lib/nl_langinfo.c

   1 /* nl_langinfo() replacement: query locale dependent information.
   2
   3    Copyright (C) 2007-2024 Free Software Foundation, Inc.
   4
   5    This file is free software: you can redistribute it and/or modify
   6    it under the terms of the GNU Lesser General Public License as
   7    published by the Free Software Foundation; either version 2.1 of the
   8    License, or (at your option) any later version.
   9
  10    This file 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.
  14
  15    You should have received a copy of the GNU Lesser General Public License
  16    along with this program.  If not, see <https://www.gnu.org/licenses/>.  */
  17
  18 #include <config.h>
  19
  20 /* Specification.  */
  21 #include <langinfo.h>
  22
  23 #include <locale.h>
  24 #include <stdlib.h>
  25 #include <string.h>
  26 #if defined _WIN32 && ! defined __CYGWIN__
  27 # define WIN32_LEAN_AND_MEAN  /* avoid including junk */
  28 # include <windows.h>
  29 # include <stdio.h>
  30 #endif
  31
  32 #if REPLACE_NL_LANGINFO && !NL_LANGINFO_MTSAFE
  33
  34 # if AVOID_ANY_THREADS
  35
  36 /* The option '--disable-threads' explicitly requests no locking.  */
  37
  38 # elif defined _WIN32 && !defined __CYGWIN__
  39
  40 #  define WIN32_LEAN_AND_MEAN  /* avoid including junk */
  41 #  include <windows.h>
  42
  43 # elif HAVE_PTHREAD_API
  44
  45 #  include <pthread.h>
  46 #  if HAVE_THREADS_H && HAVE_WEAK_SYMBOLS
  47 #   include <threads.h>
  48 #   pragma weak thrd_exit
  49 #   define c11_threads_in_use() (thrd_exit != NULL)
  50 #  else
  51 #   define c11_threads_in_use() 0
  52 #  endif
  53
  54 # elif HAVE_THREADS_H
  55
  56 #  include <threads.h>
  57
  58 # endif
  59
  60 #endif
  61
  62 /* nl_langinfo() must be multithread-safe.  To achieve this without using
  63    thread-local storage:
  64      1. We use a specific static buffer for each possible argument.
  65         So that different threads can call nl_langinfo with different arguments,
  66         without interfering.
  67      2. We use a simple strcpy or memcpy to fill this static buffer.  Filling it
  68         through, for example, strcpy + strcat would not be guaranteed to leave
  69         the buffer's contents intact if another thread is currently accessing
  70         it.  If necessary, the contents is first assembled in a stack-allocated
  71         buffer.  */
  72
  73 #if !REPLACE_NL_LANGINFO || GNULIB_defined_CODESET
  74 /* Return the codeset of the current locale, if this is easily deducible.
  75    Otherwise, return "".  */
  76 static char *
  77 ctype_codeset (void)
  78 {
  79   /* This function is only used on platforms which don't have uselocale().
  80      Therefore we don't need to look at the per-thread locale first, here.  */
  81   static char result[2 + 10 + 1];
  82   char buf[2 + 10 + 1];
  83   char locale[SETLOCALE_NULL_MAX];
  84   char *codeset;
  85   size_t codesetlen;
  86
  87   if (setlocale_null_r (LC_CTYPE, locale, sizeof (locale)))
  88     locale[0] = '\0';
  89
  90   codeset = buf;
  91   codeset[0] = '\0';
  92
  93   if (locale[0])
  94     {
  95       /* If the locale name contains an encoding after the dot, return it.  */
  96       char *dot = strchr (locale, '.');
  97
  98       if (dot)
  99         {
 100           /* Look for the possible @... trailer and remove it, if any.  */
 101           char *codeset_start = dot + 1;
 102           char const *modifier = strchr (codeset_start, '@');
 103
 104           if (! modifier)
 105             codeset = codeset_start;
 106           else
 107             {
 108               codesetlen = modifier - codeset_start;
 109               if (codesetlen < sizeof buf)
 110                 {
 111                   codeset = memcpy (buf, codeset_start, codesetlen);
 112                   codeset[codesetlen] = '\0';
 113                 }
 114             }
 115         }
 116     }
 117
 118 # if defined _WIN32 && ! defined __CYGWIN__
 119   /* If setlocale is successful, it returns the number of the
 120      codepage, as a string.  Otherwise, fall back on Windows API
 121      GetACP, which returns the locale's codepage as a number (although
 122      this doesn't change according to what the 'setlocale' call specified).
 123      Either way, prepend "CP" to make it a valid codeset name.  */
 124   codesetlen = strlen (codeset);
 125   if (0 < codesetlen && codesetlen < sizeof buf - 2)
 126     memmove (buf + 2, codeset, codesetlen + 1);
 127   else
 128     sprintf (buf + 2, "%u", GetACP ());
 129   /* For a locale name such as "French_France.65001", in Windows 10,
 130      setlocale now returns "French_France.utf8" instead.  */
 131   if (strcmp (buf + 2, "65001") == 0 || strcmp (buf + 2, "utf8") == 0)
 132     return (char *) "UTF-8";
 133   else
 134     {
 135       memcpy (buf, "CP", 2);
 136       strcpy (result, buf);
 137       return result;
 138     }
 139 # else
 140   strcpy (result, codeset);
 141   return result;
 142 #endif
 143 }
 144 #endif
 145
 146
 147 #if REPLACE_NL_LANGINFO
 148
 149 /* Override nl_langinfo with support for added nl_item values.  */
 150
 151 # undef nl_langinfo
 152
 153 /* Without locking, on Solaris 11.3, test-nl_langinfo-mt fails, with message
 154    "thread5 disturbed by threadN!", even when threadN invokes only
 155       nl_langinfo (CODESET);
 156       nl_langinfo (CRNCYSTR);
 157    Similarly on Solaris 10.  */
 158
 159 # if !NL_LANGINFO_MTSAFE /* Solaris */
 160
 161 #  define ITEMS (MAXSTRMSG + 1)
 162 #  define MAX_RESULT_LEN 80
 163
 164 static char *
 165 nl_langinfo_unlocked (nl_item item)
 166 {
 167   static char result[ITEMS][MAX_RESULT_LEN];
 168
 169   /* The result of nl_langinfo is in storage that can be overwritten by
 170      other calls to nl_langinfo.  */
 171   char *tmp = nl_langinfo (item);
 172   if (item >= 0 && item < ITEMS && tmp != NULL)
 173     {
 174       size_t tmp_len = strlen (tmp);
 175       if (tmp_len < MAX_RESULT_LEN)
 176         strcpy (result[item], tmp);
 177       else
 178         {
 179           /* Produce a truncated result.  Oh well...  */
 180           result[item][MAX_RESULT_LEN - 1] = '\0';
 181           memcpy (result[item], tmp, MAX_RESULT_LEN - 1);
 182         }
 183       return result[item];
 184     }
 185   else
 186     return tmp;
 187 }
 188
 189 /* Use a lock, so that no two threads can invoke nl_langinfo_unlocked
 190    at the same time.  */
 191
 192 /* Prohibit renaming this symbol.  */
 193 #  undef gl_get_nl_langinfo_lock
 194
 195 #  if AVOID_ANY_THREADS
 196
 197 /* The option '--disable-threads' explicitly requests no locking.  */
 198 #   define nl_langinfo_with_lock nl_langinfo_unlocked
 199
 200 #  elif defined _WIN32 && !defined __CYGWIN__
 201
 202 extern __declspec(dllimport) CRITICAL_SECTION *gl_get_nl_langinfo_lock (void);
 203
 204 static char *
 205 nl_langinfo_with_lock (nl_item item)
 206 {
 207   CRITICAL_SECTION *lock = gl_get_nl_langinfo_lock ();
 208   char *ret;
 209
 210   EnterCriticalSection (lock);
 211   ret = nl_langinfo_unlocked (item);
 212   LeaveCriticalSection (lock);
 213
 214   return ret;
 215 }
 216
 217 #  elif HAVE_PTHREAD_API
 218
 219 extern
 220 #   if defined _WIN32 || defined __CYGWIN__
 221   __declspec(dllimport)
 222 #   endif
 223   pthread_mutex_t *gl_get_nl_langinfo_lock (void);
 224
 225 #   if HAVE_WEAK_SYMBOLS /* musl libc, FreeBSD, NetBSD, OpenBSD, Haiku */
 226
 227      /* Avoid the need to link with '-lpthread'.  */
 228 #    pragma weak pthread_mutex_lock
 229 #    pragma weak pthread_mutex_unlock
 230
 231      /* Determine whether libpthread is in use.  */
 232 #    pragma weak pthread_mutexattr_gettype
 233      /* See the comments in lock.h.  */
 234 #    define pthread_in_use() \
 235        (pthread_mutexattr_gettype != NULL || c11_threads_in_use ())
 236
 237 #   else
 238 #    define pthread_in_use() 1
 239 #   endif
 240
 241 static char *
 242 nl_langinfo_with_lock (nl_item item)
 243 {
 244   if (pthread_in_use())
 245     {
 246       pthread_mutex_t *lock = gl_get_nl_langinfo_lock ();
 247       char *ret;
 248
 249       if (pthread_mutex_lock (lock))
 250         abort ();
 251       ret = nl_langinfo_unlocked (item);
 252       if (pthread_mutex_unlock (lock))
 253         abort ();
 254
 255       return ret;
 256     }
 257   else
 258     return nl_langinfo_unlocked (item);
 259 }
 260
 261 #  elif HAVE_THREADS_H
 262
 263 extern mtx_t *gl_get_nl_langinfo_lock (void);
 264
 265 static char *
 266 nl_langinfo_with_lock (nl_item item)
 267 {
 268   mtx_t *lock = gl_get_nl_langinfo_lock ();
 269   char *ret;
 270
 271   if (mtx_lock (lock) != thrd_success)
 272     abort ();
 273   ret = nl_langinfo_unlocked (item);
 274   if (mtx_unlock (lock) != thrd_success)
 275     abort ();
 276
 277   return ret;
 278 }
 279
 280 #  endif
 281
 282 # else
 283
 284 /* On other platforms, no lock is needed.  */
 285 #  define nl_langinfo_with_lock nl_langinfo
 286
 287 # endif
 288
 289 char *
 290 rpl_nl_langinfo (nl_item item)
 291 {
 292   switch (item)
 293     {
 294 # if GNULIB_defined_CODESET
 295     case CODESET:
 296       return ctype_codeset ();
 297 # endif
 298 # if GNULIB_defined_T_FMT_AMPM
 299     case T_FMT_AMPM:
 300       return (char *) "%I:%M:%S %p";
 301 # endif
 302 # if GNULIB_defined_ALTMON
 303     case ALTMON_1:
 304     case ALTMON_2:
 305     case ALTMON_3:
 306     case ALTMON_4:
 307     case ALTMON_5:
 308     case ALTMON_6:
 309     case ALTMON_7:
 310     case ALTMON_8:
 311     case ALTMON_9:
 312     case ALTMON_10:
 313     case ALTMON_11:
 314     case ALTMON_12:
 315       /* We don't ship the appropriate localizations with gnulib.  Therefore,
 316          treat ALTMON_i like MON_i.  */
 317       item = item - ALTMON_1 + MON_1;
 318       break;
 319 # endif
 320 # if GNULIB_defined_ERA
 321     case ERA:
 322       /* The format is not standardized.  In glibc it is a sequence of strings
 323          of the form "direction:offset:start_date:end_date:era_name:era_format"
 324          with an empty string at the end.  */
 325       return (char *) "";
 326     case ERA_D_FMT:
 327       /* The %Ex conversion in strftime behaves like %x if the locale does not
 328          have an alternative time format.  */
 329       item = D_FMT;
 330       break;
 331     case ERA_D_T_FMT:
 332       /* The %Ec conversion in strftime behaves like %c if the locale does not
 333          have an alternative time format.  */
 334       item = D_T_FMT;
 335       break;
 336     case ERA_T_FMT:
 337       /* The %EX conversion in strftime behaves like %X if the locale does not
 338          have an alternative time format.  */
 339       item = T_FMT;
 340       break;
 341     case ALT_DIGITS:
 342       /* The format is not standardized.  In glibc it is a sequence of 10
 343          strings, appended in memory.  */
 344       return (char *) "\0\0\0\0\0\0\0\0\0\0";
 345 # endif
 346 # if GNULIB_defined_YESEXPR || !FUNC_NL_LANGINFO_YESEXPR_WORKS
 347     case YESEXPR:
 348       return (char *) "^[yY]";
 349     case NOEXPR:
 350       return (char *) "^[nN]";
 351 # endif
 352     default:
 353       break;
 354     }
 355   return nl_langinfo_with_lock (item);
 356 }
 357
 358 #else
 359
 360 /* Provide nl_langinfo from scratch, either for native MS-Windows, or
 361    for old Unix platforms without locales, such as Linux libc5 or
 362    BeOS.  */
 363
 364 # include <time.h>
 365
 366 char *
 367 nl_langinfo (nl_item item)
 368 {
 369   char buf[100];
 370   struct tm tmm = { 0 };
 371
 372   switch (item)
 373     {
 374     /* nl_langinfo items of the LC_CTYPE category */
 375     case CODESET:
 376       {
 377         char *codeset = ctype_codeset ();
 378         if (*codeset)
 379           return codeset;
 380       }
 381 # ifdef __BEOS__
 382       return (char *) "UTF-8";
 383 # else
 384       return (char *) "ISO-8859-1";
 385 # endif
 386     /* nl_langinfo items of the LC_NUMERIC category */
 387     case RADIXCHAR:
 388       return localeconv () ->decimal_point;
 389     case THOUSEP:
 390       return localeconv () ->thousands_sep;
 391 # ifdef GROUPING
 392     case GROUPING:
 393       return localeconv () ->grouping;
 394 # endif
 395     /* nl_langinfo items of the LC_TIME category.
 396        TODO: Really use the locale.  */
 397     case D_T_FMT:
 398     case ERA_D_T_FMT:
 399       return (char *) "%a %b %e %H:%M:%S %Y";
 400     case D_FMT:
 401     case ERA_D_FMT:
 402       return (char *) "%m/%d/%y";
 403     case T_FMT:
 404     case ERA_T_FMT:
 405       return (char *) "%H:%M:%S";
 406     case T_FMT_AMPM:
 407       return (char *) "%I:%M:%S %p";
 408     case AM_STR:
 409       {
 410         static char result[80];
 411         if (!strftime (buf, sizeof result, "%p", &tmm))
 412           return (char *) "AM";
 413         strcpy (result, buf);
 414         return result;
 415       }
 416     case PM_STR:
 417       {
 418         static char result[80];
 419         tmm.tm_hour = 12;
 420         if (!strftime (buf, sizeof result, "%p", &tmm))
 421           return (char *) "PM";
 422         strcpy (result, buf);
 423         return result;
 424       }
 425     case DAY_1:
 426     case DAY_2:
 427     case DAY_3:
 428     case DAY_4:
 429     case DAY_5:
 430     case DAY_6:
 431     case DAY_7:
 432       {
 433         static char result[7][50];
 434         static char const days[][sizeof "Wednesday"] = {
 435           "Sunday", "Monday", "Tuesday", "Wednesday", "Thursday",
 436           "Friday", "Saturday"
 437         };
 438         tmm.tm_wday = item - DAY_1;
 439         if (!strftime (buf, sizeof result[0], "%A", &tmm))
 440           return (char *) days[item - DAY_1];
 441         strcpy (result[item - DAY_1], buf);
 442         return result[item - DAY_1];
 443       }
 444     case ABDAY_1:
 445     case ABDAY_2:
 446     case ABDAY_3:
 447     case ABDAY_4:
 448     case ABDAY_5:
 449     case ABDAY_6:
 450     case ABDAY_7:
 451       {
 452         static char result[7][30];
 453         static char const abdays[][sizeof "Sun"] = {
 454           "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"
 455         };
 456         tmm.tm_wday = item - ABDAY_1;
 457         if (!strftime (buf, sizeof result[0], "%a", &tmm))
 458           return (char *) abdays[item - ABDAY_1];
 459         strcpy (result[item - ABDAY_1], buf);
 460         return result[item - ABDAY_1];
 461       }
 462     {
 463       static char const months[][sizeof "September"] = {
 464         "January", "February", "March", "April", "May", "June", "July",
 465         "August", "September", "October", "November", "December"
 466       };
 467       case MON_1:
 468       case MON_2:
 469       case MON_3:
 470       case MON_4:
 471       case MON_5:
 472       case MON_6:
 473       case MON_7:
 474       case MON_8:
 475       case MON_9:
 476       case MON_10:
 477       case MON_11:
 478       case MON_12:
 479         {
 480           static char result[12][50];
 481           tmm.tm_mon = item - MON_1;
 482           if (!strftime (buf, sizeof result[0], "%B", &tmm))
 483             return (char *) months[item - MON_1];
 484           strcpy (result[item - MON_1], buf);
 485           return result[item - MON_1];
 486         }
 487       case ALTMON_1:
 488       case ALTMON_2:
 489       case ALTMON_3:
 490       case ALTMON_4:
 491       case ALTMON_5:
 492       case ALTMON_6:
 493       case ALTMON_7:
 494       case ALTMON_8:
 495       case ALTMON_9:
 496       case ALTMON_10:
 497       case ALTMON_11:
 498       case ALTMON_12:
 499         {
 500           static char result[12][50];
 501           tmm.tm_mon = item - ALTMON_1;
 502           /* The platforms without nl_langinfo() don't support strftime with
 503              %OB.  We don't even need to try.  */
 504           #if 0
 505           if (!strftime (buf, sizeof result[0], "%OB", &tmm))
 506           #endif
 507             if (!strftime (buf, sizeof result[0], "%B", &tmm))
 508               return (char *) months[item - ALTMON_1];
 509           strcpy (result[item - ALTMON_1], buf);
 510           return result[item - ALTMON_1];
 511         }
 512     }
 513     case ABMON_1:
 514     case ABMON_2:
 515     case ABMON_3:
 516     case ABMON_4:
 517     case ABMON_5:
 518     case ABMON_6:
 519     case ABMON_7:
 520     case ABMON_8:
 521     case ABMON_9:
 522     case ABMON_10:
 523     case ABMON_11:
 524     case ABMON_12:
 525       {
 526         static char result[12][30];
 527         static char const abmonths[][sizeof "Jan"] = {
 528           "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul",
 529           "Aug", "Sep", "Oct", "Nov", "Dec"
 530         };
 531         tmm.tm_mon = item - ABMON_1;
 532         if (!strftime (buf, sizeof result[0], "%b", &tmm))
 533           return (char *) abmonths[item - ABMON_1];
 534         strcpy (result[item - ABMON_1], buf);
 535         return result[item - ABMON_1];
 536       }
 537     case ERA:
 538       return (char *) "";
 539     case ALT_DIGITS:
 540       return (char *) "\0\0\0\0\0\0\0\0\0\0";
 541     /* nl_langinfo items of the LC_MONETARY category.  */
 542     case CRNCYSTR:
 543       return localeconv () ->currency_symbol;
 544 # ifdef INT_CURR_SYMBOL
 545     case INT_CURR_SYMBOL:
 546       return localeconv () ->int_curr_symbol;
 547     case MON_DECIMAL_POINT:
 548       return localeconv () ->mon_decimal_point;
 549     case MON_THOUSANDS_SEP:
 550       return localeconv () ->mon_thousands_sep;
 551     case MON_GROUPING:
 552       return localeconv () ->mon_grouping;
 553     case POSITIVE_SIGN:
 554       return localeconv () ->positive_sign;
 555     case NEGATIVE_SIGN:
 556       return localeconv () ->negative_sign;
 557     case FRAC_DIGITS:
 558       return & localeconv () ->frac_digits;
 559     case INT_FRAC_DIGITS:
 560       return & localeconv () ->int_frac_digits;
 561     case P_CS_PRECEDES:
 562       return & localeconv () ->p_cs_precedes;
 563     case N_CS_PRECEDES:
 564       return & localeconv () ->n_cs_precedes;
 565     case P_SEP_BY_SPACE:
 566       return & localeconv () ->p_sep_by_space;
 567     case N_SEP_BY_SPACE:
 568       return & localeconv () ->n_sep_by_space;
 569     case P_SIGN_POSN:
 570       return & localeconv () ->p_sign_posn;
 571     case N_SIGN_POSN:
 572       return & localeconv () ->n_sign_posn;
 573 # endif
 574     /* nl_langinfo items of the LC_MESSAGES category
 575        TODO: Really use the locale. */
 576     case YESEXPR:
 577       return (char *) "^[yY]";
 578     case NOEXPR:
 579       return (char *) "^[nN]";
 580     default:
 581       return (char *) "";
 582     }
 583 }
 584
 585 #endif