1 /*-
2 * Copyright (c) 2003-2011 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27 #include "archive_platform.h"
28 __FBSDID("$FreeBSD$");
29
30 /*
31 * Basic resizable string support, to simplify manipulating arbitrary-sized
32 * strings while minimizing heap activity.
33 *
34 * In particular, the buffer used by a string object is only grown, it
35 * never shrinks, so you can clear and reuse the same string object
36 * without incurring additional memory allocations.
37 */
38
39 #ifdef HAVE_ERRNO_H
40 #include <errno.h>
41 #endif
42 #ifdef HAVE_ICONV_H
43 #include <iconv.h>
44 #endif
45 #ifdef HAVE_LANGINFO_H
46 #include <langinfo.h>
47 #endif
48 #ifdef HAVE_LOCALCHARSET_H
49 #include <localcharset.h>
50 #endif
51 #ifdef HAVE_STDLIB_H
52 #include <stdlib.h>
53 #endif
54 #ifdef HAVE_STRING_H
55 #include <string.h>
56 #endif
57 #ifdef HAVE_WCHAR_H
58 #include <wchar.h>
59 #endif
60 #if defined(_WIN32) && !defined(__CYGWIN__)
61 #include <windows.h>
62 #include <locale.h>
63 #endif
64
65 #include "archive_endian.h"
66 #include "archive_private.h"
67 #include "archive_string.h"
68 #include "archive_string_composition.h"
69
70 #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy)
71 #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t))
72 #endif
73
74 struct archive_string_conv {
75 struct archive_string_conv *next;
76 char *from_charset;
77 char *to_charset;
78 unsigned from_cp;
79 unsigned to_cp;
80 /* Set 1 if from_charset and to_charset are the same. */
81 int same;
82 int flag;
83 #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified
84 * charset. */
85 #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from
86 * specified charset. */
87 #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */
88 #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting
89 * MBS. */
90 #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive
91 * 2.x in the wrong assumption. */
92 #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C.
93 * Before UTF-8 characters are actually
94 * processed. */
95 #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D.
96 * Before UTF-8 characters are actually
97 * processed.
98 * Currently this only for MAC OS X. */
99 #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */
100 #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */
101 #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */
102 #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */
103 #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */
104 #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */
105 #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE)
106 #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE)
107
108 #if HAVE_ICONV
109 iconv_t cd;
110 iconv_t cd_w;/* Use at archive_mstring on
111 * Windows. */
112 #endif
113 /* A temporary buffer for normalization. */
114 struct archive_string utftmp;
115 int (*converter[2])(struct archive_string *, const void *, size_t,
116 struct archive_string_conv *);
117 int nconverter;
118 };
119
120 #define CP_C_LOCALE 0 /* "C" locale only for this file. */
121 #define CP_UTF16LE 1200
122 #define CP_UTF16BE 1201
123
124 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF)
125 #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF)
126 #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF)
127 #define UNICODE_MAX 0x10FFFF
128 #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */
129 /* Set U+FFFD(Replacement character) in UTF-8. */
130 #define UTF8_SET_R_CHAR(outp) do { \
131 (outp)[0] = 0xef; \
132 (outp)[1] = 0xbf; \
133 (outp)[2] = 0xbd; \
134 } while (0)
135 #define UTF8_R_CHAR_SIZE 3
136
137 static struct archive_string_conv *find_sconv_object(struct archive *,
138 const char *, const char *);
139 static void add_sconv_object(struct archive *, struct archive_string_conv *);
140 static struct archive_string_conv *create_sconv_object(const char *,
141 const char *, unsigned, int);
142 static void free_sconv_object(struct archive_string_conv *);
143 static struct archive_string_conv *get_sconv_object(struct archive *,
144 const char *, const char *, int);
145 static unsigned make_codepage_from_charset(const char *);
146 static unsigned get_current_codepage(void);
147 static unsigned get_current_oemcp(void);
148 static size_t mbsnbytes(const void *, size_t);
149 static size_t utf16nbytes(const void *, size_t);
150 #if defined(_WIN32) && !defined(__CYGWIN__)
151 static int archive_wstring_append_from_mbs_in_codepage(
152 struct archive_wstring *, const char *, size_t,
153 struct archive_string_conv *);
154 static int archive_string_append_from_wcs_in_codepage(struct archive_string *,
155 const wchar_t *, size_t, struct archive_string_conv *);
156 static int is_big_endian(void);
157 static int strncat_in_codepage(struct archive_string *, const void *,
158 size_t, struct archive_string_conv *);
159 static int win_strncat_from_utf16be(struct archive_string *, const void *,
160 size_t, struct archive_string_conv *);
161 static int win_strncat_from_utf16le(struct archive_string *, const void *,
162 size_t, struct archive_string_conv *);
163 static int win_strncat_to_utf16be(struct archive_string *, const void *,
164 size_t, struct archive_string_conv *);
165 static int win_strncat_to_utf16le(struct archive_string *, const void *,
166 size_t, struct archive_string_conv *);
167 #endif
168 static int best_effort_strncat_from_utf16be(struct archive_string *,
169 const void *, size_t, struct archive_string_conv *);
170 static int best_effort_strncat_from_utf16le(struct archive_string *,
171 const void *, size_t, struct archive_string_conv *);
172 static int best_effort_strncat_to_utf16be(struct archive_string *,
173 const void *, size_t, struct archive_string_conv *);
174 static int best_effort_strncat_to_utf16le(struct archive_string *,
175 const void *, size_t, struct archive_string_conv *);
176 #if defined(HAVE_ICONV)
177 static int iconv_strncat_in_locale(struct archive_string *, const void *,
178 size_t, struct archive_string_conv *);
179 #endif
180 static int best_effort_strncat_in_locale(struct archive_string *,
181 const void *, size_t, struct archive_string_conv *);
182 static int _utf8_to_unicode(uint32_t *, const char *, size_t);
183 static int utf8_to_unicode(uint32_t *, const char *, size_t);
184 static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t);
185 static int cesu8_to_unicode(uint32_t *, const char *, size_t);
186 static size_t unicode_to_utf8(char *, size_t, uint32_t);
187 static int utf16_to_unicode(uint32_t *, const char *, size_t, int);
188 static size_t unicode_to_utf16be(char *, size_t, uint32_t);
189 static size_t unicode_to_utf16le(char *, size_t, uint32_t);
190 static int strncat_from_utf8_libarchive2(struct archive_string *,
191 const void *, size_t, struct archive_string_conv *);
192 static int strncat_from_utf8_to_utf8(struct archive_string *, const void *,
193 size_t, struct archive_string_conv *);
194 static int archive_string_normalize_C(struct archive_string *, const void *,
195 size_t, struct archive_string_conv *);
196 static int archive_string_normalize_D(struct archive_string *, const void *,
197 size_t, struct archive_string_conv *);
198 static int archive_string_append_unicode(struct archive_string *,
199 const void *, size_t, struct archive_string_conv *);
200
201 static struct archive_string *
archive_string_append(struct archive_string * as,const char * p,size_t s)202 archive_string_append(struct archive_string *as, const char *p, size_t s)
203 {
204 if (archive_string_ensure(as, as->length + s + 1) == NULL)
205 return (NULL);
206 memcpy(as->s + as->length, p, s);
207 as->length += s;
208 as->s[as->length] = 0;
209 return (as);
210 }
211
212 static struct archive_wstring *
archive_wstring_append(struct archive_wstring * as,const wchar_t * p,size_t s)213 archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s)
214 {
215 if (archive_wstring_ensure(as, as->length + s + 1) == NULL)
216 return (NULL);
217 wmemcpy(as->s + as->length, p, s);
218 as->length += s;
219 as->s[as->length] = 0;
220 return (as);
221 }
222
223 void
archive_string_concat(struct archive_string * dest,struct archive_string * src)224 archive_string_concat(struct archive_string *dest, struct archive_string *src)
225 {
226 if (archive_string_append(dest, src->s, src->length) == NULL)
227 __archive_errx(1, "Out of memory");
228 }
229
230 void
archive_wstring_concat(struct archive_wstring * dest,struct archive_wstring * src)231 archive_wstring_concat(struct archive_wstring *dest,
232 struct archive_wstring *src)
233 {
234 if (archive_wstring_append(dest, src->s, src->length) == NULL)
235 __archive_errx(1, "Out of memory");
236 }
237
238 void
archive_string_free(struct archive_string * as)239 archive_string_free(struct archive_string *as)
240 {
241 as->length = 0;
242 as->buffer_length = 0;
243 free(as->s);
244 as->s = NULL;
245 }
246
247 void
archive_wstring_free(struct archive_wstring * as)248 archive_wstring_free(struct archive_wstring *as)
249 {
250 as->length = 0;
251 as->buffer_length = 0;
252 free(as->s);
253 as->s = NULL;
254 }
255
256 struct archive_wstring *
archive_wstring_ensure(struct archive_wstring * as,size_t s)257 archive_wstring_ensure(struct archive_wstring *as, size_t s)
258 {
259 return (struct archive_wstring *)
260 archive_string_ensure((struct archive_string *)as,
261 s * sizeof(wchar_t));
262 }
263
264 /* Returns NULL on any allocation failure. */
265 struct archive_string *
archive_string_ensure(struct archive_string * as,size_t s)266 archive_string_ensure(struct archive_string *as, size_t s)
267 {
268 char *p;
269 size_t new_length;
270
271 /* If buffer is already big enough, don't reallocate. */
272 if (as->s && (s <= as->buffer_length))
273 return (as);
274
275 /*
276 * Growing the buffer at least exponentially ensures that
277 * append operations are always linear in the number of
278 * characters appended. Using a smaller growth rate for
279 * larger buffers reduces memory waste somewhat at the cost of
280 * a larger constant factor.
281 */
282 if (as->buffer_length < 32)
283 /* Start with a minimum 32-character buffer. */
284 new_length = 32;
285 else if (as->buffer_length < 8192)
286 /* Buffers under 8k are doubled for speed. */
287 new_length = as->buffer_length + as->buffer_length;
288 else {
289 /* Buffers 8k and over grow by at least 25% each time. */
290 new_length = as->buffer_length + as->buffer_length / 4;
291 /* Be safe: If size wraps, fail. */
292 if (new_length < as->buffer_length) {
293 /* On failure, wipe the string and return NULL. */
294 archive_string_free(as);
295 errno = ENOMEM;/* Make sure errno has ENOMEM. */
296 return (NULL);
297 }
298 }
299 /*
300 * The computation above is a lower limit to how much we'll
301 * grow the buffer. In any case, we have to grow it enough to
302 * hold the request.
303 */
304 if (new_length < s)
305 new_length = s;
306 /* Now we can reallocate the buffer. */
307 p = (char *)realloc(as->s, new_length);
308 if (p == NULL) {
309 /* On failure, wipe the string and return NULL. */
310 archive_string_free(as);
311 errno = ENOMEM;/* Make sure errno has ENOMEM. */
312 return (NULL);
313 }
314
315 as->s = p;
316 as->buffer_length = new_length;
317 return (as);
318 }
319
320 /*
321 * TODO: See if there's a way to avoid scanning
322 * the source string twice. Then test to see
323 * if it actually helps (remember that we're almost
324 * always called with pretty short arguments, so
325 * such an optimization might not help).
326 */
327 struct archive_string *
archive_strncat(struct archive_string * as,const void * _p,size_t n)328 archive_strncat(struct archive_string *as, const void *_p, size_t n)
329 {
330 size_t s;
331 const char *p, *pp;
332
333 p = (const char *)_p;
334
335 /* Like strlen(p), except won't examine positions beyond p[n]. */
336 s = 0;
337 pp = p;
338 while (s < n && *pp) {
339 pp++;
340 s++;
341 }
342 if ((as = archive_string_append(as, p, s)) == NULL)
343 __archive_errx(1, "Out of memory");
344 return (as);
345 }
346
347 struct archive_wstring *
archive_wstrncat(struct archive_wstring * as,const wchar_t * p,size_t n)348 archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n)
349 {
350 size_t s;
351 const wchar_t *pp;
352
353 /* Like strlen(p), except won't examine positions beyond p[n]. */
354 s = 0;
355 pp = p;
356 while (s < n && *pp) {
357 pp++;
358 s++;
359 }
360 if ((as = archive_wstring_append(as, p, s)) == NULL)
361 __archive_errx(1, "Out of memory");
362 return (as);
363 }
364
365 struct archive_string *
archive_strcat(struct archive_string * as,const void * p)366 archive_strcat(struct archive_string *as, const void *p)
367 {
368 /* strcat is just strncat without an effective limit.
369 * Assert that we'll never get called with a source
370 * string over 16MB.
371 * TODO: Review all uses of strcat in the source
372 * and try to replace them with strncat().
373 */
374 return archive_strncat(as, p, 0x1000000);
375 }
376
377 struct archive_wstring *
archive_wstrcat(struct archive_wstring * as,const wchar_t * p)378 archive_wstrcat(struct archive_wstring *as, const wchar_t *p)
379 {
380 /* Ditto. */
381 return archive_wstrncat(as, p, 0x1000000);
382 }
383
384 struct archive_string *
archive_strappend_char(struct archive_string * as,char c)385 archive_strappend_char(struct archive_string *as, char c)
386 {
387 if ((as = archive_string_append(as, &c, 1)) == NULL)
388 __archive_errx(1, "Out of memory");
389 return (as);
390 }
391
392 struct archive_wstring *
archive_wstrappend_wchar(struct archive_wstring * as,wchar_t c)393 archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c)
394 {
395 if ((as = archive_wstring_append(as, &c, 1)) == NULL)
396 __archive_errx(1, "Out of memory");
397 return (as);
398 }
399
400 /*
401 * Get the "current character set" name to use with iconv.
402 * On FreeBSD, the empty character set name "" chooses
403 * the correct character encoding for the current locale,
404 * so this isn't necessary.
405 * But iconv on Mac OS 10.6 doesn't seem to handle this correctly;
406 * on that system, we have to explicitly call nl_langinfo()
407 * to get the right name. Not sure about other platforms.
408 *
409 * NOTE: GNU libiconv does not recognize the character-set name
410 * which some platform nl_langinfo(CODESET) returns, so we should
411 * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv.
412 */
413 static const char *
default_iconv_charset(const char * charset)414 default_iconv_charset(const char *charset) {
415 if (charset != NULL && charset[0] != '\0')
416 return charset;
417 #if HAVE_LOCALE_CHARSET && !defined(__APPLE__)
418 /* locale_charset() is broken on Mac OS */
419 return locale_charset();
420 #elif HAVE_NL_LANGINFO
421 return nl_langinfo(CODESET);
422 #else
423 return "";
424 #endif
425 }
426
427 #if defined(_WIN32) && !defined(__CYGWIN__)
428
429 /*
430 * Convert MBS to WCS.
431 * Note: returns -1 if conversion fails.
432 */
433 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)434 archive_wstring_append_from_mbs(struct archive_wstring *dest,
435 const char *p, size_t len)
436 {
437 return archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL);
438 }
439
440 static int
archive_wstring_append_from_mbs_in_codepage(struct archive_wstring * dest,const char * s,size_t length,struct archive_string_conv * sc)441 archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
442 const char *s, size_t length, struct archive_string_conv *sc)
443 {
444 int count, ret = 0;
445 UINT from_cp;
446
447 if (sc != NULL)
448 from_cp = sc->from_cp;
449 else
450 from_cp = get_current_codepage();
451
452 if (from_cp == CP_C_LOCALE) {
453 /*
454 * "C" locale special process.
455 */
456 wchar_t *ws;
457 const unsigned char *mp;
458
459 if (NULL == archive_wstring_ensure(dest,
460 dest->length + length + 1))
461 return (-1);
462
463 ws = dest->s + dest->length;
464 mp = (const unsigned char *)s;
465 count = 0;
466 while (count < (int)length && *mp) {
467 *ws++ = (wchar_t)*mp++;
468 count++;
469 }
470 } else if (sc != NULL &&
471 (sc->flag & (SCONV_NORMALIZATION_C | SCONV_NORMALIZATION_D))) {
472 /*
473 * Normalize UTF-8 and UTF-16BE and convert it directly
474 * to UTF-16 as wchar_t.
475 */
476 struct archive_string u16;
477 int saved_flag = sc->flag;/* save current flag. */
478
479 if (is_big_endian())
480 sc->flag |= SCONV_TO_UTF16BE;
481 else
482 sc->flag |= SCONV_TO_UTF16LE;
483
484 if (sc->flag & SCONV_FROM_UTF16) {
485 /*
486 * UTF-16BE/LE NFD ===> UTF-16 NFC
487 * UTF-16BE/LE NFC ===> UTF-16 NFD
488 */
489 count = (int)utf16nbytes(s, length);
490 } else {
491 /*
492 * UTF-8 NFD ===> UTF-16 NFC
493 * UTF-8 NFC ===> UTF-16 NFD
494 */
495 count = (int)mbsnbytes(s, length);
496 }
497 u16.s = (char *)dest->s;
498 u16.length = dest->length << 1;;
499 u16.buffer_length = dest->buffer_length;
500 if (sc->flag & SCONV_NORMALIZATION_C)
501 ret = archive_string_normalize_C(&u16, s, count, sc);
502 else
503 ret = archive_string_normalize_D(&u16, s, count, sc);
504 dest->s = (wchar_t *)u16.s;
505 dest->length = u16.length >> 1;
506 dest->buffer_length = u16.buffer_length;
507 sc->flag = saved_flag;/* restore the saved flag. */
508 return (ret);
509 } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) {
510 count = (int)utf16nbytes(s, length);
511 count >>= 1; /* to be WCS length */
512 /* Allocate memory for WCS. */
513 if (NULL == archive_wstring_ensure(dest,
514 dest->length + count + 1))
515 return (-1);
516 wmemcpy(dest->s + dest->length, (const wchar_t *)s, count);
517 if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) {
518 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
519 int b;
520 for (b = 0; b < count; b++) {
521 uint16_t val = archive_le16dec(u16+b);
522 archive_be16enc(u16+b, val);
523 }
524 } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) {
525 uint16_t *u16 = (uint16_t *)(dest->s + dest->length);
526 int b;
527 for (b = 0; b < count; b++) {
528 uint16_t val = archive_be16dec(u16+b);
529 archive_le16enc(u16+b, val);
530 }
531 }
532 } else {
533 DWORD mbflag;
534 size_t buffsize;
535
536 if (sc == NULL)
537 mbflag = 0;
538 else if (sc->flag & SCONV_FROM_CHARSET) {
539 /* Do not trust the length which comes from
540 * an archive file. */
541 length = mbsnbytes(s, length);
542 mbflag = 0;
543 } else
544 mbflag = MB_PRECOMPOSED;
545
546 buffsize = dest->length + length + 1;
547 do {
548 /* Allocate memory for WCS. */
549 if (NULL == archive_wstring_ensure(dest, buffsize))
550 return (-1);
551 /* Convert MBS to WCS. */
552 count = MultiByteToWideChar(from_cp,
553 mbflag, s, (int)length, dest->s + dest->length,
554 (int)(dest->buffer_length >> 1) -1);
555 if (count == 0 &&
556 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
557 /* Expand the WCS buffer. */
558 buffsize = dest->buffer_length << 1;
559 continue;
560 }
561 if (count == 0 && length != 0)
562 ret = -1;
563 } while (0);
564 }
565 dest->length += count;
566 dest->s[dest->length] = L'\0';
567 return (ret);
568 }
569
570 #else
571
572 /*
573 * Convert MBS to WCS.
574 * Note: returns -1 if conversion fails.
575 */
576 int
archive_wstring_append_from_mbs(struct archive_wstring * dest,const char * p,size_t len)577 archive_wstring_append_from_mbs(struct archive_wstring *dest,
578 const char *p, size_t len)
579 {
580 size_t r;
581 int ret_val = 0;
582 /*
583 * No single byte will be more than one wide character,
584 * so this length estimate will always be big enough.
585 */
586 size_t wcs_length = len;
587 size_t mbs_length = len;
588 const char *mbs = p;
589 wchar_t *wcs;
590 #if HAVE_MBRTOWC
591 mbstate_t shift_state;
592
593 memset(&shift_state, 0, sizeof(shift_state));
594 #endif
595 if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1))
596 return (-1);
597 wcs = dest->s + dest->length;
598 /*
599 * We cannot use mbsrtowcs/mbstowcs here because those may convert
600 * extra MBS when strlen(p) > len and one wide character consis of
601 * multi bytes.
602 */
603 while (*mbs && mbs_length > 0) {
604 if (wcs_length == 0) {
605 dest->length = wcs - dest->s;
606 dest->s[dest->length] = L'\0';
607 wcs_length = mbs_length;
608 if (NULL == archive_wstring_ensure(dest,
609 dest->length + wcs_length + 1))
610 return (-1);
611 wcs = dest->s + dest->length;
612 }
613 #if HAVE_MBRTOWC
614 r = mbrtowc(wcs, mbs, wcs_length, &shift_state);
615 #else
616 r = mbtowc(wcs, mbs, wcs_length);
617 #endif
618 if (r == (size_t)-1 || r == (size_t)-2) {
619 ret_val = -1;
620 if (errno == EILSEQ) {
621 ++mbs;
622 --mbs_length;
623 continue;
624 } else
625 break;
626 }
627 if (r == 0 || r > mbs_length)
628 break;
629 wcs++;
630 wcs_length--;
631 mbs += r;
632 mbs_length -= r;
633 }
634 dest->length = wcs - dest->s;
635 dest->s[dest->length] = L'\0';
636 return (ret_val);
637 }
638
639 #endif
640
641 #if defined(_WIN32) && !defined(__CYGWIN__)
642
643 /*
644 * WCS ==> MBS.
645 * Note: returns -1 if conversion fails.
646 *
647 * Win32 builds use WideCharToMultiByte from the Windows API.
648 * (Maybe Cygwin should too? WideCharToMultiByte will know a
649 * lot more about local character encodings than the wcrtomb()
650 * wrapper is going to know.)
651 */
652 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)653 archive_string_append_from_wcs(struct archive_string *as,
654 const wchar_t *w, size_t len)
655 {
656 return archive_string_append_from_wcs_in_codepage(as, w, len, NULL);
657 }
658
659 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)660 archive_string_append_from_wcs_in_codepage(struct archive_string *as,
661 const wchar_t *ws, size_t len, struct archive_string_conv *sc)
662 {
663 BOOL defchar_used, *dp;
664 int count, ret = 0;
665 UINT to_cp;
666 int wslen = (int)len;
667
668 if (sc != NULL)
669 to_cp = sc->to_cp;
670 else
671 to_cp = get_current_codepage();
672
673 if (to_cp == CP_C_LOCALE) {
674 /*
675 * "C" locale special process.
676 */
677 const wchar_t *wp = ws;
678 char *p;
679
680 if (NULL == archive_string_ensure(as,
681 as->length + wslen +1))
682 return (-1);
683 p = as->s + as->length;
684 count = 0;
685 defchar_used = 0;
686 while (count < wslen && *wp) {
687 if (*wp > 255) {
688 *p++ = '?';
689 wp++;
690 defchar_used = 1;
691 } else
692 *p++ = (char)*wp++;
693 count++;
694 }
695 } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) {
696 uint16_t *u16;
697
698 if (NULL ==
699 archive_string_ensure(as, as->length + len * 2 + 2))
700 return (-1);
701 u16 = (uint16_t *)(as->s + as->length);
702 count = 0;
703 defchar_used = 0;
704 if (sc->flag & SCONV_TO_UTF16BE) {
705 while (count < (int)len && *ws) {
706 archive_be16enc(u16+count, *ws);
707 ws++;
708 count++;
709 }
710 } else {
711 while (count < (int)len && *ws) {
712 archive_le16enc(u16+count, *ws);
713 ws++;
714 count++;
715 }
716 }
717 count <<= 1; /* to be byte size */
718 } else {
719 /* Make sure the MBS buffer has plenty to set. */
720 if (NULL ==
721 archive_string_ensure(as, as->length + len * 2 + 1))
722 return (-1);
723 do {
724 defchar_used = 0;
725 if (to_cp == CP_UTF8 || sc == NULL)
726 dp = NULL;
727 else
728 dp = &defchar_used;
729 count = WideCharToMultiByte(to_cp, 0, ws, wslen,
730 as->s + as->length, (int)as->buffer_length-1, NULL, dp);
731 if (count == 0 &&
732 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
733 /* Expand the MBS buffer and retry. */
734 if (NULL == archive_string_ensure(as,
735 as->buffer_length + len))
736 return (-1);
737 continue;
738 }
739 if (count == 0)
740 ret = -1;
741 } while (0);
742 }
743 as->length += count;
744 as->s[as->length] = '\0';
745 return (defchar_used?-1:ret);
746 }
747
748 #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB)
749
750 /*
751 * Translates a wide character string into current locale character set
752 * and appends to the archive_string. Note: returns -1 if conversion
753 * fails.
754 */
755 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)756 archive_string_append_from_wcs(struct archive_string *as,
757 const wchar_t *w, size_t len)
758 {
759 /* We cannot use the standard wcstombs() here because it
760 * cannot tell us how big the output buffer should be. So
761 * I've built a loop around wcrtomb() or wctomb() that
762 * converts a character at a time and resizes the string as
763 * needed. We prefer wcrtomb() when it's available because
764 * it's thread-safe. */
765 int n, ret_val = 0;
766 char *p;
767 char *end;
768 #if HAVE_WCRTOMB
769 mbstate_t shift_state;
770
771 memset(&shift_state, 0, sizeof(shift_state));
772 #else
773 /* Clear the shift state before starting. */
774 wctomb(NULL, L'\0');
775 #endif
776 /*
777 * Allocate buffer for MBS.
778 * We need this allocation here since it is possible that
779 * as->s is still NULL.
780 */
781 if (archive_string_ensure(as, as->length + len + 1) == NULL)
782 return (-1);
783
784 p = as->s + as->length;
785 end = as->s + as->buffer_length - MB_CUR_MAX -1;
786 while (*w != L'\0' && len > 0) {
787 if (p >= end) {
788 as->length = p - as->s;
789 as->s[as->length] = '\0';
790 /* Re-allocate buffer for MBS. */
791 if (archive_string_ensure(as,
792 as->length + len * 2 + 1) == NULL)
793 return (-1);
794 p = as->s + as->length;
795 end = as->s + as->buffer_length - MB_CUR_MAX -1;
796 }
797 #if HAVE_WCRTOMB
798 n = wcrtomb(p, *w++, &shift_state);
799 #else
800 n = wctomb(p, *w++);
801 #endif
802 if (n == -1) {
803 if (errno == EILSEQ) {
804 /* Skip an illegal wide char. */
805 *p++ = '?';
806 ret_val = -1;
807 } else {
808 ret_val = -1;
809 break;
810 }
811 } else
812 p += n;
813 len--;
814 }
815 as->length = p - as->s;
816 as->s[as->length] = '\0';
817 return (ret_val);
818 }
819
820 #else /* HAVE_WCTOMB || HAVE_WCRTOMB */
821
822 /*
823 * TODO: Test if __STDC_ISO_10646__ is defined.
824 * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion
825 * one character at a time. If a non-Windows platform doesn't have
826 * either of these, fall back to the built-in UTF8 conversion.
827 */
828 int
archive_string_append_from_wcs(struct archive_string * as,const wchar_t * w,size_t len)829 archive_string_append_from_wcs(struct archive_string *as,
830 const wchar_t *w, size_t len)
831 {
832 (void)as;/* UNUSED */
833 (void)w;/* UNUSED */
834 (void)len;/* UNUSED */
835 errno = ENOSYS;
836 return (-1);
837 }
838
839 #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */
840
841 /*
842 * Find a string conversion object by a pair of 'from' charset name
843 * and 'to' charset name from an archive object.
844 * Return NULL if not found.
845 */
846 static struct archive_string_conv *
find_sconv_object(struct archive * a,const char * fc,const char * tc)847 find_sconv_object(struct archive *a, const char *fc, const char *tc)
848 {
849 struct archive_string_conv *sc;
850
851 if (a == NULL)
852 return (NULL);
853
854 for (sc = a->sconv; sc != NULL; sc = sc->next) {
855 if (strcmp(sc->from_charset, fc) == 0 &&
856 strcmp(sc->to_charset, tc) == 0)
857 break;
858 }
859 return (sc);
860 }
861
862 /*
863 * Register a string object to an archive object.
864 */
865 static void
add_sconv_object(struct archive * a,struct archive_string_conv * sc)866 add_sconv_object(struct archive *a, struct archive_string_conv *sc)
867 {
868 struct archive_string_conv **psc;
869
870 /* Add a new sconv to sconv list. */
871 psc = &(a->sconv);
872 while (*psc != NULL)
873 psc = &((*psc)->next);
874 *psc = sc;
875 }
876
877 static void
add_converter(struct archive_string_conv * sc,int (* converter)(struct archive_string *,const void *,size_t,struct archive_string_conv *))878 add_converter(struct archive_string_conv *sc, int (*converter)
879 (struct archive_string *, const void *, size_t,
880 struct archive_string_conv *))
881 {
882 if (sc == NULL || sc->nconverter >= 2)
883 __archive_errx(1, "Programing error");
884 sc->converter[sc->nconverter++] = converter;
885 }
886
887 static void
setup_converter(struct archive_string_conv * sc)888 setup_converter(struct archive_string_conv *sc)
889 {
890
891 /* Reset. */
892 sc->nconverter = 0;
893
894 /*
895 * Perform special sequence for the incorrect UTF-8 filenames
896 * made by libarchive2.x.
897 */
898 if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) {
899 add_converter(sc, strncat_from_utf8_libarchive2);
900 return;
901 }
902
903 /*
904 * Convert a string to UTF-16BE/LE.
905 */
906 if (sc->flag & SCONV_TO_UTF16) {
907 /*
908 * If the current locale is UTF-8, we can translate
909 * a UTF-8 string into a UTF-16BE string.
910 */
911 if (sc->flag & SCONV_FROM_UTF8) {
912 add_converter(sc, archive_string_append_unicode);
913 return;
914 }
915
916 #if defined(_WIN32) && !defined(__CYGWIN__)
917 if (sc->flag & SCONV_WIN_CP) {
918 if (sc->flag & SCONV_TO_UTF16BE)
919 add_converter(sc, win_strncat_to_utf16be);
920 else
921 add_converter(sc, win_strncat_to_utf16le);
922 return;
923 }
924 #endif
925
926 #if defined(HAVE_ICONV)
927 if (sc->cd != (iconv_t)-1) {
928 add_converter(sc, iconv_strncat_in_locale);
929 return;
930 }
931 #endif
932
933 if (sc->flag & SCONV_BEST_EFFORT) {
934 if (sc->flag & SCONV_TO_UTF16BE)
935 add_converter(sc,
936 best_effort_strncat_to_utf16be);
937 else
938 add_converter(sc,
939 best_effort_strncat_to_utf16le);
940 } else
941 /* Make sure we have no converter. */
942 sc->nconverter = 0;
943 return;
944 }
945
946 /*
947 * Convert a string from UTF-16BE/LE.
948 */
949 if (sc->flag & SCONV_FROM_UTF16) {
950 /*
951 * At least we should normalize a UTF-16BE string.
952 */
953 if (sc->flag & SCONV_NORMALIZATION_D)
954 add_converter(sc,archive_string_normalize_D);
955 else if (sc->flag & SCONV_NORMALIZATION_C)
956 add_converter(sc, archive_string_normalize_C);
957
958 if (sc->flag & SCONV_TO_UTF8) {
959 /*
960 * If the current locale is UTF-8, we can translate
961 * a UTF-16BE/LE string into a UTF-8 string directly.
962 */
963 if (!(sc->flag &
964 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
965 add_converter(sc,
966 archive_string_append_unicode);
967 return;
968 }
969
970 #if defined(_WIN32) && !defined(__CYGWIN__)
971 if (sc->flag & SCONV_WIN_CP) {
972 if (sc->flag & SCONV_FROM_UTF16BE)
973 add_converter(sc, win_strncat_from_utf16be);
974 else
975 add_converter(sc, win_strncat_from_utf16le);
976 return;
977 }
978 #endif
979
980 #if defined(HAVE_ICONV)
981 if (sc->cd != (iconv_t)-1) {
982 add_converter(sc, iconv_strncat_in_locale);
983 return;
984 }
985 #endif
986
987 if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
988 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE))
989 add_converter(sc, best_effort_strncat_from_utf16be);
990 else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
991 == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE))
992 add_converter(sc, best_effort_strncat_from_utf16le);
993 else
994 /* Make sure we have no converter. */
995 sc->nconverter = 0;
996 return;
997 }
998
999 if (sc->flag & SCONV_FROM_UTF8) {
1000 /*
1001 * At least we should normalize a UTF-8 string.
1002 */
1003 if (sc->flag & SCONV_NORMALIZATION_D)
1004 add_converter(sc,archive_string_normalize_D);
1005 else if (sc->flag & SCONV_NORMALIZATION_C)
1006 add_converter(sc, archive_string_normalize_C);
1007
1008 /*
1009 * Copy UTF-8 string with a check of CESU-8.
1010 * Apparently, iconv does not check surrogate pairs in UTF-8
1011 * when both from-charset and to-charset are UTF-8, and then
1012 * we use our UTF-8 copy code.
1013 */
1014 if (sc->flag & SCONV_TO_UTF8) {
1015 /*
1016 * If the current locale is UTF-8, we can translate
1017 * a UTF-16BE string into a UTF-8 string directly.
1018 */
1019 if (!(sc->flag &
1020 (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C)))
1021 add_converter(sc, strncat_from_utf8_to_utf8);
1022 return;
1023 }
1024 }
1025
1026 #if defined(_WIN32) && !defined(__CYGWIN__)
1027 /*
1028 * On Windows we can use Windows API for a string conversion.
1029 */
1030 if (sc->flag & SCONV_WIN_CP) {
1031 add_converter(sc, strncat_in_codepage);
1032 return;
1033 }
1034 #endif
1035
1036 #if HAVE_ICONV
1037 if (sc->cd != (iconv_t)-1) {
1038 add_converter(sc, iconv_strncat_in_locale);
1039 /*
1040 * iconv generally does not support UTF-8-MAC and so
1041 * we have to the output of iconv from NFC to NFD if
1042 * need.
1043 */
1044 if ((sc->flag & SCONV_FROM_CHARSET) &&
1045 (sc->flag & SCONV_TO_UTF8)) {
1046 if (sc->flag & SCONV_NORMALIZATION_D)
1047 add_converter(sc, archive_string_normalize_D);
1048 }
1049 return;
1050 }
1051 #endif
1052
1053 /*
1054 * Try conversion in the best effort or no conversion.
1055 */
1056 if ((sc->flag & SCONV_BEST_EFFORT) || sc->same)
1057 add_converter(sc, best_effort_strncat_in_locale);
1058 else
1059 /* Make sure we have no converter. */
1060 sc->nconverter = 0;
1061 }
1062
1063 /*
1064 * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE
1065 * and CP932 which are referenced in create_sconv_object().
1066 */
1067 static const char *
canonical_charset_name(const char * charset)1068 canonical_charset_name(const char *charset)
1069 {
1070 char cs[16];
1071 char *p;
1072 const char *s;
1073
1074 if (charset == NULL || charset[0] == '\0'
1075 || strlen(charset) > 15)
1076 return (charset);
1077
1078 /* Copy name to uppercase. */
1079 p = cs;
1080 s = charset;
1081 while (*s) {
1082 char c = *s++;
1083 if (c >= 'a' && c <= 'z')
1084 c -= 'a' - 'A';
1085 *p++ = c;
1086 }
1087 *p++ = '\0';
1088
1089 if (strcmp(cs, "UTF-8") == 0 ||
1090 strcmp(cs, "UTF8") == 0)
1091 return ("UTF-8");
1092 if (strcmp(cs, "UTF-16BE") == 0 ||
1093 strcmp(cs, "UTF16BE") == 0)
1094 return ("UTF-16BE");
1095 if (strcmp(cs, "UTF-16LE") == 0 ||
1096 strcmp(cs, "UTF16LE") == 0)
1097 return ("UTF-16LE");
1098 if (strcmp(cs, "CP932") == 0)
1099 return ("CP932");
1100 return (charset);
1101 }
1102
1103 /*
1104 * Create a string conversion object.
1105 */
1106 static struct archive_string_conv *
create_sconv_object(const char * fc,const char * tc,unsigned current_codepage,int flag)1107 create_sconv_object(const char *fc, const char *tc,
1108 unsigned current_codepage, int flag)
1109 {
1110 struct archive_string_conv *sc;
1111
1112 sc = calloc(1, sizeof(*sc));
1113 if (sc == NULL)
1114 return (NULL);
1115 sc->next = NULL;
1116 sc->from_charset = strdup(fc);
1117 if (sc->from_charset == NULL) {
1118 free(sc);
1119 return (NULL);
1120 }
1121 sc->to_charset = strdup(tc);
1122 if (sc->to_charset == NULL) {
1123 free(sc->from_charset);
1124 free(sc);
1125 return (NULL);
1126 }
1127 archive_string_init(&sc->utftmp);
1128
1129 if (flag & SCONV_TO_CHARSET) {
1130 /*
1131 * Convert characters from the current locale charset to
1132 * a specified charset.
1133 */
1134 sc->from_cp = current_codepage;
1135 sc->to_cp = make_codepage_from_charset(tc);
1136 #if defined(_WIN32) && !defined(__CYGWIN__)
1137 if (IsValidCodePage(sc->to_cp))
1138 flag |= SCONV_WIN_CP;
1139 #endif
1140 } else if (flag & SCONV_FROM_CHARSET) {
1141 /*
1142 * Convert characters from a specified charset to
1143 * the current locale charset.
1144 */
1145 sc->to_cp = current_codepage;
1146 sc->from_cp = make_codepage_from_charset(fc);
1147 #if defined(_WIN32) && !defined(__CYGWIN__)
1148 if (IsValidCodePage(sc->from_cp))
1149 flag |= SCONV_WIN_CP;
1150 #endif
1151 }
1152
1153 /*
1154 * Check if "from charset" and "to charset" are the same.
1155 */
1156 if (strcmp(fc, tc) == 0 ||
1157 (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp))
1158 sc->same = 1;
1159 else
1160 sc->same = 0;
1161
1162 /*
1163 * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE.
1164 */
1165 if (strcmp(tc, "UTF-8") == 0)
1166 flag |= SCONV_TO_UTF8;
1167 else if (strcmp(tc, "UTF-16BE") == 0)
1168 flag |= SCONV_TO_UTF16BE;
1169 else if (strcmp(tc, "UTF-16LE") == 0)
1170 flag |= SCONV_TO_UTF16LE;
1171 if (strcmp(fc, "UTF-8") == 0)
1172 flag |= SCONV_FROM_UTF8;
1173 else if (strcmp(fc, "UTF-16BE") == 0)
1174 flag |= SCONV_FROM_UTF16BE;
1175 else if (strcmp(fc, "UTF-16LE") == 0)
1176 flag |= SCONV_FROM_UTF16LE;
1177 #if defined(_WIN32) && !defined(__CYGWIN__)
1178 if (sc->to_cp == CP_UTF8)
1179 flag |= SCONV_TO_UTF8;
1180 else if (sc->to_cp == CP_UTF16BE)
1181 flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP;
1182 else if (sc->to_cp == CP_UTF16LE)
1183 flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP;
1184 if (sc->from_cp == CP_UTF8)
1185 flag |= SCONV_FROM_UTF8;
1186 else if (sc->from_cp == CP_UTF16BE)
1187 flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP;
1188 else if (sc->from_cp == CP_UTF16LE)
1189 flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP;
1190 #endif
1191
1192 /*
1193 * Set a flag for Unicode NFD. Usually iconv cannot correctly
1194 * handle it. So we have to translate NFD characters to NFC ones
1195 * ourselves before iconv handles. Another reason is to prevent
1196 * that the same sight of two filenames, one is NFC and other
1197 * is NFD, would be in its directory.
1198 * On Mac OS X, although its filesystem layer automatically
1199 * convert filenames to NFD, it would be useful for filename
1200 * comparing to find out the same filenames that we normalize
1201 * that to be NFD ourselves.
1202 */
1203 if ((flag & SCONV_FROM_CHARSET) &&
1204 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) {
1205 #if defined(__APPLE__)
1206 if (flag & SCONV_TO_UTF8)
1207 flag |= SCONV_NORMALIZATION_D;
1208 else
1209 #endif
1210 flag |= SCONV_NORMALIZATION_C;
1211 }
1212 #if defined(__APPLE__)
1213 /*
1214 * In case writing an archive file, make sure that a filename
1215 * going to be passed to iconv is a Unicode NFC string since
1216 * a filename in HFS Plus filesystem is a Unicode NFD one and
1217 * iconv cannot handle it with "UTF-8" charset. It is simpler
1218 * than a use of "UTF-8-MAC" charset.
1219 */
1220 if ((flag & SCONV_TO_CHARSET) &&
1221 (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1222 !(flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1223 flag |= SCONV_NORMALIZATION_C;
1224 /*
1225 * In case reading an archive file. make sure that a filename
1226 * will be passed to users is a Unicode NFD string in order to
1227 * correctly compare the filename with other one which comes
1228 * from HFS Plus filesystem.
1229 */
1230 if ((flag & SCONV_FROM_CHARSET) &&
1231 !(flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1232 (flag & SCONV_TO_UTF8))
1233 flag |= SCONV_NORMALIZATION_D;
1234 #endif
1235
1236 #if defined(HAVE_ICONV)
1237 sc->cd_w = (iconv_t)-1;
1238 /*
1239 * Create an iconv object.
1240 */
1241 if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) &&
1242 (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) ||
1243 (flag & SCONV_WIN_CP)) {
1244 /* This case we won't use iconv. */
1245 sc->cd = (iconv_t)-1;
1246 } else {
1247 sc->cd = iconv_open(tc, fc);
1248 if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) {
1249 /*
1250 * Unfortunaly, all of iconv implements do support
1251 * "CP932" character-set, so we should use "SJIS"
1252 * instead if iconv_open failed.
1253 */
1254 if (strcmp(tc, "CP932") == 0)
1255 sc->cd = iconv_open("SJIS", fc);
1256 else if (strcmp(fc, "CP932") == 0)
1257 sc->cd = iconv_open(tc, "SJIS");
1258 }
1259 #if defined(_WIN32) && !defined(__CYGWIN__)
1260 /*
1261 * archive_mstring on Windows directly convert multi-bytes
1262 * into archive_wstring in order not to depend on locale
1263 * so that you can do a I18N programing. This will be
1264 * used only in archive_mstring_copy_mbs_len_l so far.
1265 */
1266 if (flag & SCONV_FROM_CHARSET) {
1267 sc->cd_w = iconv_open("UTF-8", fc);
1268 if (sc->cd_w == (iconv_t)-1 &&
1269 (sc->flag & SCONV_BEST_EFFORT)) {
1270 if (strcmp(fc, "CP932") == 0)
1271 sc->cd_w = iconv_open("UTF-8", "SJIS");
1272 }
1273 }
1274 #endif /* _WIN32 && !__CYGWIN__ */
1275 }
1276 #endif /* HAVE_ICONV */
1277
1278 sc->flag = flag;
1279
1280 /*
1281 * Set up converters.
1282 */
1283 setup_converter(sc);
1284
1285 return (sc);
1286 }
1287
1288 /*
1289 * Free a string conversion object.
1290 */
1291 static void
free_sconv_object(struct archive_string_conv * sc)1292 free_sconv_object(struct archive_string_conv *sc)
1293 {
1294 free(sc->from_charset);
1295 free(sc->to_charset);
1296 archive_string_free(&sc->utftmp);
1297 #if HAVE_ICONV
1298 if (sc->cd != (iconv_t)-1)
1299 iconv_close(sc->cd);
1300 if (sc->cd_w != (iconv_t)-1)
1301 iconv_close(sc->cd_w);
1302 #endif
1303 free(sc);
1304 }
1305
1306 #if defined(_WIN32) && !defined(__CYGWIN__)
1307 static unsigned
my_atoi(const char * p)1308 my_atoi(const char *p)
1309 {
1310 unsigned cp;
1311
1312 cp = 0;
1313 while (*p) {
1314 if (*p >= '0' && *p <= '9')
1315 cp = cp * 10 + (*p - '0');
1316 else
1317 return (-1);
1318 p++;
1319 }
1320 return (cp);
1321 }
1322
1323 /*
1324 * Translate Charset name (as used by iconv) into CodePage (as used by Windows)
1325 * Return -1 if failed.
1326 *
1327 * Note: This translation code may be insufficient.
1328 */
1329 static struct charset {
1330 const char *name;
1331 unsigned cp;
1332 } charsets[] = {
1333 /* MUST BE SORTED! */
1334 {"ASCII", 1252},
1335 {"ASMO-708", 708},
1336 {"BIG5", 950},
1337 {"CHINESE", 936},
1338 {"CP367", 1252},
1339 {"CP819", 1252},
1340 {"CP1025", 21025},
1341 {"DOS-720", 720},
1342 {"DOS-862", 862},
1343 {"EUC-CN", 51936},
1344 {"EUC-JP", 51932},
1345 {"EUC-KR", 949},
1346 {"EUCCN", 51936},
1347 {"EUCJP", 51932},
1348 {"EUCKR", 949},
1349 {"GB18030", 54936},
1350 {"GB2312", 936},
1351 {"HEBREW", 1255},
1352 {"HZ-GB-2312", 52936},
1353 {"IBM273", 20273},
1354 {"IBM277", 20277},
1355 {"IBM278", 20278},
1356 {"IBM280", 20280},
1357 {"IBM284", 20284},
1358 {"IBM285", 20285},
1359 {"IBM290", 20290},
1360 {"IBM297", 20297},
1361 {"IBM367", 1252},
1362 {"IBM420", 20420},
1363 {"IBM423", 20423},
1364 {"IBM424", 20424},
1365 {"IBM819", 1252},
1366 {"IBM871", 20871},
1367 {"IBM880", 20880},
1368 {"IBM905", 20905},
1369 {"IBM924", 20924},
1370 {"ISO-8859-1", 28591},
1371 {"ISO-8859-13", 28603},
1372 {"ISO-8859-15", 28605},
1373 {"ISO-8859-2", 28592},
1374 {"ISO-8859-3", 28593},
1375 {"ISO-8859-4", 28594},
1376 {"ISO-8859-5", 28595},
1377 {"ISO-8859-6", 28596},
1378 {"ISO-8859-7", 28597},
1379 {"ISO-8859-8", 28598},
1380 {"ISO-8859-9", 28599},
1381 {"ISO8859-1", 28591},
1382 {"ISO8859-13", 28603},
1383 {"ISO8859-15", 28605},
1384 {"ISO8859-2", 28592},
1385 {"ISO8859-3", 28593},
1386 {"ISO8859-4", 28594},
1387 {"ISO8859-5", 28595},
1388 {"ISO8859-6", 28596},
1389 {"ISO8859-7", 28597},
1390 {"ISO8859-8", 28598},
1391 {"ISO8859-9", 28599},
1392 {"JOHAB", 1361},
1393 {"KOI8-R", 20866},
1394 {"KOI8-U", 21866},
1395 {"KS_C_5601-1987", 949},
1396 {"LATIN1", 1252},
1397 {"LATIN2", 28592},
1398 {"MACINTOSH", 10000},
1399 {"SHIFT-JIS", 932},
1400 {"SHIFT_JIS", 932},
1401 {"SJIS", 932},
1402 {"US", 1252},
1403 {"US-ASCII", 1252},
1404 {"UTF-16", 1200},
1405 {"UTF-16BE", 1201},
1406 {"UTF-16LE", 1200},
1407 {"UTF-8", CP_UTF8},
1408 {"X-EUROPA", 29001},
1409 {"X-MAC-ARABIC", 10004},
1410 {"X-MAC-CE", 10029},
1411 {"X-MAC-CHINESEIMP", 10008},
1412 {"X-MAC-CHINESETRAD", 10002},
1413 {"X-MAC-CROATIAN", 10082},
1414 {"X-MAC-CYRILLIC", 10007},
1415 {"X-MAC-GREEK", 10006},
1416 {"X-MAC-HEBREW", 10005},
1417 {"X-MAC-ICELANDIC", 10079},
1418 {"X-MAC-JAPANESE", 10001},
1419 {"X-MAC-KOREAN", 10003},
1420 {"X-MAC-ROMANIAN", 10010},
1421 {"X-MAC-THAI", 10021},
1422 {"X-MAC-TURKISH", 10081},
1423 {"X-MAC-UKRAINIAN", 10017},
1424 };
1425 static unsigned
make_codepage_from_charset(const char * charset)1426 make_codepage_from_charset(const char *charset)
1427 {
1428 char cs[16];
1429 char *p;
1430 unsigned cp;
1431 int a, b;
1432
1433 if (charset == NULL || strlen(charset) > 15)
1434 return -1;
1435
1436 /* Copy name to uppercase. */
1437 p = cs;
1438 while (*charset) {
1439 char c = *charset++;
1440 if (c >= 'a' && c <= 'z')
1441 c -= 'a' - 'A';
1442 *p++ = c;
1443 }
1444 *p++ = '\0';
1445 cp = -1;
1446
1447 /* Look it up in the table first, so that we can easily
1448 * override CP367, which we map to 1252 instead of 367. */
1449 a = 0;
1450 b = sizeof(charsets)/sizeof(charsets[0]);
1451 while (b > a) {
1452 int c = (b + a) / 2;
1453 int r = strcmp(charsets[c].name, cs);
1454 if (r < 0)
1455 a = c + 1;
1456 else if (r > 0)
1457 b = c;
1458 else
1459 return charsets[c].cp;
1460 }
1461
1462 /* If it's not in the table, try to parse it. */
1463 switch (*cs) {
1464 case 'C':
1465 if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') {
1466 cp = my_atoi(cs + 2);
1467 } else if (strcmp(cs, "CP_ACP") == 0)
1468 cp = get_current_codepage();
1469 else if (strcmp(cs, "CP_OEMCP") == 0)
1470 cp = get_current_oemcp();
1471 break;
1472 case 'I':
1473 if (cs[1] == 'B' && cs[2] == 'M' &&
1474 cs[3] >= '0' && cs[3] <= '9') {
1475 cp = my_atoi(cs + 3);
1476 }
1477 break;
1478 case 'W':
1479 if (strncmp(cs, "WINDOWS-", 8) == 0) {
1480 cp = my_atoi(cs + 8);
1481 if (cp != 874 && (cp < 1250 || cp > 1258))
1482 cp = -1;/* This may invalid code. */
1483 }
1484 break;
1485 }
1486 return (cp);
1487 }
1488
1489 /*
1490 * Return ANSI Code Page of current locale set by setlocale().
1491 */
1492 static unsigned
get_current_codepage(void)1493 get_current_codepage(void)
1494 {
1495 char *locale, *p;
1496 unsigned cp;
1497
1498 locale = setlocale(LC_CTYPE, NULL);
1499 if (locale == NULL)
1500 return (GetACP());
1501 if (locale[0] == 'C' && locale[1] == '\0')
1502 return (CP_C_LOCALE);
1503 p = strrchr(locale, '.');
1504 if (p == NULL)
1505 return (GetACP());
1506 cp = my_atoi(p+1);
1507 if (cp <= 0)
1508 return (GetACP());
1509 return (cp);
1510 }
1511
1512 /*
1513 * Translation table between Locale Name and ACP/OEMCP.
1514 */
1515 static struct {
1516 unsigned acp;
1517 unsigned ocp;
1518 const char *locale;
1519 } acp_ocp_map[] = {
1520 { 950, 950, "Chinese_Taiwan" },
1521 { 936, 936, "Chinese_People's Republic of China" },
1522 { 950, 950, "Chinese_Taiwan" },
1523 { 1250, 852, "Czech_Czech Republic" },
1524 { 1252, 850, "Danish_Denmark" },
1525 { 1252, 850, "Dutch_Netherlands" },
1526 { 1252, 850, "Dutch_Belgium" },
1527 { 1252, 437, "English_United States" },
1528 { 1252, 850, "English_Australia" },
1529 { 1252, 850, "English_Canada" },
1530 { 1252, 850, "English_New Zealand" },
1531 { 1252, 850, "English_United Kingdom" },
1532 { 1252, 437, "English_United States" },
1533 { 1252, 850, "Finnish_Finland" },
1534 { 1252, 850, "French_France" },
1535 { 1252, 850, "French_Belgium" },
1536 { 1252, 850, "French_Canada" },
1537 { 1252, 850, "French_Switzerland" },
1538 { 1252, 850, "German_Germany" },
1539 { 1252, 850, "German_Austria" },
1540 { 1252, 850, "German_Switzerland" },
1541 { 1253, 737, "Greek_Greece" },
1542 { 1250, 852, "Hungarian_Hungary" },
1543 { 1252, 850, "Icelandic_Iceland" },
1544 { 1252, 850, "Italian_Italy" },
1545 { 1252, 850, "Italian_Switzerland" },
1546 { 932, 932, "Japanese_Japan" },
1547 { 949, 949, "Korean_Korea" },
1548 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1549 { 1252, 850, "Norwegian (BokmOl)_Norway" },
1550 { 1252, 850, "Norwegian-Nynorsk_Norway" },
1551 { 1250, 852, "Polish_Poland" },
1552 { 1252, 850, "Portuguese_Portugal" },
1553 { 1252, 850, "Portuguese_Brazil" },
1554 { 1251, 866, "Russian_Russia" },
1555 { 1250, 852, "Slovak_Slovakia" },
1556 { 1252, 850, "Spanish_Spain" },
1557 { 1252, 850, "Spanish_Mexico" },
1558 { 1252, 850, "Spanish_Spain" },
1559 { 1252, 850, "Swedish_Sweden" },
1560 { 1254, 857, "Turkish_Turkey" },
1561 { 0, 0, NULL}
1562 };
1563
1564 /*
1565 * Return OEM Code Page of current locale set by setlocale().
1566 */
1567 static unsigned
get_current_oemcp(void)1568 get_current_oemcp(void)
1569 {
1570 int i;
1571 char *locale, *p;
1572 size_t len;
1573
1574 locale = setlocale(LC_CTYPE, NULL);
1575 if (locale == NULL)
1576 return (GetOEMCP());
1577 if (locale[0] == 'C' && locale[1] == '\0')
1578 return (CP_C_LOCALE);
1579
1580 p = strrchr(locale, '.');
1581 if (p == NULL)
1582 return (GetOEMCP());
1583 len = p - locale;
1584 for (i = 0; acp_ocp_map[i].acp; i++) {
1585 if (strncmp(acp_ocp_map[i].locale, locale, len) == 0)
1586 return (acp_ocp_map[i].ocp);
1587 }
1588 return (GetOEMCP());
1589 }
1590 #else
1591
1592 /*
1593 * POSIX platform does not use CodePage.
1594 */
1595
1596 static unsigned
get_current_codepage(void)1597 get_current_codepage(void)
1598 {
1599 return (-1);/* Unknown */
1600 }
1601 static unsigned
make_codepage_from_charset(const char * charset)1602 make_codepage_from_charset(const char *charset)
1603 {
1604 (void)charset; /* UNUSED */
1605 return (-1);/* Unknown */
1606 }
1607 static unsigned
get_current_oemcp(void)1608 get_current_oemcp(void)
1609 {
1610 return (-1);/* Unknown */
1611 }
1612
1613 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
1614
1615 /*
1616 * Return a string conversion object.
1617 */
1618 static struct archive_string_conv *
get_sconv_object(struct archive * a,const char * fc,const char * tc,int flag)1619 get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag)
1620 {
1621 struct archive_string_conv *sc;
1622 unsigned current_codepage;
1623
1624 /* Check if we have made the sconv object. */
1625 sc = find_sconv_object(a, fc, tc);
1626 if (sc != NULL)
1627 return (sc);
1628
1629 if (a == NULL)
1630 current_codepage = get_current_codepage();
1631 else
1632 current_codepage = a->current_codepage;
1633
1634 sc = create_sconv_object(canonical_charset_name(fc),
1635 canonical_charset_name(tc), current_codepage, flag);
1636 if (sc == NULL) {
1637 if (a != NULL)
1638 archive_set_error(a, ENOMEM,
1639 "Could not allocate memory for "
1640 "a string conversion object");
1641 return (NULL);
1642 }
1643
1644 /*
1645 * If there is no converter for current string conversion object,
1646 * we cannot handle this conversion.
1647 */
1648 if (sc->nconverter == 0) {
1649 if (a != NULL) {
1650 #if HAVE_ICONV
1651 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1652 "iconv_open failed : Cannot handle ``%s''",
1653 (flag & SCONV_TO_CHARSET)?tc:fc);
1654 #else
1655 archive_set_error(a, ARCHIVE_ERRNO_MISC,
1656 "A character-set conversion not fully supported "
1657 "on this platform");
1658 #endif
1659 }
1660 /* Failed; free a sconv object. */
1661 free_sconv_object(sc);
1662 return (NULL);
1663 }
1664
1665 /*
1666 * Success!
1667 */
1668 if (a != NULL)
1669 add_sconv_object(a, sc);
1670 return (sc);
1671 }
1672
1673 static const char *
get_current_charset(struct archive * a)1674 get_current_charset(struct archive *a)
1675 {
1676 const char *cur_charset;
1677
1678 if (a == NULL)
1679 cur_charset = default_iconv_charset("");
1680 else {
1681 cur_charset = default_iconv_charset(a->current_code);
1682 if (a->current_code == NULL) {
1683 a->current_code = strdup(cur_charset);
1684 a->current_codepage = get_current_codepage();
1685 a->current_oemcp = get_current_oemcp();
1686 }
1687 }
1688 return (cur_charset);
1689 }
1690
1691 /*
1692 * Make and Return a string conversion object.
1693 * Return NULL if the platform does not support the specified conversion
1694 * and best_effort is 0.
1695 * If best_effort is set, A string conversion object must be returned
1696 * unless memory allocation for the object fails, but the conversion
1697 * might fail when non-ASCII code is found.
1698 */
1699 struct archive_string_conv *
archive_string_conversion_to_charset(struct archive * a,const char * charset,int best_effort)1700 archive_string_conversion_to_charset(struct archive *a, const char *charset,
1701 int best_effort)
1702 {
1703 int flag = SCONV_TO_CHARSET;
1704
1705 if (best_effort)
1706 flag |= SCONV_BEST_EFFORT;
1707 return (get_sconv_object(a, get_current_charset(a), charset, flag));
1708 }
1709
1710 struct archive_string_conv *
archive_string_conversion_from_charset(struct archive * a,const char * charset,int best_effort)1711 archive_string_conversion_from_charset(struct archive *a, const char *charset,
1712 int best_effort)
1713 {
1714 int flag = SCONV_FROM_CHARSET;
1715
1716 if (best_effort)
1717 flag |= SCONV_BEST_EFFORT;
1718 return (get_sconv_object(a, charset, get_current_charset(a), flag));
1719 }
1720
1721 /*
1722 * archive_string_default_conversion_*_archive() are provided for Windows
1723 * platform because other archiver application use CP_OEMCP for
1724 * MultiByteToWideChar() and WideCharToMultiByte() for the filenames
1725 * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP
1726 * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP).
1727 * So we should make a string conversion between CP_ACP and CP_OEMCP
1728 * for compatibillty.
1729 */
1730 #if defined(_WIN32) && !defined(__CYGWIN__)
1731 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1732 archive_string_default_conversion_for_read(struct archive *a)
1733 {
1734 const char *cur_charset = get_current_charset(a);
1735 char oemcp[16];
1736
1737 /* NOTE: a check of cur_charset is unneeded but we need
1738 * that get_current_charset() has been surely called at
1739 * this time whatever C compiler optimized. */
1740 if (cur_charset != NULL &&
1741 (a->current_codepage == CP_C_LOCALE ||
1742 a->current_codepage == a->current_oemcp))
1743 return (NULL);/* no conversion. */
1744
1745 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1746 /* Make sure a null termination must be set. */
1747 oemcp[sizeof(oemcp)-1] = '\0';
1748 return (get_sconv_object(a, oemcp, cur_charset,
1749 SCONV_FROM_CHARSET));
1750 }
1751
1752 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1753 archive_string_default_conversion_for_write(struct archive *a)
1754 {
1755 const char *cur_charset = get_current_charset(a);
1756 char oemcp[16];
1757
1758 /* NOTE: a check of cur_charset is unneeded but we need
1759 * that get_current_charset() has been surely called at
1760 * this time whatever C compiler optimized. */
1761 if (cur_charset != NULL &&
1762 (a->current_codepage == CP_C_LOCALE ||
1763 a->current_codepage == a->current_oemcp))
1764 return (NULL);/* no conversion. */
1765
1766 _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp);
1767 /* Make sure a null termination must be set. */
1768 oemcp[sizeof(oemcp)-1] = '\0';
1769 return (get_sconv_object(a, cur_charset, oemcp,
1770 SCONV_TO_CHARSET));
1771 }
1772 #else
1773 struct archive_string_conv *
archive_string_default_conversion_for_read(struct archive * a)1774 archive_string_default_conversion_for_read(struct archive *a)
1775 {
1776 (void)a; /* UNUSED */
1777 return (NULL);
1778 }
1779
1780 struct archive_string_conv *
archive_string_default_conversion_for_write(struct archive * a)1781 archive_string_default_conversion_for_write(struct archive *a)
1782 {
1783 (void)a; /* UNUSED */
1784 return (NULL);
1785 }
1786 #endif
1787
1788 /*
1789 * Dispose of all character conversion objects in the archive object.
1790 */
1791 void
archive_string_conversion_free(struct archive * a)1792 archive_string_conversion_free(struct archive *a)
1793 {
1794 struct archive_string_conv *sc;
1795 struct archive_string_conv *sc_next;
1796
1797 for (sc = a->sconv; sc != NULL; sc = sc_next) {
1798 sc_next = sc->next;
1799 free_sconv_object(sc);
1800 }
1801 a->sconv = NULL;
1802 free(a->current_code);
1803 a->current_code = NULL;
1804 }
1805
1806 /*
1807 * Return a conversion charset name.
1808 */
1809 const char *
archive_string_conversion_charset_name(struct archive_string_conv * sc)1810 archive_string_conversion_charset_name(struct archive_string_conv *sc)
1811 {
1812 if (sc->flag & SCONV_TO_CHARSET)
1813 return (sc->to_charset);
1814 else
1815 return (sc->from_charset);
1816 }
1817
1818 /*
1819 * Change the behavior of a string conversion.
1820 */
1821 void
archive_string_conversion_set_opt(struct archive_string_conv * sc,int opt)1822 archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt)
1823 {
1824 switch (opt) {
1825 /*
1826 * A filename in UTF-8 was made with libarchive 2.x in a wrong
1827 * assumption that wchar_t was Unicode.
1828 * This option enables simulating the assumption in order to read
1829 * that filname correctly.
1830 */
1831 case SCONV_SET_OPT_UTF8_LIBARCHIVE2X:
1832 #if (defined(_WIN32) && !defined(__CYGWIN__)) \
1833 || defined(__STDC_ISO_10646__) || defined(__APPLE__)
1834 /*
1835 * Nothing to do for it since wchar_t on these platforms
1836 * is really Unicode.
1837 */
1838 (void)sc; /* UNUSED */
1839 #else
1840 if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) {
1841 sc->flag |= SCONV_UTF8_LIBARCHIVE_2;
1842 /* Set up string converters. */
1843 setup_converter(sc);
1844 }
1845 #endif
1846 break;
1847 case SCONV_SET_OPT_NORMALIZATION_C:
1848 if ((sc->flag & SCONV_NORMALIZATION_C) == 0) {
1849 sc->flag |= SCONV_NORMALIZATION_C;
1850 sc->flag &= ~SCONV_NORMALIZATION_D;
1851 /* Set up string converters. */
1852 setup_converter(sc);
1853 }
1854 break;
1855 case SCONV_SET_OPT_NORMALIZATION_D:
1856 #if defined(HAVE_ICONV)
1857 /*
1858 * If iconv will take the string, do not change the
1859 * setting of the normalization.
1860 */
1861 if (!(sc->flag & SCONV_WIN_CP) &&
1862 (sc->flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8)) &&
1863 !(sc->flag & (SCONV_TO_UTF16 | SCONV_TO_UTF8)))
1864 break;
1865 #endif
1866 if ((sc->flag & SCONV_NORMALIZATION_D) == 0) {
1867 sc->flag |= SCONV_NORMALIZATION_D;
1868 sc->flag &= ~SCONV_NORMALIZATION_C;
1869 /* Set up string converters. */
1870 setup_converter(sc);
1871 }
1872 break;
1873 default:
1874 break;
1875 }
1876 }
1877
1878 /*
1879 *
1880 * Copy one archive_string to another in locale conversion.
1881 *
1882 * archive_strncat_l();
1883 * archive_strncpy_l();
1884 *
1885 */
1886
1887 static size_t
mbsnbytes(const void * _p,size_t n)1888 mbsnbytes(const void *_p, size_t n)
1889 {
1890 size_t s;
1891 const char *p, *pp;
1892
1893 if (_p == NULL)
1894 return (0);
1895 p = (const char *)_p;
1896
1897 /* Like strlen(p), except won't examine positions beyond p[n]. */
1898 s = 0;
1899 pp = p;
1900 while (s < n && *pp) {
1901 pp++;
1902 s++;
1903 }
1904 return (s);
1905 }
1906
1907 static size_t
utf16nbytes(const void * _p,size_t n)1908 utf16nbytes(const void *_p, size_t n)
1909 {
1910 size_t s;
1911 const char *p, *pp;
1912
1913 if (_p == NULL)
1914 return (0);
1915 p = (const char *)_p;
1916
1917 /* Like strlen(p), except won't examine positions beyond p[n]. */
1918 s = 0;
1919 pp = p;
1920 n >>= 1;
1921 while (s < n && (pp[0] || pp[1])) {
1922 pp += 2;
1923 s++;
1924 }
1925 return (s<<1);
1926 }
1927
1928 int
archive_strncpy_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1929 archive_strncpy_l(struct archive_string *as, const void *_p, size_t n,
1930 struct archive_string_conv *sc)
1931 {
1932 as->length = 0;
1933 return (archive_strncat_l(as, _p, n, sc));
1934 }
1935
1936 int
archive_strncat_l(struct archive_string * as,const void * _p,size_t n,struct archive_string_conv * sc)1937 archive_strncat_l(struct archive_string *as, const void *_p, size_t n,
1938 struct archive_string_conv *sc)
1939 {
1940 const void *s;
1941 size_t length;
1942 int i, r = 0, r2;
1943
1944 /* We must allocate memory even if there is no data for conversion
1945 * or copy. This simulates archive_string_append behavior. */
1946 if (_p == NULL || n == 0) {
1947 int tn = 1;
1948 if (sc != NULL && (sc->flag & SCONV_TO_UTF16))
1949 tn = 2;
1950 if (archive_string_ensure(as, as->length + tn) == NULL)
1951 return (-1);
1952 as->s[as->length] = 0;
1953 if (tn == 2)
1954 as->s[as->length+1] = 0;
1955 return (0);
1956 }
1957
1958 /*
1959 * If sc is NULL, we just make a copy.
1960 */
1961 if (sc == NULL) {
1962 length = mbsnbytes(_p, n);
1963 if (archive_string_append(as, _p, length) == NULL)
1964 return (-1);/* No memory */
1965 return (0);
1966 }
1967
1968 if (sc->flag & SCONV_FROM_UTF16)
1969 length = utf16nbytes(_p, n);
1970 else
1971 length = mbsnbytes(_p, n);
1972 s = _p;
1973 i = 0;
1974 if (sc->nconverter > 1) {
1975 sc->utftmp.length = 0;
1976 r2 = sc->converter[0](&(sc->utftmp), s, length, sc);
1977 if (r2 != 0 && errno == ENOMEM)
1978 return (r2);
1979 if (r > r2)
1980 r = r2;
1981 s = sc->utftmp.s;
1982 length = sc->utftmp.length;
1983 ++i;
1984 }
1985 r2 = sc->converter[i](as, s, length, sc);
1986 if (r > r2)
1987 r = r2;
1988 return (r);
1989 }
1990
1991 #if HAVE_ICONV
1992
1993 /*
1994 * Return -1 if conversion failes.
1995 */
1996 static int
iconv_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)1997 iconv_strncat_in_locale(struct archive_string *as, const void *_p,
1998 size_t length, struct archive_string_conv *sc)
1999 {
2000 ICONV_CONST char *itp;
2001 size_t remaining;
2002 iconv_t cd;
2003 char *outp;
2004 size_t avail, bs;
2005 int return_value = 0; /* success */
2006 int to_size, from_size;
2007
2008 if (sc->flag & SCONV_TO_UTF16)
2009 to_size = 2;
2010 else
2011 to_size = 1;
2012 if (sc->flag & SCONV_FROM_UTF16)
2013 from_size = 2;
2014 else
2015 from_size = 1;
2016
2017 if (archive_string_ensure(as, as->length + length*2+to_size) == NULL)
2018 return (-1);
2019
2020 cd = sc->cd;
2021 itp = (char *)(uintptr_t)_p;
2022 remaining = length;
2023 outp = as->s + as->length;
2024 avail = as->buffer_length - as->length - to_size;
2025 while (remaining >= (size_t)from_size) {
2026 size_t result = iconv(cd, &itp, &remaining, &outp, &avail);
2027
2028 if (result != (size_t)-1)
2029 break; /* Conversion completed. */
2030
2031 if (errno == EILSEQ || errno == EINVAL) {
2032 /*
2033 * If an output charset is UTF-8 or UTF-16BE/LE,
2034 * unknown character should be U+FFFD
2035 * (replacement character).
2036 */
2037 if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) {
2038 size_t rbytes;
2039 if (sc->flag & SCONV_TO_UTF8)
2040 rbytes = UTF8_R_CHAR_SIZE;
2041 else
2042 rbytes = 2;
2043
2044 if (avail < rbytes) {
2045 as->length = outp - as->s;
2046 bs = as->buffer_length +
2047 (remaining * to_size) + rbytes;
2048 if (NULL ==
2049 archive_string_ensure(as, bs))
2050 return (-1);
2051 outp = as->s + as->length;
2052 avail = as->buffer_length
2053 - as->length - to_size;
2054 }
2055 if (sc->flag & SCONV_TO_UTF8)
2056 UTF8_SET_R_CHAR(outp);
2057 else if (sc->flag & SCONV_TO_UTF16BE)
2058 archive_be16enc(outp, UNICODE_R_CHAR);
2059 else
2060 archive_le16enc(outp, UNICODE_R_CHAR);
2061 outp += rbytes;
2062 avail -= rbytes;
2063 } else {
2064 /* Skip the illegal input bytes. */
2065 *outp++ = '?';
2066 avail--;
2067 }
2068 itp += from_size;
2069 remaining -= from_size;
2070 return_value = -1; /* failure */
2071 } else {
2072 /* E2BIG no output buffer,
2073 * Increase an output buffer. */
2074 as->length = outp - as->s;
2075 bs = as->buffer_length + remaining * 2;
2076 if (NULL == archive_string_ensure(as, bs))
2077 return (-1);
2078 outp = as->s + as->length;
2079 avail = as->buffer_length - as->length - to_size;
2080 }
2081 }
2082 as->length = outp - as->s;
2083 as->s[as->length] = 0;
2084 if (to_size == 2)
2085 as->s[as->length+1] = 0;
2086 return (return_value);
2087 }
2088
2089 #endif /* HAVE_ICONV */
2090
2091
2092 #if defined(_WIN32) && !defined(__CYGWIN__)
2093
2094 /*
2095 * Translate a string from a some CodePage to an another CodePage by
2096 * Windows APIs, and copy the result. Return -1 if conversion failes.
2097 */
2098 static int
strncat_in_codepage(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2099 strncat_in_codepage(struct archive_string *as,
2100 const void *_p, size_t length, struct archive_string_conv *sc)
2101 {
2102 const char *s = (const char *)_p;
2103 struct archive_wstring aws;
2104 size_t l;
2105 int r, saved_flag;
2106
2107 archive_string_init(&aws);
2108 saved_flag = sc->flag;
2109 sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C);
2110 r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc);
2111 sc->flag = saved_flag;
2112 if (r != 0) {
2113 archive_wstring_free(&aws);
2114 if (errno != ENOMEM)
2115 archive_string_append(as, s, length);
2116 return (-1);
2117 }
2118
2119 l = as->length;
2120 r = archive_string_append_from_wcs_in_codepage(
2121 as, aws.s, aws.length, sc);
2122 if (r != 0 && errno != ENOMEM && l == as->length)
2123 archive_string_append(as, s, length);
2124 archive_wstring_free(&aws);
2125 return (r);
2126 }
2127
2128 /*
2129 * Test whether MBS ==> WCS is okay.
2130 */
2131 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2132 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2133 {
2134 const char *p = (const char *)_p;
2135 unsigned codepage;
2136 DWORD mbflag = MB_ERR_INVALID_CHARS;
2137
2138 if (sc->flag & SCONV_FROM_CHARSET)
2139 codepage = sc->to_cp;
2140 else
2141 codepage = sc->from_cp;
2142
2143 if (codepage == CP_C_LOCALE)
2144 return (0);
2145 if (codepage != CP_UTF8)
2146 mbflag |= MB_PRECOMPOSED;
2147
2148 if (MultiByteToWideChar(codepage, mbflag, p, (int)n, NULL, 0) == 0)
2149 return (-1); /* Invalid */
2150 return (0); /* Okay */
2151 }
2152
2153 #else
2154
2155 /*
2156 * Test whether MBS ==> WCS is okay.
2157 */
2158 static int
invalid_mbs(const void * _p,size_t n,struct archive_string_conv * sc)2159 invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc)
2160 {
2161 const char *p = (const char *)_p;
2162 size_t r;
2163
2164 #if HAVE_MBRTOWC
2165 mbstate_t shift_state;
2166
2167 memset(&shift_state, 0, sizeof(shift_state));
2168 #else
2169 /* Clear the shift state before starting. */
2170 mbtowc(NULL, NULL, 0);
2171 #endif
2172 while (n) {
2173 wchar_t wc;
2174
2175 #if HAVE_MBRTOWC
2176 r = mbrtowc(&wc, p, n, &shift_state);
2177 #else
2178 r = mbtowc(&wc, p, n);
2179 #endif
2180 if (r == (size_t)-1 || r == (size_t)-2)
2181 return (-1);/* Invalid. */
2182 if (r == 0)
2183 break;
2184 p += r;
2185 n -= r;
2186 }
2187 (void)sc; /* UNUSED */
2188 return (0); /* All Okey. */
2189 }
2190
2191 #endif /* defined(_WIN32) && !defined(__CYGWIN__) */
2192
2193 /*
2194 * Basically returns -1 because we cannot make a conversion of charset
2195 * without iconv but in some cases this would return 0.
2196 * Returns 0 if all copied characters are ASCII.
2197 * Returns 0 if both from-locale and to-locale are the same and those
2198 * can be WCS with no error.
2199 */
2200 static int
best_effort_strncat_in_locale(struct archive_string * as,const void * _p,size_t length,struct archive_string_conv * sc)2201 best_effort_strncat_in_locale(struct archive_string *as, const void *_p,
2202 size_t length, struct archive_string_conv *sc)
2203 {
2204 size_t remaining;
2205 char *otp;
2206 const uint8_t *itp;
2207 size_t avail;
2208 int return_value = 0; /* success */
2209
2210 /*
2211 * If both from-locale and to-locale is the same, this makes a copy.
2212 * And then this checks all copied MBS can be WCS if so returns 0.
2213 */
2214 if (sc->same) {
2215 if (archive_string_append(as, _p, length) == NULL)
2216 return (-1);/* No memory */
2217 return (invalid_mbs(_p, length, sc));
2218 }
2219
2220 /*
2221 * If a character is ASCII, this just copies it. If not, this
2222 * assigns '?' charater instead but in UTF-8 locale this assigns
2223 * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD,
2224 * a Replacement Character in Unicode.
2225 */
2226 if (archive_string_ensure(as, as->length + length + 1) == NULL)
2227 return (-1);
2228
2229 remaining = length;
2230 itp = (const uint8_t *)_p;
2231 otp = as->s + as->length;
2232 avail = as->buffer_length - as->length -1;
2233 while (*itp && remaining > 0) {
2234 if (*itp > 127 && (sc->flag & SCONV_TO_UTF8)) {
2235 if (avail < UTF8_R_CHAR_SIZE) {
2236 as->length = otp - as->s;
2237 if (NULL == archive_string_ensure(as,
2238 as->buffer_length + remaining +
2239 UTF8_R_CHAR_SIZE))
2240 return (-1);
2241 otp = as->s + as->length;
2242 avail = as->buffer_length - as->length -1;
2243 }
2244 /*
2245 * When coping a string in UTF-8, unknown character
2246 * should be U+FFFD (replacement character).
2247 */
2248 UTF8_SET_R_CHAR(otp);
2249 otp += UTF8_R_CHAR_SIZE;
2250 avail -= UTF8_R_CHAR_SIZE;
2251 itp++;
2252 remaining--;
2253 return_value = -1;
2254 } else if (*itp > 127) {
2255 *otp++ = '?';
2256 itp++;
2257 remaining--;
2258 return_value = -1;
2259 } else {
2260 *otp++ = (char)*itp++;
2261 remaining--;
2262 }
2263 }
2264 as->length = otp - as->s;
2265 as->s[as->length] = '\0';
2266 return (return_value);
2267 }
2268
2269
2270 /*
2271 * Unicode conversion functions.
2272 * - UTF-8 <===> UTF-8 in removing surrogate pairs.
2273 * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs.
2274 * - UTF-8 made by libarchive 2.x ===> UTF-8.
2275 * - UTF-16BE <===> UTF-8.
2276 *
2277 */
2278
2279 /*
2280 * Utility to convert a single UTF-8 sequence.
2281 *
2282 * Usually return used bytes, return used byte in negative value when
2283 * a unicode character is replaced with U+FFFD.
2284 * See also http://unicode.org/review/pr-121.html Public Review Issue #121
2285 * Recommended Practice for Replacement Characters.
2286 */
2287 static int
_utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2288 _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2289 {
2290 static const char utf8_count[256] = {
2291 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */
2292 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */
2293 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */
2294 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */
2295 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */
2296 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */
2297 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */
2298 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */
2299 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */
2300 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */
2301 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */
2302 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */
2303 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */
2304 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */
2305 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */
2306 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */
2307 };
2308 int ch, i;
2309 int cnt;
2310 uint32_t wc;
2311
2312 /* Sanity check. */
2313 if (n == 0)
2314 return (0);
2315 /*
2316 * Decode 1-4 bytes depending on the value of the first byte.
2317 */
2318 ch = (unsigned char)*s;
2319 if (ch == 0)
2320 return (0); /* Standard: return 0 for end-of-string. */
2321 cnt = utf8_count[ch];
2322
2323 /* Invalide sequence or there are not plenty bytes. */
2324 if ((int)n < cnt) {
2325 cnt = (int)n;
2326 for (i = 1; i < cnt; i++) {
2327 if ((s[i] & 0xc0) != 0x80) {
2328 cnt = i;
2329 break;
2330 }
2331 }
2332 goto invalid_sequence;
2333 }
2334
2335 /* Make a Unicode code point from a single UTF-8 sequence. */
2336 switch (cnt) {
2337 case 1: /* 1 byte sequence. */
2338 *pwc = ch & 0x7f;
2339 return (cnt);
2340 case 2: /* 2 bytes sequence. */
2341 if ((s[1] & 0xc0) != 0x80) {
2342 cnt = 1;
2343 goto invalid_sequence;
2344 }
2345 *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f);
2346 return (cnt);
2347 case 3: /* 3 bytes sequence. */
2348 if ((s[1] & 0xc0) != 0x80) {
2349 cnt = 1;
2350 goto invalid_sequence;
2351 }
2352 if ((s[2] & 0xc0) != 0x80) {
2353 cnt = 2;
2354 goto invalid_sequence;
2355 }
2356 wc = ((ch & 0x0f) << 12)
2357 | ((s[1] & 0x3f) << 6)
2358 | (s[2] & 0x3f);
2359 if (wc < 0x800)
2360 goto invalid_sequence;/* Overlong sequence. */
2361 break;
2362 case 4: /* 4 bytes sequence. */
2363 if ((s[1] & 0xc0) != 0x80) {
2364 cnt = 1;
2365 goto invalid_sequence;
2366 }
2367 if ((s[2] & 0xc0) != 0x80) {
2368 cnt = 2;
2369 goto invalid_sequence;
2370 }
2371 if ((s[3] & 0xc0) != 0x80) {
2372 cnt = 3;
2373 goto invalid_sequence;
2374 }
2375 wc = ((ch & 0x07) << 18)
2376 | ((s[1] & 0x3f) << 12)
2377 | ((s[2] & 0x3f) << 6)
2378 | (s[3] & 0x3f);
2379 if (wc < 0x10000)
2380 goto invalid_sequence;/* Overlong sequence. */
2381 break;
2382 default: /* Others are all invalid sequence. */
2383 if (ch == 0xc0 || ch == 0xc1)
2384 cnt = 2;
2385 else if (ch >= 0xf5 && ch <= 0xf7)
2386 cnt = 4;
2387 else if (ch >= 0xf8 && ch <= 0xfb)
2388 cnt = 5;
2389 else if (ch == 0xfc || ch == 0xfd)
2390 cnt = 6;
2391 else
2392 cnt = 1;
2393 if ((int)n < cnt)
2394 cnt = (int)n;
2395 for (i = 1; i < cnt; i++) {
2396 if ((s[i] & 0xc0) != 0x80) {
2397 cnt = i;
2398 break;
2399 }
2400 }
2401 goto invalid_sequence;
2402 }
2403
2404 /* The code point larger than 0x10FFFF is not leagal
2405 * Unicode values. */
2406 if (wc > UNICODE_MAX)
2407 goto invalid_sequence;
2408 /* Correctly gets a Unicode, returns used bytes. */
2409 *pwc = wc;
2410 return (cnt);
2411 invalid_sequence:
2412 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2413 return (cnt * -1);
2414 }
2415
2416 static int
utf8_to_unicode(uint32_t * pwc,const char * s,size_t n)2417 utf8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2418 {
2419 int cnt;
2420
2421 cnt = _utf8_to_unicode(pwc, s, n);
2422 /* Any of Surrogate pair is not leagal Unicode values. */
2423 if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc))
2424 return (-3);
2425 return (cnt);
2426 }
2427
2428 static inline uint32_t
combine_surrogate_pair(uint32_t uc,uint32_t uc2)2429 combine_surrogate_pair(uint32_t uc, uint32_t uc2)
2430 {
2431 uc -= 0xD800;
2432 uc *= 0x400;
2433 uc += uc2 - 0xDC00;
2434 uc += 0x10000;
2435 return (uc);
2436 }
2437
2438 /*
2439 * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in
2440 * removing surrogate pairs.
2441 *
2442 * CESU-8: The Compatibility Encoding Scheme for UTF-16.
2443 *
2444 * Usually return used bytes, return used byte in negative value when
2445 * a unicode character is replaced with U+FFFD.
2446 */
2447 static int
cesu8_to_unicode(uint32_t * pwc,const char * s,size_t n)2448 cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n)
2449 {
2450 uint32_t wc = 0;
2451 int cnt;
2452
2453 cnt = _utf8_to_unicode(&wc, s, n);
2454 if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) {
2455 uint32_t wc2 = 0;
2456 if (n - 3 < 3) {
2457 /* Invalid byte sequence. */
2458 goto invalid_sequence;
2459 }
2460 cnt = _utf8_to_unicode(&wc2, s+3, n-3);
2461 if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) {
2462 /* Invalid byte sequence. */
2463 goto invalid_sequence;
2464 }
2465 wc = combine_surrogate_pair(wc, wc2);
2466 cnt = 6;
2467 } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) {
2468 /* Invalid byte sequence. */
2469 goto invalid_sequence;
2470 }
2471 *pwc = wc;
2472 return (cnt);
2473 invalid_sequence:
2474 *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */
2475 if (cnt > 0)
2476 cnt *= -1;
2477 return (cnt);
2478 }
2479
2480 /*
2481 * Convert a Unicode code point to a single UTF-8 sequence.
2482 *
2483 * NOTE:This function does not check if the Unicode is leagal or not.
2484 * Please you definitely check it before calling this.
2485 */
2486 static size_t
unicode_to_utf8(char * p,size_t remaining,uint32_t uc)2487 unicode_to_utf8(char *p, size_t remaining, uint32_t uc)
2488 {
2489 char *_p = p;
2490
2491 /* Translate code point to UTF8 */
2492 if (uc <= 0x7f) {
2493 if (remaining == 0)
2494 return (0);
2495 *p++ = (char)uc;
2496 } else if (uc <= 0x7ff) {
2497 if (remaining < 2)
2498 return (0);
2499 *p++ = 0xc0 | ((uc >> 6) & 0x1f);
2500 *p++ = 0x80 | (uc & 0x3f);
2501 } else if (uc <= 0xffff) {
2502 if (remaining < 3)
2503 return (0);
2504 *p++ = 0xe0 | ((uc >> 12) & 0x0f);
2505 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2506 *p++ = 0x80 | (uc & 0x3f);
2507 } else if (uc <= UNICODE_MAX) {
2508 if (remaining < 4)
2509 return (0);
2510 *p++ = 0xf0 | ((uc >> 18) & 0x07);
2511 *p++ = 0x80 | ((uc >> 12) & 0x3f);
2512 *p++ = 0x80 | ((uc >> 6) & 0x3f);
2513 *p++ = 0x80 | (uc & 0x3f);
2514 } else {
2515 /*
2516 * Undescribed code point should be U+FFFD
2517 * (replacement character).
2518 */
2519 if (remaining < UTF8_R_CHAR_SIZE)
2520 return (0);
2521 UTF8_SET_R_CHAR(p);
2522 p += UTF8_R_CHAR_SIZE;
2523 }
2524 return (p - _p);
2525 }
2526
2527 static int
utf16be_to_unicode(uint32_t * pwc,const char * s,size_t n)2528 utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n)
2529 {
2530 return (utf16_to_unicode(pwc, s, n, 1));
2531 }
2532
2533 static int
utf16le_to_unicode(uint32_t * pwc,const char * s,size_t n)2534 utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n)
2535 {
2536 return (utf16_to_unicode(pwc, s, n, 0));
2537 }
2538
2539 static int
utf16_to_unicode(uint32_t * pwc,const char * s,size_t n,int be)2540 utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be)
2541 {
2542 const char *utf16 = s;
2543 unsigned uc;
2544
2545 if (n == 0)
2546 return (0);
2547 if (n == 1) {
2548 /* set the Replacement Character instead. */
2549 *pwc = UNICODE_R_CHAR;
2550 return (-1);
2551 }
2552
2553 if (be)
2554 uc = archive_be16dec(utf16);
2555 else
2556 uc = archive_le16dec(utf16);
2557 utf16 += 2;
2558
2559 /* If this is a surrogate pair, assemble the full code point.*/
2560 if (IS_HIGH_SURROGATE_LA(uc)) {
2561 unsigned uc2;
2562
2563 if (n >= 4) {
2564 if (be)
2565 uc2 = archive_be16dec(utf16);
2566 else
2567 uc2 = archive_le16dec(utf16);
2568 } else
2569 uc2 = 0;
2570 if (IS_LOW_SURROGATE_LA(uc2)) {
2571 uc = combine_surrogate_pair(uc, uc2);
2572 utf16 += 2;
2573 } else {
2574 /* Undescribed code point should be U+FFFD
2575 * (replacement character). */
2576 *pwc = UNICODE_R_CHAR;
2577 return (-2);
2578 }
2579 }
2580
2581 /*
2582 * Surrogate pair values(0xd800 through 0xdfff) are only
2583 * used by UTF-16, so, after above culculation, the code
2584 * must not be surrogate values, and Unicode has no codes
2585 * larger than 0x10ffff. Thus, those are not leagal Unicode
2586 * values.
2587 */
2588 if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) {
2589 /* Undescribed code point should be U+FFFD
2590 * (replacement character). */
2591 *pwc = UNICODE_R_CHAR;
2592 return (((int)(utf16 - s)) * -1);
2593 }
2594 *pwc = uc;
2595 return ((int)(utf16 - s));
2596 }
2597
2598 static size_t
unicode_to_utf16be(char * p,size_t remaining,uint32_t uc)2599 unicode_to_utf16be(char *p, size_t remaining, uint32_t uc)
2600 {
2601 char *utf16 = p;
2602
2603 if (uc > 0xffff) {
2604 /* We have a code point that won't fit into a
2605 * wchar_t; convert it to a surrogate pair. */
2606 if (remaining < 4)
2607 return (0);
2608 uc -= 0x10000;
2609 archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2610 archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2611 return (4);
2612 } else {
2613 if (remaining < 2)
2614 return (0);
2615 archive_be16enc(utf16, uc);
2616 return (2);
2617 }
2618 }
2619
2620 static size_t
unicode_to_utf16le(char * p,size_t remaining,uint32_t uc)2621 unicode_to_utf16le(char *p, size_t remaining, uint32_t uc)
2622 {
2623 char *utf16 = p;
2624
2625 if (uc > 0xffff) {
2626 /* We have a code point that won't fit into a
2627 * wchar_t; convert it to a surrogate pair. */
2628 if (remaining < 4)
2629 return (0);
2630 uc -= 0x10000;
2631 archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800);
2632 archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00);
2633 return (4);
2634 } else {
2635 if (remaining < 2)
2636 return (0);
2637 archive_le16enc(utf16, uc);
2638 return (2);
2639 }
2640 }
2641
2642 /*
2643 * Copy UTF-8 string in checking surrogate pair.
2644 * If any surrogate pair are found, it would be canonicalized.
2645 */
2646 static int
strncat_from_utf8_to_utf8(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2647 strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p,
2648 size_t len, struct archive_string_conv *sc)
2649 {
2650 const char *s;
2651 char *p, *endp;
2652 int n, ret = 0;
2653
2654 (void)sc; /* UNUSED */
2655
2656 if (archive_string_ensure(as, as->length + len + 1) == NULL)
2657 return (-1);
2658
2659 s = (const char *)_p;
2660 p = as->s + as->length;
2661 endp = as->s + as->buffer_length -1;
2662 do {
2663 uint32_t uc;
2664 const char *ss = s;
2665 size_t w;
2666
2667 /*
2668 * Forward byte sequence until a conversion of that is needed.
2669 */
2670 while ((n = utf8_to_unicode(&uc, s, len)) > 0) {
2671 s += n;
2672 len -= n;
2673 }
2674 if (ss < s) {
2675 if (p + (s - ss) > endp) {
2676 as->length = p - as->s;
2677 if (archive_string_ensure(as,
2678 as->buffer_length + len + 1) == NULL)
2679 return (-1);
2680 p = as->s + as->length;
2681 endp = as->s + as->buffer_length -1;
2682 }
2683
2684 memcpy(p, ss, s - ss);
2685 p += s - ss;
2686 }
2687
2688 /*
2689 * If n is negative, current byte sequence needs a replacement.
2690 */
2691 if (n < 0) {
2692 if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) {
2693 /* Current byte sequence may be CESU-8. */
2694 n = cesu8_to_unicode(&uc, s, len);
2695 }
2696 if (n < 0) {
2697 ret = -1;
2698 n *= -1;/* Use a replaced unicode character. */
2699 }
2700
2701 /* Rebuild UTF-8 byte sequence. */
2702 while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) {
2703 as->length = p - as->s;
2704 if (archive_string_ensure(as,
2705 as->buffer_length + len + 1) == NULL)
2706 return (-1);
2707 p = as->s + as->length;
2708 endp = as->s + as->buffer_length -1;
2709 }
2710 p += w;
2711 s += n;
2712 len -= n;
2713 }
2714 } while (n > 0);
2715 as->length = p - as->s;
2716 as->s[as->length] = '\0';
2717 return (ret);
2718 }
2719
2720 static int
archive_string_append_unicode(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2721 archive_string_append_unicode(struct archive_string *as, const void *_p,
2722 size_t len, struct archive_string_conv *sc)
2723 {
2724 const char *s;
2725 char *p, *endp;
2726 uint32_t uc;
2727 size_t w;
2728 int n, ret = 0, ts, tm;
2729 int (*parse)(uint32_t *, const char *, size_t);
2730 size_t (*unparse)(char *, size_t, uint32_t);
2731
2732 if (sc->flag & SCONV_TO_UTF16BE) {
2733 unparse = unicode_to_utf16be;
2734 ts = 2;
2735 } else if (sc->flag & SCONV_TO_UTF16LE) {
2736 unparse = unicode_to_utf16le;
2737 ts = 2;
2738 } else if (sc->flag & SCONV_TO_UTF8) {
2739 unparse = unicode_to_utf8;
2740 ts = 1;
2741 } else {
2742 /*
2743 * This case is going to be converted to another
2744 * character-set through iconv.
2745 */
2746 if (sc->flag & SCONV_FROM_UTF16BE) {
2747 unparse = unicode_to_utf16be;
2748 ts = 2;
2749 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2750 unparse = unicode_to_utf16le;
2751 ts = 2;
2752 } else {
2753 unparse = unicode_to_utf8;
2754 ts = 1;
2755 }
2756 }
2757
2758 if (sc->flag & SCONV_FROM_UTF16BE) {
2759 parse = utf16be_to_unicode;
2760 tm = 1;
2761 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2762 parse = utf16le_to_unicode;
2763 tm = 1;
2764 } else {
2765 parse = cesu8_to_unicode;
2766 tm = ts;
2767 }
2768
2769 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2770 return (-1);
2771
2772 s = (const char *)_p;
2773 p = as->s + as->length;
2774 endp = as->s + as->buffer_length - ts;
2775 while ((n = parse(&uc, s, len)) != 0) {
2776 if (n < 0) {
2777 /* Use a replaced unicode character. */
2778 n *= -1;
2779 ret = -1;
2780 }
2781 s += n;
2782 len -= n;
2783 while ((w = unparse(p, endp - p, uc)) == 0) {
2784 /* There is not enough output buffer so
2785 * we have to expand it. */
2786 as->length = p - as->s;
2787 if (archive_string_ensure(as,
2788 as->buffer_length + len * tm + ts) == NULL)
2789 return (-1);
2790 p = as->s + as->length;
2791 endp = as->s + as->buffer_length - ts;
2792 }
2793 p += w;
2794 }
2795 as->length = p - as->s;
2796 as->s[as->length] = '\0';
2797 if (ts == 2)
2798 as->s[as->length+1] = '\0';
2799 return (ret);
2800 }
2801
2802 /*
2803 * Following Constants for Hangul compositions this information comes from
2804 * Unicode Standard Annex #15 http://unicode.org/reports/tr15/
2805 */
2806 #define HC_SBASE 0xAC00
2807 #define HC_LBASE 0x1100
2808 #define HC_VBASE 0x1161
2809 #define HC_TBASE 0x11A7
2810 #define HC_LCOUNT 19
2811 #define HC_VCOUNT 21
2812 #define HC_TCOUNT 28
2813 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT)
2814 #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT)
2815
2816 static uint32_t
get_nfc(uint32_t uc,uint32_t uc2)2817 get_nfc(uint32_t uc, uint32_t uc2)
2818 {
2819 int t, b;
2820
2821 t = 0;
2822 b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1;
2823 while (b >= t) {
2824 int m = (t + b) / 2;
2825 if (u_composition_table[m].cp1 < uc)
2826 t = m + 1;
2827 else if (u_composition_table[m].cp1 > uc)
2828 b = m - 1;
2829 else if (u_composition_table[m].cp2 < uc2)
2830 t = m + 1;
2831 else if (u_composition_table[m].cp2 > uc2)
2832 b = m - 1;
2833 else
2834 return (u_composition_table[m].nfc);
2835 }
2836 return (0);
2837 }
2838
2839 #define FDC_MAX 10 /* The maximum number of Following Decomposable
2840 * Characters. */
2841
2842 /*
2843 * Update first code point.
2844 */
2845 #define UPDATE_UC(new_uc) do { \
2846 uc = new_uc; \
2847 ucptr = NULL; \
2848 } while (0)
2849
2850 /*
2851 * Replace first code point with second code point.
2852 */
2853 #define REPLACE_UC_WITH_UC2() do { \
2854 uc = uc2; \
2855 ucptr = uc2ptr; \
2856 n = n2; \
2857 } while (0)
2858
2859 #define EXPAND_BUFFER() do { \
2860 as->length = p - as->s; \
2861 if (archive_string_ensure(as, \
2862 as->buffer_length + len * tm + ts) == NULL)\
2863 return (-1); \
2864 p = as->s + as->length; \
2865 endp = as->s + as->buffer_length - ts; \
2866 } while (0)
2867
2868 #define UNPARSE(p, endp, uc) do { \
2869 while ((w = unparse(p, (endp) - (p), uc)) == 0) {\
2870 EXPAND_BUFFER(); \
2871 } \
2872 p += w; \
2873 } while (0)
2874
2875 /*
2876 * Write first code point.
2877 * If the code point has not be changed from its original code,
2878 * this just copies it from its original buffer pointer.
2879 * If not, this converts it to UTF-8 byte sequence and copies it.
2880 */
2881 #define WRITE_UC() do { \
2882 if (ucptr) { \
2883 if (p + n > endp) \
2884 EXPAND_BUFFER(); \
2885 switch (n) { \
2886 case 4: \
2887 *p++ = *ucptr++; \
2888 /* FALL THROUGH */ \
2889 case 3: \
2890 *p++ = *ucptr++; \
2891 /* FALL THROUGH */ \
2892 case 2: \
2893 *p++ = *ucptr++; \
2894 /* FALL THROUGH */ \
2895 case 1: \
2896 *p++ = *ucptr; \
2897 break; \
2898 } \
2899 ucptr = NULL; \
2900 } else { \
2901 UNPARSE(p, endp, uc); \
2902 } \
2903 } while (0)
2904
2905 /*
2906 * Collect following decomposable code points.
2907 */
2908 #define COLLECT_CPS(start) do { \
2909 int _i; \
2910 for (_i = start; _i < FDC_MAX ; _i++) { \
2911 nx = parse(&ucx[_i], s, len); \
2912 if (nx <= 0) \
2913 break; \
2914 cx = CCC(ucx[_i]); \
2915 if (cl >= cx && cl != 228 && cx != 228)\
2916 break; \
2917 s += nx; \
2918 len -= nx; \
2919 cl = cx; \
2920 ccx[_i] = cx; \
2921 } \
2922 if (_i >= FDC_MAX) { \
2923 ret = -1; \
2924 ucx_size = FDC_MAX; \
2925 } else \
2926 ucx_size = _i; \
2927 } while (0)
2928
2929 /*
2930 * Normalize UTF-8/UTF-16BE characters to Form C and copy the result.
2931 *
2932 * TODO: Convert composition exclusions,which are never converted
2933 * from NFC,NFD,NFKC and NFKD, to Form C.
2934 */
2935 static int
archive_string_normalize_C(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)2936 archive_string_normalize_C(struct archive_string *as, const void *_p,
2937 size_t len, struct archive_string_conv *sc)
2938 {
2939 const char *s = (const char *)_p;
2940 char *p, *endp;
2941 uint32_t uc, uc2;
2942 size_t w;
2943 int always_replace, n, n2, ret = 0, spair, ts, tm;
2944 int (*parse)(uint32_t *, const char *, size_t);
2945 size_t (*unparse)(char *, size_t, uint32_t);
2946
2947 always_replace = 1;
2948 ts = 1;/* text size. */
2949 if (sc->flag & SCONV_TO_UTF16BE) {
2950 unparse = unicode_to_utf16be;
2951 ts = 2;
2952 if (sc->flag & SCONV_FROM_UTF16BE)
2953 always_replace = 0;
2954 } else if (sc->flag & SCONV_TO_UTF16LE) {
2955 unparse = unicode_to_utf16le;
2956 ts = 2;
2957 if (sc->flag & SCONV_FROM_UTF16LE)
2958 always_replace = 0;
2959 } else if (sc->flag & SCONV_TO_UTF8) {
2960 unparse = unicode_to_utf8;
2961 if (sc->flag & SCONV_FROM_UTF8)
2962 always_replace = 0;
2963 } else {
2964 /*
2965 * This case is going to be converted to another
2966 * character-set through iconv.
2967 */
2968 always_replace = 0;
2969 if (sc->flag & SCONV_FROM_UTF16BE) {
2970 unparse = unicode_to_utf16be;
2971 ts = 2;
2972 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2973 unparse = unicode_to_utf16le;
2974 ts = 2;
2975 } else {
2976 unparse = unicode_to_utf8;
2977 }
2978 }
2979
2980 if (sc->flag & SCONV_FROM_UTF16BE) {
2981 parse = utf16be_to_unicode;
2982 tm = 1;
2983 spair = 4;/* surrogate pair size in UTF-16. */
2984 } else if (sc->flag & SCONV_FROM_UTF16LE) {
2985 parse = utf16le_to_unicode;
2986 tm = 1;
2987 spair = 4;/* surrogate pair size in UTF-16. */
2988 } else {
2989 parse = cesu8_to_unicode;
2990 tm = ts;
2991 spair = 6;/* surrogate pair size in UTF-8. */
2992 }
2993
2994 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
2995 return (-1);
2996
2997 p = as->s + as->length;
2998 endp = as->s + as->buffer_length - ts;
2999 while ((n = parse(&uc, s, len)) != 0) {
3000 const char *ucptr, *uc2ptr;
3001
3002 if (n < 0) {
3003 /* Use a replaced unicode character. */
3004 UNPARSE(p, endp, uc);
3005 s += n*-1;
3006 len -= n*-1;
3007 ret = -1;
3008 continue;
3009 } else if (n == spair || always_replace)
3010 /* uc is converted from a surrogate pair.
3011 * this should be treated as a changed code. */
3012 ucptr = NULL;
3013 else
3014 ucptr = s;
3015 s += n;
3016 len -= n;
3017
3018 /* Read second code point. */
3019 while ((n2 = parse(&uc2, s, len)) > 0) {
3020 uint32_t ucx[FDC_MAX];
3021 int ccx[FDC_MAX];
3022 int cl, cx, i, nx, ucx_size;
3023 int LIndex,SIndex;
3024 uint32_t nfc;
3025
3026 if (n2 == spair || always_replace)
3027 /* uc2 is converted from a surrogate pair.
3028 * this should be treated as a changed code. */
3029 uc2ptr = NULL;
3030 else
3031 uc2ptr = s;
3032 s += n2;
3033 len -= n2;
3034
3035 /*
3036 * If current second code point is out of decomposable
3037 * code points, finding compositions is unneeded.
3038 */
3039 if (!IS_DECOMPOSABLE_BLOCK(uc2)) {
3040 WRITE_UC();
3041 REPLACE_UC_WITH_UC2();
3042 continue;
3043 }
3044
3045 /*
3046 * Try to combine current code points.
3047 */
3048 /*
3049 * We have to combine Hangul characters according to
3050 * http://uniicode.org/reports/tr15/#Hangul
3051 */
3052 if (0 <= (LIndex = uc - HC_LBASE) &&
3053 LIndex < HC_LCOUNT) {
3054 /*
3055 * Hangul Composition.
3056 * 1. Two current code points are L and V.
3057 */
3058 int VIndex = uc2 - HC_VBASE;
3059 if (0 <= VIndex && VIndex < HC_VCOUNT) {
3060 /* Make syllable of form LV. */
3061 UPDATE_UC(HC_SBASE +
3062 (LIndex * HC_VCOUNT + VIndex) *
3063 HC_TCOUNT);
3064 } else {
3065 WRITE_UC();
3066 REPLACE_UC_WITH_UC2();
3067 }
3068 continue;
3069 } else if (0 <= (SIndex = uc - HC_SBASE) &&
3070 SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) {
3071 /*
3072 * Hangul Composition.
3073 * 2. Two current code points are LV and T.
3074 */
3075 int TIndex = uc2 - HC_TBASE;
3076 if (0 < TIndex && TIndex < HC_TCOUNT) {
3077 /* Make syllable of form LVT. */
3078 UPDATE_UC(uc + TIndex);
3079 } else {
3080 WRITE_UC();
3081 REPLACE_UC_WITH_UC2();
3082 }
3083 continue;
3084 } else if ((nfc = get_nfc(uc, uc2)) != 0) {
3085 /* A composition to current code points
3086 * is found. */
3087 UPDATE_UC(nfc);
3088 continue;
3089 } else if ((cl = CCC(uc2)) == 0) {
3090 /* Clearly 'uc2' the second code point is not
3091 * a decomposable code. */
3092 WRITE_UC();
3093 REPLACE_UC_WITH_UC2();
3094 continue;
3095 }
3096
3097 /*
3098 * Collect following decomposable code points.
3099 */
3100 cx = 0;
3101 ucx[0] = uc2;
3102 ccx[0] = cl;
3103 COLLECT_CPS(1);
3104
3105 /*
3106 * Find a composed code in the collected code points.
3107 */
3108 i = 1;
3109 while (i < ucx_size) {
3110 int j;
3111
3112 if ((nfc = get_nfc(uc, ucx[i])) == 0) {
3113 i++;
3114 continue;
3115 }
3116
3117 /*
3118 * nfc is composed of uc and ucx[i].
3119 */
3120 UPDATE_UC(nfc);
3121
3122 /*
3123 * Remove ucx[i] by shifting
3124 * following code points.
3125 */
3126 for (j = i; j+1 < ucx_size; j++) {
3127 ucx[j] = ucx[j+1];
3128 ccx[j] = ccx[j+1];
3129 }
3130 ucx_size --;
3131
3132 /*
3133 * Collect following code points blocked
3134 * by ucx[i] the removed code point.
3135 */
3136 if (ucx_size > 0 && i == ucx_size &&
3137 nx > 0 && cx == cl) {
3138 cl = ccx[ucx_size-1];
3139 COLLECT_CPS(ucx_size);
3140 }
3141 /*
3142 * Restart finding a composed code with
3143 * the updated uc from the top of the
3144 * collected code points.
3145 */
3146 i = 0;
3147 }
3148
3149 /*
3150 * Apparently the current code points are not
3151 * decomposed characters or already composed.
3152 */
3153 WRITE_UC();
3154 for (i = 0; i < ucx_size; i++)
3155 UNPARSE(p, endp, ucx[i]);
3156
3157 /*
3158 * Flush out remaining canonical combining characters.
3159 */
3160 if (nx > 0 && cx == cl && len > 0) {
3161 while ((nx = parse(&ucx[0], s, len))
3162 > 0) {
3163 cx = CCC(ucx[0]);
3164 if (cl > cx)
3165 break;
3166 s += nx;
3167 len -= nx;
3168 cl = cx;
3169 UNPARSE(p, endp, ucx[0]);
3170 }
3171 }
3172 break;
3173 }
3174 if (n2 < 0) {
3175 WRITE_UC();
3176 /* Use a replaced unicode character. */
3177 UNPARSE(p, endp, uc2);
3178 s += n2*-1;
3179 len -= n2*-1;
3180 ret = -1;
3181 continue;
3182 } else if (n2 == 0) {
3183 WRITE_UC();
3184 break;
3185 }
3186 }
3187 as->length = p - as->s;
3188 as->s[as->length] = '\0';
3189 if (ts == 2)
3190 as->s[as->length+1] = '\0';
3191 return (ret);
3192 }
3193
3194 static int
get_nfd(uint32_t * cp1,uint32_t * cp2,uint32_t uc)3195 get_nfd(uint32_t *cp1, uint32_t *cp2, uint32_t uc)
3196 {
3197 int t, b;
3198
3199 /*
3200 * These are not converted to NFD on Mac OS.
3201 */
3202 if ((uc >= 0x2000 && uc <= 0x2FFF) ||
3203 (uc >= 0xF900 && uc <= 0xFAFF) ||
3204 (uc >= 0x2F800 && uc <= 0x2FAFF))
3205 return (0);
3206 /*
3207 * Those code points are not converted to NFD on Mac OS.
3208 * I do not know the reason because it is undocumented.
3209 * NFC NFD
3210 * 1109A ==> 11099 110BA
3211 * 1109C ==> 1109B 110BA
3212 * 110AB ==> 110A5 110BA
3213 */
3214 if (uc == 0x1109A || uc == 0x1109C || uc == 0x110AB)
3215 return (0);
3216
3217 t = 0;
3218 b = sizeof(u_decomposition_table)/sizeof(u_decomposition_table[0]) -1;
3219 while (b >= t) {
3220 int m = (t + b) / 2;
3221 if (u_decomposition_table[m].nfc < uc)
3222 t = m + 1;
3223 else if (u_decomposition_table[m].nfc > uc)
3224 b = m - 1;
3225 else {
3226 *cp1 = u_decomposition_table[m].cp1;
3227 *cp2 = u_decomposition_table[m].cp2;
3228 return (1);
3229 }
3230 }
3231 return (0);
3232 }
3233
3234 #define REPLACE_UC_WITH(cp) do { \
3235 uc = cp; \
3236 ucptr = NULL; \
3237 } while (0)
3238
3239 /*
3240 * Normalize UTF-8 characters to Form D and copy the result.
3241 */
3242 static int
archive_string_normalize_D(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3243 archive_string_normalize_D(struct archive_string *as, const void *_p,
3244 size_t len, struct archive_string_conv *sc)
3245 {
3246 const char *s = (const char *)_p;
3247 char *p, *endp;
3248 uint32_t uc, uc2;
3249 size_t w;
3250 int always_replace, n, n2, ret = 0, spair, ts, tm;
3251 int (*parse)(uint32_t *, const char *, size_t);
3252 size_t (*unparse)(char *, size_t, uint32_t);
3253
3254 always_replace = 1;
3255 ts = 1;/* text size. */
3256 if (sc->flag & SCONV_TO_UTF16BE) {
3257 unparse = unicode_to_utf16be;
3258 ts = 2;
3259 if (sc->flag & SCONV_FROM_UTF16BE)
3260 always_replace = 0;
3261 } else if (sc->flag & SCONV_TO_UTF16LE) {
3262 unparse = unicode_to_utf16le;
3263 ts = 2;
3264 if (sc->flag & SCONV_FROM_UTF16LE)
3265 always_replace = 0;
3266 } else if (sc->flag & SCONV_TO_UTF8) {
3267 unparse = unicode_to_utf8;
3268 if (sc->flag & SCONV_FROM_UTF8)
3269 always_replace = 0;
3270 } else {
3271 /*
3272 * This case is going to be converted to another
3273 * character-set through iconv.
3274 */
3275 always_replace = 0;
3276 if (sc->flag & SCONV_FROM_UTF16BE) {
3277 unparse = unicode_to_utf16be;
3278 ts = 2;
3279 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3280 unparse = unicode_to_utf16le;
3281 ts = 2;
3282 } else {
3283 unparse = unicode_to_utf8;
3284 }
3285 }
3286
3287 if (sc->flag & SCONV_FROM_UTF16BE) {
3288 parse = utf16be_to_unicode;
3289 tm = 1;
3290 spair = 4;/* surrogate pair size in UTF-16. */
3291 } else if (sc->flag & SCONV_FROM_UTF16LE) {
3292 parse = utf16le_to_unicode;
3293 tm = 1;
3294 spair = 4;/* surrogate pair size in UTF-16. */
3295 } else {
3296 parse = cesu8_to_unicode;
3297 tm = ts;
3298 spair = 6;/* surrogate pair size in UTF-8. */
3299 }
3300
3301 if (archive_string_ensure(as, as->length + len * tm + ts) == NULL)
3302 return (-1);
3303
3304 p = as->s + as->length;
3305 endp = as->s + as->buffer_length - ts;
3306 while ((n = parse(&uc, s, len)) != 0) {
3307 const char *ucptr;
3308 uint32_t cp1, cp2;
3309 int SIndex;
3310 struct {
3311 uint32_t uc;
3312 int ccc;
3313 } fdc[FDC_MAX];
3314 int fdi, fdj;
3315 int ccc;
3316
3317 check_first_code:
3318 if (n < 0) {
3319 /* Use a replaced unicode character. */
3320 UNPARSE(p, endp, uc);
3321 s += n*-1;
3322 len -= n*-1;
3323 ret = -1;
3324 continue;
3325 } else if (n == spair || always_replace)
3326 /* uc is converted from a surrogate pair.
3327 * this should be treated as a changed code. */
3328 ucptr = NULL;
3329 else
3330 ucptr = s;
3331 s += n;
3332 len -= n;
3333
3334 /* Hangul Decomposition. */
3335 if ((SIndex = uc - HC_SBASE) >= 0 && SIndex < HC_SCOUNT) {
3336 int L = HC_LBASE + SIndex / HC_NCOUNT;
3337 int V = HC_VBASE + (SIndex % HC_NCOUNT) / HC_TCOUNT;
3338 int T = HC_TBASE + SIndex % HC_TCOUNT;
3339
3340 REPLACE_UC_WITH(L);
3341 WRITE_UC();
3342 REPLACE_UC_WITH(V);
3343 WRITE_UC();
3344 if (T != HC_TBASE) {
3345 REPLACE_UC_WITH(T);
3346 WRITE_UC();
3347 }
3348 continue;
3349 }
3350 if (IS_DECOMPOSABLE_BLOCK(uc) && CCC(uc) != 0) {
3351 WRITE_UC();
3352 continue;
3353 }
3354
3355 fdi = 0;
3356 while (get_nfd(&cp1, &cp2, uc) && fdi < FDC_MAX) {
3357 int k;
3358
3359 for (k = fdi; k > 0; k--)
3360 fdc[k] = fdc[k-1];
3361 fdc[0].ccc = CCC(cp2);
3362 fdc[0].uc = cp2;
3363 fdi++;
3364 REPLACE_UC_WITH(cp1);
3365 }
3366
3367 /* Read following code points. */
3368 while ((n2 = parse(&uc2, s, len)) > 0 &&
3369 (ccc = CCC(uc2)) != 0 && fdi < FDC_MAX) {
3370 int j, k;
3371
3372 s += n2;
3373 len -= n2;
3374 for (j = 0; j < fdi; j++) {
3375 if (fdc[j].ccc > ccc)
3376 break;
3377 }
3378 if (j < fdi) {
3379 for (k = fdi; k > j; k--)
3380 fdc[k] = fdc[k-1];
3381 fdc[j].ccc = ccc;
3382 fdc[j].uc = uc2;
3383 } else {
3384 fdc[fdi].ccc = ccc;
3385 fdc[fdi].uc = uc2;
3386 }
3387 fdi++;
3388 }
3389
3390 WRITE_UC();
3391 for (fdj = 0; fdj < fdi; fdj++) {
3392 REPLACE_UC_WITH(fdc[fdj].uc);
3393 WRITE_UC();
3394 }
3395
3396 if (n2 == 0)
3397 break;
3398 REPLACE_UC_WITH(uc2);
3399 n = n2;
3400 goto check_first_code;
3401 }
3402 as->length = p - as->s;
3403 as->s[as->length] = '\0';
3404 if (ts == 2)
3405 as->s[as->length+1] = '\0';
3406 return (ret);
3407 }
3408
3409 /*
3410 * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption
3411 * that WCS is Unicode. It is true for several platforms but some are false.
3412 * And then people who did not use UTF-8 locale on the non Unicode WCS
3413 * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those
3414 * now cannot get right filename from libarchive 3.x and later since we
3415 * fixed the wrong assumption and it is incompatible to older its versions.
3416 * So we provide special option, "compat-2x.x", for resolving it.
3417 * That option enable the string conversion of libarchive 2.x.
3418 *
3419 * Translates the wrong UTF-8 string made by libarchive 2.x into current
3420 * locale character set and appends to the archive_string.
3421 * Note: returns -1 if conversion fails.
3422 */
3423 static int
strncat_from_utf8_libarchive2(struct archive_string * as,const void * _p,size_t len,struct archive_string_conv * sc)3424 strncat_from_utf8_libarchive2(struct archive_string *as,
3425 const void *_p, size_t len, struct archive_string_conv *sc)
3426 {
3427 const char *s;
3428 int n;
3429 char *p;
3430 char *end;
3431 uint32_t unicode;
3432 #if HAVE_WCRTOMB
3433 mbstate_t shift_state;
3434
3435 memset(&shift_state, 0, sizeof(shift_state));
3436 #else
3437 /* Clear the shift state before starting. */
3438 wctomb(NULL, L'\0');
3439 #endif
3440 (void)sc; /* UNUSED */
3441 /*
3442 * Allocate buffer for MBS.
3443 * We need this allocation here since it is possible that
3444 * as->s is still NULL.
3445 */
3446 if (archive_string_ensure(as, as->length + len + 1) == NULL)
3447 return (-1);
3448
3449 s = (const char *)_p;
3450 p = as->s + as->length;
3451 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3452 while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) {
3453 wchar_t wc;
3454
3455 if (p >= end) {
3456 as->length = p - as->s;
3457 /* Re-allocate buffer for MBS. */
3458 if (archive_string_ensure(as,
3459 as->length + len * 2 + 1) == NULL)
3460 return (-1);
3461 p = as->s + as->length;
3462 end = as->s + as->buffer_length - MB_CUR_MAX -1;
3463 }
3464
3465 /*
3466 * As libarchie 2.x, translates the UTF-8 characters into
3467 * wide-characters in the assumption that WCS is Unicode.
3468 */
3469 if (n < 0) {
3470 n *= -1;
3471 wc = L'?';
3472 } else
3473 wc = (wchar_t)unicode;
3474
3475 s += n;
3476 len -= n;
3477 /*
3478 * Translates the wide-character into the current locale MBS.
3479 */
3480 #if HAVE_WCRTOMB
3481 n = (int)wcrtomb(p, wc, &shift_state);
3482 #else
3483 n = (int)wctomb(p, wc);
3484 #endif
3485 if (n == -1)
3486 return (-1);
3487 p += n;
3488 }
3489 as->length = p - as->s;
3490 as->s[as->length] = '\0';
3491 return (0);
3492 }
3493
3494
3495 /*
3496 * Conversion functions between current locale dependent MBS and UTF-16BE.
3497 * strncat_from_utf16be() : UTF-16BE --> MBS
3498 * strncat_to_utf16be() : MBS --> UTF16BE
3499 */
3500
3501 #if defined(_WIN32) && !defined(__CYGWIN__)
3502
3503 /*
3504 * Convert a UTF-16BE/LE string to current locale and copy the result.
3505 * Return -1 if conversion failes.
3506 */
3507 static int
win_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3508 win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes,
3509 struct archive_string_conv *sc, int be)
3510 {
3511 struct archive_string tmp;
3512 const char *u16;
3513 int ll;
3514 BOOL defchar;
3515 char *mbs;
3516 size_t mbs_size, b;
3517 int ret = 0;
3518
3519 bytes &= ~1;
3520 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3521 return (-1);
3522
3523 mbs = as->s + as->length;
3524 mbs_size = as->buffer_length - as->length -1;
3525
3526 if (sc->to_cp == CP_C_LOCALE) {
3527 /*
3528 * "C" locale special process.
3529 */
3530 u16 = _p;
3531 ll = 0;
3532 for (b = 0; b < bytes; b += 2) {
3533 uint16_t val;
3534 if (be)
3535 val = archive_be16dec(u16+b);
3536 else
3537 val = archive_le16dec(u16+b);
3538 if (val > 255) {
3539 *mbs++ = '?';
3540 ret = -1;
3541 } else
3542 *mbs++ = (char)(val&0xff);
3543 ll++;
3544 }
3545 as->length += ll;
3546 as->s[as->length] = '\0';
3547 return (ret);
3548 }
3549
3550 archive_string_init(&tmp);
3551 if (be) {
3552 if (is_big_endian()) {
3553 u16 = _p;
3554 } else {
3555 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3556 return (-1);
3557 memcpy(tmp.s, _p, bytes);
3558 for (b = 0; b < bytes; b += 2) {
3559 uint16_t val = archive_be16dec(tmp.s+b);
3560 archive_le16enc(tmp.s+b, val);
3561 }
3562 u16 = tmp.s;
3563 }
3564 } else {
3565 if (!is_big_endian()) {
3566 u16 = _p;
3567 } else {
3568 if (archive_string_ensure(&tmp, bytes+2) == NULL)
3569 return (-1);
3570 memcpy(tmp.s, _p, bytes);
3571 for (b = 0; b < bytes; b += 2) {
3572 uint16_t val = archive_le16dec(tmp.s+b);
3573 archive_be16enc(tmp.s+b, val);
3574 }
3575 u16 = tmp.s;
3576 }
3577 }
3578
3579 do {
3580 defchar = 0;
3581 ll = WideCharToMultiByte(sc->to_cp, 0,
3582 (LPCWSTR)u16, (int)bytes>>1, mbs, (int)mbs_size,
3583 NULL, &defchar);
3584 if (ll == 0 &&
3585 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
3586 /* Need more buffer for MBS. */
3587 ll = WideCharToMultiByte(sc->to_cp, 0,
3588 (LPCWSTR)u16, (int)bytes, NULL, 0, NULL, NULL);
3589 if (archive_string_ensure(as, ll +1) == NULL)
3590 return (-1);
3591 mbs = as->s + as->length;
3592 mbs_size = as->buffer_length - as->length -1;
3593 continue;
3594 }
3595 } while (0);
3596 archive_string_free(&tmp);
3597 as->length += ll;
3598 as->s[as->length] = '\0';
3599 if (ll == 0 || defchar)
3600 ret = -1;
3601 return (ret);
3602 }
3603
3604 static int
win_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3605 win_strncat_from_utf16be(struct archive_string *as, const void *_p,
3606 size_t bytes, struct archive_string_conv *sc)
3607 {
3608 return (win_strncat_from_utf16(as, _p, bytes, sc, 1));
3609 }
3610
3611 static int
win_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3612 win_strncat_from_utf16le(struct archive_string *as, const void *_p,
3613 size_t bytes, struct archive_string_conv *sc)
3614 {
3615 return (win_strncat_from_utf16(as, _p, bytes, sc, 0));
3616 }
3617
3618 static int
is_big_endian(void)3619 is_big_endian(void)
3620 {
3621 uint16_t d = 1;
3622
3623 return (archive_be16dec(&d) == 1);
3624 }
3625
3626 /*
3627 * Convert a current locale string to UTF-16BE/LE and copy the result.
3628 * Return -1 if conversion failes.
3629 */
3630 static int
win_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3631 win_strncat_to_utf16(struct archive_string *as16, const void *_p,
3632 size_t length, struct archive_string_conv *sc, int bigendian)
3633 {
3634 const char *s = (const char *)_p;
3635 char *u16;
3636 size_t count, avail;
3637
3638 if (archive_string_ensure(as16,
3639 as16->length + (length + 1) * 2) == NULL)
3640 return (-1);
3641
3642 u16 = as16->s + as16->length;
3643 avail = as16->buffer_length - 2;
3644 if (sc->from_cp == CP_C_LOCALE) {
3645 /*
3646 * "C" locale special process.
3647 */
3648 count = 0;
3649 while (count < length && *s) {
3650 if (bigendian)
3651 archive_be16enc(u16, *s);
3652 else
3653 archive_le16enc(u16, *s);
3654 u16 += 2;
3655 s++;
3656 count++;
3657 }
3658 as16->length += count << 1;
3659 as16->s[as16->length] = 0;
3660 as16->s[as16->length+1] = 0;
3661 return (0);
3662 }
3663 do {
3664 count = MultiByteToWideChar(sc->from_cp,
3665 MB_PRECOMPOSED, s, (int)length, (LPWSTR)u16, (int)avail>>1);
3666 if (count == 0 &&
3667 GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
3668 /* Need more buffer for UTF-16 string */
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 continue;
3677 }
3678 } while (0);
3679 as16->length += count * 2;
3680 as16->s[as16->length] = 0;
3681 as16->s[as16->length+1] = 0;
3682 if (count == 0)
3683 return (-1);
3684
3685 if (is_big_endian()) {
3686 if (!bigendian) {
3687 while (count > 0) {
3688 uint16_t v = archive_be16dec(u16);
3689 archive_le16enc(u16, v);
3690 u16 += 2;
3691 count--;
3692 }
3693 }
3694 } else {
3695 if (bigendian) {
3696 while (count > 0) {
3697 uint16_t v = archive_le16dec(u16);
3698 archive_be16enc(u16, v);
3699 u16 += 2;
3700 count--;
3701 }
3702 }
3703 }
3704 return (0);
3705 }
3706
3707 static int
win_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3708 win_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3709 size_t length, struct archive_string_conv *sc)
3710 {
3711 return (win_strncat_to_utf16(as16, _p, length, sc, 1));
3712 }
3713
3714 static int
win_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3715 win_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3716 size_t length, struct archive_string_conv *sc)
3717 {
3718 return (win_strncat_to_utf16(as16, _p, length, sc, 0));
3719 }
3720
3721 #endif /* _WIN32 && !__CYGWIN__ */
3722
3723 /*
3724 * Do the best effort for conversions.
3725 * We cannot handle UTF-16BE character-set without such iconv,
3726 * but there is a chance if a string consists just ASCII code or
3727 * a current locale is UTF-8.
3728 */
3729
3730 /*
3731 * Convert a UTF-16BE string to current locale and copy the result.
3732 * Return -1 if conversion failes.
3733 */
3734 static int
best_effort_strncat_from_utf16(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc,int be)3735 best_effort_strncat_from_utf16(struct archive_string *as, const void *_p,
3736 size_t bytes, struct archive_string_conv *sc, int be)
3737 {
3738 const char *utf16 = (const char *)_p;
3739 char *mbs;
3740 uint32_t uc;
3741 int n, ret;
3742
3743 (void)sc; /* UNUSED */
3744 /*
3745 * Other case, we should do the best effort.
3746 * If all character are ASCII(<0x7f), we can convert it.
3747 * if not , we set a alternative character and return -1.
3748 */
3749 ret = 0;
3750 if (archive_string_ensure(as, as->length + bytes +1) == NULL)
3751 return (-1);
3752 mbs = as->s + as->length;
3753
3754 while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) {
3755 if (n < 0) {
3756 n *= -1;
3757 ret = -1;
3758 }
3759 bytes -= n;
3760 utf16 += n;
3761
3762 if (uc > 127) {
3763 /* We cannot handle it. */
3764 *mbs++ = '?';
3765 ret = -1;
3766 } else
3767 *mbs++ = (char)uc;
3768 }
3769 as->length = mbs - as->s;
3770 as->s[as->length] = '\0';
3771 return (ret);
3772 }
3773
3774 static int
best_effort_strncat_from_utf16be(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3775 best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p,
3776 size_t bytes, struct archive_string_conv *sc)
3777 {
3778 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1));
3779 }
3780
3781 static int
best_effort_strncat_from_utf16le(struct archive_string * as,const void * _p,size_t bytes,struct archive_string_conv * sc)3782 best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p,
3783 size_t bytes, struct archive_string_conv *sc)
3784 {
3785 return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0));
3786 }
3787
3788 /*
3789 * Convert a current locale string to UTF-16BE/LE and copy the result.
3790 * Return -1 if conversion failes.
3791 */
3792 static int
best_effort_strncat_to_utf16(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc,int bigendian)3793 best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p,
3794 size_t length, struct archive_string_conv *sc, int bigendian)
3795 {
3796 const char *s = (const char *)_p;
3797 char *utf16;
3798 size_t remaining;
3799 int ret;
3800
3801 (void)sc; /* UNUSED */
3802 /*
3803 * Other case, we should do the best effort.
3804 * If all character are ASCII(<0x7f), we can convert it.
3805 * if not , we set a alternative character and return -1.
3806 */
3807 ret = 0;
3808 remaining = length;
3809
3810 if (archive_string_ensure(as16,
3811 as16->length + (length + 1) * 2) == NULL)
3812 return (-1);
3813
3814 utf16 = as16->s + as16->length;
3815 while (remaining--) {
3816 unsigned c = *s++;
3817 if (c > 127) {
3818 /* We cannot handle it. */
3819 c = UNICODE_R_CHAR;
3820 ret = -1;
3821 }
3822 if (bigendian)
3823 archive_be16enc(utf16, c);
3824 else
3825 archive_le16enc(utf16, c);
3826 utf16 += 2;
3827 }
3828 as16->length = utf16 - as16->s;
3829 as16->s[as16->length] = 0;
3830 as16->s[as16->length+1] = 0;
3831 return (ret);
3832 }
3833
3834 static int
best_effort_strncat_to_utf16be(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3835 best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p,
3836 size_t length, struct archive_string_conv *sc)
3837 {
3838 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1));
3839 }
3840
3841 static int
best_effort_strncat_to_utf16le(struct archive_string * as16,const void * _p,size_t length,struct archive_string_conv * sc)3842 best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p,
3843 size_t length, struct archive_string_conv *sc)
3844 {
3845 return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0));
3846 }
3847
3848
3849 /*
3850 * Multistring operations.
3851 */
3852
3853 void
archive_mstring_clean(struct archive_mstring * aes)3854 archive_mstring_clean(struct archive_mstring *aes)
3855 {
3856 archive_wstring_free(&(aes->aes_wcs));
3857 archive_string_free(&(aes->aes_mbs));
3858 archive_string_free(&(aes->aes_utf8));
3859 archive_string_free(&(aes->aes_mbs_in_locale));
3860 aes->aes_set = 0;
3861 }
3862
3863 void
archive_mstring_copy(struct archive_mstring * dest,struct archive_mstring * src)3864 archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src)
3865 {
3866 dest->aes_set = src->aes_set;
3867 archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs));
3868 archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8));
3869 archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs));
3870 }
3871
3872 int
archive_mstring_get_utf8(struct archive * a,struct archive_mstring * aes,const char ** p)3873 archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes,
3874 const char **p)
3875 {
3876 struct archive_string_conv *sc;
3877 int r;
3878
3879 /* If we already have a UTF8 form, return that immediately. */
3880 if (aes->aes_set & AES_SET_UTF8) {
3881 *p = aes->aes_utf8.s;
3882 return (0);
3883 }
3884
3885 *p = NULL;
3886 if (aes->aes_set & AES_SET_MBS) {
3887 sc = archive_string_conversion_to_charset(a, "UTF-8", 1);
3888 if (sc == NULL)
3889 return (-1);/* Couldn't allocate memory for sc. */
3890 r = archive_strncpy_l(&(aes->aes_mbs), aes->aes_mbs.s,
3891 aes->aes_mbs.length, sc);
3892 if (a == NULL)
3893 free_sconv_object(sc);
3894 if (r == 0) {
3895 aes->aes_set |= AES_SET_UTF8;
3896 *p = aes->aes_utf8.s;
3897 return (0);/* success. */
3898 } else
3899 return (-1);/* failure. */
3900 }
3901 return (0);/* success. */
3902 }
3903
3904 int
archive_mstring_get_mbs(struct archive * a,struct archive_mstring * aes,const char ** p)3905 archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes,
3906 const char **p)
3907 {
3908 int r, ret = 0;
3909
3910 (void)a; /* UNUSED */
3911 /* If we already have an MBS form, return that immediately. */
3912 if (aes->aes_set & AES_SET_MBS) {
3913 *p = aes->aes_mbs.s;
3914 return (ret);
3915 }
3916
3917 *p = NULL;
3918 /* If there's a WCS form, try converting with the native locale. */
3919 if (aes->aes_set & AES_SET_WCS) {
3920 archive_string_empty(&(aes->aes_mbs));
3921 r = archive_string_append_from_wcs(&(aes->aes_mbs),
3922 aes->aes_wcs.s, aes->aes_wcs.length);
3923 *p = aes->aes_mbs.s;
3924 if (r == 0) {
3925 aes->aes_set |= AES_SET_MBS;
3926 return (ret);
3927 } else
3928 ret = -1;
3929 }
3930
3931 /*
3932 * Only a UTF-8 form cannot avail because its conversion already
3933 * failed at archive_mstring_update_utf8().
3934 */
3935 return (ret);
3936 }
3937
3938 int
archive_mstring_get_wcs(struct archive * a,struct archive_mstring * aes,const wchar_t ** wp)3939 archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes,
3940 const wchar_t **wp)
3941 {
3942 int r, ret = 0;
3943
3944 (void)a;/* UNUSED */
3945 /* Return WCS form if we already have it. */
3946 if (aes->aes_set & AES_SET_WCS) {
3947 *wp = aes->aes_wcs.s;
3948 return (ret);
3949 }
3950
3951 *wp = NULL;
3952 /* Try converting MBS to WCS using native locale. */
3953 if (aes->aes_set & AES_SET_MBS) {
3954 archive_wstring_empty(&(aes->aes_wcs));
3955 r = archive_wstring_append_from_mbs(&(aes->aes_wcs),
3956 aes->aes_mbs.s, aes->aes_mbs.length);
3957 if (r == 0) {
3958 aes->aes_set |= AES_SET_WCS;
3959 *wp = aes->aes_wcs.s;
3960 } else
3961 ret = -1;/* failure. */
3962 }
3963 return (ret);
3964 }
3965
3966 int
archive_mstring_get_mbs_l(struct archive_mstring * aes,const char ** p,size_t * length,struct archive_string_conv * sc)3967 archive_mstring_get_mbs_l(struct archive_mstring *aes,
3968 const char **p, size_t *length, struct archive_string_conv *sc)
3969 {
3970 int r, ret = 0;
3971
3972 #if defined(_WIN32) && !defined(__CYGWIN__)
3973 /*
3974 * Internationalization programing on Windows must use Wide
3975 * characters because Windows platform cannot make locale UTF-8.
3976 */
3977 if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) {
3978 archive_string_empty(&(aes->aes_mbs_in_locale));
3979 r = archive_string_append_from_wcs_in_codepage(
3980 &(aes->aes_mbs_in_locale), aes->aes_wcs.s,
3981 aes->aes_wcs.length, sc);
3982 if (r == 0) {
3983 *p = aes->aes_mbs_in_locale.s;
3984 if (length != NULL)
3985 *length = aes->aes_mbs_in_locale.length;
3986 return (0);
3987 } else if (errno == ENOMEM)
3988 return (-1);
3989 else
3990 ret = -1;
3991 }
3992 #endif
3993
3994 /* If there is not an MBS form but is a WCS form, try converting
3995 * with the native locale to be used for translating it to specified
3996 * character-set. */
3997 if ((aes->aes_set & AES_SET_MBS) == 0 &&
3998 (aes->aes_set & AES_SET_WCS) != 0) {
3999 archive_string_empty(&(aes->aes_mbs));
4000 r = archive_string_append_from_wcs(&(aes->aes_mbs),
4001 aes->aes_wcs.s, aes->aes_wcs.length);
4002 if (r == 0)
4003 aes->aes_set |= AES_SET_MBS;
4004 else if (errno == ENOMEM)
4005 return (-1);
4006 else
4007 ret = -1;
4008 }
4009 /* If we already have an MBS form, use it to be translated to
4010 * specified character-set. */
4011 if (aes->aes_set & AES_SET_MBS) {
4012 if (sc == NULL) {
4013 /* Conversion is unneeded. */
4014 *p = aes->aes_mbs.s;
4015 if (length != NULL)
4016 *length = aes->aes_mbs.length;
4017 return (0);
4018 }
4019 ret = archive_strncpy_l(&(aes->aes_mbs_in_locale),
4020 aes->aes_mbs.s, aes->aes_mbs.length, sc);
4021 *p = aes->aes_mbs_in_locale.s;
4022 if (length != NULL)
4023 *length = aes->aes_mbs_in_locale.length;
4024 } else {
4025 *p = NULL;
4026 if (length != NULL)
4027 *length = 0;
4028 }
4029 return (ret);
4030 }
4031
4032 int
archive_mstring_copy_mbs(struct archive_mstring * aes,const char * mbs)4033 archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs)
4034 {
4035 if (mbs == NULL) {
4036 aes->aes_set = 0;
4037 return (0);
4038 }
4039 return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs)));
4040 }
4041
4042 int
archive_mstring_copy_mbs_len(struct archive_mstring * aes,const char * mbs,size_t len)4043 archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs,
4044 size_t len)
4045 {
4046 if (mbs == NULL) {
4047 aes->aes_set = 0;
4048 return (0);
4049 }
4050 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4051 archive_strncpy(&(aes->aes_mbs), mbs, len);
4052 archive_string_empty(&(aes->aes_utf8));
4053 archive_wstring_empty(&(aes->aes_wcs));
4054 return (0);
4055 }
4056
4057 int
archive_mstring_copy_wcs(struct archive_mstring * aes,const wchar_t * wcs)4058 archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs)
4059 {
4060 return archive_mstring_copy_wcs_len(aes, wcs,
4061 wcs == NULL ? 0 : wcslen(wcs));
4062 }
4063
4064 int
archive_mstring_copy_wcs_len(struct archive_mstring * aes,const wchar_t * wcs,size_t len)4065 archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs,
4066 size_t len)
4067 {
4068 if (wcs == NULL) {
4069 aes->aes_set = 0;
4070 }
4071 aes->aes_set = AES_SET_WCS; /* Only WCS form set. */
4072 archive_string_empty(&(aes->aes_mbs));
4073 archive_string_empty(&(aes->aes_utf8));
4074 archive_wstrncpy(&(aes->aes_wcs), wcs, len);
4075 return (0);
4076 }
4077
4078 int
archive_mstring_copy_mbs_len_l(struct archive_mstring * aes,const char * mbs,size_t len,struct archive_string_conv * sc)4079 archive_mstring_copy_mbs_len_l(struct archive_mstring *aes,
4080 const char *mbs, size_t len, struct archive_string_conv *sc)
4081 {
4082 int r;
4083
4084 if (mbs == NULL) {
4085 aes->aes_set = 0;
4086 return (0);
4087 }
4088 archive_string_empty(&(aes->aes_mbs));
4089 archive_wstring_empty(&(aes->aes_wcs));
4090 archive_string_empty(&(aes->aes_utf8));
4091 #if defined(_WIN32) && !defined(__CYGWIN__)
4092 /*
4093 * Internationalization programing on Windows must use Wide
4094 * characters because Windows platform cannot make locale UTF-8.
4095 */
4096 if (sc == NULL) {
4097 if (archive_string_append(&(aes->aes_mbs),
4098 mbs, mbsnbytes(mbs, len)) == NULL) {
4099 aes->aes_set = 0;
4100 r = -1;
4101 } else {
4102 aes->aes_set = AES_SET_MBS;
4103 r = 0;
4104 }
4105 #if defined(HAVE_ICONV)
4106 } else if (sc != NULL && sc->cd_w != (iconv_t)-1) {
4107 /*
4108 * This case happens only when MultiByteToWideChar() cannot
4109 * handle sc->from_cp, and we have to iconv in order to
4110 * translate character-set to wchar_t,UTF-16.
4111 */
4112 iconv_t cd = sc->cd;
4113 unsigned from_cp;
4114 int flag;
4115
4116 /*
4117 * Translate multi-bytes from some character-set to UTF-8.
4118 */
4119 sc->cd = sc->cd_w;
4120 r = archive_strncpy_l(&(aes->aes_utf8), mbs, len, sc);
4121 sc->cd = cd;
4122 if (r != 0) {
4123 aes->aes_set = 0;
4124 return (r);
4125 }
4126 aes->aes_set = AES_SET_UTF8;
4127
4128 /*
4129 * Append the UTF-8 string into wstring.
4130 */
4131 flag = sc->flag;
4132 sc->flag &= ~(SCONV_NORMALIZATION_C
4133 | SCONV_TO_UTF16| SCONV_FROM_UTF16);
4134 from_cp = sc->from_cp;
4135 sc->from_cp = CP_UTF8;
4136 r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs),
4137 aes->aes_utf8.s, aes->aes_utf8.length, sc);
4138 sc->flag = flag;
4139 sc->from_cp = from_cp;
4140 if (r == 0)
4141 aes->aes_set |= AES_SET_WCS;
4142 #endif
4143 } else {
4144 r = archive_wstring_append_from_mbs_in_codepage(
4145 &(aes->aes_wcs), mbs, len, sc);
4146 if (r == 0)
4147 aes->aes_set = AES_SET_WCS;
4148 else
4149 aes->aes_set = 0;
4150 }
4151 #else
4152 r = archive_strncpy_l(&(aes->aes_mbs), mbs, len, sc);
4153 if (r == 0)
4154 aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */
4155 else
4156 aes->aes_set = 0;
4157 #endif
4158 return (r);
4159 }
4160
4161 /*
4162 * The 'update' form tries to proactively update all forms of
4163 * this string (WCS and MBS) and returns an error if any of
4164 * them fail. This is used by the 'pax' handler, for instance,
4165 * to detect and report character-conversion failures early while
4166 * still allowing clients to get potentially useful values from
4167 * the more tolerant lazy conversions. (get_mbs and get_wcs will
4168 * strive to give the user something useful, so you can get hopefully
4169 * usable values even if some of the character conversions are failing.)
4170 */
4171 int
archive_mstring_update_utf8(struct archive * a,struct archive_mstring * aes,const char * utf8)4172 archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes,
4173 const char *utf8)
4174 {
4175 struct archive_string_conv *sc;
4176 int r;
4177
4178 if (utf8 == NULL) {
4179 aes->aes_set = 0;
4180 return (0); /* Succeeded in clearing everything. */
4181 }
4182
4183 /* Save the UTF8 string. */
4184 archive_strcpy(&(aes->aes_utf8), utf8);
4185
4186 /* Empty the mbs and wcs strings. */
4187 archive_string_empty(&(aes->aes_mbs));
4188 archive_wstring_empty(&(aes->aes_wcs));
4189
4190 aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */
4191
4192 /* Try converting UTF-8 to MBS, return false on failure. */
4193 sc = archive_string_conversion_from_charset(a, "UTF-8", 1);
4194 if (sc == NULL)
4195 return (-1);/* Couldn't allocate memory for sc. */
4196 r = archive_strcpy_l(&(aes->aes_mbs), utf8, sc);
4197 if (a == NULL)
4198 free_sconv_object(sc);
4199 if (r != 0)
4200 return (-1);
4201 aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */
4202
4203 /* Try converting MBS to WCS, return false on failure. */
4204 if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s,
4205 aes->aes_mbs.length))
4206 return (-1);
4207 aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS;
4208
4209 /* All conversions succeeded. */
4210 return (0);
4211 }
4212