1 /* -*- buffer-read-only: t -*- vi: set ro: */ 2 /* DO NOT EDIT! GENERATED AUTOMATICALLY! */ 3 /* Determine a canonical name for the current locale's character encoding. 4 5 Copyright (C) 2000-2006, 2008-2010 Free Software Foundation, Inc. 6 7 This program is free software; you can redistribute it and/or modify 8 it under the terms of the GNU General Public License as published by 9 the Free Software Foundation; either version 3, or (at your option) 10 any later version. 11 12 This program is distributed in the hope that it will be useful, 13 but WITHOUT ANY WARRANTY; without even the implied warranty of 14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 15 GNU General Public License for more details. 16 17 You should have received a copy of the GNU General Public License along 18 with this program; if not, write to the Free Software Foundation, 19 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ 20 21 /* Written by Bruno Haible <bruno@clisp.org>. */ 22 23 #include <config.h> 24 25 /* Specification. */ 26 #include "localcharset.h" 27 28 #include <fcntl.h> 29 #include <stddef.h> 30 #include <stdio.h> 31 #include <string.h> 32 #include <stdlib.h> 33 34 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET 35 # define DARWIN7 /* Darwin 7 or newer, i.e. MacOS X 10.3 or newer */ 36 #endif 37 38 #if defined _WIN32 || defined __WIN32__ 39 # define WIN32_NATIVE 40 #endif 41 42 #if defined __EMX__ 43 /* Assume EMX program runs on OS/2, even if compiled under DOS. */ 44 # ifndef OS2 45 # define OS2 46 # endif 47 #endif 48 49 #if !defined WIN32_NATIVE 50 # include <unistd.h> 51 # if HAVE_LANGINFO_CODESET 52 # include <langinfo.h> 53 # else 54 # if 0 /* see comment below */ 55 # include <locale.h> 56 # endif 57 # endif 58 # ifdef __CYGWIN__ 59 # define WIN32_LEAN_AND_MEAN 60 # include <windows.h> 61 # endif 62 #elif defined WIN32_NATIVE 63 # define WIN32_LEAN_AND_MEAN 64 # include <windows.h> 65 #endif 66 #if defined OS2 67 # define INCL_DOS 68 # include <os2.h> 69 #endif 70 71 #if ENABLE_RELOCATABLE 72 # include "relocatable.h" 73 #else 74 # define relocate(pathname) (pathname) 75 #endif 76 77 /* Get LIBDIR. */ 78 #ifndef LIBDIR 79 # include "configmake.h" 80 #endif 81 82 /* Define O_NOFOLLOW to 0 on platforms where it does not exist. */ 83 #ifndef O_NOFOLLOW 84 # define O_NOFOLLOW 0 85 #endif 86 87 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__ 88 /* Win32, Cygwin, OS/2, DOS */ 89 # define ISSLASH(C) ((C) == '/' || (C) == '\\') 90 #endif 91 92 #ifndef DIRECTORY_SEPARATOR 93 # define DIRECTORY_SEPARATOR '/' 94 #endif 95 96 #ifndef ISSLASH 97 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR) 98 #endif 99 100 #if HAVE_DECL_GETC_UNLOCKED 101 # undef getc 102 # define getc getc_unlocked 103 #endif 104 105 /* The following static variable is declared 'volatile' to avoid a 106 possible multithread problem in the function get_charset_aliases. If we 107 are running in a threaded environment, and if two threads initialize 108 'charset_aliases' simultaneously, both will produce the same value, 109 and everything will be ok if the two assignments to 'charset_aliases' 110 are atomic. But I don't know what will happen if the two assignments mix. */ 111 #if __STDC__ != 1 112 # define volatile /* empty */ 113 #endif 114 /* Pointer to the contents of the charset.alias file, if it has already been 115 read, else NULL. Its format is: 116 ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0' */ 117 static const char * volatile charset_aliases; 118 119 /* Return a pointer to the contents of the charset.alias file. */ 120 static const char * 121 get_charset_aliases (void) 122 { 123 const char *cp; 124 125 cp = charset_aliases; 126 if (cp == NULL) 127 { 128 #if !(defined DARWIN7 || defined VMS || defined WIN32_NATIVE || defined __CYGWIN__) 129 const char *dir; 130 const char *base = "charset.alias"; 131 char *file_name; 132 133 /* Make it possible to override the charset.alias location. This is 134 necessary for running the testsuite before "make install". */ 135 dir = getenv ("CHARSETALIASDIR"); 136 if (dir == NULL || dir[0] == '\0') 137 dir = relocate (LIBDIR); 138 139 /* Concatenate dir and base into freshly allocated file_name. */ 140 { 141 size_t dir_len = strlen (dir); 142 size_t base_len = strlen (base); 143 int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1])); 144 file_name = (char *) malloc (dir_len + add_slash + base_len + 1); 145 if (file_name != NULL) 146 { 147 memcpy (file_name, dir, dir_len); 148 if (add_slash) 149 file_name[dir_len] = DIRECTORY_SEPARATOR; 150 memcpy (file_name + dir_len + add_slash, base, base_len + 1); 151 } 152 } 153 154 if (file_name == NULL) 155 /* Out of memory. Treat the file as empty. */ 156 cp = ""; 157 else 158 { 159 int fd; 160 161 /* Open the file. Reject symbolic links on platforms that support 162 O_NOFOLLOW. This is a security feature. Without it, an attacker 163 could retrieve parts of the contents (namely, the tail of the 164 first line that starts with "* ") of an arbitrary file by placing 165 a symbolic link to that file under the name "charset.alias" in 166 some writable directory and defining the environment variable 167 CHARSETALIASDIR to point to that directory. */ 168 fd = open (file_name, 169 O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0)); 170 if (fd < 0) 171 /* File not found. Treat it as empty. */ 172 cp = ""; 173 else 174 { 175 FILE *fp; 176 177 fp = fdopen (fd, "r"); 178 if (fp == NULL) 179 { 180 /* Out of memory. Treat the file as empty. */ 181 close (fd); 182 cp = ""; 183 } 184 else 185 { 186 /* Parse the file's contents. */ 187 char *res_ptr = NULL; 188 size_t res_size = 0; 189 190 for (;;) 191 { 192 int c; 193 char buf1[50+1]; 194 char buf2[50+1]; 195 size_t l1, l2; 196 char *old_res_ptr; 197 198 c = getc (fp); 199 if (c == EOF) 200 break; 201 if (c == '\n' || c == ' ' || c == '\t') 202 continue; 203 if (c == '#') 204 { 205 /* Skip comment, to end of line. */ 206 do 207 c = getc (fp); 208 while (!(c == EOF || c == '\n')); 209 if (c == EOF) 210 break; 211 continue; 212 } 213 ungetc (c, fp); 214 if (fscanf (fp, "%50s %50s", buf1, buf2) < 2) 215 break; 216 l1 = strlen (buf1); 217 l2 = strlen (buf2); 218 old_res_ptr = res_ptr; 219 if (res_size == 0) 220 { 221 res_size = l1 + 1 + l2 + 1; 222 res_ptr = (char *) malloc (res_size + 1); 223 } 224 else 225 { 226 res_size += l1 + 1 + l2 + 1; 227 res_ptr = (char *) realloc (res_ptr, res_size + 1); 228 } 229 if (res_ptr == NULL) 230 { 231 /* Out of memory. */ 232 res_size = 0; 233 if (old_res_ptr != NULL) 234 free (old_res_ptr); 235 break; 236 } 237 strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1); 238 strcpy (res_ptr + res_size - (l2 + 1), buf2); 239 } 240 fclose (fp); 241 if (res_size == 0) 242 cp = ""; 243 else 244 { 245 *(res_ptr + res_size) = '\0'; 246 cp = res_ptr; 247 } 248 } 249 } 250 251 free (file_name); 252 } 253 254 #else 255 256 # if defined DARWIN7 257 /* To avoid the trouble of installing a file that is shared by many 258 GNU packages -- many packaging systems have problems with this --, 259 simply inline the aliases here. */ 260 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0" 261 "ISO8859-2" "\0" "ISO-8859-2" "\0" 262 "ISO8859-4" "\0" "ISO-8859-4" "\0" 263 "ISO8859-5" "\0" "ISO-8859-5" "\0" 264 "ISO8859-7" "\0" "ISO-8859-7" "\0" 265 "ISO8859-9" "\0" "ISO-8859-9" "\0" 266 "ISO8859-13" "\0" "ISO-8859-13" "\0" 267 "ISO8859-15" "\0" "ISO-8859-15" "\0" 268 "KOI8-R" "\0" "KOI8-R" "\0" 269 "KOI8-U" "\0" "KOI8-U" "\0" 270 "CP866" "\0" "CP866" "\0" 271 "CP949" "\0" "CP949" "\0" 272 "CP1131" "\0" "CP1131" "\0" 273 "CP1251" "\0" "CP1251" "\0" 274 "eucCN" "\0" "GB2312" "\0" 275 "GB2312" "\0" "GB2312" "\0" 276 "eucJP" "\0" "EUC-JP" "\0" 277 "eucKR" "\0" "EUC-KR" "\0" 278 "Big5" "\0" "BIG5" "\0" 279 "Big5HKSCS" "\0" "BIG5-HKSCS" "\0" 280 "GBK" "\0" "GBK" "\0" 281 "GB18030" "\0" "GB18030" "\0" 282 "SJIS" "\0" "SHIFT_JIS" "\0" 283 "ARMSCII-8" "\0" "ARMSCII-8" "\0" 284 "PT154" "\0" "PT154" "\0" 285 /*"ISCII-DEV" "\0" "?" "\0"*/ 286 "*" "\0" "UTF-8" "\0"; 287 # endif 288 289 # if defined VMS 290 /* To avoid the troubles of an extra file charset.alias_vms in the 291 sources of many GNU packages, simply inline the aliases here. */ 292 /* The list of encodings is taken from the OpenVMS 7.3-1 documentation 293 "Compaq C Run-Time Library Reference Manual for OpenVMS systems" 294 section 10.7 "Handling Different Character Sets". */ 295 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0" 296 "ISO8859-2" "\0" "ISO-8859-2" "\0" 297 "ISO8859-5" "\0" "ISO-8859-5" "\0" 298 "ISO8859-7" "\0" "ISO-8859-7" "\0" 299 "ISO8859-8" "\0" "ISO-8859-8" "\0" 300 "ISO8859-9" "\0" "ISO-8859-9" "\0" 301 /* Japanese */ 302 "eucJP" "\0" "EUC-JP" "\0" 303 "SJIS" "\0" "SHIFT_JIS" "\0" 304 "DECKANJI" "\0" "DEC-KANJI" "\0" 305 "SDECKANJI" "\0" "EUC-JP" "\0" 306 /* Chinese */ 307 "eucTW" "\0" "EUC-TW" "\0" 308 "DECHANYU" "\0" "DEC-HANYU" "\0" 309 "DECHANZI" "\0" "GB2312" "\0" 310 /* Korean */ 311 "DECKOREAN" "\0" "EUC-KR" "\0"; 312 # endif 313 314 # if defined WIN32_NATIVE || defined __CYGWIN__ 315 /* To avoid the troubles of installing a separate file in the same 316 directory as the DLL and of retrieving the DLL's directory at 317 runtime, simply inline the aliases here. */ 318 319 cp = "CP936" "\0" "GBK" "\0" 320 "CP1361" "\0" "JOHAB" "\0" 321 "CP20127" "\0" "ASCII" "\0" 322 "CP20866" "\0" "KOI8-R" "\0" 323 "CP20936" "\0" "GB2312" "\0" 324 "CP21866" "\0" "KOI8-RU" "\0" 325 "CP28591" "\0" "ISO-8859-1" "\0" 326 "CP28592" "\0" "ISO-8859-2" "\0" 327 "CP28593" "\0" "ISO-8859-3" "\0" 328 "CP28594" "\0" "ISO-8859-4" "\0" 329 "CP28595" "\0" "ISO-8859-5" "\0" 330 "CP28596" "\0" "ISO-8859-6" "\0" 331 "CP28597" "\0" "ISO-8859-7" "\0" 332 "CP28598" "\0" "ISO-8859-8" "\0" 333 "CP28599" "\0" "ISO-8859-9" "\0" 334 "CP28605" "\0" "ISO-8859-15" "\0" 335 "CP38598" "\0" "ISO-8859-8" "\0" 336 "CP51932" "\0" "EUC-JP" "\0" 337 "CP51936" "\0" "GB2312" "\0" 338 "CP51949" "\0" "EUC-KR" "\0" 339 "CP51950" "\0" "EUC-TW" "\0" 340 "CP54936" "\0" "GB18030" "\0" 341 "CP65001" "\0" "UTF-8" "\0"; 342 # endif 343 #endif 344 345 charset_aliases = cp; 346 } 347 348 return cp; 349 } 350 351 /* Determine the current locale's character encoding, and canonicalize it 352 into one of the canonical names listed in config.charset. 353 The result must not be freed; it is statically allocated. 354 If the canonical name cannot be determined, the result is a non-canonical 355 name. */ 356 357 #ifdef STATIC 358 STATIC 359 #endif 360 const char * 361 locale_charset (void) 362 { 363 const char *codeset; 364 const char *aliases; 365 366 #if !(defined WIN32_NATIVE || defined OS2) 367 368 # if HAVE_LANGINFO_CODESET 369 370 /* Most systems support nl_langinfo (CODESET) nowadays. */ 371 codeset = nl_langinfo (CODESET); 372 373 # ifdef __CYGWIN__ 374 /* Cygwin < 1.7 does not have locales. nl_langinfo (CODESET) always 375 returns "US-ASCII". Return the suffix of the locale name from the 376 environment variables (if present) or the codepage as a number. */ 377 if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0) 378 { 379 const char *locale; 380 static char buf[2 + 10 + 1]; 381 382 locale = getenv ("LC_ALL"); 383 if (locale == NULL || locale[0] == '\0') 384 { 385 locale = getenv ("LC_CTYPE"); 386 if (locale == NULL || locale[0] == '\0') 387 locale = getenv ("LANG"); 388 } 389 if (locale != NULL && locale[0] != '\0') 390 { 391 /* If the locale name contains an encoding after the dot, return 392 it. */ 393 const char *dot = strchr (locale, '.'); 394 395 if (dot != NULL) 396 { 397 const char *modifier; 398 399 dot++; 400 /* Look for the possible @... trailer and remove it, if any. */ 401 modifier = strchr (dot, '@'); 402 if (modifier == NULL) 403 return dot; 404 if (modifier - dot < sizeof (buf)) 405 { 406 memcpy (buf, dot, modifier - dot); 407 buf [modifier - dot] = '\0'; 408 return buf; 409 } 410 } 411 } 412 413 /* Woe32 has a function returning the locale's codepage as a number: 414 GetACP(). This encoding is used by Cygwin, unless the user has set 415 the environment variable CYGWIN=codepage:oem (which very few people 416 do). 417 Output directed to console windows needs to be converted (to 418 GetOEMCP() if the console is using a raster font, or to 419 GetConsoleOutputCP() if it is using a TrueType font). Cygwin does 420 this conversion transparently (see winsup/cygwin/fhandler_console.cc), 421 converting to GetConsoleOutputCP(). This leads to correct results, 422 except when SetConsoleOutputCP has been called and a raster font is 423 in use. */ 424 sprintf (buf, "CP%u", GetACP ()); 425 codeset = buf; 426 } 427 # endif 428 429 # else 430 431 /* On old systems which lack it, use setlocale or getenv. */ 432 const char *locale = NULL; 433 434 /* But most old systems don't have a complete set of locales. Some 435 (like SunOS 4 or DJGPP) have only the C locale. Therefore we don't 436 use setlocale here; it would return "C" when it doesn't support the 437 locale name the user has set. */ 438 # if 0 439 locale = setlocale (LC_CTYPE, NULL); 440 # endif 441 if (locale == NULL || locale[0] == '\0') 442 { 443 locale = getenv ("LC_ALL"); 444 if (locale == NULL || locale[0] == '\0') 445 { 446 locale = getenv ("LC_CTYPE"); 447 if (locale == NULL || locale[0] == '\0') 448 locale = getenv ("LANG"); 449 } 450 } 451 452 /* On some old systems, one used to set locale = "iso8859_1". On others, 453 you set it to "language_COUNTRY.charset". In any case, we resolve it 454 through the charset.alias file. */ 455 codeset = locale; 456 457 # endif 458 459 #elif defined WIN32_NATIVE 460 461 static char buf[2 + 10 + 1]; 462 463 /* Woe32 has a function returning the locale's codepage as a number: 464 GetACP(). 465 When the output goes to a console window, it needs to be provided in 466 GetOEMCP() encoding if the console is using a raster font, or in 467 GetConsoleOutputCP() encoding if it is using a TrueType font. 468 But in GUI programs and for output sent to files and pipes, GetACP() 469 encoding is the best bet. */ 470 sprintf (buf, "CP%u", GetACP ()); 471 codeset = buf; 472 473 #elif defined OS2 474 475 const char *locale; 476 static char buf[2 + 10 + 1]; 477 ULONG cp[3]; 478 ULONG cplen; 479 480 /* Allow user to override the codeset, as set in the operating system, 481 with standard language environment variables. */ 482 locale = getenv ("LC_ALL"); 483 if (locale == NULL || locale[0] == '\0') 484 { 485 locale = getenv ("LC_CTYPE"); 486 if (locale == NULL || locale[0] == '\0') 487 locale = getenv ("LANG"); 488 } 489 if (locale != NULL && locale[0] != '\0') 490 { 491 /* If the locale name contains an encoding after the dot, return it. */ 492 const char *dot = strchr (locale, '.'); 493 494 if (dot != NULL) 495 { 496 const char *modifier; 497 498 dot++; 499 /* Look for the possible @... trailer and remove it, if any. */ 500 modifier = strchr (dot, '@'); 501 if (modifier == NULL) 502 return dot; 503 if (modifier - dot < sizeof (buf)) 504 { 505 memcpy (buf, dot, modifier - dot); 506 buf [modifier - dot] = '\0'; 507 return buf; 508 } 509 } 510 511 /* Resolve through the charset.alias file. */ 512 codeset = locale; 513 } 514 else 515 { 516 /* OS/2 has a function returning the locale's codepage as a number. */ 517 if (DosQueryCp (sizeof (cp), cp, &cplen)) 518 codeset = ""; 519 else 520 { 521 sprintf (buf, "CP%u", cp[0]); 522 codeset = buf; 523 } 524 } 525 526 #endif 527 528 if (codeset == NULL) 529 /* The canonical name cannot be determined. */ 530 codeset = ""; 531 532 /* Resolve alias. */ 533 for (aliases = get_charset_aliases (); 534 *aliases != '\0'; 535 aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1) 536 if (strcmp (codeset, aliases) == 0 537 || (aliases[0] == '*' && aliases[1] == '\0')) 538 { 539 codeset = aliases + strlen (aliases) + 1; 540 break; 541 } 542 543 /* Don't return an empty string. GNU libc and GNU libiconv interpret 544 the empty string as denoting "the locale's character encoding", 545 thus GNU libiconv would call this function a second time. */ 546 if (codeset[0] == '\0') 547 codeset = "ASCII"; 548 549 return codeset; 550 } 551