1 /* $NetBSD: gettext.c,v 1.14 2003/03/09 01:02:34 lukem 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 1.14 2003/03/09 01:02:34 lukem Exp $"); 33 34 #include <sys/types.h> 35 #include <sys/param.h> 36 #include <sys/stat.h> 37 #include <sys/mman.h> 38 #include <sys/uio.h> 39 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 "pathnames.h" 52 53 static const char *lookup_category __P((int)); 54 static const char *split_locale __P((const char *)); 55 static const char *lookup_mofile __P((char *, size_t, const char *, 56 const char *, const char *, const char *, struct domainbinding *)); 57 static u_int32_t flip __P((u_int32_t, u_int32_t)); 58 static int validate __P((void *, struct mohandle *)); 59 static int mapit __P((const char *, struct domainbinding *)); 60 static int unmapit __P((struct domainbinding *)); 61 static const char *lookup_hash __P((const char *, struct domainbinding *)); 62 static const char *lookup_bsearch __P((const char *, struct domainbinding *)); 63 static const char *lookup __P((const char *, struct domainbinding *)); 64 static const char *get_lang_env(const char *); 65 66 /* 67 * shortcut functions. the main implementation resides in dcngettext(). 68 */ 69 char * 70 gettext(msgid) 71 const char *msgid; 72 { 73 74 return dcngettext(NULL, msgid, NULL, 1UL, LC_MESSAGES); 75 } 76 77 char * 78 dgettext(domainname, msgid) 79 const char *domainname; 80 const char *msgid; 81 { 82 83 return dcngettext(domainname, msgid, NULL, 1UL, LC_MESSAGES); 84 } 85 86 char * 87 dcgettext(domainname, msgid, category) 88 const char *domainname; 89 const char *msgid; 90 int category; 91 { 92 93 return dcngettext(domainname, msgid, NULL, 1UL, category); 94 } 95 96 char * 97 ngettext(msgid1, msgid2, n) 98 const char *msgid1; 99 const char *msgid2; 100 unsigned long int n; 101 { 102 103 return dcngettext(NULL, msgid1, msgid2, n, LC_MESSAGES); 104 } 105 106 char * 107 dngettext(domainname, msgid1, msgid2, n) 108 const char *domainname; 109 const char *msgid1; 110 const char *msgid2; 111 unsigned long int n; 112 { 113 114 return dcngettext(domainname, msgid1, msgid2, n, LC_MESSAGES); 115 } 116 117 /* 118 * dcngettext() - 119 * lookup internationalized message on database locale/category/domainname 120 * (like ja_JP.eucJP/LC_MESSAGES/domainname). 121 * if n equals to 1, internationalized message will be looked up for msgid1. 122 * otherwise, message will be looked up for msgid2. 123 * if the lookup fails, the function will return msgid1 or msgid2 as is. 124 * 125 * Even though the return type is "char *", caller should not rewrite the 126 * region pointed to by the return value (should be "const char *", but can't 127 * change it for compatibility with other implementations). 128 * 129 * by default (if domainname == NULL), domainname is taken from the value set 130 * by textdomain(). usually name of the application (like "ls") is used as 131 * domainname. category is usually LC_MESSAGES. 132 * 133 * the code reads in *.mo files generated by GNU gettext. *.mo is a host- 134 * endian encoded file. both endians are supported here, as the files are in 135 * /usr/share/locale! (or we should move those files into /usr/libdata) 136 */ 137 138 static const char * 139 lookup_category(category) 140 int category; 141 { 142 143 switch (category) { 144 case LC_COLLATE: return "LC_COLLATE"; 145 case LC_CTYPE: return "LC_CTYPE"; 146 case LC_MONETARY: return "LC_MONETARY"; 147 case LC_NUMERIC: return "LC_NUMERIC"; 148 case LC_TIME: return "LC_TIME"; 149 case LC_MESSAGES: return "LC_MESSAGES"; 150 } 151 return NULL; 152 } 153 154 /* 155 * XPG syntax: language[_territory[.codeset]][@modifier] 156 * XXX boundary check on "result" is lacking 157 */ 158 static const char * 159 split_locale(lname) 160 const char *lname; 161 { 162 char buf[BUFSIZ], tmp[BUFSIZ]; 163 char *l, *t, *c, *m; 164 static char result[BUFSIZ]; 165 166 memset(result, 0, sizeof(result)); 167 168 if (strlen(lname) + 1 > sizeof(buf)) { 169 fail: 170 return lname; 171 } 172 173 strlcpy(buf, lname, sizeof(buf)); 174 m = strrchr(buf, '@'); 175 if (m) 176 *m++ = '\0'; 177 c = strrchr(buf, '.'); 178 if (c) 179 *c++ = '\0'; 180 t = strrchr(buf, '_'); 181 if (t) 182 *t++ = '\0'; 183 l = buf; 184 if (strlen(l) == 0) 185 goto fail; 186 if (c && !t) 187 goto fail; 188 189 if (m) { 190 if (t) { 191 if (c) { 192 snprintf(tmp, sizeof(tmp), "%s_%s.%s@%s", 193 l, t, c, m); 194 strlcat(result, tmp, sizeof(result)); 195 strlcat(result, ":", sizeof(result)); 196 } 197 snprintf(tmp, sizeof(tmp), "%s_%s@%s", l, t, m); 198 strlcat(result, tmp, sizeof(result)); 199 strlcat(result, ":", sizeof(result)); 200 } 201 snprintf(tmp, sizeof(tmp), "%s@%s", l, m); 202 strlcat(result, tmp, sizeof(result)); 203 strlcat(result, ":", sizeof(result)); 204 } 205 if (t) { 206 if (c) { 207 snprintf(tmp, sizeof(tmp), "%s_%s.%s", l, t, c); 208 strlcat(result, tmp, sizeof(result)); 209 strlcat(result, ":", sizeof(result)); 210 } 211 snprintf(tmp, sizeof(tmp), "%s_%s", l, t); 212 strlcat(result, tmp, sizeof(result)); 213 strlcat(result, ":", sizeof(result)); 214 } 215 strlcat(result, l, sizeof(result)); 216 217 return result; 218 } 219 220 static const char * 221 lookup_mofile(buf, len, dir, lpath, category, domainname, db) 222 char *buf; 223 size_t len; 224 const char *dir; 225 const char *lpath; /* list of locales to be tried */ 226 const char *category; 227 const char *domainname; 228 struct domainbinding *db; 229 { 230 struct stat st; 231 char *p, *q; 232 char lpath_tmp[BUFSIZ]; 233 234 strlcpy(lpath_tmp, lpath, sizeof(lpath_tmp)); 235 q = lpath_tmp; 236 /* CONSTCOND */ 237 while (1) { 238 p = strsep(&q, ":"); 239 if (!p) 240 break; 241 if (!*p) 242 continue; 243 244 /* don't mess with default locales */ 245 if (strcmp(p, "C") == 0 || strcmp(p, "POSIX") == 0) 246 return NULL; 247 248 /* validate pathname */ 249 if (strchr(p, '/') || strchr(category, '/')) 250 continue; 251 #if 1 /*?*/ 252 if (strchr(domainname, '/')) 253 continue; 254 #endif 255 256 snprintf(buf, len, "%s/%s/%s/%s.mo", dir, p, 257 category, domainname); 258 if (stat(buf, &st) < 0) 259 continue; 260 if ((st.st_mode & S_IFMT) != S_IFREG) 261 continue; 262 263 if (mapit(buf, db) == 0) 264 return buf; 265 } 266 267 return NULL; 268 } 269 270 static u_int32_t 271 flip(v, magic) 272 u_int32_t v; 273 u_int32_t magic; 274 { 275 276 if (magic == MO_MAGIC) 277 return v; 278 else if (magic == MO_MAGIC_SWAPPED) { 279 v = ((v >> 24) & 0xff) | ((v >> 8) & 0xff00) | 280 ((v << 8) & 0xff0000) | ((v << 24) & 0xff000000); 281 return v; 282 } else { 283 abort(); 284 /*NOTREACHED*/ 285 } 286 } 287 288 static int 289 validate(arg, mohandle) 290 void *arg; 291 struct mohandle *mohandle; 292 { 293 char *p; 294 295 p = (char *)arg; 296 if (p < (char *)mohandle->addr || 297 p > (char *)mohandle->addr + mohandle->len) 298 return 0; 299 else 300 return 1; 301 } 302 303 int 304 mapit(path, db) 305 const char *path; 306 struct domainbinding *db; 307 { 308 int fd; 309 struct stat st; 310 char *base; 311 u_int32_t magic, revision; 312 struct moentry *otable, *ttable; 313 struct moentry_h *p; 314 struct mo *mo; 315 size_t l; 316 int i; 317 char *v; 318 struct mohandle *mohandle = &db->mohandle; 319 320 if (mohandle->addr && mohandle->addr != MAP_FAILED && 321 mohandle->mo.mo_magic) 322 return 0; /*already opened*/ 323 324 unmapit(db); 325 326 #if 0 327 if (secure_path(path) != 0) 328 goto fail; 329 #endif 330 if (stat(path, &st) < 0) 331 goto fail; 332 if ((st.st_mode & S_IFMT) != S_IFREG || st.st_size > GETTEXT_MMAP_MAX) 333 goto fail; 334 fd = open(path, O_RDONLY); 335 if (fd < 0) 336 goto fail; 337 if (read(fd, &magic, sizeof(magic)) != sizeof(magic) || 338 (magic != MO_MAGIC && magic != MO_MAGIC_SWAPPED)) { 339 close(fd); 340 goto fail; 341 } 342 if (read(fd, &revision, sizeof(revision)) != sizeof(revision) || 343 flip(revision, magic) != MO_REVISION) { 344 close(fd); 345 goto fail; 346 } 347 mohandle->addr = mmap(NULL, (size_t)st.st_size, PROT_READ, 348 MAP_FILE | MAP_SHARED, fd, (off_t)0); 349 if (!mohandle->addr || mohandle->addr == MAP_FAILED) { 350 close(fd); 351 goto fail; 352 } 353 close(fd); 354 mohandle->len = (size_t)st.st_size; 355 356 base = mohandle->addr; 357 mo = (struct mo *)mohandle->addr; 358 359 /* flip endian. do not flip magic number! */ 360 mohandle->mo.mo_magic = mo->mo_magic; 361 mohandle->mo.mo_revision = flip(mo->mo_revision, magic); 362 mohandle->mo.mo_nstring = flip(mo->mo_nstring, magic); 363 364 /* validate otable/ttable */ 365 otable = (struct moentry *)(base + flip(mo->mo_otable, magic)); 366 ttable = (struct moentry *)(base + flip(mo->mo_ttable, magic)); 367 if (!validate(otable, mohandle) || 368 !validate(&otable[mohandle->mo.mo_nstring], mohandle)) { 369 unmapit(db); 370 goto fail; 371 } 372 if (!validate(ttable, mohandle) || 373 !validate(&ttable[mohandle->mo.mo_nstring], mohandle)) { 374 unmapit(db); 375 goto fail; 376 } 377 378 /* allocate [ot]table, and convert to normal pointer representation. */ 379 l = sizeof(struct moentry_h) * mohandle->mo.mo_nstring; 380 mohandle->mo.mo_otable = (struct moentry_h *)malloc(l); 381 if (!mohandle->mo.mo_otable) { 382 unmapit(db); 383 goto fail; 384 } 385 mohandle->mo.mo_ttable = (struct moentry_h *)malloc(l); 386 if (!mohandle->mo.mo_ttable) { 387 unmapit(db); 388 goto fail; 389 } 390 p = mohandle->mo.mo_otable; 391 for (i = 0; i < mohandle->mo.mo_nstring; i++) { 392 p[i].len = flip(otable[i].len, magic); 393 p[i].off = base + flip(otable[i].off, magic); 394 395 if (!validate(p[i].off, mohandle) || 396 !validate(p[i].off + p[i].len + 1, mohandle)) { 397 unmapit(db); 398 goto fail; 399 } 400 } 401 p = mohandle->mo.mo_ttable; 402 for (i = 0; i < mohandle->mo.mo_nstring; i++) { 403 p[i].len = flip(ttable[i].len, magic); 404 p[i].off = base + flip(ttable[i].off, magic); 405 406 if (!validate(p[i].off, mohandle) || 407 !validate(p[i].off + p[i].len + 1, mohandle)) { 408 unmapit(db); 409 goto fail; 410 } 411 } 412 413 /* grab MIME-header and charset field */ 414 mohandle->mo.mo_header = lookup("", db); 415 if (mohandle->mo.mo_header) 416 v = strstr(mohandle->mo.mo_header, "charset="); 417 else 418 v = NULL; 419 if (v) { 420 mohandle->mo.mo_charset = strdup(v + 8); 421 if (!mohandle->mo.mo_charset) 422 goto fail; 423 v = strchr(mohandle->mo.mo_charset, '\n'); 424 if (v) 425 *v = '\0'; 426 } 427 428 /* 429 * XXX check charset, reject it if we are unable to support the charset 430 * with the current locale. 431 * for example, if we are using euc-jp locale and we are looking at 432 * *.mo file encoded by euc-kr (charset=euc-kr), we should reject 433 * the *.mo file as we cannot support it. 434 */ 435 436 return 0; 437 438 fail: 439 return -1; 440 } 441 442 static int 443 unmapit(db) 444 struct domainbinding *db; 445 { 446 struct mohandle *mohandle = &db->mohandle; 447 448 /* unmap if there's already mapped region */ 449 if (mohandle->addr && mohandle->addr != MAP_FAILED) 450 munmap(mohandle->addr, mohandle->len); 451 mohandle->addr = NULL; 452 if (mohandle->mo.mo_otable) 453 free(mohandle->mo.mo_otable); 454 if (mohandle->mo.mo_ttable) 455 free(mohandle->mo.mo_ttable); 456 if (mohandle->mo.mo_charset) 457 free(mohandle->mo.mo_charset); 458 memset(&mohandle->mo, 0, sizeof(mohandle->mo)); 459 return 0; 460 } 461 462 /* ARGSUSED */ 463 static const char * 464 lookup_hash(msgid, db) 465 const char *msgid; 466 struct domainbinding *db; 467 { 468 469 /* 470 * XXX should try a hashed lookup here, but to do so, we need to 471 * look inside the GPL'ed *.c and re-implement... 472 */ 473 return NULL; 474 } 475 476 static const char * 477 lookup_bsearch(msgid, db) 478 const char *msgid; 479 struct domainbinding *db; 480 { 481 int top, bottom, middle, omiddle; 482 int n; 483 struct mohandle *mohandle = &db->mohandle; 484 485 top = 0; 486 bottom = mohandle->mo.mo_nstring; 487 omiddle = -1; 488 /* CONSTCOND */ 489 while (1) { 490 if (top > bottom) 491 break; 492 middle = (top + bottom) / 2; 493 /* avoid possible infinite loop, when the data is not sorted */ 494 if (omiddle == middle) 495 break; 496 if (middle < 0 || middle >= mohandle->mo.mo_nstring) 497 break; 498 499 n = strcmp(msgid, mohandle->mo.mo_otable[middle].off); 500 if (n == 0) 501 return (const char *)mohandle->mo.mo_ttable[middle].off; 502 else if (n < 0) 503 bottom = middle; 504 else 505 top = middle; 506 omiddle = middle; 507 } 508 509 return NULL; 510 } 511 512 static const char * 513 lookup(msgid, db) 514 const char *msgid; 515 struct domainbinding *db; 516 { 517 const char *v; 518 519 v = lookup_hash(msgid, db); 520 if (v) 521 return v; 522 523 return lookup_bsearch(msgid, db); 524 } 525 526 static const char *get_lang_env(const char *category_name) 527 { 528 const char *lang; 529 530 /* 1. see LANGUAGE variable first. */ 531 lang = getenv("LANGUAGE"); 532 if (lang) 533 return lang; 534 535 /* 2. if LANGUAGE isn't set, see LC_ALL, LC_xxx, LANG. */ 536 lang = getenv("LC_ALL"); 537 if (!lang) 538 lang = getenv(category_name); 539 if (!lang) 540 lang = getenv("LANG"); 541 542 if (!lang) 543 return 0; /* error */ 544 545 return split_locale(lang); 546 } 547 548 char * 549 dcngettext(domainname, msgid1, msgid2, n, category) 550 const char *domainname; 551 const char *msgid1; 552 const char *msgid2; 553 unsigned long int n; 554 int category; 555 { 556 const char *msgid; 557 char path[PATH_MAX]; 558 const char *lpath; 559 static char olpath[PATH_MAX]; 560 const char *cname = NULL; 561 const char *v; 562 static char *ocname = NULL; 563 static char *odomainname = NULL; 564 struct domainbinding *db; 565 566 msgid = (n == 1) ? msgid1 : msgid2; 567 if (msgid == NULL) 568 return NULL; 569 570 if (!domainname) 571 domainname = __current_domainname; 572 cname = lookup_category(category); 573 if (!domainname || !cname) 574 goto fail; 575 576 lpath = get_lang_env(cname); 577 if (!lpath) 578 goto fail; 579 580 for (db = __bindings; db; db = db->next) 581 if (strcmp(db->domainname, domainname) == 0) 582 break; 583 if (!db) { 584 if (!bindtextdomain(domainname, _PATH_TEXTDOMAIN)) 585 goto fail; 586 db = __bindings; 587 } 588 589 /* resolve relative path */ 590 /* XXX not necessary? */ 591 if (db->path[0] != '/') { 592 char buf[PATH_MAX]; 593 594 if (getcwd(buf, sizeof(buf)) == 0) 595 goto fail; 596 if (strlcat(buf, "/", sizeof(buf)) >= sizeof(buf)) 597 goto fail; 598 if (strlcat(buf, db->path, sizeof(buf)) >= sizeof(buf)) 599 goto fail; 600 strcpy(db->path, buf); 601 } 602 603 /* don't bother looking it up if the values are the same */ 604 if (odomainname && strcmp(domainname, odomainname) == 0 && 605 ocname && strcmp(cname, ocname) == 0 && strcmp(lpath, olpath) == 0 && 606 db->mohandle.mo.mo_magic) 607 goto found; 608 609 /* try to find appropriate file, from $LANGUAGE */ 610 if (lookup_mofile(path, sizeof(path), db->path, lpath, cname, 611 domainname, db) == NULL) 612 goto fail; 613 614 if (odomainname) 615 free(odomainname); 616 if (ocname) 617 free(ocname); 618 odomainname = strdup(domainname); 619 ocname = strdup(cname); 620 if (!odomainname || !ocname) { 621 if (odomainname) 622 free(odomainname); 623 if (ocname) 624 free(ocname); 625 odomainname = ocname = NULL; 626 } 627 else 628 strlcpy(olpath, lpath, sizeof(olpath)); 629 630 found: 631 v = lookup(msgid, db); 632 if (v) { 633 /* 634 * XXX call iconv() here, if translated text is encoded 635 * differently from currently-selected encoding (locale). 636 * look at Content-type header in *.mo file, in string obtained 637 * by gettext(""). 638 */ 639 640 /* 641 * Given the amount of printf-format security issues, it may 642 * be a good idea to validate if the original msgid and the 643 * translated message format string carry the same printf-like 644 * format identifiers. 645 */ 646 647 msgid = v; 648 } 649 650 fail: 651 /* LINTED const cast */ 652 return (char *)msgid; 653 } 654