1/* Determine a canonical name for the current locale's character encoding.
2
3   Copyright (C) 2000-2006 Free Software Foundation, Inc.
4
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)
8   any later version.
9
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.
14
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,
18   USA.  */
19
20/* Written by Bruno Haible <bruno@clisp.org>.  */
21
22#ifdef HAVE_CONFIG_H
23# include <config.h>
24#endif
25
26/* Specification.  */
27#include "localcharset.h"
28
29#if HAVE_STDDEF_H
30# include <stddef.h>
31#endif
32
33#include <stdio.h>
34#if HAVE_STRING_H
35# include <string.h>
36#else
37# include <strings.h>
38#endif
39#if HAVE_STDLIB_H
40# include <stdlib.h>
41#endif
42
43#if defined _WIN32 || defined __WIN32__
44# define WIN32_NATIVE
45#endif
46
47#if defined __EMX__
48/* Assume EMX program runs on OS/2, even if compiled under DOS.  */
49# define OS2
50#endif
51
52#if !defined WIN32_NATIVE
53# if HAVE_LANGINFO_CODESET
54#  include <langinfo.h>
55# else
56#  if HAVE_SETLOCALE
57#   include <locale.h>
58#  endif
59# endif
60# ifdef __CYGWIN__
61#  define WIN32_LEAN_AND_MEAN
62#  include <windows.h>
63# endif
64#elif defined WIN32_NATIVE
65# define WIN32_LEAN_AND_MEAN
66# include <windows.h>
67#endif
68#if defined OS2
69# define INCL_DOS
70# include <os2.h>
71#endif
72
73#if ENABLE_RELOCATABLE
74# include "relocatable.h"
75#else
76# define relocate(pathname) (pathname)
77#endif
78
79#if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
80  /* Win32, Cygwin, OS/2, DOS */
81# define ISSLASH(C) ((C) == '/' || (C) == '\\')
82#endif
83
84#ifndef DIRECTORY_SEPARATOR
85# define DIRECTORY_SEPARATOR '/'
86#endif
87
88#ifndef ISSLASH
89# define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
90#endif
91
92#if HAVE_DECL_GETC_UNLOCKED
93# undef getc
94# define getc getc_unlocked
95#endif
96
97/* The following static variable is declared 'volatile' to avoid a
98   possible multithread problem in the function get_charset_aliases. If we
99   are running in a threaded environment, and if two threads initialize
100   'charset_aliases' simultaneously, both will produce the same value,
101   and everything will be ok if the two assignments to 'charset_aliases'
102   are atomic. But I don't know what will happen if the two assignments mix.  */
103#if __STDC__ != 1
104# define volatile /* empty */
105#endif
106/* Pointer to the contents of the charset.alias file, if it has already been
107   read, else NULL.  Its format is:
108   ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0'  */
109static const char * volatile charset_aliases;
110
111/* Return a pointer to the contents of the charset.alias file.  */
112static const char *
113get_charset_aliases (void)
114{
115  const char *cp;
116
117  cp = charset_aliases;
118  if (cp == NULL)
119    {
120#if !(defined VMS || defined WIN32_NATIVE || defined __CYGWIN__)
121      FILE *fp;
122      const char *dir;
123      const char *base = "charset.alias";
124      char *file_name;
125
126      /* Make it possible to override the charset.alias location.  This is
127	 necessary for running the testsuite before "make install".  */
128      dir = getenv ("CHARSETALIASDIR");
129      if (dir == NULL || dir[0] == '\0')
130	dir = relocate (LIBDIR);
131
132      /* Concatenate dir and base into freshly allocated file_name.  */
133      {
134	size_t dir_len = strlen (dir);
135	size_t base_len = strlen (base);
136	int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
137	file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
138	if (file_name != NULL)
139	  {
140	    memcpy (file_name, dir, dir_len);
141	    if (add_slash)
142	      file_name[dir_len] = DIRECTORY_SEPARATOR;
143	    memcpy (file_name + dir_len + add_slash, base, base_len + 1);
144	  }
145      }
146
147      if (file_name == NULL || (fp = fopen (file_name, "r")) == NULL)
148	/* Out of memory or file not found, treat it as empty.  */
149	cp = "";
150      else
151	{
152	  /* Parse the file's contents.  */
153	  char *res_ptr = NULL;
154	  size_t res_size = 0;
155
156	  for (;;)
157	    {
158	      int c;
159	      char buf1[50+1];
160	      char buf2[50+1];
161	      size_t l1, l2;
162	      char *old_res_ptr;
163
164	      c = getc (fp);
165	      if (c == EOF)
166		break;
167	      if (c == '\n' || c == ' ' || c == '\t')
168		continue;
169	      if (c == '#')
170		{
171		  /* Skip comment, to end of line.  */
172		  do
173		    c = getc (fp);
174		  while (!(c == EOF || c == '\n'));
175		  if (c == EOF)
176		    break;
177		  continue;
178		}
179	      ungetc (c, fp);
180	      if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
181		break;
182	      l1 = strlen (buf1);
183	      l2 = strlen (buf2);
184	      old_res_ptr = res_ptr;
185	      if (res_size == 0)
186		{
187		  res_size = l1 + 1 + l2 + 1;
188		  res_ptr = (char *) malloc (res_size + 1);
189		}
190	      else
191		{
192		  res_size += l1 + 1 + l2 + 1;
193		  res_ptr = (char *) realloc (res_ptr, res_size + 1);
194		}
195	      if (res_ptr == NULL)
196		{
197		  /* Out of memory. */
198		  res_size = 0;
199		  if (old_res_ptr != NULL)
200		    free (old_res_ptr);
201		  break;
202		}
203	      strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
204	      strcpy (res_ptr + res_size - (l2 + 1), buf2);
205	    }
206	  fclose (fp);
207	  if (res_size == 0)
208	    cp = "";
209	  else
210	    {
211	      *(res_ptr + res_size) = '\0';
212	      cp = res_ptr;
213	    }
214	}
215
216      if (file_name != NULL)
217	free (file_name);
218
219#else
220
221# if defined VMS
222      /* To avoid the troubles of an extra file charset.alias_vms in the
223	 sources of many GNU packages, simply inline the aliases here.  */
224      /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
225	 "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
226	 section 10.7 "Handling Different Character Sets".  */
227      cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
228	   "ISO8859-2" "\0" "ISO-8859-2" "\0"
229	   "ISO8859-5" "\0" "ISO-8859-5" "\0"
230	   "ISO8859-7" "\0" "ISO-8859-7" "\0"
231	   "ISO8859-8" "\0" "ISO-8859-8" "\0"
232	   "ISO8859-9" "\0" "ISO-8859-9" "\0"
233	   /* Japanese */
234	   "eucJP" "\0" "EUC-JP" "\0"
235	   "SJIS" "\0" "SHIFT_JIS" "\0"
236	   "DECKANJI" "\0" "DEC-KANJI" "\0"
237	   "SDECKANJI" "\0" "EUC-JP" "\0"
238	   /* Chinese */
239	   "eucTW" "\0" "EUC-TW" "\0"
240	   "DECHANYU" "\0" "DEC-HANYU" "\0"
241	   "DECHANZI" "\0" "GB2312" "\0"
242	   /* Korean */
243	   "DECKOREAN" "\0" "EUC-KR" "\0";
244# endif
245
246# if defined WIN32_NATIVE || defined __CYGWIN__
247      /* To avoid the troubles of installing a separate file in the same
248	 directory as the DLL and of retrieving the DLL's directory at
249	 runtime, simply inline the aliases here.  */
250
251      cp = "CP936" "\0" "GBK" "\0"
252	   "CP1361" "\0" "JOHAB" "\0"
253	   "CP20127" "\0" "ASCII" "\0"
254	   "CP20866" "\0" "KOI8-R" "\0"
255	   "CP20936" "\0" "GB2312" "\0"
256	   "CP21866" "\0" "KOI8-RU" "\0"
257	   "CP28591" "\0" "ISO-8859-1" "\0"
258	   "CP28592" "\0" "ISO-8859-2" "\0"
259	   "CP28593" "\0" "ISO-8859-3" "\0"
260	   "CP28594" "\0" "ISO-8859-4" "\0"
261	   "CP28595" "\0" "ISO-8859-5" "\0"
262	   "CP28596" "\0" "ISO-8859-6" "\0"
263	   "CP28597" "\0" "ISO-8859-7" "\0"
264	   "CP28598" "\0" "ISO-8859-8" "\0"
265	   "CP28599" "\0" "ISO-8859-9" "\0"
266	   "CP28605" "\0" "ISO-8859-15" "\0"
267	   "CP38598" "\0" "ISO-8859-8" "\0"
268	   "CP51932" "\0" "EUC-JP" "\0"
269	   "CP51936" "\0" "GB2312" "\0"
270	   "CP51949" "\0" "EUC-KR" "\0"
271	   "CP51950" "\0" "EUC-TW" "\0"
272	   "CP54936" "\0" "GB18030" "\0"
273	   "CP65001" "\0" "UTF-8" "\0";
274# endif
275#endif
276
277      charset_aliases = cp;
278    }
279
280  return cp;
281}
282
283/* Determine the current locale's character encoding, and canonicalize it
284   into one of the canonical names listed in config.charset.
285   The result must not be freed; it is statically allocated.
286   If the canonical name cannot be determined, the result is a non-canonical
287   name.  */
288
289#ifdef STATIC
290STATIC
291#endif
292const char *
293locale_charset (void)
294{
295  const char *codeset;
296  const char *aliases;
297
298#if !(defined WIN32_NATIVE || defined OS2)
299
300# if HAVE_LANGINFO_CODESET
301
302  /* Most systems support nl_langinfo (CODESET) nowadays.  */
303  codeset = nl_langinfo (CODESET);
304
305#  ifdef __CYGWIN__
306  /* Cygwin 2006 does not have locales.  nl_langinfo (CODESET) always
307     returns "US-ASCII".  As long as this is not fixed, return the suffix
308     of the locale name from the environment variables (if present) or
309     the codepage as a number.  */
310  if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
311    {
312      const char *locale;
313      static char buf[2 + 10 + 1];
314
315      locale = getenv ("LC_ALL");
316      if (locale == NULL || locale[0] == '\0')
317	{
318	  locale = getenv ("LC_CTYPE");
319	  if (locale == NULL || locale[0] == '\0')
320	    locale = getenv ("LANG");
321	}
322      if (locale != NULL && locale[0] != '\0')
323	{
324	  /* If the locale name contains an encoding after the dot, return
325	     it.  */
326	  const char *dot = strchr (locale, '.');
327
328	  if (dot != NULL)
329	    {
330	      const char *modifier;
331
332	      dot++;
333	      /* Look for the possible @... trailer and remove it, if any.  */
334	      modifier = strchr (dot, '@');
335	      if (modifier == NULL)
336		return dot;
337	      if (modifier - dot < sizeof (buf))
338		{
339		  memcpy (buf, dot, modifier - dot);
340		  buf [modifier - dot] = '\0';
341		  return buf;
342		}
343	    }
344	}
345
346      /* Woe32 has a function returning the locale's codepage as a number.  */
347      sprintf (buf, "CP%u", GetACP ());
348      codeset = buf;
349    }
350#  endif
351
352# else
353
354  /* On old systems which lack it, use setlocale or getenv.  */
355  const char *locale = NULL;
356
357  /* But most old systems don't have a complete set of locales.  Some
358     (like SunOS 4 or DJGPP) have only the C locale.  Therefore we don't
359     use setlocale here; it would return "C" when it doesn't support the
360     locale name the user has set.  */
361#  if HAVE_SETLOCALE && 0
362  locale = setlocale (LC_CTYPE, NULL);
363#  endif
364  if (locale == NULL || locale[0] == '\0')
365    {
366      locale = getenv ("LC_ALL");
367      if (locale == NULL || locale[0] == '\0')
368	{
369	  locale = getenv ("LC_CTYPE");
370	  if (locale == NULL || locale[0] == '\0')
371	    locale = getenv ("LANG");
372	}
373    }
374
375  /* On some old systems, one used to set locale = "iso8859_1". On others,
376     you set it to "language_COUNTRY.charset". In any case, we resolve it
377     through the charset.alias file.  */
378  codeset = locale;
379
380# endif
381
382#elif defined WIN32_NATIVE
383
384  static char buf[2 + 10 + 1];
385
386  /* Woe32 has a function returning the locale's codepage as a number.  */
387  sprintf (buf, "CP%u", GetACP ());
388  codeset = buf;
389
390#elif defined OS2
391
392  const char *locale;
393  static char buf[2 + 10 + 1];
394  ULONG cp[3];
395  ULONG cplen;
396
397  /* Allow user to override the codeset, as set in the operating system,
398     with standard language environment variables.  */
399  locale = getenv ("LC_ALL");
400  if (locale == NULL || locale[0] == '\0')
401    {
402      locale = getenv ("LC_CTYPE");
403      if (locale == NULL || locale[0] == '\0')
404	locale = getenv ("LANG");
405    }
406  if (locale != NULL && locale[0] != '\0')
407    {
408      /* If the locale name contains an encoding after the dot, return it.  */
409      const char *dot = strchr (locale, '.');
410
411      if (dot != NULL)
412	{
413	  const char *modifier;
414
415	  dot++;
416	  /* Look for the possible @... trailer and remove it, if any.  */
417	  modifier = strchr (dot, '@');
418	  if (modifier == NULL)
419	    return dot;
420	  if (modifier - dot < sizeof (buf))
421	    {
422	      memcpy (buf, dot, modifier - dot);
423	      buf [modifier - dot] = '\0';
424	      return buf;
425	    }
426	}
427
428      /* Resolve through the charset.alias file.  */
429      codeset = locale;
430    }
431  else
432    {
433      /* OS/2 has a function returning the locale's codepage as a number.  */
434      if (DosQueryCp (sizeof (cp), cp, &cplen))
435	codeset = "";
436      else
437	{
438	  sprintf (buf, "CP%u", cp[0]);
439	  codeset = buf;
440	}
441    }
442
443#endif
444
445  if (codeset == NULL)
446    /* The canonical name cannot be determined.  */
447    codeset = "";
448
449  /* Resolve alias. */
450  for (aliases = get_charset_aliases ();
451       *aliases != '\0';
452       aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
453    if (strcmp (codeset, aliases) == 0
454	|| (aliases[0] == '*' && aliases[1] == '\0'))
455      {
456	codeset = aliases + strlen (aliases) + 1;
457	break;
458      }
459
460  /* Don't return an empty string.  GNU libc and GNU libiconv interpret
461     the empty string as denoting "the locale's character encoding",
462     thus GNU libiconv would call this function a second time.  */
463  if (codeset[0] == '\0')
464    codeset = "ASCII";
465
466  return codeset;
467}
468