271 if (sc->flag & SCONV_FROM_UTF16BE) {
3272 unparse = unicode_to_utf16be;
3273 ts = 2;
3274 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3275 unparse = unicode_to_utf16le;
3276 ts = 2;
3277 } else {
3278 unparse = unicode_to_utf8;
3279 }
3280 }
3281
3282 if (sc->flag & SCONV_FROM_UTF16BE) {
3283 parse = utf16be_to_unicode;
3284 tm = 1;
3285 spair = 4;/* surrogate pair size in UTF-16. */
3286 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3287 parse = utf16le_to_unicode;
3288 tm = 1;
3289 spair = 4;/* surrogate pair size in UTF-16. */
3290 } else {
3291 parse = cesu8_to_unicode;
3292 tm = ts;
3293 spair = 6;/* surrogate pair size in UTF-8. */
3294 }
3295
3296 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3297 return (-1);
3298
3299 p = as->s + as->length;
3300 endp = as->s + as->buffer_length - ts;
3301 while ((n = parse(&uc, s, len)) != 0) {
3302 const char *ucptr;
3303 uint32_t cp1, cp2;
3304 int SIndex;
3305 struct {
3306 uint32_t uc;
3307 int ccc;
3308 } fdc[FDC_MAX];
3309 int fdi, fdj;
3310 int ccc;
3311
3312 check_first_code:
3313 if (n < 0) {
3314 /* Use a replaced unicode character. */
3315 UNPARSE(p, endp, uc);
3316 s += n*-1;
3317 len -= n*-1;
3318 ret = -1;
3319 continue;
3320 } else if (n == spair || always_replace)
3321 /* uc is converted from a surrogate pair.
3322 * this should be treated as a changed code. */
3323 ucptr = NULL;
3324 else
3325 ucptr = s;
3326 s += n;
3327 len -= n;
3328
3329 /* Hangul Decomposition. */
3330 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3331 int L = HC_LBASE + SIndex / HC_NCOUNT;
3332 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3333 int T = HC_TBASE + SIndex % HC_TCOUNT;
3334
3335 REPLACE_UC_WITH(L);
3336 WRITE_UC();
3337 REPLACE_UC_WITH(V);
3338 WRITE_UC();
3339 if (T != HC_TBASE) {
3340 REPLACE_UC_WITH(T);
3341 WRITE_UC();
3342 }
3343 continue;
3344 }
3345 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3346 WRITE_UC();
3347 continue;
3348 }
3349
3350 fdi = 0;
3351 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3352 int k;
3353
3354 for (k = fdi; k > 0; k--)
3355 fdc[k] = fdc[k-1];
3356 fdc[0].ccc = CCC(cp2);
3357 fdc[0].uc = cp2;
3358 fdi++;
3359 REPLACE_UC_WITH(cp1);
3360 }
3361
3362 /* Read following code points. */
3363 while ((n2 = parse(&uc2, s, len)) > 0 &&
3364 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3365 int j, k;
3366
3367 s += n2;
3368 len -= n2;
3369 for (j = 0; j < fdi; j++) {
3370 if (fdc[j].ccc > ccc)
3371 break;
3372 }
3373 if (j < fdi) {
3374 for (k = fdi; k > j; k--)
3375 fdc[k] = fdc[k-1];
3376 fdc[j].ccc = ccc;
3377 fdc[j].uc = uc2;
3378 } else {
3379 fdc[fdi].ccc = ccc;
3380 fdc[fdi].uc = uc2;
3381 }
3382 fdi++;
3383 }
3384
3385 WRITE_UC();
3386 for (fdj = 0; fdj < fdi; fdj++) {
3387 REPLACE_UC_WITH(fdc[fdj].uc);
3388 WRITE_UC();
3389 }
3390
3391 if (n2 == 0)
3392 break;
3393 REPLACE_UC_WITH(uc2);
3394 n = n2;
3395 goto check_first_code;
3396 }
3397 as->length = p - as->s;
3398 as->s[as->length] = '\0';
3399 if (ts == 2)
3400 as->s[as->length+1] = '\0';
3401 return (ret);
3402 }
3403
3404 /*
3405 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3406 * that WCS is Unicode. It is true for several platforms but some are false.
3407 * And then people who did not use UTF-8 locale on the non Unicode WCS
3408 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3409 * now cannot get right filename from libarchive 3.x and later since we
3410 * fixed the wrong assumption and it is incompatible to older its versions.
3411 * So we provide special option, "compat-2x.x", for resolving it.
3412 * That option enable the string conversion of libarchive 2.x.
3413 *
3414 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3415 * locale character set and appends to the archive_string.
3416 * Note: returns -1 if conversion fails.
3417 */
3418 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3419 strncat_from_utf8_libarchive2(struct archive_string *as,
3420 const void *_p, size_t len, struct archive_string_conv *sc)
3421 {
3422 const char *s;
3423 int n;
3424 char *p;
3425 char *end;
3426 uint32_t unicode;
3427 #if HAVE_WCRTOMB
3428 mbstate_t shift_state;
3429
3430 memset(&shift_state, 0, sizeof(shift_state));
3431 #else
3432 /* Clear the shift state before starting. */
3433 wctomb(NULL, L'\0');
3434 #endif
3435 (void)sc; /* UNUSED */
3436 /*
3437 * Allocate buffer for MBS.
3438 * We need this allocation here since it is possible that
3439 * as->s is still NULL.
3440 */
3441 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3442 return (-1);
3443
3444 s = (const char *)_p;
3445 p = as->s + as->length;
3446 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3447 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3448 wchar_t wc;
3449
3450 if (p >= end) {
3451 as->length = p - as->s;
3452 /* Re-allocate buffer for MBS. */
3453 if (archive_string_ensure(as,
3454 as->length + max(len * 2,
3455 (size_t)MB_CUR_MAX) + 1) == NULL)
3456 return (-1);
3457 p = as->s + as->length;
3458 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3459 }
3460
3461 /*
3462 * As libarchive 2.x, translates the UTF-8 characters into
3463 * wide-characters in the assumption that WCS is Unicode.
3464 */
3465 if (n < 0) {
3466 n *= -1;
3467 wc = L'?';
3468 } else
3469 wc = (wchar_t)unicode;
3470
3471 s += n;
3472 len -= n;
3473 /*
3474 * Translates the wide-character into the current locale MBS.
3475 */
3476 #if HAVE_WCRTOMB
3477 n = (int)wcrtomb(p, wc, &shift_state);
3478 #else
3479 n = (int)wctomb(p, wc);
3480 #endif
3481 if (n == -1)
3482 return (-1);
3483 p += n;
3484 }
3485 as->length = p - as->s;
3486 as->s[as->length] = '\0';
3487 return (0);
3488 }
3489
3490
3491 /*
3492 * Conversion functions between current locale dependent MBS and UTF-16BE.
3493 * strncat_from_utf16be() : UTF-16BE --> MBS
3494 * strncat_to_utf16be() : MBS --> UTF16BE
3495 */
3496
3497 #if defined(_WIN32) && !defined(__CYGWIN__)
3498
3499 /*
3500 * Convert a UTF-16BE/LE string to current locale and copy the result.
3501 * Return -1 if conversion fails.
3502 */
3503 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3504 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3505 struct archive_string_conv *sc, int be)
3506 {
3507 struct archive_string tmp;
3508 const char *u16;
3509 int ll;
3510 BOOL defchar;
3511 char *mbs;
3512 size_t mbs_size, b;
3513 int ret = 0;
3514
3515 bytes &= ~1;
3516 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3517 return (-1);
3518
3519 mbs = as->s + as->length;
3520 mbs_size = as->buffer_length - as->length -1;
3521
3522 if (sc->to_cp == CP_C_LOCALE) {
3523 /*
3524 * "C" locale special process.
3525 */
3526 u16 = _p;
3527 ll = 0;
3528 for (b = 0; b < bytes; b += 2) {
3529 uint16_t val;
3530 if (be)
3531 val = archive_be16dec(u16+b);
3532 else
3533 val = archive_le16dec(u16+b);
3534 if (val > 255) {
3535 *mbs++ = '?';
3536 ret = -1;
3537 } else
3538 *mbs++ = (char)(val&0xff);
3539 ll++;
3540 }
3541 as->length += ll;
3542 as->s[as->length] = '\0';
3543 return (ret);
3544 }
3545
3546 archive_string_init(&tmp);
3547 if (be) {
3548 if (is_big_endian()) {
3549 u16 = _p;
3550 } else {
3551 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3552 return (-1);
3553 memcpy(tmp.s, _p, bytes);
3554 for (b = 0; b < bytes; b += 2) {
3555 uint16_t val = archive_be16dec(tmp.s+b);
3556 archive_le16enc(tmp.s+b, val);
3557 }
3558 u16 = tmp.s;
3559 }
3560 } else {
3561 if (!is_big_endian()) {
3562 u16 = _p;
3563 } else {
3564 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3565 return (-1);
3566 memcpy(tmp.s, _p, bytes);
3567 for (b = 0; b < bytes; b += 2) {
3568 uint16_t val = archive_le16dec(tmp.s+b);
3569 archive_be16enc(tmp.s+b, val);
3570 }
3571 u16 = tmp.s;
3572 }
3573 }
3574
3575 do {
3576 defchar = 0;
3577 ll = WideCharToMultiByte(sc->to_cp, 0,
3578 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3579 NULL, &defchar);
3580 /* Exit loop if we succeeded */
3581 if (ll != 0 ||
3582 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3583 break;
3584 }
3585 /* Else expand buffer and loop to try again. */
3586 ll = WideCharToMultiByte(sc->to_cp, 0,
3587 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3588 if (archive_string_ensure(as, ll +1) == NULL)
3589 return (-1);
3590 mbs = as->s + as->length;
3591 mbs_size = as->buffer_length - as->length -1;
3592 } while (1);
3593 archive_string_free(&tmp);
3594 as->length += ll;
3595 as->s[as->length] = '\0';
3596 if (ll == 0 || defchar)
3597 ret = -1;
3598 return (ret);
3599 }
3600
3601 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3602 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3603 size_t bytes, struct archive_string_conv *sc)
3604 {
3605 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3606 }
3607
3608 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3609 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3610 size_t bytes, struct archive_string_conv *sc)
3611 {
3612 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3613 }
3614
3615 static int
is_big_endian(void)3616 is_big_endian(void)
3617 {
3618 uint16_t d = 1;
3619
3620 return (archive_be16dec(&d) == 1);
3621 }
3622
3623 /*
3624 * Convert a current locale string to UTF-16BE/LE and copy the result.
3625 * Return -1 if conversion fails.
3626 */
3627 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3628 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3629 size_t length, struct archive_string_conv *sc, int bigendian)
3630 {
3631 const char *s = (const char *)_p;
3632 char *u16;
3633 size_t count, avail;
3634
3635 if (archive_string_ensure(as16,
3636 as16->length + (length + 1) * 2) == NULL)
3637 return (-1);
3638
3639 u16 = as16->s + as16->length;
3640 avail = as16->buffer_length - 2;
3641 if (sc->from_cp == CP_C_LOCALE) {
3642 /*
3643 * "C" locale special process.
3644 */
3645 count = 0;
3646 while (count < length && *s) {
3647 if (bigendian)
3648 archive_be16enc(u16, *s);
3649 else
3650 archive_le16enc(u16, *s);
3651 u16 += 2;
3652 s++;
3653 count++;
3654 }
3655 as16->length += count << 1;
3656 as16->s[as16->length] = 0;
3657 as16->s[as16->length+1] = 0;
3658 return (0);
3659 }
3660 do {
3661 count = MultiByteToWideChar(sc->from_cp,
3662 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3663 /* Exit loop if we succeeded */
3664 if (count != 0 ||
3665 GetLastError() != ERROR_INSUFFICIENT_BUFFER) {
3666 break;
3667 }
3668 /* Expand buffer and try again */
3669 count = MultiByteToWideChar(sc->from_cp,
3670 MB_PRECOMPOSED, s, (int)length, NULL, 0);
3671 if (archive_string_ensure(as16, (count +1) * 2)
3672 == NULL)
3673 return (-1);
3674 u16 = as16->s + as16->length;
3675 avail = as16->buffer_length - 2;
3676 } while (1);
3677 as16->length += count * 2;
3678 as16->s[as16->length] = 0;
3679 as16->s[as16->length+1] = 0;
3680 if (count == 0)
3681 return (-1);
3682
3683 if (is_big_endian()) {
3684 if (!bigendian) {
3685 while (count > 0) {
3686 uint16_t v = archive_be16dec(u16);
3687 archive_le16enc(u16, v);
3688 u16 += 2;
3689 count--;
3690 }
3691 }
3692 } else {
3693 if (bigendian) {
3694 while (count > 0) {
3695 uint16_t v = archive_le16dec(u16);
3696 archive_be16enc(u16, v);
3697 u16 += 2;
3698 count--;
3699 }
3700 }
3701 }
3702 return (0);
3703 }
3704
3705 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3706 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3707 size_t length, struct archive_string_conv *sc)
3708 {
3709 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3710 }
3711
3712 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3713 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3714 size_t length, struct archive_string_conv *sc)
3715 {
3716 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3717 }
3718
3719 #endif /* _WIN32 && !__CYGWIN__ */
3720
3721 /*
3722 * Do the best effort for conversions.
3723 * We cannot handle UTF-16BE character-set without such iconv,
3724 * but there is a chance if a string consists just ASCII code or
3725 * a current locale is UTF-8.
3726 */
3727
3728 /*
3729 * Convert a UTF-16BE string to current locale and copy the result.
3730 * Return -1 if conversion fails.
3731 */
3732 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3733 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3734 size_t bytes, struct archive_string_conv *sc, int be)
3735 {
3736 const char *utf16 = (const char *)_p;
3737 char *mbs;
3738 uint32_t uc;
3739 int n, ret;
3740
3741 (void)sc; /* UNUSED */
3742 /*
3743 * Other case, we should do the best effort.
3744 * If all character are ASCII(<0x7f), we can convert it.
3745 * if not , we set a alternative character and return -1.
3746 */
3747 ret = 0;
3748 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3749 return (-1);
3750 mbs = as->s + as->length;
3751
3752 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3753 if (n < 0) {
3754 n *= -1;
3755 ret = -1;
3756 }
3757 bytes -= n;
3758 utf16 += n;
3759
3760 if (uc > 127) {
3761 /* We cannot handle it. */
3762 *mbs++ = '?';
3763 ret = -1;
3764 } else
3765 *mbs++ = (char)uc;
3766 }
3767 as->length = mbs - as->s;
3768 as->s[as->length] = '\0';
3769 return (ret);
3770 }
3771
3772 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3773 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3774 size_t bytes, struct archive_string_conv *sc)
3775 {
3776 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3777 }
3778
3779 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3780 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3781 size_t bytes, struct archive_string_conv *sc)
3782 {
3783 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3784 }
3785
3786 /*
3787 * Convert a current locale string to UTF-16BE/LE and copy the result.
3788 * Return -1 if conversion fails.
3789 */
3790 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3791 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3792 size_t length, struct archive_string_conv *sc, int bigendian)
3793 {
3794 const char *s = (const char *)_p;
3795 char *utf16;
3796 size_t remaining;
3797 int ret;
3798
3799 (void)sc; /* UNUSED */
3800 /*
3801 * Other case, we should do the best effort.
3802 * If all character are ASCII(<0x7f), we can convert it.
3803 * if not , we set a alternative character and return -1.
3804 */
3805 ret = 0;
3806 remaining = length;
3807
3808 if (archive_string_ensure(as16,
3809 as16->length + (length + 1) * 2) == NULL)
3810 return (-1);
3811
3812 utf16 = as16->s + as16->length;
3813 while (remaining--) {
3814 unsigned c = *s++;
3815 if (c > 127) {
3816 /* We cannot handle it. */
3817 c = UNICODE_R_CHAR;
3818 ret = -1;
3819 }
3820 if (bigendian)
3821 archive_be16enc(utf16, c);
3822 else
3823 archive_le16enc(utf16, c);
3824 utf16 += 2;
3825 }
3826 as16->length = utf16 - as16->s;
3827 as16->s[as16->length] = 0;
3828 as16->s[as16->length+1] = 0;
3829 return (ret);
3830 }
3831
3832 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3833 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3834 size_t length, struct archive_string_conv *sc)
3835 {
3836 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3837 }
3838
3839 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3840 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3841 size_t length, struct archive_string_conv *sc)
3842 {
3843 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3844 }
3845
3846
3847 /*
3848 * Multistring operations.
3849 */
3850
3851 void
archive_mstring_clean(struct archive_mstring * aes)3852 archive_mstring_clean(struct archive_mstring *aes)
3853 {
3854 archive_wstring_free(&(aes->aes_wcs));
3855 archive_string_free(&(aes->aes_mbs));
3856 archive_string_free(&(aes->aes_utf8));
3857 archive_string_free(&(aes->aes_mbs_in_locale));
3858 aes->aes_set = 0;
3859 }
3860
3861 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3862 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3863 {
3864 dest->aes_set = src->aes_set;
3865 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3866 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3867 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3868 }
3869
3870 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3871 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3872 const char **p)
3873 {
3874 struct archive_string_conv *sc;
3875 int r;
3876
3877 /* If we already have a UTF8 form, return that immediately. */
3878 if (aes->aes_set & AES_SET_UTF8) {
3879 *p = aes->aes_utf8.s;
3880 return (0);
3881 }
3882
3883 *p = NULL;
3884 /* Try converting WCS to MBS first if MBS does not exist yet. */
3885 if ((aes->aes_set & AES_SET_MBS) == 0) {
3886 const char *pm; /* unused */
3887 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
3888 }
3889 if (aes->aes_set & AES_SET_MBS) {
3890 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3891 if (sc == NULL)
3892 return (-1);/* Couldn't allocate memory for sc. */
3893 r = archive_strncpy_l(&(aes->aes_utf8), aes->aes_mbs.s,
3894 aes->aes_mbs.length, sc);
3895 if (a == NULL)
3896 free_sconv_object(sc);
3897 if (r == 0) {
3898 aes->aes_set |= AES_SET_UTF8;
3899 *p = aes->aes_utf8.s;
3900 return (0);/* success. */
3901 } else
3902 return (-1);/* failure. */
3903 }
3904 return (0);/* success. */
3905 }
3906
3907 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3908 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3909 const char **p)
3910 {
3911 struct archive_string_conv *sc;
3912 int r, ret = 0;
3913
3914 /* If we already have an MBS form, return that immediately. */
3915 if (aes->aes_set & AES_SET_MBS) {
3916 *p = aes->aes_mbs.s;
3917 return (ret);
3918 }
3919
3920 *p = NULL;
3921 /* If there's a WCS form, try converting with the native locale. */
3922 if (aes->aes_set & AES_SET_WCS) {
3923 archive_string_empty(&(aes->aes_mbs));
3924 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3925 aes->aes_wcs.s, aes->aes_wcs.length);
3926 *p = aes->aes_mbs.s;
3927 if (r == 0) {
3928 aes->aes_set |= AES_SET_MBS;
3929 return (ret);
3930 } else
3931 ret = -1;
3932 }
3933
3934 /* If there's a UTF-8 form, try converting with the native locale. */
3935 if (aes->aes_set & AES_SET_UTF8) {
3936 archive_string_empty(&(aes->aes_mbs));
3937 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
3938 if (sc == NULL)
3939 return (-1);/* Couldn't allocate memory for sc. */
3940 r = archive_strncpy_l(&(aes->aes_mbs),
3941 aes->aes_utf8.s, aes->aes_utf8.length, sc);
3942 if (a == NULL)
3943 free_sconv_object(sc);
3944 *p = aes->aes_mbs.s;
3945 if (r == 0) {
3946 aes->aes_set |= AES_SET_MBS;
3947 ret = 0;/* success; overwrite previous error. */
3948 } else
3949 ret = -1;/* failure. */
3950 }
3951 return (ret);
3952 }
3953
3954 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)3955 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3956 const wchar_t **wp)
3957 {
3958 int r, ret = 0;
3959
3960 (void)a;/* UNUSED */
3961 /* Return WCS form if we already have it. */
3962 if (aes->aes_set & AES_SET_WCS) {
3963 *wp = aes->aes_wcs.s;
3964 return (ret);
3965 }
3966
3967 *wp = NULL;
3968 /* Try converting UTF8 to MBS first if MBS does not exist yet. */
3969 if ((aes->aes_set & AES_SET_MBS) == 0) {
3970 const char *p; /* unused */
3971 archive_mstring_get_mbs(a, aes, &p); /* ignore errors, we'll handle it later */
3972 }
3973 /* Try converting MBS to WCS using native locale. */
3974 if (aes->aes_set & AES_SET_MBS) {
3975 archive_wstring_empty(&(aes->aes_wcs));
3976 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
3977 aes->aes_mbs.s, aes->aes_mbs.length);
3978 if (r == 0) {
3979 aes->aes_set |= AES_SET_WCS;
3980 *wp = aes->aes_wcs.s;
3981 } else
3982 ret = -1;/* failure. */
3983 }
3984 return (ret);
3985 }
3986
3987 int
archive_mstring_get_mbs_l(struct archive * a,struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)3988 archive_mstring_get_mbs_l(struct archive *a, struct archive_mstring *aes,
3989 const char **p, size_t *length, struct archive_string_conv *sc)
3990 {
3991 int r, ret = 0;
3992
3993 (void)r; /* UNUSED */
3994 #if defined(_WIN32) && !defined(__CYGWIN__)
3995 /*
3996 * Internationalization programming on Windows must use Wide
3997 * characters because Windows platform cannot make locale UTF-8.
3998 */
3999 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
4000 archive_string_empty(&(aes->aes_mbs_in_locale));
4001 r = archive_string_append_from_wcs_in_codepage(
4002 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
4003 aes->aes_wcs.length, sc);
4004 if (r == 0) {
4005 *p = aes->aes_mbs_in_locale.s;
4006 if (length != NULL)
4007 *length = aes->aes_mbs_in_locale.length;
4008 return (0);
4009 } else if (errno == ENOMEM)
4010 return (-1);
4011 else
4012 ret = -1;
4013 }
4014 #endif
4015
4016 /* If there is not an MBS form but there is a WCS or UTF8 form, try converting
4017 * with the native locale to be used for translating it to specified
4018 * character-set. */
4019 if ((aes->aes_set & AES_SET_MBS) == 0) {
4020 const char *pm; /* unused */
4021 archive_mstring_get_mbs(a, aes, &pm); /* ignore errors, we'll handle it later */
4022 }
4023 /* If we already have an MBS form, use it to be translated to
4024 * specified character-set. */
4025 if (aes->aes_set & AES_SET_MBS) {
4026 if (sc == NULL) {
4027 /* Conversion is unneeded. */
4028 *p = aes->aes_mbs.s;
4029 if (length != NULL)
4030 *length = aes->aes_mbs.length;
4031 return (0);
4032 }
4033 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4034 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4035 *p = aes->aes_mbs_in_locale.s;
4036 if (length != NULL)
4037 *length = aes->aes_mbs_in_locale.length;
4038 } else {
4039 *p = NULL;
4040 if (length != NULL)
4041 *length = 0;
4042 }
4043 return (ret);
4044 }
4045
4046 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4047 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4048 {
4049 if (mbs == NULL) {
4050 aes->aes_set = 0;
4051 return (0);
4052 }
4053 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4054 }
4055
4056 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4057 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4058 size_t len)
4059 {
4060 if (mbs == NULL) {
4061 aes->aes_set = 0;
4062 return (0);
4063 }
4064 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4065 archive_strncpy(&(aes->aes_mbs), mbs, len);
4066 archive_string_empty(&(aes->aes_utf8));
4067 archive_wstring_empty(&(aes->aes_wcs));
4068 return (0);
4069 }
4070
4071 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4072 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4073 {
4074 return archive_mstring_copy_wcs_len(aes, wcs,
4075 wcs == NULL ? 0 : wcslen(wcs));
4076 }
4077
4078 int
archive_mstring_copy_utf8(struct archive_mstring * aes,const char * utf8)4079 archive_mstring_copy_utf8(struct archive_mstring *aes, const char *utf8)
4080 {
4081 if (utf8 == NULL) {
4082 aes->aes_set = 0;
4083 return (0);
4084 }
4085 aes->aes_set = AES_SET_UTF8;
4086 archive_string_empty(&(aes->aes_mbs));
4087 archive_string_empty(&(aes->aes_wcs));
4088 archive_strncpy(&(aes->aes_utf8), utf8, strlen(utf8));
4089 return (int)strlen(utf8);
4090 }
4091
4092 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4093 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4094 size_t len)
4095 {
4096 if (wcs == NULL) {
4097 aes->aes_set = 0;
4098 return (0);
4099 }
4100 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4101 archive_string_empty(&(aes->aes_mbs));
4102 archive_string_empty(&(aes->aes_utf8));
4103 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4104 return (0);
4105 }
4106
4107 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4108 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4109 const char *mbs, size_t len, struct archive_string_conv *sc)
4110 {
4111 int r;
4112
4113 if (mbs == NULL) {
4114 aes->aes_set = 0;
4115 return (0);
4116 }
4117 archive_string_empty(&(aes->aes_mbs));
4118 archive_wstring_empty(&(aes->aes_wcs));
4119 archive_string_empty(&(aes->aes_utf8));
4120 #if defined(_WIN32) && !defined(__CYGWIN__)
4121 /*
4122 * Internationalization programming on Windows must use Wide
4123 * characters because Windows platform cannot make locale UTF-8.
4124 */
4125 if (sc == NULL) {
4126 if (archive_string_append(&(aes->aes_mbs),
4127 mbs, mbsnbytes(mbs, len)) == NULL) {
4128 aes->aes_set = 0;
4129 r = -1;
4130 } else {
4131 aes->aes_set = AES_SET_MBS;
4132 r = 0;
4133 }
4134 #if defined(HAVE_ICONV)
4135 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4136 /*
4137 * This case happens only when MultiByteToWideChar() cannot
4138 * handle sc->from_cp, and we have to iconv in order to
4139 * translate character-set to wchar_t,UTF-16.
4140 */
4141 iconv_t cd = sc->cd;
4142 unsigned from_cp;
4143 int flag;
4144
4145 /*
4146 * Translate multi-bytes from some character-set to UTF-8.
4147 */
4148 sc->cd = sc->cd_w;
4149 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4150 sc->cd = cd;
4151 if (r != 0) {
4152 aes->aes_set = 0;
4153 return (r);
4154 }
4155 aes->aes_set = AES_SET_UTF8;
4156
4157 /*
4158 * Append the UTF-8 string into wstring.
4159 */
4160 flag = sc->flag;
4161 sc->flag &= ~(SCONV_NORMALIZATION_C
4162 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4163 from_cp = sc->from_cp;
4164 sc->from_cp = CP_UTF8;
4165 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4166 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4167 sc->flag = flag;
4168 sc->from_cp = from_cp;
4169 if (r == 0)
4170 aes->aes_set |= AES_SET_WCS;
4171 #endif
4172 } else {
4173 r = archive_wstring_append_from_mbs_in_codepage(
4174 &(aes->aes_wcs), mbs, len, sc);
4175 if (r == 0)
4176 aes->aes_set = AES_SET_WCS;
4177 else
4178 aes->aes_set = 0;
4179 }
4180 #else
4181 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4182 if (r == 0)
4183 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4184 else
4185 aes->aes_set = 0;
4186 #endif
4187 return (r);
4188 }
4189
4190 /*
4191 * The 'update' form tries to proactively update all forms of
4192 * this string (WCS and MBS) and returns an error if any of
4193 * them fail. This is used by the 'pax' handler, for instance,
4194 * to detect and report character-conversion failures early while
4195 * still allowing clients to get potentially useful values from
4196 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4197 * strive to give the user something useful, so you can get hopefully
4198 * usable values even if some of the character conversions are failing.)
4199 */
4200 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4201 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4202 const char *utf8)
4203 {
4204 struct archive_string_conv *sc;
4205 int r;
4206
4207 if (utf8 == NULL) {
4208 aes->aes_set = 0;
4209 return (0); /* Succeeded in clearing everything. */
4210 }
4211
4212 /* Save the UTF8 string. */
4213 archive_strcpy(&(aes->aes_utf8), utf8);
4214
4215 /* Empty the mbs and wcs strings. */
4216 archive_string_empty(&(aes->aes_mbs));
4217 archive_wstring_empty(&(aes->aes_wcs));
4218
4219 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4220
4221 /* Try converting UTF-8 to MBS, return false on failure. */
4222 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4223 if (sc == NULL)
4224 return (-1);/* Couldn't allocate memory for sc. */
4225 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4226 if (a == NULL)
4227 free_sconv_object(sc);
4228 if (r != 0)
4229 return (-1);
4230 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4231
4232 /* Try converting MBS to WCS, return false on failure. */
4233 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4234 aes->aes_mbs.length))
4235 return (-1);
4236 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4237
4238 /* All conversions succeeded. */
4239 return (0);
4240 }
4241