1 /* Determine a canonical name for the current locale's character encoding.
3 Copyright (C) 2000-2006, 2008-2014 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 Lesser 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 Lesser General Public License for more details.
15 You should have received a copy of the GNU Lesser General Public License along
16 with this program; if not, see <http://www.gnu.org/licenses/>. */
18 /* Written by Bruno Haible <bruno@clisp.org>. */
23 #include "localcharset.h"
31 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET
32 # define DARWIN7 /* Darwin 7 or newer, i.e. Mac OS X 10.3 or newer */
35 #if defined _WIN32 || defined __WIN32__
36 # define WINDOWS_NATIVE
41 /* Assume EMX program runs on OS/2, even if compiled under DOS. */
47 #if !defined WINDOWS_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 WINDOWS_NATIVE
61 # define WIN32_LEAN_AND_MEAN
69 /* For MB_CUR_MAX_L */
74 #if ENABLE_RELOCATABLE
75 # include "relocatable.h"
77 # define relocate(pathname) (pathname)
82 # include "configmake.h"
85 /* Define O_NOFOLLOW to 0 on platforms where it does not exist. */
90 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
91 /* Native Windows, Cygwin, OS/2, DOS */
92 # define ISSLASH(C) ((C) == '/' || (C) == '\\')
95 #ifndef DIRECTORY_SEPARATOR
96 # define DIRECTORY_SEPARATOR '/'
100 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
103 #if HAVE_DECL_GETC_UNLOCKED
105 # define getc getc_unlocked
108 /* The following static variable is declared 'volatile' to avoid a
109 possible multithread problem in the function get_charset_aliases. If we
110 are running in a threaded environment, and if two threads initialize
111 'charset_aliases' simultaneously, both will produce the same value,
112 and everything will be ok if the two assignments to 'charset_aliases'
113 are atomic. But I don't know what will happen if the two assignments mix. */
115 # define volatile /* empty */
117 /* Pointer to the contents of the charset.alias file, if it has already been
118 read, else NULL. Its format is:
119 ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0' */
120 static const char * volatile charset_aliases;
122 /* Return a pointer to the contents of the charset.alias file. */
124 get_charset_aliases (void)
128 cp = charset_aliases;
131 #if !(defined DARWIN7 || defined VMS || defined WINDOWS_NATIVE || defined __CYGWIN__)
133 const char *base = "charset.alias";
136 /* Make it possible to override the charset.alias location. This is
137 necessary for running the testsuite before "make install". */
138 dir = getenv ("CHARSETALIASDIR");
139 if (dir == NULL || dir[0] == '\0')
140 dir = relocate (LIBDIR);
142 /* Concatenate dir and base into freshly allocated file_name. */
144 size_t dir_len = strlen (dir);
145 size_t base_len = strlen (base);
146 int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
147 file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
148 if (file_name != NULL)
150 memcpy (file_name, dir, dir_len);
152 file_name[dir_len] = DIRECTORY_SEPARATOR;
153 memcpy (file_name + dir_len + add_slash, base, base_len + 1);
157 if (file_name == NULL)
158 /* Out of memory. Treat the file as empty. */
164 /* Open the file. Reject symbolic links on platforms that support
165 O_NOFOLLOW. This is a security feature. Without it, an attacker
166 could retrieve parts of the contents (namely, the tail of the
167 first line that starts with "* ") of an arbitrary file by placing
168 a symbolic link to that file under the name "charset.alias" in
169 some writable directory and defining the environment variable
170 CHARSETALIASDIR to point to that directory. */
171 fd = open (file_name,
172 O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0));
174 /* File not found. Treat it as empty. */
180 fp = fdopen (fd, "r");
183 /* Out of memory. Treat the file as empty. */
189 /* Parse the file's contents. */
190 char *res_ptr = NULL;
204 if (c == '\n' || c == ' ' || c == '\t')
208 /* Skip comment, to end of line. */
211 while (!(c == EOF || c == '\n'));
217 if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
221 old_res_ptr = res_ptr;
224 res_size = l1 + 1 + l2 + 1;
225 res_ptr = (char *) malloc (res_size + 1);
229 res_size += l1 + 1 + l2 + 1;
230 res_ptr = (char *) realloc (res_ptr, res_size + 1);
239 strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
240 strcpy (res_ptr + res_size - (l2 + 1), buf2);
247 *(res_ptr + res_size) = '\0';
259 /* To avoid the trouble of installing a file that is shared by many
260 GNU packages -- many packaging systems have problems with this --,
261 simply inline the aliases here. */
262 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
263 "ISO8859-2" "\0" "ISO-8859-2" "\0"
264 "ISO8859-4" "\0" "ISO-8859-4" "\0"
265 "ISO8859-5" "\0" "ISO-8859-5" "\0"
266 "ISO8859-7" "\0" "ISO-8859-7" "\0"
267 "ISO8859-9" "\0" "ISO-8859-9" "\0"
268 "ISO8859-13" "\0" "ISO-8859-13" "\0"
269 "ISO8859-15" "\0" "ISO-8859-15" "\0"
270 "KOI8-R" "\0" "KOI8-R" "\0"
271 "KOI8-U" "\0" "KOI8-U" "\0"
272 "CP866" "\0" "CP866" "\0"
273 "CP949" "\0" "CP949" "\0"
274 "CP1131" "\0" "CP1131" "\0"
275 "CP1251" "\0" "CP1251" "\0"
276 "eucCN" "\0" "GB2312" "\0"
277 "GB2312" "\0" "GB2312" "\0"
278 "eucJP" "\0" "EUC-JP" "\0"
279 "eucKR" "\0" "EUC-KR" "\0"
280 "Big5" "\0" "BIG5" "\0"
281 "Big5HKSCS" "\0" "BIG5-HKSCS" "\0"
282 "GBK" "\0" "GBK" "\0"
283 "GB18030" "\0" "GB18030" "\0"
284 "SJIS" "\0" "SHIFT_JIS" "\0"
285 "ARMSCII-8" "\0" "ARMSCII-8" "\0"
286 "PT154" "\0" "PT154" "\0"
287 /*"ISCII-DEV" "\0" "?" "\0"*/
288 "*" "\0" "UTF-8" "\0";
292 /* To avoid the troubles of an extra file charset.alias_vms in the
293 sources of many GNU packages, simply inline the aliases here. */
294 /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
295 "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
296 section 10.7 "Handling Different Character Sets". */
297 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
298 "ISO8859-2" "\0" "ISO-8859-2" "\0"
299 "ISO8859-5" "\0" "ISO-8859-5" "\0"
300 "ISO8859-7" "\0" "ISO-8859-7" "\0"
301 "ISO8859-8" "\0" "ISO-8859-8" "\0"
302 "ISO8859-9" "\0" "ISO-8859-9" "\0"
304 "eucJP" "\0" "EUC-JP" "\0"
305 "SJIS" "\0" "SHIFT_JIS" "\0"
306 "DECKANJI" "\0" "DEC-KANJI" "\0"
307 "SDECKANJI" "\0" "EUC-JP" "\0"
309 "eucTW" "\0" "EUC-TW" "\0"
310 "DECHANYU" "\0" "DEC-HANYU" "\0"
311 "DECHANZI" "\0" "GB2312" "\0"
313 "DECKOREAN" "\0" "EUC-KR" "\0";
316 # if defined WINDOWS_NATIVE || defined __CYGWIN__
317 /* To avoid the troubles of installing a separate file in the same
318 directory as the DLL and of retrieving the DLL's directory at
319 runtime, simply inline the aliases here. */
321 cp = "CP936" "\0" "GBK" "\0"
322 "CP1361" "\0" "JOHAB" "\0"
323 "CP20127" "\0" "ASCII" "\0"
324 "CP20866" "\0" "KOI8-R" "\0"
325 "CP20936" "\0" "GB2312" "\0"
326 "CP21866" "\0" "KOI8-RU" "\0"
327 "CP28591" "\0" "ISO-8859-1" "\0"
328 "CP28592" "\0" "ISO-8859-2" "\0"
329 "CP28593" "\0" "ISO-8859-3" "\0"
330 "CP28594" "\0" "ISO-8859-4" "\0"
331 "CP28595" "\0" "ISO-8859-5" "\0"
332 "CP28596" "\0" "ISO-8859-6" "\0"
333 "CP28597" "\0" "ISO-8859-7" "\0"
334 "CP28598" "\0" "ISO-8859-8" "\0"
335 "CP28599" "\0" "ISO-8859-9" "\0"
336 "CP28605" "\0" "ISO-8859-15" "\0"
337 "CP38598" "\0" "ISO-8859-8" "\0"
338 "CP51932" "\0" "EUC-JP" "\0"
339 "CP51936" "\0" "GB2312" "\0"
340 "CP51949" "\0" "EUC-KR" "\0"
341 "CP51950" "\0" "EUC-TW" "\0"
342 "CP54936" "\0" "GB18030" "\0"
343 "CP65001" "\0" "UTF-8" "\0";
347 charset_aliases = cp;
353 /* Determine the current locale's character encoding, and canonicalize it
354 into one of the canonical names listed in config.charset.
355 The result must not be freed; it is statically allocated.
356 If the canonical name cannot be determined, the result is a non-canonical
363 locale_charset (void)
368 #if !(defined WINDOWS_NATIVE || defined OS2)
370 # if HAVE_LANGINFO_CODESET
372 /* Most systems support nl_langinfo (CODESET) nowadays. */
373 codeset = nl_langinfo (CODESET);
376 /* Cygwin < 1.7 does not have locales. nl_langinfo (CODESET) always
377 returns "US-ASCII". Return the suffix of the locale name from the
378 environment variables (if present) or the codepage as a number. */
379 if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
382 static char buf[2 + 10 + 1];
384 locale = getenv ("LC_ALL");
385 if (locale == NULL || locale[0] == '\0')
387 locale = getenv ("LC_CTYPE");
388 if (locale == NULL || locale[0] == '\0')
389 locale = getenv ("LANG");
391 if (locale != NULL && locale[0] != '\0')
393 /* If the locale name contains an encoding after the dot, return
395 const char *dot = strchr (locale, '.');
399 const char *modifier;
402 /* Look for the possible @... trailer and remove it, if any. */
403 modifier = strchr (dot, '@');
404 if (modifier == NULL)
406 if (modifier - dot < sizeof (buf))
408 memcpy (buf, dot, modifier - dot);
409 buf [modifier - dot] = '\0';
415 /* The Windows API has a function returning the locale's codepage as a
416 number: GetACP(). This encoding is used by Cygwin, unless the user
417 has set the environment variable CYGWIN=codepage:oem (which very few
419 Output directed to console windows needs to be converted (to
420 GetOEMCP() if the console is using a raster font, or to
421 GetConsoleOutputCP() if it is using a TrueType font). Cygwin does
422 this conversion transparently (see winsup/cygwin/fhandler_console.cc),
423 converting to GetConsoleOutputCP(). This leads to correct results,
424 except when SetConsoleOutputCP has been called and a raster font is
426 sprintf (buf, "CP%u", GetACP ());
433 /* On old systems which lack it, use setlocale or getenv. */
434 const char *locale = NULL;
436 /* But most old systems don't have a complete set of locales. Some
437 (like SunOS 4 or DJGPP) have only the C locale. Therefore we don't
438 use setlocale here; it would return "C" when it doesn't support the
439 locale name the user has set. */
441 locale = setlocale (LC_CTYPE, NULL);
443 if (locale == NULL || locale[0] == '\0')
445 locale = getenv ("LC_ALL");
446 if (locale == NULL || locale[0] == '\0')
448 locale = getenv ("LC_CTYPE");
449 if (locale == NULL || locale[0] == '\0')
450 locale = getenv ("LANG");
454 /* On some old systems, one used to set locale = "iso8859_1". On others,
455 you set it to "language_COUNTRY.charset". In any case, we resolve it
456 through the charset.alias file. */
461 #elif defined WINDOWS_NATIVE
463 static char buf[2 + 10 + 1];
465 /* The Windows API has a function returning the locale's codepage as
466 a number, but the value doesn't change according to what the
467 'setlocale' call specified. So we use it as a last resort, in
468 case the string returned by 'setlocale' doesn't specify the
470 char *current_locale = setlocale (LC_ALL, NULL);
473 /* If they set different locales for different categories,
474 'setlocale' will return a semi-colon separated list of locale
475 values. To make sure we use the correct one, we choose LC_CTYPE. */
476 if (strchr (current_locale, ';'))
477 current_locale = setlocale (LC_CTYPE, NULL);
479 pdot = strrchr (current_locale, '.');
481 sprintf (buf, "CP%s", pdot + 1);
484 /* The Windows API has a function returning the locale's codepage as a
486 When the output goes to a console window, it needs to be provided in
487 GetOEMCP() encoding if the console is using a raster font, or in
488 GetConsoleOutputCP() encoding if it is using a TrueType font.
489 But in GUI programs and for output sent to files and pipes, GetACP()
490 encoding is the best bet. */
491 sprintf (buf, "CP%u", GetACP ());
498 static char buf[2 + 10 + 1];
502 /* Allow user to override the codeset, as set in the operating system,
503 with standard language environment variables. */
504 locale = getenv ("LC_ALL");
505 if (locale == NULL || locale[0] == '\0')
507 locale = getenv ("LC_CTYPE");
508 if (locale == NULL || locale[0] == '\0')
509 locale = getenv ("LANG");
511 if (locale != NULL && locale[0] != '\0')
513 /* If the locale name contains an encoding after the dot, return it. */
514 const char *dot = strchr (locale, '.');
518 const char *modifier;
521 /* Look for the possible @... trailer and remove it, if any. */
522 modifier = strchr (dot, '@');
523 if (modifier == NULL)
525 if (modifier - dot < sizeof (buf))
527 memcpy (buf, dot, modifier - dot);
528 buf [modifier - dot] = '\0';
533 /* Resolve through the charset.alias file. */
538 /* OS/2 has a function returning the locale's codepage as a number. */
539 if (DosQueryCp (sizeof (cp), cp, &cplen))
543 sprintf (buf, "CP%u", cp[0]);
551 /* The canonical name cannot be determined. */
555 for (aliases = get_charset_aliases ();
557 aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
558 if (strcmp (codeset, aliases) == 0
559 || (aliases[0] == '*' && aliases[1] == '\0'))
561 codeset = aliases + strlen (aliases) + 1;
565 /* Don't return an empty string. GNU libc and GNU libiconv interpret
566 the empty string as denoting "the locale's character encoding",
567 thus GNU libiconv would call this function a second time. */
568 if (codeset[0] == '\0')
572 /* Mac OS X sets MB_CUR_MAX to 1 when LC_ALL=C, and "UTF-8"
573 (the default codeset) does not work when MB_CUR_MAX is 1. */
574 if (strcmp (codeset, "UTF-8") == 0 && MB_CUR_MAX_L (uselocale (NULL)) <= 1)