gettext.c revision 1.31 1 /* $NetBSD: gettext.c,v 1.31 2019/10/03 16:35:57 christos Exp $ */
2
3 /*-
4 * Copyright (c) 2000, 2001 Citrus Project,
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 * $Citrus: xpg4dl/FreeBSD/lib/libintl/gettext.c,v 1.31 2001/09/27 15:18:45 yamt Exp $
29 */
30
31 #include <sys/cdefs.h>
32 __RCSID("$NetBSD: gettext.c,v 1.31 2019/10/03 16:35:57 christos Exp $");
33
34 #include <sys/param.h>
35 #include <sys/stat.h>
36 #include <sys/mman.h>
37 #include <sys/uio.h>
38
39 #include <assert.h>
40 #include <fcntl.h>
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <unistd.h>
44 #include <string.h>
45 #if 0
46 #include <util.h>
47 #endif
48 #include <libintl.h>
49 #include <locale.h>
50 #include "libintl_local.h"
51 #include "plural_parser.h"
52 #include "pathnames.h"
53
54 /* GNU gettext added a hack to add some context to messages. If a message is
55 * used in multiple locations, it needs some amount of context to make the
56 * translation clear to translators. GNU gettext, rather than modifying the
57 * message format, concatenates the context, \004 and the message id.
58 */
59 #define MSGCTXT_ID_SEPARATOR '\004'
60
61 static const char *pgettext_impl(const char *, const char *, const char *,
62 const char *, unsigned long int, int);
63 static char *concatenate_ctxt_id(const char *, const char *);
64 static const char *lookup_category(int);
65 static const char *split_locale(const char *);
66 static const char *lookup_mofile(char *, size_t, const char *, const char *,
67 const char *, const char *,
68 struct domainbinding *);
69 static uint32_t flip(uint32_t, uint32_t);
70 static int validate(void *, struct mohandle *);
71 static int mapit(const char *, struct domainbinding *);
72 static int unmapit(struct domainbinding *);
73 static const char *lookup_hash(const char *, struct domainbinding *, size_t *);
74 static const char *lookup_bsearch(const char *, struct domainbinding *,
75 size_t *);
76 static const char *lookup(const char *, struct domainbinding *, size_t *);
77 static const char *get_lang_env(const char *);
78
79 /*
80 * shortcut functions. the main implementation resides in dcngettext().
81 */
82 char *
83 gettext(const char *msgid)
84 {
85
86 return dcngettext(NULL, msgid, NULL, 1UL, LC_MESSAGES);
87 }
88
89 char *
90 dgettext(const char *domainname, const char *msgid)
91 {
92
93 return dcngettext(domainname, msgid, NULL, 1UL, LC_MESSAGES);
94 }
95
96 char *
97 dcgettext(const char *domainname, const char *msgid, int category)
98 {
99
100 return dcngettext(domainname, msgid, NULL, 1UL, category);
101 }
102
103 char *
104 ngettext(const char *msgid1, const char *msgid2, unsigned long int n)
105 {
106
107 return dcngettext(NULL, msgid1, msgid2, n, LC_MESSAGES);
108 }
109
110 char *
111 dngettext(const char *domainname, const char *msgid1, const char *msgid2,
112 unsigned long int n)
113 {
114
115 return dcngettext(domainname, msgid1, msgid2, n, LC_MESSAGES);
116 }
117
118 const char *
119 pgettext(const char *msgctxt, const char *msgid)
120 {
121
122 return pgettext_impl(NULL, msgctxt, msgid, NULL, 1UL, LC_MESSAGES);
123 }
124
125 const char *
126 dpgettext(const char *domainname, const char *msgctxt, const char *msgid)
127 {
128
129 return pgettext_impl(domainname, msgctxt, msgid, NULL, 1UL, LC_MESSAGES);
130 }
131
132 const char *
133 dcpgettext(const char *domainname, const char *msgctxt, const char *msgid,
134 int category)
135 {
136
137 return pgettext_impl(domainname, msgctxt, msgid, NULL, 1UL, category);
138 }
139
140 const char *
141 npgettext(const char *msgctxt, const char *msgid1, const char *msgid2,
142 unsigned long int n)
143 {
144
145 return pgettext_impl(NULL, msgctxt, msgid1, msgid2, n, LC_MESSAGES);
146 }
147
148 const char *
149 dnpgettext(const char *domainname, const char *msgctxt, const char *msgid1,
150 const char *msgid2, unsigned long int n)
151 {
152
153 return pgettext_impl(domainname, msgctxt, msgid1, msgid2, n, LC_MESSAGES);
154 }
155
156 const char *
157 dcnpgettext(const char *domainname, const char *msgctxt, const char *msgid1,
158 const char *msgid2, unsigned long int n, int category)
159 {
160
161 return pgettext_impl(domainname, msgctxt, msgid1, msgid2, n, category);
162 }
163
164 static const char *
165 pgettext_impl(const char *domainname, const char *msgctxt, const char *msgid1,
166 const char *msgid2, unsigned long int n, int category)
167 {
168 char *msgctxt_id;
169 char *translation;
170 char *p;
171
172 if ((msgctxt_id = concatenate_ctxt_id(msgctxt, msgid1)) == NULL)
173 return msgid1;
174
175 translation = dcngettext(domainname, msgctxt_id,
176 msgid2, n, category);
177 free(msgctxt_id);
178
179 p = strchr(translation, '\004');
180 if (p)
181 return p + 1;
182 return translation;
183 }
184
185 /*
186 * dcngettext() -
187 * lookup internationalized message on database locale/category/domainname
188 * (like ja_JP.eucJP/LC_MESSAGES/domainname).
189 * if n equals to 1, internationalized message will be looked up for msgid1.
190 * otherwise, message will be looked up for msgid2.
191 * if the lookup fails, the function will return msgid1 or msgid2 as is.
192 *
193 * Even though the return type is "char *", caller should not rewrite the
194 * region pointed to by the return value (should be "const char *", but can't
195 * change it for compatibility with other implementations).
196 *
197 * by default (if domainname == NULL), domainname is taken from the value set
198 * by textdomain(). usually name of the application (like "ls") is used as
199 * domainname. category is usually LC_MESSAGES.
200 *
201 * the code reads in *.mo files generated by GNU gettext. *.mo is a host-
202 * endian encoded file. both endians are supported here, as the files are in
203 * /usr/share/locale! (or we should move those files into /usr/libdata)
204 */
205
206 static char *
207 concatenate_ctxt_id(const char *msgctxt, const char *msgid)
208 {
209 char *ret;
210
211 if (asprintf(&ret, "%s%c%s", msgctxt, MSGCTXT_ID_SEPARATOR, msgid) == -1)
212 return NULL;
213
214 return ret;
215 }
216
217 static const char *
218 lookup_category(int category)
219 {
220
221 switch (category) {
222 case LC_COLLATE: return "LC_COLLATE";
223 case LC_CTYPE: return "LC_CTYPE";
224 case LC_MONETARY: return "LC_MONETARY";
225 case LC_NUMERIC: return "LC_NUMERIC";
226 case LC_TIME: return "LC_TIME";
227 case LC_MESSAGES: return "LC_MESSAGES";
228 }
229 return NULL;
230 }
231
232 #define MAXBUFLEN 1024
233 /*
234 * XPG syntax: language[_territory[.codeset]][@modifier]
235 * XXX boundary check on "result" is lacking
236 */
237 static const char *
238 split_locale(const char *lname)
239 {
240 char buf[MAXBUFLEN], tmp[2 * MAXBUFLEN];
241 char *l, *t, *c, *m;
242 static char result[4 * MAXBUFLEN];
243
244 memset(result, 0, sizeof(result));
245
246 if (strlen(lname) + 1 > sizeof(buf)) {
247 fail:
248 return lname;
249 }
250
251 strlcpy(buf, lname, sizeof(buf));
252 m = strrchr(buf, '@');
253 if (m)
254 *m++ = '\0';
255 c = strrchr(buf, '.');
256 if (c)
257 *c++ = '\0';
258 t = strrchr(buf, '_');
259 if (t)
260 *t++ = '\0';
261 l = buf;
262 if (strlen(l) == 0)
263 goto fail;
264 if (c && !t)
265 goto fail;
266
267 if (m) {
268 if (t) {
269 if (c) {
270 snprintf(tmp, sizeof(tmp), "%s_%s.%s@%s",
271 l, t, c, m);
272 strlcat(result, tmp, sizeof(result));
273 strlcat(result, ":", sizeof(result));
274 }
275 snprintf(tmp, sizeof(tmp), "%s_%s@%s", l, t, m);
276 strlcat(result, tmp, sizeof(result));
277 strlcat(result, ":", sizeof(result));
278 }
279 snprintf(tmp, sizeof(tmp), "%s@%s", l, m);
280 strlcat(result, tmp, sizeof(result));
281 strlcat(result, ":", sizeof(result));
282 }
283 if (t) {
284 if (c) {
285 snprintf(tmp, sizeof(tmp), "%s_%s.%s", l, t, c);
286 strlcat(result, tmp, sizeof(result));
287 strlcat(result, ":", sizeof(result));
288 }
289 snprintf(tmp, sizeof(tmp), "%s_%s", l, t);
290 strlcat(result, tmp, sizeof(result));
291 strlcat(result, ":", sizeof(result));
292 }
293 strlcat(result, l, sizeof(result));
294
295 return result;
296 }
297
298 static const char *
299 lookup_mofile(char *buf, size_t len, const char *dir, const char *lpath,
300 const char *category, const char *domainname,
301 struct domainbinding *db)
302 {
303 struct stat st;
304 char *p, *q;
305 char lpath_tmp[BUFSIZ];
306
307 /*
308 * LANGUAGE is a colon separated list of locale names.
309 */
310
311 strlcpy(lpath_tmp, lpath, sizeof(lpath_tmp));
312 q = lpath_tmp;
313 /* CONSTCOND */
314 while (1) {
315 p = strsep(&q, ":");
316 if (!p)
317 break;
318 if (!*p)
319 continue;
320
321 /* don't mess with default locales */
322 if (strcmp(p, "C") == 0 || strcmp(p, "POSIX") == 0)
323 return NULL;
324
325 /* validate pathname */
326 if (strchr(p, '/') || strchr(category, '/'))
327 continue;
328 #if 1 /*?*/
329 if (strchr(domainname, '/'))
330 continue;
331 #endif
332
333 int rv = snprintf(buf, len, "%s/%s/%s/%s.mo", dir, p,
334 category, domainname);
335 if (rv > (int)len)
336 return NULL;
337 if (stat(buf, &st) < 0)
338 continue;
339 if ((st.st_mode & S_IFMT) != S_IFREG)
340 continue;
341
342 if (mapit(buf, db) == 0)
343 return buf;
344 }
345
346 return NULL;
347 }
348
349 static uint32_t
350 flip(uint32_t v, uint32_t magic)
351 {
352
353 if (magic == MO_MAGIC)
354 return v;
355 else if (magic == MO_MAGIC_SWAPPED) {
356 v = ((v >> 24) & 0xff) | ((v >> 8) & 0xff00) |
357 ((v << 8) & 0xff0000) | ((v << 24) & 0xff000000);
358 return v;
359 } else {
360 abort();
361 /*NOTREACHED*/
362 }
363 }
364
365 static int
366 validate(void *arg, struct mohandle *mohandle)
367 {
368 char *p;
369
370 p = (char *)arg;
371 if (p < (char *)mohandle->addr ||
372 p > (char *)mohandle->addr + mohandle->len)
373 return 0;
374 else
375 return 1;
376 }
377
378 /*
379 * calculate the step value if the hash value is conflicted.
380 */
381 static __inline uint32_t
382 calc_collision_step(uint32_t hashval, uint32_t hashsize)
383 {
384 _DIAGASSERT(hashsize>2);
385 return (hashval % (hashsize - 2)) + 1;
386 }
387
388 /*
389 * calculate the next index while conflicting.
390 */
391 static __inline uint32_t
392 calc_next_index(uint32_t curidx, uint32_t hashsize, uint32_t step)
393 {
394 return curidx+step - (curidx >= hashsize-step ? hashsize : 0);
395 }
396
397 static int
398 get_sysdep_string_table(struct mosysdepstr_h **table_h, uint32_t *ofstable,
399 uint32_t nstrings, uint32_t magic, char *base)
400 {
401 unsigned int i;
402 int j, count;
403 size_t l;
404 struct mosysdepstr *table;
405
406 for (i=0; i<nstrings; i++) {
407 /* get mosysdepstr record */
408 /* LINTED: ignore the alignment problem. */
409 table = (struct mosysdepstr *)(base + flip(ofstable[i], magic));
410 /* count number of segments */
411 count = 0;
412 while (flip(table->segs[count++].ref, magic) != MO_LASTSEG)
413 ;
414 /* get table */
415 l = sizeof(struct mosysdepstr_h) +
416 sizeof(struct mosysdepsegentry_h) * (count-1);
417 table_h[i] = (struct mosysdepstr_h *)malloc(l);
418 if (!table_h[i])
419 return -1;
420 memset(table_h[i], 0, l);
421 table_h[i]->off = (const char *)(base + flip(table->off, magic));
422 for (j=0; j<count; j++) {
423 table_h[i]->segs[j].len =
424 flip(table->segs[j].len, magic);
425 table_h[i]->segs[j].ref =
426 flip(table->segs[j].ref, magic);
427 }
428 /* LINTED: ignore the alignment problem. */
429 table = (struct mosysdepstr *)&table->segs[count];
430 }
431 return 0;
432 }
433
434 static int
435 expand_sysdep(struct mohandle *mohandle, struct mosysdepstr_h *str)
436 {
437 int i;
438 const char *src;
439 char *dst;
440
441 /* check whether already expanded */
442 if (str->expanded)
443 return 0;
444
445 /* calc total length */
446 str->expanded_len = 1;
447 for (i=0; /*CONSTCOND*/1; i++) {
448 str->expanded_len += str->segs[i].len;
449 if (str->segs[i].ref == MO_LASTSEG)
450 break;
451 str->expanded_len +=
452 mohandle->mo.mo_sysdep_segs[str->segs[i].ref].len;
453 }
454 /* expand */
455 str->expanded = malloc(str->expanded_len);
456 if (!str->expanded)
457 return -1;
458 src = str->off;
459 dst = str->expanded;
460 for (i=0; /*CONSTCOND*/1; i++) {
461 memcpy(dst, src, str->segs[i].len);
462 src += str->segs[i].len;
463 dst += str->segs[i].len;
464 if (str->segs[i].ref == MO_LASTSEG)
465 break;
466 memcpy(dst, mohandle->mo.mo_sysdep_segs[str->segs[i].ref].str,
467 mohandle->mo.mo_sysdep_segs[str->segs[i].ref].len);
468 dst += mohandle->mo.mo_sysdep_segs[str->segs[i].ref].len;
469 }
470 *dst = '\0';
471
472 return 0;
473 }
474
475 static void
476 insert_to_hash(uint32_t *htable, uint32_t hsize, const char *str, uint32_t ref)
477 {
478 uint32_t hashval, idx, step;
479
480 hashval = __intl_string_hash(str);
481 step = calc_collision_step(hashval, hsize);
482 idx = hashval % hsize;
483
484 while (htable[idx])
485 idx = calc_next_index(idx, hsize, step);
486
487 htable[idx] = ref;
488 }
489
490 static int
491 setup_sysdep_stuffs(struct mo *mo, struct mohandle *mohandle, char *base)
492 {
493 uint32_t magic;
494 struct moentry *stable;
495 size_t l;
496 unsigned int i;
497 char *v;
498 uint32_t *ofstable;
499
500 magic = mo->mo_magic;
501
502 mohandle->mo.mo_sysdep_nsegs = flip(mo->mo_sysdep_nsegs, magic);
503 mohandle->mo.mo_sysdep_nstring = flip(mo->mo_sysdep_nstring, magic);
504
505 if (mohandle->mo.mo_sysdep_nstring == 0)
506 return 0;
507
508 /* check hash size */
509 if (mohandle->mo.mo_hsize <= 2 ||
510 mohandle->mo.mo_hsize <
511 (mohandle->mo.mo_nstring + mohandle->mo.mo_sysdep_nstring))
512 return -1;
513
514 /* get sysdep segments */
515 l = sizeof(struct mosysdepsegs_h) * mohandle->mo.mo_sysdep_nsegs;
516 mohandle->mo.mo_sysdep_segs = (struct mosysdepsegs_h *)malloc(l);
517 if (!mohandle->mo.mo_sysdep_segs)
518 return -1;
519 /* LINTED: ignore the alignment problem. */
520 stable = (struct moentry *)(base + flip(mo->mo_sysdep_segoff, magic));
521 for (i=0; i<mohandle->mo.mo_sysdep_nsegs; i++) {
522 v = base + flip(stable[i].off, magic);
523 mohandle->mo.mo_sysdep_segs[i].str =
524 __intl_sysdep_get_string_by_tag(
525 v,
526 &mohandle->mo.mo_sysdep_segs[i].len);
527 }
528
529 /* get sysdep string table */
530 mohandle->mo.mo_sysdep_otable =
531 (struct mosysdepstr_h **)calloc(mohandle->mo.mo_sysdep_nstring,
532 sizeof(struct mosysdepstr_h *));
533 if (!mohandle->mo.mo_sysdep_otable)
534 return -1;
535 /* LINTED: ignore the alignment problem. */
536 ofstable = (uint32_t *)(base + flip(mo->mo_sysdep_otable, magic));
537 if (get_sysdep_string_table(mohandle->mo.mo_sysdep_otable, ofstable,
538 mohandle->mo.mo_sysdep_nstring, magic,
539 base))
540 return -1;
541 mohandle->mo.mo_sysdep_ttable =
542 (struct mosysdepstr_h **)calloc(mohandle->mo.mo_sysdep_nstring,
543 sizeof(struct mosysdepstr_h *));
544 if (!mohandle->mo.mo_sysdep_ttable)
545 return -1;
546 /* LINTED: ignore the alignment problem. */
547 ofstable = (uint32_t *)(base + flip(mo->mo_sysdep_ttable, magic));
548 if (get_sysdep_string_table(mohandle->mo.mo_sysdep_ttable, ofstable,
549 mohandle->mo.mo_sysdep_nstring, magic,
550 base))
551 return -1;
552
553 /* update hash */
554 for (i=0; i<mohandle->mo.mo_sysdep_nstring; i++) {
555 if (expand_sysdep(mohandle, mohandle->mo.mo_sysdep_otable[i]))
556 return -1;
557 insert_to_hash(mohandle->mo.mo_htable,
558 mohandle->mo.mo_hsize,
559 mohandle->mo.mo_sysdep_otable[i]->expanded,
560 (i+1) | MO_HASH_SYSDEP_MASK);
561 }
562
563 return 0;
564 }
565
566 int
567 mapit(const char *path, struct domainbinding *db)
568 {
569 int fd;
570 struct stat st;
571 char *base;
572 uint32_t magic, revision, flags = 0;
573 struct moentry *otable, *ttable;
574 const uint32_t *htable;
575 struct moentry_h *p;
576 struct mo *mo;
577 size_t l, headerlen;
578 unsigned int i;
579 char *v;
580 struct mohandle *mohandle = &db->mohandle;
581
582 if (mohandle->addr && mohandle->addr != MAP_FAILED &&
583 mohandle->mo.mo_magic)
584 return 0; /*already opened*/
585
586 unmapit(db);
587
588 #if 0
589 if (secure_path(path) != 0)
590 goto fail;
591 #endif
592 if (stat(path, &st) < 0)
593 goto fail;
594 if ((st.st_mode & S_IFMT) != S_IFREG || st.st_size > GETTEXT_MMAP_MAX)
595 goto fail;
596 fd = open(path, O_RDONLY);
597 if (fd < 0)
598 goto fail;
599 if (read(fd, &magic, sizeof(magic)) != sizeof(magic) ||
600 (magic != MO_MAGIC && magic != MO_MAGIC_SWAPPED)) {
601 close(fd);
602 goto fail;
603 }
604 if (read(fd, &revision, sizeof(revision)) != sizeof(revision)) {
605 close(fd);
606 goto fail;
607 }
608 switch (flip(revision, magic)) {
609 case MO_MAKE_REV(0, 0):
610 break;
611 case MO_MAKE_REV(0, 1):
612 case MO_MAKE_REV(1, 1):
613 flags |= MO_F_SYSDEP;
614 break;
615 default:
616 close(fd);
617 goto fail;
618 }
619 mohandle->addr = mmap(NULL, (size_t)st.st_size, PROT_READ,
620 MAP_FILE | MAP_SHARED, fd, (off_t)0);
621 if (!mohandle->addr || mohandle->addr == MAP_FAILED) {
622 close(fd);
623 goto fail;
624 }
625 close(fd);
626 mohandle->len = (size_t)st.st_size;
627
628 base = mohandle->addr;
629 mo = (struct mo *)mohandle->addr;
630
631 /* flip endian. do not flip magic number! */
632 mohandle->mo.mo_magic = mo->mo_magic;
633 mohandle->mo.mo_revision = flip(mo->mo_revision, magic);
634 mohandle->mo.mo_nstring = flip(mo->mo_nstring, magic);
635 mohandle->mo.mo_hsize = flip(mo->mo_hsize, magic);
636 mohandle->mo.mo_flags = flags;
637
638 /* validate otable/ttable */
639 /* LINTED: ignore the alignment problem. */
640 otable = (struct moentry *)(base + flip(mo->mo_otable, magic));
641 /* LINTED: ignore the alignment problem. */
642 ttable = (struct moentry *)(base + flip(mo->mo_ttable, magic));
643 if (!validate(otable, mohandle) ||
644 !validate(&otable[mohandle->mo.mo_nstring], mohandle)) {
645 unmapit(db);
646 goto fail;
647 }
648 if (!validate(ttable, mohandle) ||
649 !validate(&ttable[mohandle->mo.mo_nstring], mohandle)) {
650 unmapit(db);
651 goto fail;
652 }
653
654 /* allocate [ot]table, and convert to normal pointer representation. */
655 l = sizeof(struct moentry_h) * mohandle->mo.mo_nstring;
656 mohandle->mo.mo_otable = (struct moentry_h *)malloc(l);
657 if (!mohandle->mo.mo_otable) {
658 unmapit(db);
659 goto fail;
660 }
661 mohandle->mo.mo_ttable = (struct moentry_h *)malloc(l);
662 if (!mohandle->mo.mo_ttable) {
663 unmapit(db);
664 goto fail;
665 }
666 p = mohandle->mo.mo_otable;
667 for (i = 0; i < mohandle->mo.mo_nstring; i++) {
668 p[i].len = flip(otable[i].len, magic);
669 p[i].off = base + flip(otable[i].off, magic);
670
671 if (!validate(p[i].off, mohandle) ||
672 !validate(p[i].off + p[i].len + 1, mohandle)) {
673 unmapit(db);
674 goto fail;
675 }
676 }
677 p = mohandle->mo.mo_ttable;
678 for (i = 0; i < mohandle->mo.mo_nstring; i++) {
679 p[i].len = flip(ttable[i].len, magic);
680 p[i].off = base + flip(ttable[i].off, magic);
681
682 if (!validate(p[i].off, mohandle) ||
683 !validate(p[i].off + p[i].len + 1, mohandle)) {
684 unmapit(db);
685 goto fail;
686 }
687 }
688 /* allocate htable, and convert it to the host order. */
689 if (mohandle->mo.mo_hsize > 2) {
690 l = sizeof(uint32_t) * mohandle->mo.mo_hsize;
691 mohandle->mo.mo_htable = (uint32_t *)malloc(l);
692 if (!mohandle->mo.mo_htable) {
693 unmapit(db);
694 goto fail;
695 }
696 /* LINTED: ignore the alignment problem. */
697 htable = (const uint32_t *)(base+flip(mo->mo_hoffset, magic));
698 for (i=0; i < mohandle->mo.mo_hsize; i++) {
699 mohandle->mo.mo_htable[i] = flip(htable[i], magic);
700 if (mohandle->mo.mo_htable[i] >=
701 mohandle->mo.mo_nstring+1) {
702 /* illegal string number. */
703 unmapit(db);
704 goto fail;
705 }
706 }
707 }
708 /* grab MIME-header and charset field */
709 mohandle->mo.mo_header = lookup("", db, &headerlen);
710 if (mohandle->mo.mo_header)
711 v = strstr(mohandle->mo.mo_header, "charset=");
712 else
713 v = NULL;
714 if (v) {
715 mohandle->mo.mo_charset = strdup(v + 8);
716 if (!mohandle->mo.mo_charset)
717 goto fail;
718 v = strchr(mohandle->mo.mo_charset, '\n');
719 if (v)
720 *v = '\0';
721 }
722 if (!mohandle->mo.mo_header ||
723 _gettext_parse_plural(&mohandle->mo.mo_plural,
724 &mohandle->mo.mo_nplurals,
725 mohandle->mo.mo_header, headerlen))
726 mohandle->mo.mo_plural = NULL;
727
728 /*
729 * XXX check charset, reject it if we are unable to support the charset
730 * with the current locale.
731 * for example, if we are using euc-jp locale and we are looking at
732 * *.mo file encoded by euc-kr (charset=euc-kr), we should reject
733 * the *.mo file as we cannot support it.
734 */
735
736 /* system dependent string support */
737 if ((mohandle->mo.mo_flags & MO_F_SYSDEP) != 0) {
738 if (setup_sysdep_stuffs(mo, mohandle, base)) {
739 unmapit(db);
740 goto fail;
741 }
742 }
743
744 return 0;
745
746 fail:
747 return -1;
748 }
749
750 static void
751 free_sysdep_table(struct mosysdepstr_h **table, uint32_t nstring)
752 {
753
754 if (! table)
755 return;
756
757 for (uint32_t i = 0; i < nstring; i++) {
758 if (table[i]) {
759 free(table[i]->expanded);
760 free(table[i]);
761 }
762 }
763 free(table);
764 }
765
766 static int
767 unmapit(struct domainbinding *db)
768 {
769 struct mohandle *mohandle = &db->mohandle;
770
771 /* unmap if there's already mapped region */
772 if (mohandle->addr && mohandle->addr != MAP_FAILED)
773 munmap(mohandle->addr, mohandle->len);
774 mohandle->addr = NULL;
775 free(mohandle->mo.mo_otable);
776 free(mohandle->mo.mo_ttable);
777 free(mohandle->mo.mo_charset);
778 free(mohandle->mo.mo_htable);
779 free(mohandle->mo.mo_sysdep_segs);
780 free_sysdep_table(mohandle->mo.mo_sysdep_otable,
781 mohandle->mo.mo_sysdep_nstring);
782 free_sysdep_table(mohandle->mo.mo_sysdep_ttable,
783 mohandle->mo.mo_sysdep_nstring);
784 _gettext_free_plural(mohandle->mo.mo_plural);
785 memset(&mohandle->mo, 0, sizeof(mohandle->mo));
786 return 0;
787 }
788
789 /* ARGSUSED */
790 static const char *
791 lookup_hash(const char *msgid, struct domainbinding *db, size_t *rlen)
792 {
793 struct mohandle *mohandle = &db->mohandle;
794 uint32_t idx, hashval, step, strno;
795 size_t len;
796 struct mosysdepstr_h *sysdep_otable, *sysdep_ttable;
797
798 if (mohandle->mo.mo_hsize <= 2 || mohandle->mo.mo_htable == NULL)
799 return NULL;
800
801 hashval = __intl_string_hash(msgid);
802 step = calc_collision_step(hashval, mohandle->mo.mo_hsize);
803 idx = hashval % mohandle->mo.mo_hsize;
804 len = strlen(msgid);
805 while (/*CONSTCOND*/1) {
806 strno = mohandle->mo.mo_htable[idx];
807 if (strno == 0) {
808 /* unexpected miss */
809 return NULL;
810 }
811 strno--;
812 if ((strno & MO_HASH_SYSDEP_MASK) == 0) {
813 /* system independent strings */
814 if (len <= mohandle->mo.mo_otable[strno].len &&
815 !strcmp(msgid, mohandle->mo.mo_otable[strno].off)) {
816 /* hit */
817 if (rlen)
818 *rlen =
819 mohandle->mo.mo_ttable[strno].len;
820 return mohandle->mo.mo_ttable[strno].off;
821 }
822 } else {
823 /* system dependent strings */
824 strno &= ~MO_HASH_SYSDEP_MASK;
825 sysdep_otable = mohandle->mo.mo_sysdep_otable[strno];
826 sysdep_ttable = mohandle->mo.mo_sysdep_ttable[strno];
827 if (len <= sysdep_otable->expanded_len &&
828 !strcmp(msgid, sysdep_otable->expanded)) {
829 /* hit */
830 if (expand_sysdep(mohandle, sysdep_ttable))
831 /* memory exhausted */
832 return NULL;
833 if (rlen)
834 *rlen = sysdep_ttable->expanded_len;
835 return sysdep_ttable->expanded;
836 }
837 }
838 idx = calc_next_index(idx, mohandle->mo.mo_hsize, step);
839 }
840 /*NOTREACHED*/
841 }
842
843 static const char *
844 lookup_bsearch(const char *msgid, struct domainbinding *db, size_t *rlen)
845 {
846 int top, bottom, middle, omiddle;
847 int n;
848 struct mohandle *mohandle = &db->mohandle;
849
850 top = 0;
851 bottom = mohandle->mo.mo_nstring;
852 omiddle = -1;
853 /* CONSTCOND */
854 while (1) {
855 if (top > bottom)
856 break;
857 middle = (top + bottom) / 2;
858 /* avoid possible infinite loop, when the data is not sorted */
859 if (omiddle == middle)
860 break;
861 if ((size_t)middle >= mohandle->mo.mo_nstring)
862 break;
863
864 n = strcmp(msgid, mohandle->mo.mo_otable[middle].off);
865 if (n == 0) {
866 if (rlen)
867 *rlen = mohandle->mo.mo_ttable[middle].len;
868 return (const char *)mohandle->mo.mo_ttable[middle].off;
869 }
870 else if (n < 0)
871 bottom = middle;
872 else
873 top = middle;
874 omiddle = middle;
875 }
876
877 return NULL;
878 }
879
880 static const char *
881 lookup(const char *msgid, struct domainbinding *db, size_t *rlen)
882 {
883 const char *v;
884
885 v = lookup_hash(msgid, db, rlen);
886 if (v)
887 return v;
888
889 return lookup_bsearch(msgid, db, rlen);
890 }
891
892 static const char *
893 get_lang_env(const char *category_name)
894 {
895 const char *lang;
896
897 /*
898 * 1. see LANGUAGE variable first.
899 *
900 * LANGUAGE is a GNU extension.
901 * It's a colon separated list of locale names.
902 */
903 lang = getenv("LANGUAGE");
904 if (lang)
905 return lang;
906
907 /*
908 * 2. if LANGUAGE isn't set, see LC_ALL, LC_xxx, LANG.
909 *
910 * It's essentially setlocale(LC_xxx, NULL).
911 */
912 lang = getenv("LC_ALL");
913 if (!lang)
914 lang = getenv(category_name);
915 if (!lang)
916 lang = getenv("LANG");
917
918 if (!lang)
919 return 0; /* error */
920
921 return split_locale(lang);
922 }
923
924 static const char *
925 get_indexed_string(const char *str, size_t len, unsigned long idx)
926 {
927 while (idx > 0) {
928 if (len <= 1)
929 return str;
930 if (*str == '\0')
931 idx--;
932 if (len > 0) {
933 str++;
934 len--;
935 }
936 }
937 return str;
938 }
939
940 #define _NGETTEXT_DEFAULT(msgid1, msgid2, n) \
941 ((char *)__UNCONST((n) == 1 ? (msgid1) : (msgid2)))
942
943 char *
944 dcngettext(const char *domainname, const char *msgid1, const char *msgid2,
945 unsigned long int n, int category)
946 {
947 const char *msgid;
948 char path[PATH_MAX+1];
949 const char *lpath;
950 static char olpath[PATH_MAX];
951 const char *cname = NULL;
952 const char *v;
953 static char *ocname = NULL;
954 static char *odomainname = NULL;
955 struct domainbinding *db;
956 unsigned long plural_index = 0;
957 size_t len;
958
959 if (!domainname)
960 domainname = __current_domainname;
961 cname = lookup_category(category);
962 if (!domainname || !cname)
963 goto fail;
964
965 lpath = get_lang_env(cname);
966 if (!lpath)
967 goto fail;
968
969 for (db = __bindings; db; db = db->next)
970 if (strcmp(db->domainname, domainname) == 0)
971 break;
972 if (!db) {
973 if (!bindtextdomain(domainname, _PATH_TEXTDOMAIN))
974 goto fail;
975 db = __bindings;
976 }
977
978 /* resolve relative path */
979 /* XXX not necessary? */
980 if (db->path[0] != '/') {
981 char buf[PATH_MAX];
982
983 if (getcwd(buf, sizeof(buf)) == 0)
984 goto fail;
985 if (strlcat(buf, "/", sizeof(buf)) >= sizeof(buf))
986 goto fail;
987 if (strlcat(buf, db->path, sizeof(buf)) >= sizeof(buf))
988 goto fail;
989 strlcpy(db->path, buf, sizeof(db->path));
990 }
991
992 /* don't bother looking it up if the values are the same */
993 if (odomainname && strcmp(domainname, odomainname) == 0 &&
994 ocname && strcmp(cname, ocname) == 0 && strcmp(lpath, olpath) == 0 &&
995 db->mohandle.mo.mo_magic)
996 goto found;
997
998 /* try to find appropriate file, from $LANGUAGE */
999 if (lookup_mofile(path, sizeof(path), db->path, lpath, cname,
1000 domainname, db) == NULL)
1001 goto fail;
1002
1003 free(odomainname);
1004 free(ocname);
1005
1006 odomainname = strdup(domainname);
1007 ocname = strdup(cname);
1008 if (!odomainname || !ocname) {
1009 free(odomainname);
1010 free(ocname);
1011
1012 odomainname = ocname = NULL;
1013 }
1014 else
1015 strlcpy(olpath, lpath, sizeof(olpath));
1016
1017 found:
1018 if (db->mohandle.mo.mo_plural) {
1019 plural_index =
1020 _gettext_calculate_plural(db->mohandle.mo.mo_plural, n);
1021 if (plural_index >= db->mohandle.mo.mo_nplurals)
1022 plural_index = 0;
1023 msgid = msgid1;
1024 } else
1025 msgid = _NGETTEXT_DEFAULT(msgid1, msgid2, n);
1026
1027 if (msgid == NULL)
1028 return NULL;
1029
1030 v = lookup(msgid, db, &len);
1031 if (v) {
1032 if (db->mohandle.mo.mo_plural)
1033 v = get_indexed_string(v, len, plural_index);
1034 /*
1035 * convert the translated message's encoding.
1036 *
1037 * special case:
1038 * a result of gettext("") shouldn't need any conversion.
1039 */
1040 if (msgid[0])
1041 v = __gettext_iconv(v, db);
1042
1043 /*
1044 * Given the amount of printf-format security issues, it may
1045 * be a good idea to validate if the original msgid and the
1046 * translated message format string carry the same printf-like
1047 * format identifiers.
1048 */
1049
1050 msgid = v;
1051 }
1052
1053 return (char *)__UNCONST(msgid);
1054
1055 fail:
1056 return _NGETTEXT_DEFAULT(msgid1, msgid2, n);
1057 }
1058