1 /* Determine a canonical name for the current locale's character encoding.
3 Copyright (C) 2000-2006, 2008-2010 Free Software Foundation, Inc.
5 This program is free software; you can redistribute it and/or modify it
6 under the terms of the GNU Library General Public License as published
7 by the Free Software Foundation; either version 2, or (at your option)
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Library General Public License for more details.
15 You should have received a copy of the GNU Library General Public
16 License along with this program; if not, write to the Free Software
17 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
20 /* Written by Bruno Haible <bruno@clisp.org>. */
25 #include "localcharset.h"
33 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET
34 # define DARWIN7 /* Darwin 7 or newer, i.e. MacOS X 10.3 or newer */
37 #if defined _WIN32 || defined __WIN32__
42 /* Assume EMX program runs on OS/2, even if compiled under DOS. */
48 #if !defined WIN32_NATIVE
50 # if HAVE_LANGINFO_CODESET
51 # include <langinfo.h>
53 # if 0 /* see comment below */
58 # define WIN32_LEAN_AND_MEAN
61 #elif defined WIN32_NATIVE
62 # define WIN32_LEAN_AND_MEAN
70 #if ENABLE_RELOCATABLE
71 # include "relocatable.h"
73 # define relocate(pathname) (pathname)
78 # include "configmake.h"
81 /* Define O_NOFOLLOW to 0 on platforms where it does not exist. */
86 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
87 /* Win32, Cygwin, OS/2, DOS */
88 # define ISSLASH(C) ((C) == '/' || (C) == '\\')
91 #ifndef DIRECTORY_SEPARATOR
92 # define DIRECTORY_SEPARATOR '/'
96 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
99 #if HAVE_DECL_GETC_UNLOCKED
101 # define getc getc_unlocked
104 /* The following static variable is declared 'volatile' to avoid a
105 possible multithread problem in the function get_charset_aliases. If we
106 are running in a threaded environment, and if two threads initialize
107 'charset_aliases' simultaneously, both will produce the same value,
108 and everything will be ok if the two assignments to 'charset_aliases'
109 are atomic. But I don't know what will happen if the two assignments mix. */
111 # define volatile /* empty */
113 /* Pointer to the contents of the charset.alias file, if it has already been
114 read, else NULL. Its format is:
115 ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0' */
116 static const char * volatile charset_aliases;
118 /* Return a pointer to the contents of the charset.alias file. */
120 get_charset_aliases (void)
124 cp = charset_aliases;
127 #if !(defined DARWIN7 || defined VMS || defined WIN32_NATIVE || defined __CYGWIN__)
129 const char *base = "charset.alias";
132 /* Make it possible to override the charset.alias location. This is
133 necessary for running the testsuite before "make install". */
134 dir = getenv ("CHARSETALIASDIR");
135 if (dir == NULL || dir[0] == '\0')
136 dir = relocate (LIBDIR);
138 /* Concatenate dir and base into freshly allocated file_name. */
140 size_t dir_len = strlen (dir);
141 size_t base_len = strlen (base);
142 int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
143 file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
144 if (file_name != NULL)
146 memcpy (file_name, dir, dir_len);
148 file_name[dir_len] = DIRECTORY_SEPARATOR;
149 memcpy (file_name + dir_len + add_slash, base, base_len + 1);
153 if (file_name == NULL)
154 /* Out of memory. Treat the file as empty. */
160 /* Open the file. Reject symbolic links on platforms that support
161 O_NOFOLLOW. This is a security feature. Without it, an attacker
162 could retrieve parts of the contents (namely, the tail of the
163 first line that starts with "* ") of an arbitrary file by placing
164 a symbolic link to that file under the name "charset.alias" in
165 some writable directory and defining the environment variable
166 CHARSETALIASDIR to point to that directory. */
167 fd = open (file_name,
168 O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0));
170 /* File not found. Treat it as empty. */
176 fp = fdopen (fd, "r");
179 /* Out of memory. Treat the file as empty. */
185 /* Parse the file's contents. */
186 char *res_ptr = NULL;
200 if (c == '\n' || c == ' ' || c == '\t')
204 /* Skip comment, to end of line. */
207 while (!(c == EOF || c == '\n'));
213 if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
217 old_res_ptr = res_ptr;
220 res_size = l1 + 1 + l2 + 1;
221 res_ptr = (char *) malloc (res_size + 1);
225 res_size += l1 + 1 + l2 + 1;
226 res_ptr = (char *) realloc (res_ptr, res_size + 1);
232 if (old_res_ptr != NULL)
236 strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
237 strcpy (res_ptr + res_size - (l2 + 1), buf2);
244 *(res_ptr + res_size) = '\0';
256 /* To avoid the trouble of installing a file that is shared by many
257 GNU packages -- many packaging systems have problems with this --,
258 simply inline the aliases here. */
259 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
260 "ISO8859-2" "\0" "ISO-8859-2" "\0"
261 "ISO8859-4" "\0" "ISO-8859-4" "\0"
262 "ISO8859-5" "\0" "ISO-8859-5" "\0"
263 "ISO8859-7" "\0" "ISO-8859-7" "\0"
264 "ISO8859-9" "\0" "ISO-8859-9" "\0"
265 "ISO8859-13" "\0" "ISO-8859-13" "\0"
266 "ISO8859-15" "\0" "ISO-8859-15" "\0"
267 "KOI8-R" "\0" "KOI8-R" "\0"
268 "KOI8-U" "\0" "KOI8-U" "\0"
269 "CP866" "\0" "CP866" "\0"
270 "CP949" "\0" "CP949" "\0"
271 "CP1131" "\0" "CP1131" "\0"
272 "CP1251" "\0" "CP1251" "\0"
273 "eucCN" "\0" "GB2312" "\0"
274 "GB2312" "\0" "GB2312" "\0"
275 "eucJP" "\0" "EUC-JP" "\0"
276 "eucKR" "\0" "EUC-KR" "\0"
277 "Big5" "\0" "BIG5" "\0"
278 "Big5HKSCS" "\0" "BIG5-HKSCS" "\0"
279 "GBK" "\0" "GBK" "\0"
280 "GB18030" "\0" "GB18030" "\0"
281 "SJIS" "\0" "SHIFT_JIS" "\0"
282 "ARMSCII-8" "\0" "ARMSCII-8" "\0"
283 "PT154" "\0" "PT154" "\0"
284 /*"ISCII-DEV" "\0" "?" "\0"*/
285 "*" "\0" "UTF-8" "\0";
289 /* To avoid the troubles of an extra file charset.alias_vms in the
290 sources of many GNU packages, simply inline the aliases here. */
291 /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
292 "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
293 section 10.7 "Handling Different Character Sets". */
294 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
295 "ISO8859-2" "\0" "ISO-8859-2" "\0"
296 "ISO8859-5" "\0" "ISO-8859-5" "\0"
297 "ISO8859-7" "\0" "ISO-8859-7" "\0"
298 "ISO8859-8" "\0" "ISO-8859-8" "\0"
299 "ISO8859-9" "\0" "ISO-8859-9" "\0"
301 "eucJP" "\0" "EUC-JP" "\0"
302 "SJIS" "\0" "SHIFT_JIS" "\0"
303 "DECKANJI" "\0" "DEC-KANJI" "\0"
304 "SDECKANJI" "\0" "EUC-JP" "\0"
306 "eucTW" "\0" "EUC-TW" "\0"
307 "DECHANYU" "\0" "DEC-HANYU" "\0"
308 "DECHANZI" "\0" "GB2312" "\0"
310 "DECKOREAN" "\0" "EUC-KR" "\0";
313 # if defined WIN32_NATIVE || defined __CYGWIN__
314 /* To avoid the troubles of installing a separate file in the same
315 directory as the DLL and of retrieving the DLL's directory at
316 runtime, simply inline the aliases here. */
318 cp = "CP936" "\0" "GBK" "\0"
319 "CP1361" "\0" "JOHAB" "\0"
320 "CP20127" "\0" "ASCII" "\0"
321 "CP20866" "\0" "KOI8-R" "\0"
322 "CP20936" "\0" "GB2312" "\0"
323 "CP21866" "\0" "KOI8-RU" "\0"
324 "CP28591" "\0" "ISO-8859-1" "\0"
325 "CP28592" "\0" "ISO-8859-2" "\0"
326 "CP28593" "\0" "ISO-8859-3" "\0"
327 "CP28594" "\0" "ISO-8859-4" "\0"
328 "CP28595" "\0" "ISO-8859-5" "\0"
329 "CP28596" "\0" "ISO-8859-6" "\0"
330 "CP28597" "\0" "ISO-8859-7" "\0"
331 "CP28598" "\0" "ISO-8859-8" "\0"
332 "CP28599" "\0" "ISO-8859-9" "\0"
333 "CP28605" "\0" "ISO-8859-15" "\0"
334 "CP38598" "\0" "ISO-8859-8" "\0"
335 "CP51932" "\0" "EUC-JP" "\0"
336 "CP51936" "\0" "GB2312" "\0"
337 "CP51949" "\0" "EUC-KR" "\0"
338 "CP51950" "\0" "EUC-TW" "\0"
339 "CP54936" "\0" "GB18030" "\0"
340 "CP65001" "\0" "UTF-8" "\0";
344 charset_aliases = cp;
350 /* Determine the current locale's character encoding, and canonicalize it
351 into one of the canonical names listed in config.charset.
352 The result must not be freed; it is statically allocated.
353 If the canonical name cannot be determined, the result is a non-canonical
360 locale_charset (void)
365 #if !(defined WIN32_NATIVE || defined OS2)
367 # if HAVE_LANGINFO_CODESET
369 /* Most systems support nl_langinfo (CODESET) nowadays. */
370 codeset = nl_langinfo (CODESET);
373 /* Cygwin < 1.7 does not have locales. nl_langinfo (CODESET) always
374 returns "US-ASCII". Return the suffix of the locale name from the
375 environment variables (if present) or the codepage as a number. */
376 if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
379 static char buf[2 + 10 + 1];
381 locale = getenv ("LC_ALL");
382 if (locale == NULL || locale[0] == '\0')
384 locale = getenv ("LC_CTYPE");
385 if (locale == NULL || locale[0] == '\0')
386 locale = getenv ("LANG");
388 if (locale != NULL && locale[0] != '\0')
390 /* If the locale name contains an encoding after the dot, return
392 const char *dot = strchr (locale, '.');
396 const char *modifier;
399 /* Look for the possible @... trailer and remove it, if any. */
400 modifier = strchr (dot, '@');
401 if (modifier == NULL)
403 if (modifier - dot < sizeof (buf))
405 memcpy (buf, dot, modifier - dot);
406 buf [modifier - dot] = '\0';
412 /* Woe32 has a function returning the locale's codepage as a number:
413 GetACP(). This encoding is used by Cygwin, unless the user has set
414 the environment variable CYGWIN=codepage:oem (which very few people
416 Output directed to console windows needs to be converted (to
417 GetOEMCP() if the console is using a raster font, or to
418 GetConsoleOutputCP() if it is using a TrueType font). Cygwin does
419 this conversion transparently (see winsup/cygwin/fhandler_console.cc),
420 converting to GetConsoleOutputCP(). This leads to correct results,
421 except when SetConsoleOutputCP has been called and a raster font is
423 sprintf (buf, "CP%u", GetACP ());
430 /* On old systems which lack it, use setlocale or getenv. */
431 const char *locale = NULL;
433 /* But most old systems don't have a complete set of locales. Some
434 (like SunOS 4 or DJGPP) have only the C locale. Therefore we don't
435 use setlocale here; it would return "C" when it doesn't support the
436 locale name the user has set. */
438 locale = setlocale (LC_CTYPE, NULL);
440 if (locale == NULL || locale[0] == '\0')
442 locale = getenv ("LC_ALL");
443 if (locale == NULL || locale[0] == '\0')
445 locale = getenv ("LC_CTYPE");
446 if (locale == NULL || locale[0] == '\0')
447 locale = getenv ("LANG");
451 /* On some old systems, one used to set locale = "iso8859_1". On others,
452 you set it to "language_COUNTRY.charset". In any case, we resolve it
453 through the charset.alias file. */
458 #elif defined WIN32_NATIVE
460 static char buf[2 + 10 + 1];
462 /* Woe32 has a function returning the locale's codepage as a number:
464 When the output goes to a console window, it needs to be provided in
465 GetOEMCP() encoding if the console is using a raster font, or in
466 GetConsoleOutputCP() encoding if it is using a TrueType font.
467 But in GUI programs and for output sent to files and pipes, GetACP()
468 encoding is the best bet. */
469 sprintf (buf, "CP%u", GetACP ());
475 static char buf[2 + 10 + 1];
479 /* Allow user to override the codeset, as set in the operating system,
480 with standard language environment variables. */
481 locale = getenv ("LC_ALL");
482 if (locale == NULL || locale[0] == '\0')
484 locale = getenv ("LC_CTYPE");
485 if (locale == NULL || locale[0] == '\0')
486 locale = getenv ("LANG");
488 if (locale != NULL && locale[0] != '\0')
490 /* If the locale name contains an encoding after the dot, return it. */
491 const char *dot = strchr (locale, '.');
495 const char *modifier;
498 /* Look for the possible @... trailer and remove it, if any. */
499 modifier = strchr (dot, '@');
500 if (modifier == NULL)
502 if (modifier - dot < sizeof (buf))
504 memcpy (buf, dot, modifier - dot);
505 buf [modifier - dot] = '\0';
510 /* Resolve through the charset.alias file. */
515 /* OS/2 has a function returning the locale's codepage as a number. */
516 if (DosQueryCp (sizeof (cp), cp, &cplen))
520 sprintf (buf, "CP%u", cp[0]);
528 /* The canonical name cannot be determined. */
532 for (aliases = get_charset_aliases ();
534 aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
535 if (strcmp (codeset, aliases) == 0
536 || (aliases[0] == '*' && aliases[1] == '\0'))
538 codeset = aliases + strlen (aliases) + 1;
542 /* Don't return an empty string. GNU libc and GNU libiconv interpret
543 the empty string as denoting "the locale's character encoding",
544 thus GNU libiconv would call this function a second time. */
545 if (codeset[0] == '\0')