1 /* $NetBSD: gettext.c,v 1.11 2001/12/09 11:11:01 yamt 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 #if defined(LIBC_SCCS) && !defined(lint) 33 __RCSID("$NetBSD: gettext.c,v 1.11 2001/12/09 11:11:01 yamt Exp $"); 34 #endif /* LIBC_SCCS and not lint */ 35 36 #include <sys/types.h> 37 #include <sys/param.h> 38 #include <sys/stat.h> 39 #include <sys/mman.h> 40 #include <sys/uio.h> 41 42 #include <fcntl.h> 43 #include <stdio.h> 44 #include <stdlib.h> 45 #include <unistd.h> 46 #include <string.h> 47 #if 0 48 #include <util.h> 49 #endif 50 #include <libintl.h> 51 #include <locale.h> 52 #include "libintl_local.h" 53 #include "pathnames.h" 54 55 static const char *lookup_category __P((int)); 56 static const char *split_locale __P((const char *)); 57 static const char *lookup_mofile __P((char *, size_t, const char *, 58 const char *, const char *, const char *, struct domainbinding *)); 59 static u_int32_t flip __P((u_int32_t, u_int32_t)); 60 static int validate __P((void *, struct mohandle *)); 61 static int mapit __P((const char *, struct domainbinding *)); 62 static int unmapit __P((struct domainbinding *)); 63 static const char *lookup_hash __P((const char *, struct domainbinding *)); 64 static const char *lookup_bsearch __P((const char *, struct domainbinding *)); 65 static const char *lookup __P((const char *, struct domainbinding *)); 66 static const char *get_lang_env(const char *); 67 68 /* 69 * shortcut functions. the main implementation resides in dcngettext(). 70 */ 71 char * 72 gettext(msgid) 73 const char *msgid; 74 { 75 76 return dcngettext(NULL, msgid, NULL, 1UL, LC_MESSAGES); 77 } 78 79 char * 80 dgettext(domainname, msgid) 81 const char *domainname; 82 const char *msgid; 83 { 84 85 return dcngettext(domainname, msgid, NULL, 1UL, LC_MESSAGES); 86 } 87 88 char * 89 dcgettext(domainname, msgid, category) 90 const char *domainname; 91 const char *msgid; 92 int category; 93 { 94 95 return dcngettext(domainname, msgid, NULL, 1UL, category); 96 } 97 98 char * 99 ngettext(msgid1, msgid2, n) 100 const char *msgid1; 101 const char *msgid2; 102 unsigned long int n; 103 { 104 105 return dcngettext(NULL, msgid1, msgid2, n, LC_MESSAGES); 106 } 107 108 char * 109 dngettext(domainname, msgid1, msgid2, n) 110 const char *domainname; 111 const char *msgid1; 112 const char *msgid2; 113 unsigned long int n; 114 { 115 116 return dcngettext(domainname, msgid1, msgid2, n, LC_MESSAGES); 117 } 118 119 /* 120 * dcngettext() - 121 * lookup internationalized message on database locale/category/domainname 122 * (like ja_JP.eucJP/LC_MESSAGES/domainname). 123 * if n equals to 1, internationalized message will be looked up for msgid1. 124 * otherwise, message will be looked up for msgid2. 125 * if the lookup fails, the function will return msgid1 or msgid2 as is. 126 * 127 * Even though the return type is "char *", caller should not rewrite the 128 * region pointed to by the return value (should be "const char *", but can't 129 * change it for compatibility with other implementations). 130 * 131 * by default (if domainname == NULL), domainname is taken from the value set 132 * by textdomain(). usually name of the application (like "ls") is used as 133 * domainname. category is usually LC_MESSAGES. 134 * 135 * the code reads in *.mo files generated by GNU gettext. *.mo is a host- 136 * endian encoded file. both endians are supported here, as the files are in 137 * /usr/share/locale! (or we should move those files into /usr/libdata) 138 */ 139 140 static const char * 141 lookup_category(category) 142 int category; 143 { 144 145 switch (category) { 146 case LC_COLLATE: return "LC_COLLATE"; 147 case LC_CTYPE: return "LC_CTYPE"; 148 case LC_MONETARY: return "LC_MONETARY"; 149 case LC_NUMERIC: return "LC_NUMERIC"; 150 case LC_TIME: return "LC_TIME"; 151 case LC_MESSAGES: return "LC_MESSAGES"; 152 } 153 return NULL; 154 } 155 156 /* 157 * XPG syntax: language[_territory[.codeset]][@modifier] 158 * XXX boundary check on "result" is lacking 159 */ 160 static const char * 161 split_locale(lname) 162 const char *lname; 163 { 164 char buf[BUFSIZ], tmp[BUFSIZ]; 165 char *l, *t, *c, *m; 166 static char result[BUFSIZ]; 167 168 memset(result, 0, sizeof(result)); 169 170 if (strlen(lname) + 1 > sizeof(buf)) { 171 fail: 172 return lname; 173 } 174 175 strlcpy(buf, lname, sizeof(buf)); 176 m = strrchr(buf, '@'); 177 if (m) 178 *m++ = '\0'; 179 c = strrchr(buf, '.'); 180 if (c) 181 *c++ = '\0'; 182 t = strrchr(buf, '_'); 183 if (t) 184 *t++ = '\0'; 185 l = buf; 186 if (strlen(l) == 0) 187 goto fail; 188 if (c && !t) 189 goto fail; 190 191 if (m) { 192 if (t) { 193 if (c) { 194 snprintf(tmp, sizeof(tmp), "%s_%s.%s@%s", 195 l, t, c, m); 196 strlcat(result, tmp, sizeof(result)); 197 strlcat(result, ":", sizeof(result)); 198 } 199 snprintf(tmp, sizeof(tmp), "%s_%s@%s", l, t, m); 200 strlcat(result, tmp, sizeof(result)); 201 strlcat(result, ":", sizeof(result)); 202 } 203 snprintf(tmp, sizeof(tmp), "%s@%s", l, m); 204 strlcat(result, tmp, sizeof(result)); 205 strlcat(result, ":", sizeof(result)); 206 } 207 if (t) { 208 if (c) { 209 snprintf(tmp, sizeof(tmp), "%s_%s.%s", l, t, c); 210 strlcat(result, tmp, sizeof(result)); 211 strlcat(result, ":", sizeof(result)); 212 } 213 strlcat(result, tmp, sizeof(result)); 214 strlcat(result, ":", sizeof(result)); 215 } 216 strlcat(result, l, sizeof(result)); 217 218 return result; 219 } 220 221 static const char * 222 lookup_mofile(buf, len, dir, lpath, category, domainname, db) 223 char *buf; 224 size_t len; 225 const char *dir; 226 const char *lpath; /* list of locales to be tried */ 227 const char *category; 228 const char *domainname; 229 struct domainbinding *db; 230 { 231 struct stat st; 232 char *p, *q; 233 char lpath_tmp[BUFSIZ]; 234 235 strlcpy(lpath_tmp, lpath, sizeof(lpath_tmp)); 236 q = lpath_tmp; 237 /* CONSTCOND */ 238 while (1) { 239 p = strsep(&q, ":"); 240 if (!p) 241 break; 242 if (!*p) 243 continue; 244 245 /* don't mess with default locales */ 246 if (strcmp(p, "C") == 0 || strcmp(p, "POSIX") == 0) 247 return NULL; 248 249 /* validate pathname */ 250 if (strchr(p, '/') || strchr(category, '/')) 251 continue; 252 #if 1 /*?*/ 253 if (strchr(domainname, '/')) 254 continue; 255 #endif 256 257 snprintf(buf, len, "%s/%s/%s/%s.mo", dir, p, 258 category, domainname); 259 if (stat(buf, &st) < 0) 260 continue; 261 if ((st.st_mode & S_IFMT) != S_IFREG) 262 continue; 263 264 if (mapit(buf, db) == 0) 265 return buf; 266 } 267 268 return NULL; 269 } 270 271 static u_int32_t 272 flip(v, magic) 273 u_int32_t v; 274 u_int32_t magic; 275 { 276 277 if (magic == MO_MAGIC) 278 return v; 279 else if (magic == MO_MAGIC_SWAPPED) { 280 v = ((v >> 24) & 0xff) | ((v >> 8) & 0xff00) | 281 ((v << 8) & 0xff0000) | ((v << 24) & 0xff000000); 282 return v; 283 } else { 284 abort(); 285 /*NOTREACHED*/ 286 } 287 } 288 289 static int 290 validate(arg, mohandle) 291 void *arg; 292 struct mohandle *mohandle; 293 { 294 char *p; 295 296 p = (char *)arg; 297 if (p < (char *)mohandle->addr || 298 p > (char *)mohandle->addr + mohandle->len) 299 return 0; 300 else 301 return 1; 302 } 303 304 int 305 mapit(path, db) 306 const char *path; 307 struct domainbinding *db; 308 { 309 int fd; 310 struct stat st; 311 char *base; 312 u_int32_t magic, revision; 313 struct moentry *otable, *ttable; 314 struct moentry_h *p; 315 struct mo *mo; 316 size_t l; 317 int i; 318 char *v; 319 struct mohandle *mohandle = &db->mohandle; 320 321 if (mohandle->addr && mohandle->addr != MAP_FAILED && 322 mohandle->mo.mo_magic) 323 return 0; /*already opened*/ 324 325 unmapit(db); 326 327 #if 0 328 if (secure_path(path) != 0) 329 goto fail; 330 #endif 331 if (stat(path, &st) < 0) 332 goto fail; 333 if ((st.st_mode & S_IFMT) != S_IFREG || st.st_size > GETTEXT_MMAP_MAX) 334 goto fail; 335 fd = open(path, O_RDONLY); 336 if (fd < 0) 337 goto fail; 338 if (read(fd, &magic, sizeof(magic)) != sizeof(magic) || 339 (magic != MO_MAGIC && magic != MO_MAGIC_SWAPPED)) { 340 close(fd); 341 goto fail; 342 } 343 if (read(fd, &revision, sizeof(revision)) != sizeof(revision) || 344 flip(revision, magic) != MO_REVISION) { 345 close(fd); 346 goto fail; 347 } 348 mohandle->addr = mmap(NULL, (size_t)st.st_size, PROT_READ, 349 MAP_FILE | MAP_SHARED, fd, (off_t)0); 350 if (!mohandle->addr || mohandle->addr == MAP_FAILED) { 351 close(fd); 352 goto fail; 353 } 354 close(fd); 355 mohandle->len = (size_t)st.st_size; 356 357 base = mohandle->addr; 358 mo = (struct mo *)mohandle->addr; 359 360 /* flip endian. do not flip magic number! */ 361 mohandle->mo.mo_magic = mo->mo_magic; 362 mohandle->mo.mo_revision = flip(mo->mo_revision, magic); 363 mohandle->mo.mo_nstring = flip(mo->mo_nstring, magic); 364 365 /* validate otable/ttable */ 366 otable = (struct moentry *)(base + flip(mo->mo_otable, magic)); 367 ttable = (struct moentry *)(base + flip(mo->mo_ttable, magic)); 368 if (!validate(otable, mohandle) || 369 !validate(&otable[mohandle->mo.mo_nstring], mohandle)) { 370 unmapit(db); 371 goto fail; 372 } 373 if (!validate(ttable, mohandle) || 374 !validate(&ttable[mohandle->mo.mo_nstring], mohandle)) { 375 unmapit(db); 376 goto fail; 377 } 378 379 /* allocate [ot]table, and convert to normal pointer representation. */ 380 l = sizeof(struct moentry_h) * mohandle->mo.mo_nstring; 381 mohandle->mo.mo_otable = (struct moentry_h *)malloc(l); 382 if (!mohandle->mo.mo_otable) { 383 unmapit(db); 384 goto fail; 385 } 386 mohandle->mo.mo_ttable = (struct moentry_h *)malloc(l); 387 if (!mohandle->mo.mo_ttable) { 388 unmapit(db); 389 goto fail; 390 } 391 p = mohandle->mo.mo_otable; 392 for (i = 0; i < mohandle->mo.mo_nstring; i++) { 393 p[i].len = flip(otable[i].len, magic); 394 p[i].off = base + flip(otable[i].off, magic); 395 396 if (!validate(p[i].off, mohandle) || 397 !validate(p[i].off + p[i].len + 1, mohandle)) { 398 unmapit(db); 399 goto fail; 400 } 401 } 402 p = mohandle->mo.mo_ttable; 403 for (i = 0; i < mohandle->mo.mo_nstring; i++) { 404 p[i].len = flip(ttable[i].len, magic); 405 p[i].off = base + flip(ttable[i].off, magic); 406 407 if (!validate(p[i].off, mohandle) || 408 !validate(p[i].off + p[i].len + 1, mohandle)) { 409 unmapit(db); 410 goto fail; 411 } 412 } 413 414 /* grab MIME-header and charset field */ 415 mohandle->mo.mo_header = lookup("", db); 416 if (mohandle->mo.mo_header) 417 v = strstr(mohandle->mo.mo_header, "charset="); 418 else 419 v = NULL; 420 if (v) { 421 mohandle->mo.mo_charset = strdup(v + 8); 422 if (!mohandle->mo.mo_charset) 423 goto fail; 424 v = strchr(mohandle->mo.mo_charset, '\n'); 425 if (v) 426 *v = '\0'; 427 } 428 429 /* 430 * XXX check charset, reject it if we are unable to support the charset 431 * with the current locale. 432 * for example, if we are using euc-jp locale and we are looking at 433 * *.mo file encoded by euc-kr (charset=euc-kr), we should reject 434 * the *.mo file as we cannot support it. 435 */ 436 437 return 0; 438 439 fail: 440 return -1; 441 } 442 443 static int 444 unmapit(db) 445 struct domainbinding *db; 446 { 447 struct mohandle *mohandle = &db->mohandle; 448 449 /* unmap if there's already mapped region */ 450 if (mohandle->addr && mohandle->addr != MAP_FAILED) 451 munmap(mohandle->addr, mohandle->len); 452 mohandle->addr = NULL; 453 if (mohandle->mo.mo_otable) 454 free(mohandle->mo.mo_otable); 455 if (mohandle->mo.mo_ttable) 456 free(mohandle->mo.mo_ttable); 457 if (mohandle->mo.mo_charset) 458 free(mohandle->mo.mo_charset); 459 memset(&mohandle->mo, 0, sizeof(mohandle->mo)); 460 return 0; 461 } 462 463 /* ARGSUSED */ 464 static const char * 465 lookup_hash(msgid, db) 466 const char *msgid; 467 struct domainbinding *db; 468 { 469 470 /* 471 * XXX should try a hashed lookup here, but to do so, we need to 472 * look inside the GPL'ed *.c and re-implement... 473 */ 474 return NULL; 475 } 476 477 static const char * 478 lookup_bsearch(msgid, db) 479 const char *msgid; 480 struct domainbinding *db; 481 { 482 int top, bottom, middle, omiddle; 483 int n; 484 struct mohandle *mohandle = &db->mohandle; 485 486 top = 0; 487 bottom = mohandle->mo.mo_nstring; 488 omiddle = -1; 489 /* CONSTCOND */ 490 while (1) { 491 if (top > bottom) 492 break; 493 middle = (top + bottom) / 2; 494 /* avoid possible infinite loop, when the data is not sorted */ 495 if (omiddle == middle) 496 break; 497 if (middle < 0 || middle >= mohandle->mo.mo_nstring) 498 break; 499 500 n = strcmp(msgid, mohandle->mo.mo_otable[middle].off); 501 if (n == 0) 502 return (const char *)mohandle->mo.mo_ttable[middle].off; 503 else if (n < 0) 504 bottom = middle; 505 else 506 top = middle; 507 omiddle = middle; 508 } 509 510 return NULL; 511 } 512 513 static const char * 514 lookup(msgid, db) 515 const char *msgid; 516 struct domainbinding *db; 517 { 518 const char *v; 519 520 v = lookup_hash(msgid, db); 521 if (v) 522 return v; 523 524 return lookup_bsearch(msgid, db); 525 } 526 527 static const char *get_lang_env(const char *category_name) 528 { 529 const char *lang; 530 531 /* 1. see LANGUAGE variable first. */ 532 lang = getenv("LANGUAGE"); 533 if (lang) 534 return lang; 535 536 /* 2. if LANGUAGE isn't set, see LC_ALL, LC_xxx, LANG. */ 537 lang = getenv(category_name); 538 if (!lang) 539 lang = getenv("LC_ALL"); 540 if (!lang) 541 lang = getenv("LANG"); 542 543 if (!lang) 544 return 0; /* error */ 545 546 return split_locale(lang); 547 } 548 549 char * 550 dcngettext(domainname, msgid1, msgid2, n, category) 551 const char *domainname; 552 const char *msgid1; 553 const char *msgid2; 554 unsigned long int n; 555 int category; 556 { 557 const char *msgid; 558 char path[PATH_MAX]; 559 const char *lpath; 560 static char olpath[PATH_MAX]; 561 const char *cname = NULL; 562 const char *v; 563 static char *ocname = NULL; 564 static char *odomainname = NULL; 565 struct domainbinding *db; 566 567 msgid = (n == 1) ? msgid1 : msgid2; 568 if (msgid == NULL) 569 return NULL; 570 571 if (!domainname) 572 domainname = __current_domainname; 573 cname = lookup_category(category); 574 if (!domainname || !cname) 575 goto fail; 576 577 lpath = get_lang_env(cname); 578 if (!lpath) 579 goto fail; 580 581 for (db = __bindings; db; db = db->next) 582 if (strcmp(db->domainname, domainname) == 0) 583 break; 584 if (!db) { 585 if (!bindtextdomain(domainname, _PATH_TEXTDOMAIN)) 586 goto fail; 587 db = __bindings; 588 } 589 590 /* resolve relative path */ 591 /* XXX not necessary? */ 592 if (db->path[0] != '/') { 593 char buf[PATH_MAX]; 594 595 if (getcwd(buf, sizeof(buf)) == 0) 596 goto fail; 597 if (strlcat(buf, "/", sizeof(buf)) >= sizeof(buf)) 598 goto fail; 599 if (strlcat(buf, db->path, sizeof(buf)) >= sizeof(buf)) 600 goto fail; 601 strcpy(db->path, buf); 602 } 603 604 /* don't bother looking it up if the values are the same */ 605 if (odomainname && strcmp(domainname, odomainname) == 0 && 606 ocname && strcmp(cname, ocname) == 0 && strcmp(lpath, olpath) == 0 && 607 db->mohandle.mo.mo_magic) 608 goto found; 609 610 /* try to find appropriate file, from $LANGUAGE */ 611 if (lookup_mofile(path, sizeof(path), db->path, lpath, cname, 612 domainname, db) == NULL) 613 goto fail; 614 615 if (odomainname) 616 free(odomainname); 617 if (ocname) 618 free(ocname); 619 odomainname = strdup(domainname); 620 ocname = strdup(cname); 621 if (!odomainname || !ocname) { 622 if (odomainname) 623 free(odomainname); 624 if (ocname) 625 free(ocname); 626 odomainname = ocname = NULL; 627 } 628 else 629 strlcpy(olpath, lpath, sizeof(olpath)); 630 631 found: 632 v = lookup(msgid, db); 633 if (v) { 634 /* 635 * XXX call iconv() here, if translated text is encoded 636 * differently from currently-selected encoding (locale). 637 * look at Content-type header in *.mo file, in string obtained 638 * by gettext(""). 639 */ 640 641 /* 642 * Given the amount of printf-format security issues, it may 643 * be a good idea to validate if the original msgid and the 644 * translated message format string carry the same printf-like 645 * format identifiers. 646 */ 647 648 msgid = v; 649 } 650 651 fail: 652 /* LINTED const cast */ 653 return (char *)msgid; 654 } 655