1 | /* Implementation of the internal dcigettext function. |
2 | Copyright (C) 1995-2018 Free Software Foundation, Inc. |
3 | |
4 | This program is free software: you can redistribute it and/or modify |
5 | it under the terms of the GNU Lesser General Public License as published by |
6 | the Free Software Foundation; either version 2.1 of the License, or |
7 | (at your option) any later version. |
8 | |
9 | This program is distributed in the hope that it will be useful, |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
12 | GNU Lesser General Public License for more details. |
13 | |
14 | You should have received a copy of the GNU Lesser General Public License |
15 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ |
16 | |
17 | /* Tell glibc's <string.h> to provide a prototype for mempcpy(). |
18 | This must come before <config.h> because <config.h> may include |
19 | <features.h>, and once <features.h> has been included, it's too late. */ |
20 | #ifndef _GNU_SOURCE |
21 | # define _GNU_SOURCE 1 |
22 | #endif |
23 | |
24 | #ifdef HAVE_CONFIG_H |
25 | # include <config.h> |
26 | #endif |
27 | |
28 | #include <sys/types.h> |
29 | |
30 | #ifdef __GNUC__ |
31 | # define alloca __builtin_alloca |
32 | # define HAVE_ALLOCA 1 |
33 | #else |
34 | # ifdef _MSC_VER |
35 | # include <malloc.h> |
36 | # define alloca _alloca |
37 | # else |
38 | # if defined HAVE_ALLOCA_H || defined _LIBC |
39 | # include <alloca.h> |
40 | # else |
41 | # ifdef _AIX |
42 | #pragma alloca |
43 | # else |
44 | # ifndef alloca |
45 | char *alloca (); |
46 | # endif |
47 | # endif |
48 | # endif |
49 | # endif |
50 | #endif |
51 | |
52 | #include <errno.h> |
53 | #ifndef errno |
54 | extern int errno; |
55 | #endif |
56 | #ifndef __set_errno |
57 | # define __set_errno(val) errno = (val) |
58 | #endif |
59 | |
60 | #include <stddef.h> |
61 | #include <stdlib.h> |
62 | #include <stdio.h> |
63 | #include <string.h> |
64 | |
65 | #if defined HAVE_UNISTD_H || defined _LIBC |
66 | # include <unistd.h> |
67 | #endif |
68 | |
69 | #include <locale.h> |
70 | |
71 | #ifdef _LIBC |
72 | /* Guess whether integer division by zero raises signal SIGFPE. |
73 | Set to 1 only if you know for sure. In case of doubt, set to 0. */ |
74 | # if defined __alpha__ || defined __arm__ || defined __i386__ \ |
75 | || defined __m68k__ || defined __s390__ |
76 | # define INTDIV0_RAISES_SIGFPE 1 |
77 | # else |
78 | # define INTDIV0_RAISES_SIGFPE 0 |
79 | # endif |
80 | #endif |
81 | #if !INTDIV0_RAISES_SIGFPE |
82 | # include <signal.h> |
83 | #endif |
84 | |
85 | #if defined HAVE_SYS_PARAM_H || defined _LIBC |
86 | # include <sys/param.h> |
87 | #endif |
88 | |
89 | #if !defined _LIBC |
90 | # include "localcharset.h" |
91 | #endif |
92 | |
93 | #include "gettextP.h" |
94 | #include "plural-exp.h" |
95 | #ifdef _LIBC |
96 | # include <libintl.h> |
97 | #else |
98 | # ifdef IN_LIBGLOCALE |
99 | # include <libintl.h> |
100 | # endif |
101 | # include "libgnuintl.h" |
102 | #endif |
103 | #include "hash-string.h" |
104 | |
105 | /* Handle multi-threaded applications. */ |
106 | #ifdef _LIBC |
107 | # include <libc-lock.h> |
108 | # define gl_rwlock_define_initialized __libc_rwlock_define_initialized |
109 | # define gl_rwlock_rdlock __libc_rwlock_rdlock |
110 | # define gl_rwlock_wrlock __libc_rwlock_wrlock |
111 | # define gl_rwlock_unlock __libc_rwlock_unlock |
112 | #else |
113 | # include "lock.h" |
114 | #endif |
115 | |
116 | /* Alignment of types. */ |
117 | #if defined __GNUC__ && __GNUC__ >= 2 |
118 | # define alignof(TYPE) __alignof__ (TYPE) |
119 | #else |
120 | # define alignof(TYPE) \ |
121 | ((int) &((struct { char dummy1; TYPE dummy2; } *) 0)->dummy2) |
122 | #endif |
123 | |
124 | /* Some compilers, like SunOS4 cc, don't have offsetof in <stddef.h>. */ |
125 | #ifndef offsetof |
126 | # define offsetof(type,ident) ((size_t)&(((type*)0)->ident)) |
127 | #endif |
128 | |
129 | /* @@ end of prolog @@ */ |
130 | |
131 | #ifdef _LIBC |
132 | /* Rename the non ANSI C functions. This is required by the standard |
133 | because some ANSI C functions will require linking with this object |
134 | file and the name space must not be polluted. */ |
135 | # define strdup __strdup |
136 | # define getcwd __getcwd |
137 | # ifndef stpcpy |
138 | # define stpcpy __stpcpy |
139 | # endif |
140 | # define tfind __tfind |
141 | #else |
142 | # if !defined HAVE_GETCWD |
143 | char *getwd (); |
144 | # define getcwd(buf, max) getwd (buf) |
145 | # else |
146 | # if VMS |
147 | # define getcwd(buf, max) (getcwd) (buf, max, 0) |
148 | # else |
149 | char *getcwd (); |
150 | # endif |
151 | # endif |
152 | # ifndef HAVE_STPCPY |
153 | static char *stpcpy (char *dest, const char *src); |
154 | # endif |
155 | # ifndef HAVE_MEMPCPY |
156 | static void *mempcpy (void *dest, const void *src, size_t n); |
157 | # endif |
158 | #endif |
159 | |
160 | /* Use a replacement if the system does not provide the `tsearch' function |
161 | family. */ |
162 | #if defined HAVE_TSEARCH || defined _LIBC |
163 | # include <search.h> |
164 | #else |
165 | # define tsearch libintl_tsearch |
166 | # define tfind libintl_tfind |
167 | # define tdelete libintl_tdelete |
168 | # define twalk libintl_twalk |
169 | # include "tsearch.h" |
170 | #endif |
171 | |
172 | #ifdef _LIBC |
173 | # define tsearch __tsearch |
174 | #endif |
175 | |
176 | /* Amount to increase buffer size by in each try. */ |
177 | #define PATH_INCR 32 |
178 | |
179 | /* The following is from pathmax.h. */ |
180 | /* Non-POSIX BSD systems might have gcc's limits.h, which doesn't define |
181 | PATH_MAX but might cause redefinition warnings when sys/param.h is |
182 | later included (as on MORE/BSD 4.3). */ |
183 | #if defined _POSIX_VERSION || (defined HAVE_LIMITS_H && !defined __GNUC__) |
184 | # include <limits.h> |
185 | #endif |
186 | |
187 | #ifndef _POSIX_PATH_MAX |
188 | # define _POSIX_PATH_MAX 255 |
189 | #endif |
190 | |
191 | #if !defined PATH_MAX && defined _PC_PATH_MAX |
192 | # define PATH_MAX (__pathconf ("/", _PC_PATH_MAX) < 1 ? 1024 : __pathconf ("/", _PC_PATH_MAX)) |
193 | #endif |
194 | |
195 | /* Don't include sys/param.h if it already has been. */ |
196 | #if defined HAVE_SYS_PARAM_H && !defined PATH_MAX && !defined MAXPATHLEN |
197 | # include <sys/param.h> |
198 | #endif |
199 | |
200 | #if !defined PATH_MAX && defined MAXPATHLEN |
201 | # define PATH_MAX MAXPATHLEN |
202 | #endif |
203 | |
204 | #ifndef PATH_MAX |
205 | # define PATH_MAX _POSIX_PATH_MAX |
206 | #endif |
207 | |
208 | /* Pathname support. |
209 | ISSLASH(C) tests whether C is a directory separator character. |
210 | IS_ABSOLUTE_PATH(P) tests whether P is an absolute path. If it is not, |
211 | it may be concatenated to a directory pathname. |
212 | IS_PATH_WITH_DIR(P) tests whether P contains a directory specification. |
213 | */ |
214 | #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__ |
215 | /* Win32, Cygwin, OS/2, DOS */ |
216 | # define ISSLASH(C) ((C) == '/' || (C) == '\\') |
217 | # define HAS_DEVICE(P) \ |
218 | ((((P)[0] >= 'A' && (P)[0] <= 'Z') || ((P)[0] >= 'a' && (P)[0] <= 'z')) \ |
219 | && (P)[1] == ':') |
220 | # define IS_ABSOLUTE_PATH(P) (ISSLASH ((P)[0]) || HAS_DEVICE (P)) |
221 | # define IS_PATH_WITH_DIR(P) \ |
222 | (strchr (P, '/') != NULL || strchr (P, '\\') != NULL || HAS_DEVICE (P)) |
223 | #else |
224 | /* Unix */ |
225 | # define ISSLASH(C) ((C) == '/') |
226 | # define IS_ABSOLUTE_PATH(P) ISSLASH ((P)[0]) |
227 | # define IS_PATH_WITH_DIR(P) (strchr (P, '/') != NULL) |
228 | #endif |
229 | |
230 | /* Whether to support different locales in different threads. */ |
231 | #if defined _LIBC || HAVE_USELOCALE || defined IN_LIBGLOCALE |
232 | # define HAVE_PER_THREAD_LOCALE |
233 | #endif |
234 | |
235 | /* This is the type used for the search tree where known translations |
236 | are stored. */ |
237 | struct known_translation_t |
238 | { |
239 | /* Domain in which to search. */ |
240 | const char *domainname; |
241 | |
242 | /* The category. */ |
243 | int category; |
244 | |
245 | #ifdef HAVE_PER_THREAD_LOCALE |
246 | /* Name of the relevant locale category, or "" for the global locale. */ |
247 | const char *localename; |
248 | #endif |
249 | |
250 | #ifdef IN_LIBGLOCALE |
251 | /* The character encoding. */ |
252 | const char *encoding; |
253 | #endif |
254 | |
255 | /* State of the catalog counter at the point the string was found. */ |
256 | int counter; |
257 | |
258 | /* Catalog where the string was found. */ |
259 | struct loaded_l10nfile *domain; |
260 | |
261 | /* And finally the translation. */ |
262 | const char *translation; |
263 | size_t translation_length; |
264 | |
265 | /* Pointer to the string in question. */ |
266 | union |
267 | { |
268 | char appended[ZERO]; /* used if domain != NULL */ |
269 | const char *ptr; /* used if domain == NULL */ |
270 | } |
271 | msgid; |
272 | }; |
273 | |
274 | gl_rwlock_define_initialized (static, tree_lock) |
275 | |
276 | /* Root of the search tree with known translations. */ |
277 | static void *root; |
278 | |
279 | /* Function to compare two entries in the table of known translations. */ |
280 | static int |
281 | transcmp (const void *p1, const void *p2) |
282 | { |
283 | const struct known_translation_t *s1; |
284 | const struct known_translation_t *s2; |
285 | int result; |
286 | |
287 | s1 = (const struct known_translation_t *) p1; |
288 | s2 = (const struct known_translation_t *) p2; |
289 | |
290 | result = strcmp (s1->domain != NULL ? s1->msgid.appended : s1->msgid.ptr, |
291 | s2->domain != NULL ? s2->msgid.appended : s2->msgid.ptr); |
292 | if (result == 0) |
293 | { |
294 | result = strcmp (s1->domainname, s2->domainname); |
295 | if (result == 0) |
296 | { |
297 | #ifdef HAVE_PER_THREAD_LOCALE |
298 | result = strcmp (s1->localename, s2->localename); |
299 | if (result == 0) |
300 | #endif |
301 | { |
302 | #ifdef IN_LIBGLOCALE |
303 | result = strcmp (s1->encoding, s2->encoding); |
304 | if (result == 0) |
305 | #endif |
306 | /* We compare the category last (though this is the cheapest |
307 | operation) since it is hopefully always the same (namely |
308 | LC_MESSAGES). */ |
309 | result = s1->category - s2->category; |
310 | } |
311 | } |
312 | } |
313 | |
314 | return result; |
315 | } |
316 | |
317 | /* Name of the default domain used for gettext(3) prior any call to |
318 | textdomain(3). The default value for this is "messages". */ |
319 | const char _nl_default_default_domain[] attribute_hidden = "messages" ; |
320 | |
321 | #ifndef IN_LIBGLOCALE |
322 | /* Value used as the default domain for gettext(3). */ |
323 | const char *_nl_current_default_domain attribute_hidden |
324 | = _nl_default_default_domain; |
325 | #endif |
326 | |
327 | /* Contains the default location of the message catalogs. */ |
328 | #if defined __EMX__ |
329 | extern const char _nl_default_dirname[]; |
330 | #else |
331 | # ifdef _LIBC |
332 | extern const char _nl_default_dirname[]; |
333 | libc_hidden_proto (_nl_default_dirname) |
334 | # endif |
335 | const char _nl_default_dirname[] = LOCALEDIR; |
336 | # ifdef _LIBC |
337 | libc_hidden_data_def (_nl_default_dirname) |
338 | # endif |
339 | #endif |
340 | |
341 | #ifndef IN_LIBGLOCALE |
342 | /* List with bindings of specific domains created by bindtextdomain() |
343 | calls. */ |
344 | struct binding *_nl_domain_bindings; |
345 | #endif |
346 | |
347 | /* Prototypes for local functions. */ |
348 | static char *plural_lookup (struct loaded_l10nfile *domain, |
349 | unsigned long int n, |
350 | const char *translation, size_t translation_len); |
351 | |
352 | #ifdef IN_LIBGLOCALE |
353 | static const char *guess_category_value (int category, |
354 | const char *categoryname, |
355 | const char *localename); |
356 | #else |
357 | static const char *guess_category_value (int category, |
358 | const char *categoryname); |
359 | #endif |
360 | |
361 | #ifdef _LIBC |
362 | # include "../locale/localeinfo.h" |
363 | # define category_to_name(category) \ |
364 | _nl_category_names.str + _nl_category_name_idxs[category] |
365 | #else |
366 | static const char *category_to_name (int category); |
367 | #endif |
368 | #if (defined _LIBC || HAVE_ICONV) && !defined IN_LIBGLOCALE |
369 | static const char *get_output_charset (struct binding *domainbinding); |
370 | #endif |
371 | |
372 | |
373 | /* For those losing systems which don't have `alloca' we have to add |
374 | some additional code emulating it. */ |
375 | #ifdef HAVE_ALLOCA |
376 | /* Nothing has to be done. */ |
377 | # define freea(p) /* nothing */ |
378 | # define ADD_BLOCK(list, address) /* nothing */ |
379 | # define FREE_BLOCKS(list) /* nothing */ |
380 | #else |
381 | struct block_list |
382 | { |
383 | void *address; |
384 | struct block_list *next; |
385 | }; |
386 | # define ADD_BLOCK(list, addr) \ |
387 | do { \ |
388 | struct block_list *newp = (struct block_list *) malloc (sizeof (*newp)); \ |
389 | /* If we cannot get a free block we cannot add the new element to \ |
390 | the list. */ \ |
391 | if (newp != NULL) { \ |
392 | newp->address = (addr); \ |
393 | newp->next = (list); \ |
394 | (list) = newp; \ |
395 | } \ |
396 | } while (0) |
397 | # define FREE_BLOCKS(list) \ |
398 | do { \ |
399 | while (list != NULL) { \ |
400 | struct block_list *old = list; \ |
401 | list = list->next; \ |
402 | free (old->address); \ |
403 | free (old); \ |
404 | } \ |
405 | } while (0) |
406 | # undef alloca |
407 | # define alloca(size) (malloc (size)) |
408 | # define freea(p) free (p) |
409 | #endif /* have alloca */ |
410 | |
411 | |
412 | #ifdef _LIBC |
413 | /* List of blocks allocated for translations. */ |
414 | typedef struct transmem_list |
415 | { |
416 | struct transmem_list *next; |
417 | char data[ZERO]; |
418 | } transmem_block_t; |
419 | static struct transmem_list *transmem_list; |
420 | #else |
421 | typedef unsigned char transmem_block_t; |
422 | #endif |
423 | |
424 | |
425 | /* Names for the libintl functions are a problem. They must not clash |
426 | with existing names and they should follow ANSI C. But this source |
427 | code is also used in GNU C Library where the names have a __ |
428 | prefix. So we have to make a difference here. */ |
429 | #ifdef _LIBC |
430 | # define DCIGETTEXT __dcigettext |
431 | #else |
432 | # define DCIGETTEXT libintl_dcigettext |
433 | #endif |
434 | |
435 | /* Lock variable to protect the global data in the gettext implementation. */ |
436 | gl_rwlock_define_initialized (, _nl_state_lock attribute_hidden) |
437 | |
438 | /* Checking whether the binaries runs SUID must be done and glibc provides |
439 | easier methods therefore we make a difference here. */ |
440 | #ifdef _LIBC |
441 | # define ENABLE_SECURE __libc_enable_secure |
442 | # define DETERMINE_SECURE |
443 | #else |
444 | # ifndef HAVE_GETUID |
445 | # define getuid() 0 |
446 | # endif |
447 | # ifndef HAVE_GETGID |
448 | # define getgid() 0 |
449 | # endif |
450 | # ifndef HAVE_GETEUID |
451 | # define geteuid() getuid() |
452 | # endif |
453 | # ifndef HAVE_GETEGID |
454 | # define getegid() getgid() |
455 | # endif |
456 | static int enable_secure; |
457 | # define ENABLE_SECURE (enable_secure == 1) |
458 | # define DETERMINE_SECURE \ |
459 | if (enable_secure == 0) \ |
460 | { \ |
461 | if (getuid () != geteuid () || getgid () != getegid ()) \ |
462 | enable_secure = 1; \ |
463 | else \ |
464 | enable_secure = -1; \ |
465 | } |
466 | #endif |
467 | |
468 | /* Get the function to evaluate the plural expression. */ |
469 | #include "eval-plural.h" |
470 | |
471 | /* Look up MSGID in the DOMAINNAME message catalog for the current |
472 | CATEGORY locale and, if PLURAL is nonzero, search over string |
473 | depending on the plural form determined by N. */ |
474 | #ifdef IN_LIBGLOCALE |
475 | char * |
476 | gl_dcigettext (const char *domainname, |
477 | const char *msgid1, const char *msgid2, |
478 | int plural, unsigned long int n, |
479 | int category, |
480 | const char *localename, const char *encoding) |
481 | #else |
482 | char * |
483 | DCIGETTEXT (const char *domainname, const char *msgid1, const char *msgid2, |
484 | int plural, unsigned long int n, int category) |
485 | #endif |
486 | { |
487 | #ifndef HAVE_ALLOCA |
488 | struct block_list *block_list = NULL; |
489 | #endif |
490 | struct loaded_l10nfile *domain; |
491 | struct binding *binding; |
492 | const char *categoryname; |
493 | const char *categoryvalue; |
494 | const char *dirname; |
495 | char *xdirname = NULL; |
496 | char *xdomainname; |
497 | char *single_locale; |
498 | char *retval; |
499 | size_t retlen; |
500 | int saved_errno; |
501 | struct known_translation_t search; |
502 | struct known_translation_t **foundp = NULL; |
503 | #if defined HAVE_PER_THREAD_LOCALE && !defined IN_LIBGLOCALE |
504 | const char *localename; |
505 | #endif |
506 | size_t domainname_len; |
507 | |
508 | /* If no real MSGID is given return NULL. */ |
509 | if (msgid1 == NULL) |
510 | return NULL; |
511 | |
512 | #ifdef _LIBC |
513 | if (category < 0 || category >= __LC_LAST || category == LC_ALL) |
514 | /* Bogus. */ |
515 | return (plural == 0 |
516 | ? (char *) msgid1 |
517 | /* Use the Germanic plural rule. */ |
518 | : n == 1 ? (char *) msgid1 : (char *) msgid2); |
519 | #endif |
520 | |
521 | /* Preserve the `errno' value. */ |
522 | saved_errno = errno; |
523 | |
524 | #ifdef _LIBC |
525 | __libc_rwlock_define (extern, __libc_setlocale_lock attribute_hidden) |
526 | __libc_rwlock_rdlock (__libc_setlocale_lock); |
527 | #endif |
528 | |
529 | gl_rwlock_rdlock (_nl_state_lock); |
530 | |
531 | /* If DOMAINNAME is NULL, we are interested in the default domain. If |
532 | CATEGORY is not LC_MESSAGES this might not make much sense but the |
533 | definition left this undefined. */ |
534 | if (domainname == NULL) |
535 | domainname = _nl_current_default_domain; |
536 | |
537 | /* OS/2 specific: backward compatibility with older libintl versions */ |
538 | #ifdef LC_MESSAGES_COMPAT |
539 | if (category == LC_MESSAGES_COMPAT) |
540 | category = LC_MESSAGES; |
541 | #endif |
542 | |
543 | /* Try to find the translation among those which we found at |
544 | some time. */ |
545 | search.domain = NULL; |
546 | search.msgid.ptr = msgid1; |
547 | search.domainname = domainname; |
548 | search.category = category; |
549 | #ifdef HAVE_PER_THREAD_LOCALE |
550 | # ifndef IN_LIBGLOCALE |
551 | # ifdef _LIBC |
552 | localename = __current_locale_name (category); |
553 | # else |
554 | categoryname = category_to_name (category); |
555 | # define CATEGORYNAME_INITIALIZED |
556 | localename = _nl_locale_name_thread_unsafe (category, categoryname); |
557 | if (localename == NULL) |
558 | localename = "" ; |
559 | # endif |
560 | # endif |
561 | search.localename = localename; |
562 | # ifdef IN_LIBGLOCALE |
563 | search.encoding = encoding; |
564 | # endif |
565 | |
566 | /* Since tfind/tsearch manage a balanced tree, concurrent tfind and |
567 | tsearch calls can be fatal. */ |
568 | gl_rwlock_rdlock (tree_lock); |
569 | |
570 | foundp = (struct known_translation_t **) tfind (&search, &root, transcmp); |
571 | |
572 | gl_rwlock_unlock (tree_lock); |
573 | |
574 | if (foundp != NULL && (*foundp)->counter == _nl_msg_cat_cntr) |
575 | { |
576 | /* Now deal with plural. */ |
577 | if (plural) |
578 | retval = plural_lookup ((*foundp)->domain, n, (*foundp)->translation, |
579 | (*foundp)->translation_length); |
580 | else |
581 | retval = (char *) (*foundp)->translation; |
582 | |
583 | gl_rwlock_unlock (_nl_state_lock); |
584 | # ifdef _LIBC |
585 | __libc_rwlock_unlock (__libc_setlocale_lock); |
586 | # endif |
587 | __set_errno (saved_errno); |
588 | return retval; |
589 | } |
590 | #endif |
591 | |
592 | /* See whether this is a SUID binary or not. */ |
593 | DETERMINE_SECURE; |
594 | |
595 | /* First find matching binding. */ |
596 | #ifdef IN_LIBGLOCALE |
597 | /* We can use a trivial binding, since _nl_find_msg will ignore it anyway, |
598 | and _nl_load_domain and _nl_find_domain just pass it through. */ |
599 | binding = NULL; |
600 | dirname = bindtextdomain (domainname, NULL); |
601 | #else |
602 | for (binding = _nl_domain_bindings; binding != NULL; binding = binding->next) |
603 | { |
604 | int compare = strcmp (domainname, binding->domainname); |
605 | if (compare == 0) |
606 | /* We found it! */ |
607 | break; |
608 | if (compare < 0) |
609 | { |
610 | /* It is not in the list. */ |
611 | binding = NULL; |
612 | break; |
613 | } |
614 | } |
615 | |
616 | if (binding == NULL) |
617 | dirname = _nl_default_dirname; |
618 | else |
619 | { |
620 | dirname = binding->dirname; |
621 | #endif |
622 | if (!IS_ABSOLUTE_PATH (dirname)) |
623 | { |
624 | /* We have a relative path. Make it absolute now. */ |
625 | char *cwd = getcwd (NULL, 0); |
626 | if (cwd == NULL) |
627 | /* We cannot get the current working directory. Don't |
628 | signal an error but simply return the default |
629 | string. */ |
630 | goto return_untranslated; |
631 | int ret = __asprintf (&xdirname, "%s/%s" , cwd, dirname); |
632 | free (cwd); |
633 | if (ret < 0) |
634 | return NULL; |
635 | dirname = xdirname; |
636 | } |
637 | #ifndef IN_LIBGLOCALE |
638 | } |
639 | #endif |
640 | |
641 | /* Now determine the symbolic name of CATEGORY and its value. */ |
642 | #ifndef CATEGORYNAME_INITIALIZED |
643 | categoryname = category_to_name (category); |
644 | #endif |
645 | #ifdef IN_LIBGLOCALE |
646 | categoryvalue = guess_category_value (category, categoryname, localename); |
647 | #else |
648 | categoryvalue = guess_category_value (category, categoryname); |
649 | #endif |
650 | |
651 | domainname_len = strlen (domainname); |
652 | xdomainname = (char *) alloca (strlen (categoryname) |
653 | + domainname_len + 5); |
654 | ADD_BLOCK (block_list, xdomainname); |
655 | |
656 | stpcpy ((char *) mempcpy (stpcpy (stpcpy (xdomainname, categoryname), "/" ), |
657 | domainname, domainname_len), |
658 | ".mo" ); |
659 | |
660 | /* Creating working area. */ |
661 | single_locale = (char *) alloca (strlen (categoryvalue) + 1); |
662 | ADD_BLOCK (block_list, single_locale); |
663 | |
664 | |
665 | /* Search for the given string. This is a loop because we perhaps |
666 | got an ordered list of languages to consider for the translation. */ |
667 | while (1) |
668 | { |
669 | /* Make CATEGORYVALUE point to the next element of the list. */ |
670 | while (categoryvalue[0] != '\0' && categoryvalue[0] == ':') |
671 | ++categoryvalue; |
672 | if (categoryvalue[0] == '\0') |
673 | { |
674 | /* The whole contents of CATEGORYVALUE has been searched but |
675 | no valid entry has been found. We solve this situation |
676 | by implicitly appending a "C" entry, i.e. no translation |
677 | will take place. */ |
678 | single_locale[0] = 'C'; |
679 | single_locale[1] = '\0'; |
680 | } |
681 | else |
682 | { |
683 | char *cp = single_locale; |
684 | while (categoryvalue[0] != '\0' && categoryvalue[0] != ':') |
685 | *cp++ = *categoryvalue++; |
686 | *cp = '\0'; |
687 | |
688 | /* When this is a SUID binary we must not allow accessing files |
689 | outside the dedicated directories. */ |
690 | if (ENABLE_SECURE && IS_PATH_WITH_DIR (single_locale)) |
691 | /* Ingore this entry. */ |
692 | continue; |
693 | } |
694 | |
695 | /* If the current locale value is C (or POSIX) we don't load a |
696 | domain. Return the MSGID. */ |
697 | if (strcmp (single_locale, "C" ) == 0 |
698 | || strcmp (single_locale, "POSIX" ) == 0) |
699 | break; |
700 | |
701 | /* Find structure describing the message catalog matching the |
702 | DOMAINNAME and CATEGORY. */ |
703 | domain = _nl_find_domain (dirname, single_locale, xdomainname, binding); |
704 | |
705 | if (domain != NULL) |
706 | { |
707 | #if defined IN_LIBGLOCALE |
708 | retval = _nl_find_msg (domain, binding, encoding, msgid1, &retlen); |
709 | #else |
710 | retval = _nl_find_msg (domain, binding, msgid1, 1, &retlen); |
711 | #endif |
712 | |
713 | if (retval == NULL) |
714 | { |
715 | int cnt; |
716 | |
717 | for (cnt = 0; domain->successor[cnt] != NULL; ++cnt) |
718 | { |
719 | #if defined IN_LIBGLOCALE |
720 | retval = _nl_find_msg (domain->successor[cnt], binding, |
721 | encoding, msgid1, &retlen); |
722 | #else |
723 | retval = _nl_find_msg (domain->successor[cnt], binding, |
724 | msgid1, 1, &retlen); |
725 | #endif |
726 | |
727 | /* Resource problems are not fatal, instead we return no |
728 | translation. */ |
729 | if (__builtin_expect (retval == (char *) -1, 0)) |
730 | goto return_untranslated; |
731 | |
732 | if (retval != NULL) |
733 | { |
734 | domain = domain->successor[cnt]; |
735 | break; |
736 | } |
737 | } |
738 | } |
739 | |
740 | /* Returning -1 means that some resource problem exists |
741 | (likely memory) and that the strings could not be |
742 | converted. Return the original strings. */ |
743 | if (__builtin_expect (retval == (char *) -1, 0)) |
744 | break; |
745 | |
746 | if (retval != NULL) |
747 | { |
748 | /* Found the translation of MSGID1 in domain DOMAIN: |
749 | starting at RETVAL, RETLEN bytes. */ |
750 | free (xdirname); |
751 | FREE_BLOCKS (block_list); |
752 | if (foundp == NULL) |
753 | { |
754 | /* Create a new entry and add it to the search tree. */ |
755 | size_t msgid_len; |
756 | size_t size; |
757 | struct known_translation_t *newp; |
758 | |
759 | msgid_len = strlen (msgid1) + 1; |
760 | size = offsetof (struct known_translation_t, msgid) |
761 | + msgid_len + domainname_len + 1; |
762 | #ifdef HAVE_PER_THREAD_LOCALE |
763 | size += strlen (localename) + 1; |
764 | #endif |
765 | newp = (struct known_translation_t *) malloc (size); |
766 | if (newp != NULL) |
767 | { |
768 | char *new_domainname; |
769 | #ifdef HAVE_PER_THREAD_LOCALE |
770 | char *new_localename; |
771 | #endif |
772 | |
773 | new_domainname = |
774 | (char *) mempcpy (newp->msgid.appended, msgid1, |
775 | msgid_len); |
776 | memcpy (new_domainname, domainname, domainname_len + 1); |
777 | #ifdef HAVE_PER_THREAD_LOCALE |
778 | new_localename = new_domainname + domainname_len + 1; |
779 | strcpy (new_localename, localename); |
780 | #endif |
781 | newp->domainname = new_domainname; |
782 | newp->category = category; |
783 | #ifdef HAVE_PER_THREAD_LOCALE |
784 | newp->localename = new_localename; |
785 | #endif |
786 | #ifdef IN_LIBGLOCALE |
787 | newp->encoding = encoding; |
788 | #endif |
789 | newp->counter = _nl_msg_cat_cntr; |
790 | newp->domain = domain; |
791 | newp->translation = retval; |
792 | newp->translation_length = retlen; |
793 | |
794 | gl_rwlock_wrlock (tree_lock); |
795 | |
796 | /* Insert the entry in the search tree. */ |
797 | foundp = (struct known_translation_t **) |
798 | tsearch (newp, &root, transcmp); |
799 | |
800 | gl_rwlock_unlock (tree_lock); |
801 | |
802 | if (foundp == NULL |
803 | || __builtin_expect (*foundp != newp, 0)) |
804 | /* The insert failed. */ |
805 | free (newp); |
806 | } |
807 | } |
808 | else |
809 | { |
810 | /* We can update the existing entry. */ |
811 | (*foundp)->counter = _nl_msg_cat_cntr; |
812 | (*foundp)->domain = domain; |
813 | (*foundp)->translation = retval; |
814 | (*foundp)->translation_length = retlen; |
815 | } |
816 | |
817 | __set_errno (saved_errno); |
818 | |
819 | /* Now deal with plural. */ |
820 | if (plural) |
821 | retval = plural_lookup (domain, n, retval, retlen); |
822 | |
823 | gl_rwlock_unlock (_nl_state_lock); |
824 | #ifdef _LIBC |
825 | __libc_rwlock_unlock (__libc_setlocale_lock); |
826 | #endif |
827 | return retval; |
828 | } |
829 | } |
830 | } |
831 | |
832 | return_untranslated: |
833 | /* Return the untranslated MSGID. */ |
834 | free (xdirname); |
835 | FREE_BLOCKS (block_list); |
836 | gl_rwlock_unlock (_nl_state_lock); |
837 | #ifdef _LIBC |
838 | __libc_rwlock_unlock (__libc_setlocale_lock); |
839 | #endif |
840 | #ifndef _LIBC |
841 | if (!ENABLE_SECURE) |
842 | { |
843 | extern void _nl_log_untranslated (const char *logfilename, |
844 | const char *domainname, |
845 | const char *msgid1, const char *msgid2, |
846 | int plural); |
847 | const char *logfilename = getenv ("GETTEXT_LOG_UNTRANSLATED" ); |
848 | |
849 | if (logfilename != NULL && logfilename[0] != '\0') |
850 | _nl_log_untranslated (logfilename, domainname, msgid1, msgid2, plural); |
851 | } |
852 | #endif |
853 | __set_errno (saved_errno); |
854 | return (plural == 0 |
855 | ? (char *) msgid1 |
856 | /* Use the Germanic plural rule. */ |
857 | : n == 1 ? (char *) msgid1 : (char *) msgid2); |
858 | } |
859 | |
860 | |
861 | /* Look up the translation of msgid within DOMAIN_FILE and DOMAINBINDING. |
862 | Return it if found. Return NULL if not found or in case of a conversion |
863 | failure (problem in the particular message catalog). Return (char *) -1 |
864 | in case of a memory allocation failure during conversion (only if |
865 | ENCODING != NULL resp. CONVERT == true). */ |
866 | char * |
867 | #ifdef IN_LIBGLOCALE |
868 | _nl_find_msg (struct loaded_l10nfile *domain_file, |
869 | struct binding *domainbinding, const char *encoding, |
870 | const char *msgid, |
871 | size_t *lengthp) |
872 | #else |
873 | _nl_find_msg (struct loaded_l10nfile *domain_file, |
874 | struct binding *domainbinding, |
875 | const char *msgid, int convert, |
876 | size_t *lengthp) |
877 | #endif |
878 | { |
879 | struct loaded_domain *domain; |
880 | nls_uint32 nstrings; |
881 | size_t act; |
882 | char *result; |
883 | size_t resultlen; |
884 | |
885 | if (domain_file->decided <= 0) |
886 | _nl_load_domain (domain_file, domainbinding); |
887 | |
888 | if (domain_file->data == NULL) |
889 | return NULL; |
890 | |
891 | domain = (struct loaded_domain *) domain_file->data; |
892 | |
893 | nstrings = domain->nstrings; |
894 | |
895 | /* Locate the MSGID and its translation. */ |
896 | if (domain->hash_tab != NULL) |
897 | { |
898 | /* Use the hashing table. */ |
899 | nls_uint32 len = strlen (msgid); |
900 | nls_uint32 hash_val = __hash_string (msgid); |
901 | nls_uint32 idx = hash_val % domain->hash_size; |
902 | nls_uint32 incr = 1 + (hash_val % (domain->hash_size - 2)); |
903 | |
904 | while (1) |
905 | { |
906 | nls_uint32 nstr = |
907 | W (domain->must_swap_hash_tab, domain->hash_tab[idx]); |
908 | |
909 | if (nstr == 0) |
910 | /* Hash table entry is empty. */ |
911 | return NULL; |
912 | |
913 | nstr--; |
914 | |
915 | /* Compare msgid with the original string at index nstr. |
916 | We compare the lengths with >=, not ==, because plural entries |
917 | are represented by strings with an embedded NUL. */ |
918 | if (nstr < nstrings |
919 | ? W (domain->must_swap, domain->orig_tab[nstr].length) >= len |
920 | && (strcmp (msgid, |
921 | domain->data + W (domain->must_swap, |
922 | domain->orig_tab[nstr].offset)) |
923 | == 0) |
924 | : domain->orig_sysdep_tab[nstr - nstrings].length > len |
925 | && (strcmp (msgid, |
926 | domain->orig_sysdep_tab[nstr - nstrings].pointer) |
927 | == 0)) |
928 | { |
929 | act = nstr; |
930 | goto found; |
931 | } |
932 | |
933 | if (idx >= domain->hash_size - incr) |
934 | idx -= domain->hash_size - incr; |
935 | else |
936 | idx += incr; |
937 | } |
938 | /* NOTREACHED */ |
939 | } |
940 | else |
941 | { |
942 | /* Try the default method: binary search in the sorted array of |
943 | messages. */ |
944 | size_t top, bottom; |
945 | |
946 | bottom = 0; |
947 | top = nstrings; |
948 | while (bottom < top) |
949 | { |
950 | int cmp_val; |
951 | |
952 | act = (bottom + top) / 2; |
953 | cmp_val = strcmp (msgid, (domain->data |
954 | + W (domain->must_swap, |
955 | domain->orig_tab[act].offset))); |
956 | if (cmp_val < 0) |
957 | top = act; |
958 | else if (cmp_val > 0) |
959 | bottom = act + 1; |
960 | else |
961 | goto found; |
962 | } |
963 | /* No translation was found. */ |
964 | return NULL; |
965 | } |
966 | |
967 | found: |
968 | /* The translation was found at index ACT. If we have to convert the |
969 | string to use a different character set, this is the time. */ |
970 | if (act < nstrings) |
971 | { |
972 | result = (char *) |
973 | (domain->data + W (domain->must_swap, domain->trans_tab[act].offset)); |
974 | resultlen = W (domain->must_swap, domain->trans_tab[act].length) + 1; |
975 | } |
976 | else |
977 | { |
978 | result = (char *) domain->trans_sysdep_tab[act - nstrings].pointer; |
979 | resultlen = domain->trans_sysdep_tab[act - nstrings].length; |
980 | } |
981 | |
982 | #if defined _LIBC || HAVE_ICONV |
983 | # ifdef IN_LIBGLOCALE |
984 | if (encoding != NULL) |
985 | # else |
986 | if (convert) |
987 | # endif |
988 | { |
989 | /* We are supposed to do a conversion. */ |
990 | # ifndef IN_LIBGLOCALE |
991 | const char *encoding = get_output_charset (domainbinding); |
992 | # endif |
993 | size_t nconversions; |
994 | struct converted_domain *convd; |
995 | size_t i; |
996 | |
997 | /* Protect against reallocation of the table. */ |
998 | gl_rwlock_rdlock (domain->conversions_lock); |
999 | |
1000 | /* Search whether a table with converted translations for this |
1001 | encoding has already been allocated. */ |
1002 | nconversions = domain->nconversions; |
1003 | convd = NULL; |
1004 | |
1005 | for (i = nconversions; i > 0; ) |
1006 | { |
1007 | i--; |
1008 | if (strcmp (domain->conversions[i].encoding, encoding) == 0) |
1009 | { |
1010 | convd = &domain->conversions[i]; |
1011 | break; |
1012 | } |
1013 | } |
1014 | |
1015 | gl_rwlock_unlock (domain->conversions_lock); |
1016 | |
1017 | if (convd == NULL) |
1018 | { |
1019 | /* We have to allocate a new conversions table. */ |
1020 | gl_rwlock_wrlock (domain->conversions_lock); |
1021 | nconversions = domain->nconversions; |
1022 | |
1023 | /* Maybe in the meantime somebody added the translation. |
1024 | Recheck. */ |
1025 | for (i = nconversions; i > 0; ) |
1026 | { |
1027 | i--; |
1028 | if (strcmp (domain->conversions[i].encoding, encoding) == 0) |
1029 | { |
1030 | convd = &domain->conversions[i]; |
1031 | goto found_convd; |
1032 | } |
1033 | } |
1034 | |
1035 | { |
1036 | /* Allocate a table for the converted translations for this |
1037 | encoding. */ |
1038 | struct converted_domain *new_conversions = |
1039 | (struct converted_domain *) |
1040 | (domain->conversions != NULL |
1041 | ? realloc (domain->conversions, |
1042 | (nconversions + 1) * sizeof (struct converted_domain)) |
1043 | : malloc ((nconversions + 1) * sizeof (struct converted_domain))); |
1044 | |
1045 | if (__builtin_expect (new_conversions == NULL, 0)) |
1046 | { |
1047 | /* Nothing we can do, no more memory. We cannot use the |
1048 | translation because it might be encoded incorrectly. */ |
1049 | unlock_fail: |
1050 | gl_rwlock_unlock (domain->conversions_lock); |
1051 | return (char *) -1; |
1052 | } |
1053 | |
1054 | domain->conversions = new_conversions; |
1055 | |
1056 | /* Copy the 'encoding' string to permanent storage. */ |
1057 | encoding = strdup (encoding); |
1058 | if (__builtin_expect (encoding == NULL, 0)) |
1059 | /* Nothing we can do, no more memory. We cannot use the |
1060 | translation because it might be encoded incorrectly. */ |
1061 | goto unlock_fail; |
1062 | |
1063 | convd = &new_conversions[nconversions]; |
1064 | convd->encoding = encoding; |
1065 | |
1066 | /* Find out about the character set the file is encoded with. |
1067 | This can be found (in textual form) in the entry "". If this |
1068 | entry does not exist or if this does not contain the 'charset=' |
1069 | information, we will assume the charset matches the one the |
1070 | current locale and we don't have to perform any conversion. */ |
1071 | # ifdef _LIBC |
1072 | convd->conv = (__gconv_t) -1; |
1073 | # else |
1074 | # if HAVE_ICONV |
1075 | convd->conv = (iconv_t) -1; |
1076 | # endif |
1077 | # endif |
1078 | { |
1079 | char *nullentry; |
1080 | size_t nullentrylen; |
1081 | |
1082 | /* Get the header entry. This is a recursion, but it doesn't |
1083 | reallocate domain->conversions because we pass |
1084 | encoding = NULL or convert = 0, respectively. */ |
1085 | nullentry = |
1086 | # ifdef IN_LIBGLOCALE |
1087 | _nl_find_msg (domain_file, domainbinding, NULL, "" , |
1088 | &nullentrylen); |
1089 | # else |
1090 | _nl_find_msg (domain_file, domainbinding, "" , 0, &nullentrylen); |
1091 | # endif |
1092 | |
1093 | /* Resource problems are fatal. If we continue onwards we will |
1094 | only attempt to calloc a new conv_tab and fail later. */ |
1095 | if (__builtin_expect (nullentry == (char *) -1, 0)) |
1096 | return (char *) -1; |
1097 | |
1098 | if (nullentry != NULL) |
1099 | { |
1100 | const char *charsetstr; |
1101 | |
1102 | charsetstr = strstr (nullentry, "charset=" ); |
1103 | if (charsetstr != NULL) |
1104 | { |
1105 | size_t len; |
1106 | char *charset; |
1107 | const char *outcharset; |
1108 | |
1109 | charsetstr += strlen ("charset=" ); |
1110 | len = strcspn (charsetstr, " \t\n" ); |
1111 | |
1112 | charset = (char *) alloca (len + 1); |
1113 | # if defined _LIBC || HAVE_MEMPCPY |
1114 | *((char *) mempcpy (charset, charsetstr, len)) = '\0'; |
1115 | # else |
1116 | memcpy (charset, charsetstr, len); |
1117 | charset[len] = '\0'; |
1118 | # endif |
1119 | |
1120 | outcharset = encoding; |
1121 | |
1122 | # ifdef _LIBC |
1123 | /* We always want to use transliteration. */ |
1124 | outcharset = norm_add_slashes (outcharset, "TRANSLIT" ); |
1125 | charset = norm_add_slashes (charset, "" ); |
1126 | int r = __gconv_open (outcharset, charset, &convd->conv, |
1127 | GCONV_AVOID_NOCONV); |
1128 | if (__builtin_expect (r != __GCONV_OK, 0)) |
1129 | { |
1130 | /* If the output encoding is the same there is |
1131 | nothing to do. Otherwise do not use the |
1132 | translation at all. */ |
1133 | if (__builtin_expect (r != __GCONV_NULCONV, 1)) |
1134 | { |
1135 | gl_rwlock_unlock (domain->conversions_lock); |
1136 | free ((char *) encoding); |
1137 | return NULL; |
1138 | } |
1139 | |
1140 | convd->conv = (__gconv_t) -1; |
1141 | } |
1142 | # else |
1143 | # if HAVE_ICONV |
1144 | /* When using GNU libc >= 2.2 or GNU libiconv >= 1.5, |
1145 | we want to use transliteration. */ |
1146 | # if (((__GLIBC__ == 2 && __GLIBC_MINOR__ >= 2) || __GLIBC__ > 2) \ |
1147 | && !defined __UCLIBC__) \ |
1148 | || _LIBICONV_VERSION >= 0x0105 |
1149 | if (strchr (outcharset, '/') == NULL) |
1150 | { |
1151 | char *tmp; |
1152 | |
1153 | len = strlen (outcharset); |
1154 | tmp = (char *) alloca (len + 10 + 1); |
1155 | memcpy (tmp, outcharset, len); |
1156 | memcpy (tmp + len, "//TRANSLIT" , 10 + 1); |
1157 | outcharset = tmp; |
1158 | |
1159 | convd->conv = iconv_open (outcharset, charset); |
1160 | |
1161 | freea (outcharset); |
1162 | } |
1163 | else |
1164 | # endif |
1165 | convd->conv = iconv_open (outcharset, charset); |
1166 | # endif |
1167 | # endif |
1168 | |
1169 | freea (charset); |
1170 | } |
1171 | } |
1172 | } |
1173 | convd->conv_tab = NULL; |
1174 | /* Here domain->conversions is still == new_conversions. */ |
1175 | domain->nconversions++; |
1176 | } |
1177 | |
1178 | found_convd: |
1179 | gl_rwlock_unlock (domain->conversions_lock); |
1180 | } |
1181 | |
1182 | if ( |
1183 | # ifdef _LIBC |
1184 | convd->conv != (__gconv_t) -1 |
1185 | # else |
1186 | # if HAVE_ICONV |
1187 | convd->conv != (iconv_t) -1 |
1188 | # endif |
1189 | # endif |
1190 | ) |
1191 | { |
1192 | /* We are supposed to do a conversion. First allocate an |
1193 | appropriate table with the same structure as the table |
1194 | of translations in the file, where we can put the pointers |
1195 | to the converted strings in. |
1196 | There is a slight complication with plural entries. They |
1197 | are represented by consecutive NUL terminated strings. We |
1198 | handle this case by converting RESULTLEN bytes, including |
1199 | NULs. */ |
1200 | |
1201 | /* This lock primarily protects the memory management variables |
1202 | freemem, freemem_size. It also protects write accesses to |
1203 | convd->conv_tab. It's not worth using a separate lock (such |
1204 | as domain->conversions_lock) for this purpose, because when |
1205 | modifying convd->conv_tab, we also need to lock freemem, |
1206 | freemem_size for most of the time. */ |
1207 | __libc_lock_define_initialized (static, lock) |
1208 | |
1209 | if (__builtin_expect (convd->conv_tab == NULL, 0)) |
1210 | { |
1211 | __libc_lock_lock (lock); |
1212 | if (convd->conv_tab == NULL) |
1213 | { |
1214 | convd->conv_tab = |
1215 | (char **) calloc (nstrings + domain->n_sysdep_strings, |
1216 | sizeof (char *)); |
1217 | if (convd->conv_tab != NULL) |
1218 | goto not_translated_yet; |
1219 | /* Mark that we didn't succeed allocating a table. */ |
1220 | convd->conv_tab = (char **) -1; |
1221 | } |
1222 | __libc_lock_unlock (lock); |
1223 | } |
1224 | |
1225 | if (__builtin_expect (convd->conv_tab == (char **) -1, 0)) |
1226 | /* Nothing we can do, no more memory. We cannot use the |
1227 | translation because it might be encoded incorrectly. */ |
1228 | return (char *) -1; |
1229 | |
1230 | if (convd->conv_tab[act] == NULL) |
1231 | { |
1232 | /* We haven't used this string so far, so it is not |
1233 | translated yet. Do this now. */ |
1234 | /* We use a bit more efficient memory handling. |
1235 | We allocate always larger blocks which get used over |
1236 | time. This is faster than many small allocations. */ |
1237 | # define INITIAL_BLOCK_SIZE 4080 |
1238 | static unsigned char *freemem; |
1239 | static size_t freemem_size; |
1240 | |
1241 | const unsigned char *inbuf; |
1242 | unsigned char *outbuf; |
1243 | int malloc_count; |
1244 | # ifndef _LIBC |
1245 | transmem_block_t *transmem_list; |
1246 | # endif |
1247 | |
1248 | __libc_lock_lock (lock); |
1249 | not_translated_yet: |
1250 | |
1251 | inbuf = (const unsigned char *) result; |
1252 | outbuf = freemem + sizeof (size_t); |
1253 | # ifndef _LIBC |
1254 | transmem_list = NULL; |
1255 | # endif |
1256 | |
1257 | malloc_count = 0; |
1258 | while (1) |
1259 | { |
1260 | transmem_block_t *newmem; |
1261 | # ifdef _LIBC |
1262 | size_t non_reversible; |
1263 | int res; |
1264 | |
1265 | if (freemem_size < sizeof (size_t)) |
1266 | goto resize_freemem; |
1267 | |
1268 | res = __gconv (convd->conv, |
1269 | &inbuf, inbuf + resultlen, |
1270 | &outbuf, |
1271 | outbuf + freemem_size - sizeof (size_t), |
1272 | &non_reversible); |
1273 | |
1274 | if (res == __GCONV_OK || res == __GCONV_EMPTY_INPUT) |
1275 | break; |
1276 | |
1277 | if (res != __GCONV_FULL_OUTPUT) |
1278 | { |
1279 | /* We should not use the translation at all, it |
1280 | is incorrectly encoded. */ |
1281 | __libc_lock_unlock (lock); |
1282 | return NULL; |
1283 | } |
1284 | |
1285 | inbuf = (const unsigned char *) result; |
1286 | # else |
1287 | # if HAVE_ICONV |
1288 | const char *inptr = (const char *) inbuf; |
1289 | size_t inleft = resultlen; |
1290 | char *outptr = (char *) outbuf; |
1291 | size_t outleft; |
1292 | |
1293 | if (freemem_size < sizeof (size_t)) |
1294 | goto resize_freemem; |
1295 | |
1296 | outleft = freemem_size - sizeof (size_t); |
1297 | if (iconv (convd->conv, |
1298 | (ICONV_CONST char **) &inptr, &inleft, |
1299 | &outptr, &outleft) |
1300 | != (size_t) (-1)) |
1301 | { |
1302 | outbuf = (unsigned char *) outptr; |
1303 | break; |
1304 | } |
1305 | if (errno != E2BIG) |
1306 | { |
1307 | __libc_lock_unlock (lock); |
1308 | return NULL; |
1309 | } |
1310 | # endif |
1311 | # endif |
1312 | |
1313 | resize_freemem: |
1314 | /* We must allocate a new buffer or resize the old one. */ |
1315 | if (malloc_count > 0) |
1316 | { |
1317 | ++malloc_count; |
1318 | freemem_size = malloc_count * INITIAL_BLOCK_SIZE; |
1319 | newmem = (transmem_block_t *) realloc (transmem_list, |
1320 | freemem_size); |
1321 | # ifdef _LIBC |
1322 | if (newmem != NULL) |
1323 | transmem_list = newmem; |
1324 | else |
1325 | { |
1326 | struct transmem_list *old = transmem_list; |
1327 | |
1328 | transmem_list = transmem_list->next; |
1329 | free (old); |
1330 | } |
1331 | # endif |
1332 | } |
1333 | else |
1334 | { |
1335 | malloc_count = 1; |
1336 | freemem_size = INITIAL_BLOCK_SIZE; |
1337 | newmem = (transmem_block_t *) malloc (freemem_size); |
1338 | # ifdef _LIBC |
1339 | if (newmem != NULL) |
1340 | { |
1341 | /* Add the block to the list of blocks we have to free |
1342 | at some point. */ |
1343 | newmem->next = transmem_list; |
1344 | transmem_list = newmem; |
1345 | } |
1346 | /* Fall through and return -1. */ |
1347 | # endif |
1348 | } |
1349 | if (__builtin_expect (newmem == NULL, 0)) |
1350 | { |
1351 | freemem = NULL; |
1352 | freemem_size = 0; |
1353 | __libc_lock_unlock (lock); |
1354 | return (char *) -1; |
1355 | } |
1356 | |
1357 | # ifdef _LIBC |
1358 | freemem = (unsigned char *) newmem->data; |
1359 | freemem_size -= offsetof (struct transmem_list, data); |
1360 | # else |
1361 | transmem_list = newmem; |
1362 | freemem = newmem; |
1363 | # endif |
1364 | |
1365 | outbuf = freemem + sizeof (size_t); |
1366 | } |
1367 | |
1368 | /* We have now in our buffer a converted string. Put this |
1369 | into the table of conversions. */ |
1370 | *(size_t *) freemem = outbuf - freemem - sizeof (size_t); |
1371 | convd->conv_tab[act] = (char *) freemem; |
1372 | /* Shrink freemem, but keep it aligned. */ |
1373 | freemem_size -= outbuf - freemem; |
1374 | freemem = outbuf; |
1375 | freemem += freemem_size & (alignof (size_t) - 1); |
1376 | freemem_size = freemem_size & ~ (alignof (size_t) - 1); |
1377 | |
1378 | __libc_lock_unlock (lock); |
1379 | } |
1380 | |
1381 | /* Now convd->conv_tab[act] contains the translation of all |
1382 | the plural variants. */ |
1383 | result = convd->conv_tab[act] + sizeof (size_t); |
1384 | resultlen = *(size_t *) convd->conv_tab[act]; |
1385 | } |
1386 | } |
1387 | |
1388 | /* The result string is converted. */ |
1389 | |
1390 | #endif /* _LIBC || HAVE_ICONV */ |
1391 | |
1392 | *lengthp = resultlen; |
1393 | return result; |
1394 | } |
1395 | |
1396 | |
1397 | /* Look up a plural variant. */ |
1398 | static char * |
1399 | plural_lookup (struct loaded_l10nfile *domain, unsigned long int n, |
1400 | const char *translation, size_t translation_len) |
1401 | { |
1402 | struct loaded_domain *domaindata = (struct loaded_domain *) domain->data; |
1403 | unsigned long int index; |
1404 | const char *p; |
1405 | |
1406 | index = plural_eval (domaindata->plural, n); |
1407 | if (index >= domaindata->nplurals) |
1408 | /* This should never happen. It means the plural expression and the |
1409 | given maximum value do not match. */ |
1410 | index = 0; |
1411 | |
1412 | /* Skip INDEX strings at TRANSLATION. */ |
1413 | p = translation; |
1414 | while (index-- > 0) |
1415 | { |
1416 | #ifdef _LIBC |
1417 | p = __rawmemchr (p, '\0'); |
1418 | #else |
1419 | p = strchr (p, '\0'); |
1420 | #endif |
1421 | /* And skip over the NUL byte. */ |
1422 | p++; |
1423 | |
1424 | if (p >= translation + translation_len) |
1425 | /* This should never happen. It means the plural expression |
1426 | evaluated to a value larger than the number of variants |
1427 | available for MSGID1. */ |
1428 | return (char *) translation; |
1429 | } |
1430 | return (char *) p; |
1431 | } |
1432 | |
1433 | #ifndef _LIBC |
1434 | /* Return string representation of locale CATEGORY. */ |
1435 | static const char * |
1436 | category_to_name (int category) |
1437 | { |
1438 | const char *retval; |
1439 | |
1440 | switch (category) |
1441 | { |
1442 | #ifdef LC_COLLATE |
1443 | case LC_COLLATE: |
1444 | retval = "LC_COLLATE" ; |
1445 | break; |
1446 | #endif |
1447 | #ifdef LC_CTYPE |
1448 | case LC_CTYPE: |
1449 | retval = "LC_CTYPE" ; |
1450 | break; |
1451 | #endif |
1452 | #ifdef LC_MONETARY |
1453 | case LC_MONETARY: |
1454 | retval = "LC_MONETARY" ; |
1455 | break; |
1456 | #endif |
1457 | #ifdef LC_NUMERIC |
1458 | case LC_NUMERIC: |
1459 | retval = "LC_NUMERIC" ; |
1460 | break; |
1461 | #endif |
1462 | #ifdef LC_TIME |
1463 | case LC_TIME: |
1464 | retval = "LC_TIME" ; |
1465 | break; |
1466 | #endif |
1467 | #ifdef LC_MESSAGES |
1468 | case LC_MESSAGES: |
1469 | retval = "LC_MESSAGES" ; |
1470 | break; |
1471 | #endif |
1472 | #ifdef LC_RESPONSE |
1473 | case LC_RESPONSE: |
1474 | retval = "LC_RESPONSE" ; |
1475 | break; |
1476 | #endif |
1477 | #ifdef LC_ALL |
1478 | case LC_ALL: |
1479 | /* This might not make sense but is perhaps better than any other |
1480 | value. */ |
1481 | retval = "LC_ALL" ; |
1482 | break; |
1483 | #endif |
1484 | default: |
1485 | /* If you have a better idea for a default value let me know. */ |
1486 | retval = "LC_XXX" ; |
1487 | } |
1488 | |
1489 | return retval; |
1490 | } |
1491 | #endif |
1492 | |
1493 | /* Guess value of current locale from value of the environment variables |
1494 | or system-dependent defaults. */ |
1495 | static const char * |
1496 | #ifdef IN_LIBGLOCALE |
1497 | guess_category_value (int category, const char *categoryname, |
1498 | const char *locale) |
1499 | |
1500 | #else |
1501 | guess_category_value (int category, const char *categoryname) |
1502 | #endif |
1503 | { |
1504 | const char *language; |
1505 | #ifndef IN_LIBGLOCALE |
1506 | const char *locale; |
1507 | # ifndef _LIBC |
1508 | const char *language_default; |
1509 | int locale_defaulted; |
1510 | # endif |
1511 | #endif |
1512 | |
1513 | /* We use the settings in the following order: |
1514 | 1. The value of the environment variable 'LANGUAGE'. This is a GNU |
1515 | extension. Its value can be a colon-separated list of locale names. |
1516 | 2. The value of the environment variable 'LC_ALL', 'LC_xxx', or 'LANG'. |
1517 | More precisely, the first among these that is set to a non-empty value. |
1518 | This is how POSIX specifies it. The value is a single locale name. |
1519 | 3. A system-dependent preference list of languages. Its value can be a |
1520 | colon-separated list of locale names. |
1521 | 4. A system-dependent default locale name. |
1522 | This way: |
1523 | - System-dependent settings can be overridden by environment variables. |
1524 | - If the system provides both a list of languages and a default locale, |
1525 | the former is used. */ |
1526 | |
1527 | #ifndef IN_LIBGLOCALE |
1528 | /* Fetch the locale name, through the POSIX method of looking to `LC_ALL', |
1529 | `LC_xxx', and `LANG'. On some systems this can be done by the |
1530 | `setlocale' function itself. */ |
1531 | # ifdef _LIBC |
1532 | locale = __current_locale_name (category); |
1533 | # else |
1534 | locale_defaulted = 0; |
1535 | # if HAVE_USELOCALE |
1536 | locale = _nl_locale_name_thread_unsafe (category, categoryname); |
1537 | if (locale == NULL) |
1538 | # endif |
1539 | { |
1540 | locale = _nl_locale_name_posix (category, categoryname); |
1541 | if (locale == NULL) |
1542 | { |
1543 | locale = _nl_locale_name_default (); |
1544 | locale_defaulted = 1; |
1545 | } |
1546 | } |
1547 | # endif |
1548 | #endif |
1549 | |
1550 | /* Ignore LANGUAGE and its system-dependent analogon if the locale is set |
1551 | to "C" because |
1552 | 1. "C" locale usually uses the ASCII encoding, and most international |
1553 | messages use non-ASCII characters. These characters get displayed |
1554 | as question marks (if using glibc's iconv()) or as invalid 8-bit |
1555 | characters (because other iconv()s refuse to convert most non-ASCII |
1556 | characters to ASCII). In any case, the output is ugly. |
1557 | 2. The precise output of some programs in the "C" locale is specified |
1558 | by POSIX and should not depend on environment variables like |
1559 | "LANGUAGE" or system-dependent information. We allow such programs |
1560 | to use gettext(). */ |
1561 | if (strcmp (locale, "C" ) == 0) |
1562 | return locale; |
1563 | |
1564 | /* The highest priority value is the value of the 'LANGUAGE' environment |
1565 | variable. */ |
1566 | language = getenv ("LANGUAGE" ); |
1567 | if (language != NULL && language[0] != '\0') |
1568 | return language; |
1569 | #if !defined IN_LIBGLOCALE && !defined _LIBC |
1570 | /* The next priority value is the locale name, if not defaulted. */ |
1571 | if (locale_defaulted) |
1572 | { |
1573 | /* The next priority value is the default language preferences list. */ |
1574 | language_default = _nl_language_preferences_default (); |
1575 | if (language_default != NULL) |
1576 | return language_default; |
1577 | } |
1578 | /* The least priority value is the locale name, if defaulted. */ |
1579 | #endif |
1580 | return locale; |
1581 | } |
1582 | |
1583 | #if (defined _LIBC || HAVE_ICONV) && !defined IN_LIBGLOCALE |
1584 | /* Returns the output charset. */ |
1585 | static const char * |
1586 | get_output_charset (struct binding *domainbinding) |
1587 | { |
1588 | /* The output charset should normally be determined by the locale. But |
1589 | sometimes the locale is not used or not correctly set up, so we provide |
1590 | a possibility for the user to override this: the OUTPUT_CHARSET |
1591 | environment variable. Moreover, the value specified through |
1592 | bind_textdomain_codeset overrides both. */ |
1593 | if (domainbinding != NULL && domainbinding->codeset != NULL) |
1594 | return domainbinding->codeset; |
1595 | else |
1596 | { |
1597 | /* For speed reasons, we look at the value of OUTPUT_CHARSET only |
1598 | once. This is a user variable that is not supposed to change |
1599 | during a program run. */ |
1600 | static char *output_charset_cache; |
1601 | static int output_charset_cached; |
1602 | |
1603 | if (!output_charset_cached) |
1604 | { |
1605 | const char *value = getenv ("OUTPUT_CHARSET" ); |
1606 | |
1607 | if (value != NULL && value[0] != '\0') |
1608 | { |
1609 | size_t len = strlen (value) + 1; |
1610 | char *value_copy = (char *) malloc (len); |
1611 | |
1612 | if (value_copy != NULL) |
1613 | memcpy (value_copy, value, len); |
1614 | output_charset_cache = value_copy; |
1615 | } |
1616 | output_charset_cached = 1; |
1617 | } |
1618 | |
1619 | if (output_charset_cache != NULL) |
1620 | return output_charset_cache; |
1621 | else |
1622 | { |
1623 | # ifdef _LIBC |
1624 | return _NL_CURRENT (LC_CTYPE, CODESET); |
1625 | # else |
1626 | # if HAVE_ICONV |
1627 | return locale_charset (); |
1628 | # endif |
1629 | # endif |
1630 | } |
1631 | } |
1632 | } |
1633 | #endif |
1634 | |
1635 | /* @@ begin of epilog @@ */ |
1636 | |
1637 | /* We don't want libintl.a to depend on any other library. So we |
1638 | avoid the non-standard function stpcpy. In GNU C Library this |
1639 | function is available, though. Also allow the symbol HAVE_STPCPY |
1640 | to be defined. */ |
1641 | #if !_LIBC && !HAVE_STPCPY |
1642 | static char * |
1643 | stpcpy (char *dest, const char *src) |
1644 | { |
1645 | while ((*dest++ = *src++) != '\0') |
1646 | /* Do nothing. */ ; |
1647 | return dest - 1; |
1648 | } |
1649 | #endif |
1650 | |
1651 | #if !_LIBC && !HAVE_MEMPCPY |
1652 | static void * |
1653 | mempcpy (void *dest, const void *src, size_t n) |
1654 | { |
1655 | return (void *) ((char *) memcpy (dest, src, n) + n); |
1656 | } |
1657 | #endif |
1658 | |
1659 | #if !_LIBC && !HAVE_TSEARCH |
1660 | # include "tsearch.c" |
1661 | #endif |
1662 | |
1663 | |
1664 | #ifdef _LIBC |
1665 | /* If we want to free all resources we have to do some work at |
1666 | program's end. */ |
1667 | libc_freeres_fn (free_mem) |
1668 | { |
1669 | void *old; |
1670 | |
1671 | while (_nl_domain_bindings != NULL) |
1672 | { |
1673 | struct binding *oldp = _nl_domain_bindings; |
1674 | _nl_domain_bindings = _nl_domain_bindings->next; |
1675 | if (oldp->dirname != _nl_default_dirname) |
1676 | /* Yes, this is a pointer comparison. */ |
1677 | free (oldp->dirname); |
1678 | free (oldp->codeset); |
1679 | free (oldp); |
1680 | } |
1681 | |
1682 | if (_nl_current_default_domain != _nl_default_default_domain) |
1683 | /* Yes, again a pointer comparison. */ |
1684 | free ((char *) _nl_current_default_domain); |
1685 | |
1686 | /* Remove the search tree with the known translations. */ |
1687 | __tdestroy (root, free); |
1688 | root = NULL; |
1689 | |
1690 | while (transmem_list != NULL) |
1691 | { |
1692 | old = transmem_list; |
1693 | transmem_list = transmem_list->next; |
1694 | free (old); |
1695 | } |
1696 | } |
1697 | #endif |
1698 | |