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, 2008-2014 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, see <http://www.gnu.org/licenses/>. */
19
20 /* Written by Bruno Haible <bruno@clisp.org>. */
21
22 #include <config.h>
23
24 /* Specification. */
25 #include "localcharset.h"
26
27 #include <fcntl.h>
28 #include <stddef.h>
29 #include <stdio.h>
30 #include <string.h>
31 #include <stdlib.h>
32
33 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET
34 # define DARWIN7 /* Darwin 7 or newer, i.e. Mac OS X 10.3 or newer */
35 #endif
36
37 #if defined _WIN32 || defined __WIN32__
38 # define WINDOWS_NATIVE
39 #endif
40
41 #if defined __EMX__
42 /* Assume EMX program runs on OS/2, even if compiled under DOS. */
43 # ifndef OS2
44 # define OS2
45 # endif
46 #endif
47
48 #if !defined WINDOWS_NATIVE
49 # include <unistd.h>
50 # if HAVE_LANGINFO_CODESET
51 # include <langinfo.h>
52 # else
53 # if 0 /* see comment below */
54 # include <locale.h>
55 # endif
56 # endif
57 # ifdef __CYGWIN__
58 # define WIN32_LEAN_AND_MEAN
59 # include <windows.h>
60 # endif
61 #elif defined WINDOWS_NATIVE
62 # define WIN32_LEAN_AND_MEAN
63 # include <windows.h>
64 #endif
65 #if defined OS2
66 # define INCL_DOS
67 # include <os2.h>
68 #endif
69
70 /* For MB_CUR_MAX_L */
71 #if defined DARWIN7
72 # include <xlocale.h>
73 #endif
74
75 #if ENABLE_RELOCATABLE
76 # include "relocatable.h"
77 #else
78 # define relocate(pathname) (pathname)
79 #endif
80
81 /* Get LIBDIR. */
82 #ifndef LIBDIR
83 # include "configmake.h"
84 #endif
85
86 /* Define O_NOFOLLOW to 0 on platforms where it does not exist. */
87 #ifndef O_NOFOLLOW
88 # define O_NOFOLLOW 0
89 #endif
90
91 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
92 /* Native Windows, Cygwin, OS/2, DOS */
93 # define ISSLASH(C) ((C) == '/' || (C) == '\\')
94 #endif
95
96 #ifndef DIRECTORY_SEPARATOR
97 # define DIRECTORY_SEPARATOR '/'
98 #endif
99
100 #ifndef ISSLASH
101 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
102 #endif
103
104 #if HAVE_DECL_GETC_UNLOCKED
105 # undef getc
106 # define getc getc_unlocked
107 #endif
108
109 /* The following static variable is declared 'volatile' to avoid a
110 possible multithread problem in the function get_charset_aliases. If we
111 are running in a threaded environment, and if two threads initialize
112 'charset_aliases' simultaneously, both will produce the same value,
113 and everything will be ok if the two assignments to 'charset_aliases'
114 are atomic. But I don't know what will happen if the two assignments mix. */
115 #if __STDC__ != 1
116 # define volatile /* empty */
117 #endif
118 /* Pointer to the contents of the charset.alias file, if it has already been
119 read, else NULL. Its format is:
120 ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0' */
121 static const char * volatile charset_aliases;
122
123 /* Return a pointer to the contents of the charset.alias file. */
124 static const char *
get_charset_aliases(void)125 get_charset_aliases (void)
126 {
127 const char *cp;
128
129 cp = charset_aliases;
130 if (cp == NULL)
131 {
132 #if !(defined DARWIN7 || defined VMS || defined WINDOWS_NATIVE || defined __CYGWIN__)
133 const char *dir;
134 const char *base = "charset.alias";
135 char *file_name;
136
137 /* Make it possible to override the charset.alias location. This is
138 necessary for running the testsuite before "make install". */
139 dir = getenv ("CHARSETALIASDIR");
140 if (dir == NULL || dir[0] == '\0')
141 dir = relocate (LIBDIR);
142
143 /* Concatenate dir and base into freshly allocated file_name. */
144 {
145 size_t dir_len = strlen (dir);
146 size_t base_len = strlen (base);
147 int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
148 file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
149 if (file_name != NULL)
150 {
151 memcpy (file_name, dir, dir_len);
152 if (add_slash)
153 file_name[dir_len] = DIRECTORY_SEPARATOR;
154 memcpy (file_name + dir_len + add_slash, base, base_len + 1);
155 }
156 }
157
158 if (file_name == NULL)
159 /* Out of memory. Treat the file as empty. */
160 cp = "";
161 else
162 {
163 int fd;
164
165 /* Open the file. Reject symbolic links on platforms that support
166 O_NOFOLLOW. This is a security feature. Without it, an attacker
167 could retrieve parts of the contents (namely, the tail of the
168 first line that starts with "* ") of an arbitrary file by placing
169 a symbolic link to that file under the name "charset.alias" in
170 some writable directory and defining the environment variable
171 CHARSETALIASDIR to point to that directory. */
172 fd = open (file_name,
173 O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0));
174 if (fd < 0)
175 /* File not found. Treat it as empty. */
176 cp = "";
177 else
178 {
179 FILE *fp;
180
181 fp = fdopen (fd, "r");
182 if (fp == NULL)
183 {
184 /* Out of memory. Treat the file as empty. */
185 close (fd);
186 cp = "";
187 }
188 else
189 {
190 /* Parse the file's contents. */
191 char *res_ptr = NULL;
192 size_t res_size = 0;
193
194 for (;;)
195 {
196 int c;
197 char buf1[50+1];
198 char buf2[50+1];
199 size_t l1, l2;
200 char *old_res_ptr;
201
202 c = getc (fp);
203 if (c == EOF)
204 break;
205 if (c == '\n' || c == ' ' || c == '\t')
206 continue;
207 if (c == '#')
208 {
209 /* Skip comment, to end of line. */
210 do
211 c = getc (fp);
212 while (!(c == EOF || c == '\n'));
213 if (c == EOF)
214 break;
215 continue;
216 }
217 ungetc (c, fp);
218 if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
219 break;
220 l1 = strlen (buf1);
221 l2 = strlen (buf2);
222 old_res_ptr = res_ptr;
223 if (res_size == 0)
224 {
225 res_size = l1 + 1 + l2 + 1;
226 res_ptr = (char *) malloc (res_size + 1);
227 }
228 else
229 {
230 res_size += l1 + 1 + l2 + 1;
231 res_ptr = (char *) realloc (res_ptr, res_size + 1);
232 }
233 if (res_ptr == NULL)
234 {
235 /* Out of memory. */
236 res_size = 0;
237 free (old_res_ptr);
238 break;
239 }
240 strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
241 strcpy (res_ptr + res_size - (l2 + 1), buf2);
242 }
243 fclose (fp);
244 if (res_size == 0)
245 cp = "";
246 else
247 {
248 *(res_ptr + res_size) = '\0';
249 cp = res_ptr;
250 }
251 }
252 }
253
254 free (file_name);
255 }
256
257 #else
258
259 # if defined DARWIN7
260 /* To avoid the trouble of installing a file that is shared by many
261 GNU packages -- many packaging systems have problems with this --,
262 simply inline the aliases here. */
263 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
264 "ISO8859-2" "\0" "ISO-8859-2" "\0"
265 "ISO8859-4" "\0" "ISO-8859-4" "\0"
266 "ISO8859-5" "\0" "ISO-8859-5" "\0"
267 "ISO8859-7" "\0" "ISO-8859-7" "\0"
268 "ISO8859-9" "\0" "ISO-8859-9" "\0"
269 "ISO8859-13" "\0" "ISO-8859-13" "\0"
270 "ISO8859-15" "\0" "ISO-8859-15" "\0"
271 "KOI8-R" "\0" "KOI8-R" "\0"
272 "KOI8-U" "\0" "KOI8-U" "\0"
273 "CP866" "\0" "CP866" "\0"
274 "CP949" "\0" "CP949" "\0"
275 "CP1131" "\0" "CP1131" "\0"
276 "CP1251" "\0" "CP1251" "\0"
277 "eucCN" "\0" "GB2312" "\0"
278 "GB2312" "\0" "GB2312" "\0"
279 "eucJP" "\0" "EUC-JP" "\0"
280 "eucKR" "\0" "EUC-KR" "\0"
281 "Big5" "\0" "BIG5" "\0"
282 "Big5HKSCS" "\0" "BIG5-HKSCS" "\0"
283 "GBK" "\0" "GBK" "\0"
284 "GB18030" "\0" "GB18030" "\0"
285 "SJIS" "\0" "SHIFT_JIS" "\0"
286 "ARMSCII-8" "\0" "ARMSCII-8" "\0"
287 "PT154" "\0" "PT154" "\0"
288 /*"ISCII-DEV" "\0" "?" "\0"*/
289 "*" "\0" "UTF-8" "\0";
290 # endif
291
292 # if defined VMS
293 /* To avoid the troubles of an extra file charset.alias_vms in the
294 sources of many GNU packages, simply inline the aliases here. */
295 /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
296 "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
297 section 10.7 "Handling Different Character Sets". */
298 cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
299 "ISO8859-2" "\0" "ISO-8859-2" "\0"
300 "ISO8859-5" "\0" "ISO-8859-5" "\0"
301 "ISO8859-7" "\0" "ISO-8859-7" "\0"
302 "ISO8859-8" "\0" "ISO-8859-8" "\0"
303 "ISO8859-9" "\0" "ISO-8859-9" "\0"
304 /* Japanese */
305 "eucJP" "\0" "EUC-JP" "\0"
306 "SJIS" "\0" "SHIFT_JIS" "\0"
307 "DECKANJI" "\0" "DEC-KANJI" "\0"
308 "SDECKANJI" "\0" "EUC-JP" "\0"
309 /* Chinese */
310 "eucTW" "\0" "EUC-TW" "\0"
311 "DECHANYU" "\0" "DEC-HANYU" "\0"
312 "DECHANZI" "\0" "GB2312" "\0"
313 /* Korean */
314 "DECKOREAN" "\0" "EUC-KR" "\0";
315 # endif
316
317 # if defined WINDOWS_NATIVE || defined __CYGWIN__
318 /* To avoid the troubles of installing a separate file in the same
319 directory as the DLL and of retrieving the DLL's directory at
320 runtime, simply inline the aliases here. */
321
322 cp = "CP936" "\0" "GBK" "\0"
323 "CP1361" "\0" "JOHAB" "\0"
324 "CP20127" "\0" "ASCII" "\0"
325 "CP20866" "\0" "KOI8-R" "\0"
326 "CP20936" "\0" "GB2312" "\0"
327 "CP21866" "\0" "KOI8-RU" "\0"
328 "CP28591" "\0" "ISO-8859-1" "\0"
329 "CP28592" "\0" "ISO-8859-2" "\0"
330 "CP28593" "\0" "ISO-8859-3" "\0"
331 "CP28594" "\0" "ISO-8859-4" "\0"
332 "CP28595" "\0" "ISO-8859-5" "\0"
333 "CP28596" "\0" "ISO-8859-6" "\0"
334 "CP28597" "\0" "ISO-8859-7" "\0"
335 "CP28598" "\0" "ISO-8859-8" "\0"
336 "CP28599" "\0" "ISO-8859-9" "\0"
337 "CP28605" "\0" "ISO-8859-15" "\0"
338 "CP38598" "\0" "ISO-8859-8" "\0"
339 "CP51932" "\0" "EUC-JP" "\0"
340 "CP51936" "\0" "GB2312" "\0"
341 "CP51949" "\0" "EUC-KR" "\0"
342 "CP51950" "\0" "EUC-TW" "\0"
343 "CP54936" "\0" "GB18030" "\0"
344 "CP65001" "\0" "UTF-8" "\0";
345 # endif
346 #endif
347
348 charset_aliases = cp;
349 }
350
351 return cp;
352 }
353
354 /* Determine the current locale's character encoding, and canonicalize it
355 into one of the canonical names listed in config.charset.
356 The result must not be freed; it is statically allocated.
357 If the canonical name cannot be determined, the result is a non-canonical
358 name. */
359
360 #ifdef STATIC
361 STATIC
362 #endif
363 const char *
locale_charset(void)364 locale_charset (void)
365 {
366 const char *codeset;
367 const char *aliases;
368
369 #if !(defined WINDOWS_NATIVE || defined OS2)
370
371 # if HAVE_LANGINFO_CODESET
372
373 /* Most systems support nl_langinfo (CODESET) nowadays. */
374 codeset = nl_langinfo (CODESET);
375
376 # ifdef __CYGWIN__
377 /* Cygwin < 1.7 does not have locales. nl_langinfo (CODESET) always
378 returns "US-ASCII". Return the suffix of the locale name from the
379 environment variables (if present) or the codepage as a number. */
380 if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
381 {
382 const char *locale;
383 static char buf[2 + 10 + 1];
384
385 locale = getenv ("LC_ALL");
386 if (locale == NULL || locale[0] == '\0')
387 {
388 locale = getenv ("LC_CTYPE");
389 if (locale == NULL || locale[0] == '\0')
390 locale = getenv ("LANG");
391 }
392 if (locale != NULL && locale[0] != '\0')
393 {
394 /* If the locale name contains an encoding after the dot, return
395 it. */
396 const char *dot = strchr (locale, '.');
397
398 if (dot != NULL)
399 {
400 const char *modifier;
401
402 dot++;
403 /* Look for the possible @... trailer and remove it, if any. */
404 modifier = strchr (dot, '@');
405 if (modifier == NULL)
406 return dot;
407 if (modifier - dot < sizeof (buf))
408 {
409 memcpy (buf, dot, modifier - dot);
410 buf [modifier - dot] = '\0';
411 return buf;
412 }
413 }
414 }
415
416 /* The Windows API has a function returning the locale's codepage as a
417 number: GetACP(). This encoding is used by Cygwin, unless the user
418 has set the environment variable CYGWIN=codepage:oem (which very few
419 people do).
420 Output directed to console windows needs to be converted (to
421 GetOEMCP() if the console is using a raster font, or to
422 GetConsoleOutputCP() if it is using a TrueType font). Cygwin does
423 this conversion transparently (see winsup/cygwin/fhandler_console.cc),
424 converting to GetConsoleOutputCP(). This leads to correct results,
425 except when SetConsoleOutputCP has been called and a raster font is
426 in use. */
427 sprintf (buf, "CP%u", GetACP ());
428 codeset = buf;
429 }
430 # endif
431
432 # else
433
434 /* On old systems which lack it, use setlocale or getenv. */
435 const char *locale = NULL;
436
437 /* But most old systems don't have a complete set of locales. Some
438 (like SunOS 4 or DJGPP) have only the C locale. Therefore we don't
439 use setlocale here; it would return "C" when it doesn't support the
440 locale name the user has set. */
441 # if 0
442 locale = setlocale (LC_CTYPE, NULL);
443 # endif
444 if (locale == NULL || locale[0] == '\0')
445 {
446 locale = getenv ("LC_ALL");
447 if (locale == NULL || locale[0] == '\0')
448 {
449 locale = getenv ("LC_CTYPE");
450 if (locale == NULL || locale[0] == '\0')
451 locale = getenv ("LANG");
452 }
453 }
454
455 /* On some old systems, one used to set locale = "iso8859_1". On others,
456 you set it to "language_COUNTRY.charset". In any case, we resolve it
457 through the charset.alias file. */
458 codeset = locale;
459
460 # endif
461
462 #elif defined WINDOWS_NATIVE
463
464 static char buf[2 + 10 + 1];
465
466 /* The Windows API has a function returning the locale's codepage as a
467 number: GetACP().
468 When the output goes to a console window, it needs to be provided in
469 GetOEMCP() encoding if the console is using a raster font, or in
470 GetConsoleOutputCP() encoding if it is using a TrueType font.
471 But in GUI programs and for output sent to files and pipes, GetACP()
472 encoding is the best bet. */
473 sprintf (buf, "CP%u", GetACP ());
474 codeset = buf;
475
476 #elif defined OS2
477
478 const char *locale;
479 static char buf[2 + 10 + 1];
480 ULONG cp[3];
481 ULONG cplen;
482
483 /* Allow user to override the codeset, as set in the operating system,
484 with standard language environment variables. */
485 locale = getenv ("LC_ALL");
486 if (locale == NULL || locale[0] == '\0')
487 {
488 locale = getenv ("LC_CTYPE");
489 if (locale == NULL || locale[0] == '\0')
490 locale = getenv ("LANG");
491 }
492 if (locale != NULL && locale[0] != '\0')
493 {
494 /* If the locale name contains an encoding after the dot, return it. */
495 const char *dot = strchr (locale, '.');
496
497 if (dot != NULL)
498 {
499 const char *modifier;
500
501 dot++;
502 /* Look for the possible @... trailer and remove it, if any. */
503 modifier = strchr (dot, '@');
504 if (modifier == NULL)
505 return dot;
506 if (modifier - dot < sizeof (buf))
507 {
508 memcpy (buf, dot, modifier - dot);
509 buf [modifier - dot] = '\0';
510 return buf;
511 }
512 }
513
514 /* Resolve through the charset.alias file. */
515 codeset = locale;
516 }
517 else
518 {
519 /* OS/2 has a function returning the locale's codepage as a number. */
520 if (DosQueryCp (sizeof (cp), cp, &cplen))
521 codeset = "";
522 else
523 {
524 sprintf (buf, "CP%u", cp[0]);
525 codeset = buf;
526 }
527 }
528
529 #endif
530
531 if (codeset == NULL)
532 /* The canonical name cannot be determined. */
533 codeset = "";
534
535 /* Resolve alias. */
536 for (aliases = get_charset_aliases ();
537 *aliases != '\0';
538 aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
539 if (strcmp (codeset, aliases) == 0
540 || (aliases[0] == '*' && aliases[1] == '\0'))
541 {
542 codeset = aliases + strlen (aliases) + 1;
543 break;
544 }
545
546 /* Don't return an empty string. GNU libc and GNU libiconv interpret
547 the empty string as denoting "the locale's character encoding",
548 thus GNU libiconv would call this function a second time. */
549 if (codeset[0] == '\0')
550 codeset = "ASCII";
551
552 #ifdef DARWIN7
553 /* Mac OS X sets MB_CUR_MAX to 1 when LC_ALL=C, and "UTF-8"
554 (the default codeset) does not work when MB_CUR_MAX is 1. */
555 if (strcmp (codeset, "UTF-8") == 0 && MB_CUR_MAX_L (uselocale (NULL)) <= 1)
556 codeset = "ASCII";
557 #endif
558
559 return codeset;
560 }
561