1 /* Determine a canonical name for the current locale's character encoding.
3 Copyright (C) 2000-2006, 2008-2011 Free Software Foundation, Inc.
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 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
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License along
16 with this program; if not, write to the Free Software Foundation,
17 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
19 /* Written by Bruno Haible <bruno@clisp.org>. */
24 #include "localcharset.h"
32 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET
33 # define DARWIN7 /* Darwin 7 or newer, i.e. MacOS X 10.3 or newer */
36 #if defined _WIN32 || defined __WIN32__
41 /* Assume EMX program runs on OS/2, even if compiled under DOS. */
47 #if !defined WIN32_NATIVE
49 # if HAVE_LANGINFO_CODESET
50 # include <langinfo.h>
52 # if 0 /* see comment below */
57 # define WIN32_LEAN_AND_MEAN
60 #elif defined WIN32_NATIVE
61 # define WIN32_LEAN_AND_MEAN
69 #if ENABLE_RELOCATABLE
70 # include "relocatable.h"
72 # define relocate(pathname) (pathname)
77 # include "configmake.h"
80 /* Define O_NOFOLLOW to 0 on platforms where it does not exist. */
85 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
86 /* Win32, Cygwin, OS/2, DOS */
87 # define ISSLASH(C) ((C) == '/' || (C) == '\\')
90 #ifndef DIRECTORY_SEPARATOR
91 # define DIRECTORY_SEPARATOR '/'
95 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
98 #if HAVE_DECL_GETC_UNLOCKED
100 # define getc getc_unlocked
103 /* The following static variable is declared 'volatile' to avoid a
104 possible multithread problem in the function get_charset_aliases. If we
105 are running in a threaded environment, and if two threads initialize
106 'charset_aliases' simultaneously, both will produce the same value,
107 and everything will be ok if the two assignments to 'charset_aliases'
108 are atomic. But I don't know what will happen if the two assignments mix. */
110 # define volatile /* empty */
112 /* Pointer to the contents of the charset.alias file, if it has already been
113 read, else NULL. Its format is:
114 ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0' */
115 static const char * volatile charset_aliases;
117 /* Return a pointer to the contents of the charset.alias file. */
119 get_charset_aliases (void)
123 cp = charset_aliases;
126 #if !(defined DARWIN7 || defined VMS || defined WIN32_NATIVE || defined __CYGWIN__)
128 const char *base = "charset.alias";
131 /* Make it possible to override the charset.alias location. This is
132 necessary for running the testsuite before "make install". */
133 dir = getenv ("CHARSETALIASDIR");
134 if (dir == NULL || dir[0] == '\0')
135 dir = relocate (LIBDIR);
137 /* Concatenate dir and base into freshly allocated file_name. */
139 size_t dir_len = strlen (dir);
140 size_t base_len = strlen (base);
141 int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
142 file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
143 if (file_name != NULL)
145 memcpy (file_name, dir, dir_len);
147 file_name[dir_len] = DIRECTORY_SEPARATOR;
148 memcpy (file_name + dir_len + add_slash, base, base_len + 1);
152 if (file_name == NULL)
153 /* Out of memory. Treat the file as empty. */
159 /* Open the file. Reject symbolic links on platforms that support
160 O_NOFOLLOW. This is a security feature. Without it, an attacker
161 could retrieve parts of the contents (namely, the tail of the
162 first line that starts with "* ") of an arbitrary file by placing
163 a symbolic link to that file under the name "charset.alias" in
164 some writable directory and defining the environment variable
165 CHARSETALIASDIR to point to that directory. */
166 fd = open (file_name,
167 O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0));
169 /* File not found. Treat it as empty. */
175 fp = fdopen (fd, "r");
178 /* Out of memory. Treat the file as empty. */
184 /* Parse the file's contents. */
185 char *res_ptr = NULL;
199 if (c == '\n' || c == ' ' || c == '\t')
203 /* Skip comment, to end of line. */
206 while (!(c == EOF || c == '\n'));
212 if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
216 old_res_ptr = res_ptr;
219 res_size = l1 + 1 + l2 + 1;
220 res_ptr = (char *) malloc (res_size + 1);
224 res_size += l1 + 1 + l2 + 1;
225 res_ptr = (char *) realloc (res_ptr, res_size + 1);
234 strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
235 strcpy (res_ptr + res_size - (l2 + 1), buf2);
242 *(res_ptr + res_size) = '\0';
254 /* To avoid the trouble of installing a file that is shared by many
255 GNU packages -- many packaging systems have problems with this --,
256 simply inline the aliases here. */
257 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
258 "ISO8859-2" "\0" "ISO-8859-2" "\0"
259 "ISO8859-4" "\0" "ISO-8859-4" "\0"
260 "ISO8859-5" "\0" "ISO-8859-5" "\0"
261 "ISO8859-7" "\0" "ISO-8859-7" "\0"
262 "ISO8859-9" "\0" "ISO-8859-9" "\0"
263 "ISO8859-13" "\0" "ISO-8859-13" "\0"
264 "ISO8859-15" "\0" "ISO-8859-15" "\0"
265 "KOI8-R" "\0" "KOI8-R" "\0"
266 "KOI8-U" "\0" "KOI8-U" "\0"
267 "CP866" "\0" "CP866" "\0"
268 "CP949" "\0" "CP949" "\0"
269 "CP1131" "\0" "CP1131" "\0"
270 "CP1251" "\0" "CP1251" "\0"
271 "eucCN" "\0" "GB2312" "\0"
272 "GB2312" "\0" "GB2312" "\0"
273 "eucJP" "\0" "EUC-JP" "\0"
274 "eucKR" "\0" "EUC-KR" "\0"
275 "Big5" "\0" "BIG5" "\0"
276 "Big5HKSCS" "\0" "BIG5-HKSCS" "\0"
277 "GBK" "\0" "GBK" "\0"
278 "GB18030" "\0" "GB18030" "\0"
279 "SJIS" "\0" "SHIFT_JIS" "\0"
280 "ARMSCII-8" "\0" "ARMSCII-8" "\0"
281 "PT154" "\0" "PT154" "\0"
282 /*"ISCII-DEV" "\0" "?" "\0"*/
283 "*" "\0" "UTF-8" "\0";
287 /* To avoid the troubles of an extra file charset.alias_vms in the
288 sources of many GNU packages, simply inline the aliases here. */
289 /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
290 "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
291 section 10.7 "Handling Different Character Sets". */
292 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
293 "ISO8859-2" "\0" "ISO-8859-2" "\0"
294 "ISO8859-5" "\0" "ISO-8859-5" "\0"
295 "ISO8859-7" "\0" "ISO-8859-7" "\0"
296 "ISO8859-8" "\0" "ISO-8859-8" "\0"
297 "ISO8859-9" "\0" "ISO-8859-9" "\0"
299 "eucJP" "\0" "EUC-JP" "\0"
300 "SJIS" "\0" "SHIFT_JIS" "\0"
301 "DECKANJI" "\0" "DEC-KANJI" "\0"
302 "SDECKANJI" "\0" "EUC-JP" "\0"
304 "eucTW" "\0" "EUC-TW" "\0"
305 "DECHANYU" "\0" "DEC-HANYU" "\0"
306 "DECHANZI" "\0" "GB2312" "\0"
308 "DECKOREAN" "\0" "EUC-KR" "\0";
311 # if defined WIN32_NATIVE || defined __CYGWIN__
312 /* To avoid the troubles of installing a separate file in the same
313 directory as the DLL and of retrieving the DLL's directory at
314 runtime, simply inline the aliases here. */
316 cp = "CP936" "\0" "GBK" "\0"
317 "CP1361" "\0" "JOHAB" "\0"
318 "CP20127" "\0" "ASCII" "\0"
319 "CP20866" "\0" "KOI8-R" "\0"
320 "CP20936" "\0" "GB2312" "\0"
321 "CP21866" "\0" "KOI8-RU" "\0"
322 "CP28591" "\0" "ISO-8859-1" "\0"
323 "CP28592" "\0" "ISO-8859-2" "\0"
324 "CP28593" "\0" "ISO-8859-3" "\0"
325 "CP28594" "\0" "ISO-8859-4" "\0"
326 "CP28595" "\0" "ISO-8859-5" "\0"
327 "CP28596" "\0" "ISO-8859-6" "\0"
328 "CP28597" "\0" "ISO-8859-7" "\0"
329 "CP28598" "\0" "ISO-8859-8" "\0"
330 "CP28599" "\0" "ISO-8859-9" "\0"
331 "CP28605" "\0" "ISO-8859-15" "\0"
332 "CP38598" "\0" "ISO-8859-8" "\0"
333 "CP51932" "\0" "EUC-JP" "\0"
334 "CP51936" "\0" "GB2312" "\0"
335 "CP51949" "\0" "EUC-KR" "\0"
336 "CP51950" "\0" "EUC-TW" "\0"
337 "CP54936" "\0" "GB18030" "\0"
338 "CP65001" "\0" "UTF-8" "\0";
342 charset_aliases = cp;
348 /* Determine the current locale's character encoding, and canonicalize it
349 into one of the canonical names listed in config.charset.
350 The result must not be freed; it is statically allocated.
351 If the canonical name cannot be determined, the result is a non-canonical
358 locale_charset (void)
363 #if !(defined WIN32_NATIVE || defined OS2)
365 # if HAVE_LANGINFO_CODESET
367 /* Most systems support nl_langinfo (CODESET) nowadays. */
368 codeset = nl_langinfo (CODESET);
371 /* Cygwin < 1.7 does not have locales. nl_langinfo (CODESET) always
372 returns "US-ASCII". Return the suffix of the locale name from the
373 environment variables (if present) or the codepage as a number. */
374 if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
377 static char buf[2 + 10 + 1];
379 locale = getenv ("LC_ALL");
380 if (locale == NULL || locale[0] == '\0')
382 locale = getenv ("LC_CTYPE");
383 if (locale == NULL || locale[0] == '\0')
384 locale = getenv ("LANG");
386 if (locale != NULL && locale[0] != '\0')
388 /* If the locale name contains an encoding after the dot, return
390 const char *dot = strchr (locale, '.');
394 const char *modifier;
397 /* Look for the possible @... trailer and remove it, if any. */
398 modifier = strchr (dot, '@');
399 if (modifier == NULL)
401 if (modifier - dot < sizeof (buf))
403 memcpy (buf, dot, modifier - dot);
404 buf [modifier - dot] = '\0';
410 /* Woe32 has a function returning the locale's codepage as a number:
411 GetACP(). This encoding is used by Cygwin, unless the user has set
412 the environment variable CYGWIN=codepage:oem (which very few people
414 Output directed to console windows needs to be converted (to
415 GetOEMCP() if the console is using a raster font, or to
416 GetConsoleOutputCP() if it is using a TrueType font). Cygwin does
417 this conversion transparently (see winsup/cygwin/fhandler_console.cc),
418 converting to GetConsoleOutputCP(). This leads to correct results,
419 except when SetConsoleOutputCP has been called and a raster font is
421 sprintf (buf, "CP%u", GetACP ());
428 /* On old systems which lack it, use setlocale or getenv. */
429 const char *locale = NULL;
431 /* But most old systems don't have a complete set of locales. Some
432 (like SunOS 4 or DJGPP) have only the C locale. Therefore we don't
433 use setlocale here; it would return "C" when it doesn't support the
434 locale name the user has set. */
436 locale = setlocale (LC_CTYPE, NULL);
438 if (locale == NULL || locale[0] == '\0')
440 locale = getenv ("LC_ALL");
441 if (locale == NULL || locale[0] == '\0')
443 locale = getenv ("LC_CTYPE");
444 if (locale == NULL || locale[0] == '\0')
445 locale = getenv ("LANG");
449 /* On some old systems, one used to set locale = "iso8859_1". On others,
450 you set it to "language_COUNTRY.charset". In any case, we resolve it
451 through the charset.alias file. */
456 #elif defined WIN32_NATIVE
458 static char buf[2 + 10 + 1];
460 /* Woe32 has a function returning the locale's codepage as a number:
462 When the output goes to a console window, it needs to be provided in
463 GetOEMCP() encoding if the console is using a raster font, or in
464 GetConsoleOutputCP() encoding if it is using a TrueType font.
465 But in GUI programs and for output sent to files and pipes, GetACP()
466 encoding is the best bet. */
467 sprintf (buf, "CP%u", GetACP ());
473 static char buf[2 + 10 + 1];
477 /* Allow user to override the codeset, as set in the operating system,
478 with standard language environment variables. */
479 locale = getenv ("LC_ALL");
480 if (locale == NULL || locale[0] == '\0')
482 locale = getenv ("LC_CTYPE");
483 if (locale == NULL || locale[0] == '\0')
484 locale = getenv ("LANG");
486 if (locale != NULL && locale[0] != '\0')
488 /* If the locale name contains an encoding after the dot, return it. */
489 const char *dot = strchr (locale, '.');
493 const char *modifier;
496 /* Look for the possible @... trailer and remove it, if any. */
497 modifier = strchr (dot, '@');
498 if (modifier == NULL)
500 if (modifier - dot < sizeof (buf))
502 memcpy (buf, dot, modifier - dot);
503 buf [modifier - dot] = '\0';
508 /* Resolve through the charset.alias file. */
513 /* OS/2 has a function returning the locale's codepage as a number. */
514 if (DosQueryCp (sizeof (cp), cp, &cplen))
518 sprintf (buf, "CP%u", cp[0]);
526 /* The canonical name cannot be determined. */
530 for (aliases = get_charset_aliases ();
532 aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
533 if (strcmp (codeset, aliases) == 0
534 || (aliases[0] == '*' && aliases[1] == '\0'))
536 codeset = aliases + strlen (aliases) + 1;
540 /* Don't return an empty string. GNU libc and GNU libiconv interpret
541 the empty string as denoting "the locale's character encoding",
542 thus GNU libiconv would call this function a second time. */
543 if (codeset[0] == '\0')