| 1 | /* locale.c |
| 2 | * |
| 3 | * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, |
| 4 | * 2002, 2003, 2005, 2006, 2007, 2008 by Larry Wall and others |
| 5 | * |
| 6 | * You may distribute under the terms of either the GNU General Public |
| 7 | * License or the Artistic License, as specified in the README file. |
| 8 | * |
| 9 | */ |
| 10 | |
| 11 | /* |
| 12 | * A Elbereth Gilthoniel, |
| 13 | * silivren penna míriel |
| 14 | * o menel aglar elenath! |
| 15 | * Na-chaered palan-díriel |
| 16 | * o galadhremmin ennorath, |
| 17 | * Fanuilos, le linnathon |
| 18 | * nef aear, si nef aearon! |
| 19 | * |
| 20 | * [p.238 of _The Lord of the Rings_, II/i: "Many Meetings"] |
| 21 | */ |
| 22 | |
| 23 | /* utility functions for handling locale-specific stuff like what |
| 24 | * character represents the decimal point. |
| 25 | * |
| 26 | * All C programs have an underlying locale. Perl code generally doesn't pay |
| 27 | * any attention to it except within the scope of a 'use locale'. For most |
| 28 | * categories, it accomplishes this by just using different operations if it is |
| 29 | * in such scope than if not. However, various libc functions called by Perl |
| 30 | * are affected by the LC_NUMERIC category, so there are macros in perl.h that |
| 31 | * are used to toggle between the current locale and the C locale depending on |
| 32 | * the desired behavior of those functions at the moment. And, LC_MESSAGES is |
| 33 | * switched to the C locale for outputting the message unless within the scope |
| 34 | * of 'use locale'. |
| 35 | * |
| 36 | * This code now has multi-thread-safe locale handling on systems that support |
| 37 | * that. This is completely transparent to most XS code. On earlier systems, |
| 38 | * it would be possible to emulate thread-safe locales, but this likely would |
| 39 | * involve a lot of locale switching, and would require XS code changes. |
| 40 | * Macros could be written so that the code wouldn't have to know which type of |
| 41 | * system is being used. It's unlikely that we would ever do that, since most |
| 42 | * modern systems support thread-safe locales, but there was code written to |
| 43 | * this end, and is retained, #ifdef'd out. |
| 44 | */ |
| 45 | |
| 46 | #include "EXTERN.h" |
| 47 | #define PERL_IN_LOCALE_C |
| 48 | #include "perl_langinfo.h" |
| 49 | #include "perl.h" |
| 50 | |
| 51 | #include "reentr.h" |
| 52 | |
| 53 | #ifdef I_WCHAR |
| 54 | # include <wchar.h> |
| 55 | #endif |
| 56 | #ifdef I_WCTYPE |
| 57 | # include <wctype.h> |
| 58 | #endif |
| 59 | |
| 60 | /* If the environment says to, we can output debugging information during |
| 61 | * initialization. This is done before option parsing, and before any thread |
| 62 | * creation, so can be a file-level static */ |
| 63 | #if ! defined(DEBUGGING) || defined(PERL_GLOBAL_STRUCT) |
| 64 | # define debug_initialization 0 |
| 65 | # define DEBUG_INITIALIZATION_set(v) |
| 66 | #else |
| 67 | static bool debug_initialization = FALSE; |
| 68 | # define DEBUG_INITIALIZATION_set(v) (debug_initialization = v) |
| 69 | #endif |
| 70 | |
| 71 | |
| 72 | /* Returns the Unix errno portion; ignoring any others. This is a macro here |
| 73 | * instead of putting it into perl.h, because unclear to khw what should be |
| 74 | * done generally. */ |
| 75 | #define GET_ERRNO saved_errno |
| 76 | |
| 77 | /* strlen() of a literal string constant. We might want this more general, |
| 78 | * but using it in just this file for now. A problem with more generality is |
| 79 | * the compiler warnings about comparing unlike signs */ |
| 80 | #define STRLENs(s) (sizeof("" s "") - 1) |
| 81 | |
| 82 | /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the |
| 83 | * return of setlocale(), then this is extremely likely to be the C or POSIX |
| 84 | * locale. However, the output of setlocale() is documented to be opaque, but |
| 85 | * the odds are extremely small that it would return these two strings for some |
| 86 | * other locale. Note that VMS in these two locales includes many non-ASCII |
| 87 | * characters as controls and punctuation (below are hex bytes): |
| 88 | * cntrl: 84-97 9B-9F |
| 89 | * punct: A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD |
| 90 | * Oddly, none there are listed as alphas, though some represent alphabetics |
| 91 | * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */ |
| 92 | #define isNAME_C_OR_POSIX(name) \ |
| 93 | ( (name) != NULL \ |
| 94 | && (( *(name) == 'C' && (*(name + 1)) == '\0') \ |
| 95 | || strEQ((name), "POSIX"))) |
| 96 | |
| 97 | #ifdef USE_LOCALE |
| 98 | |
| 99 | /* This code keeps a LRU cache of the UTF-8ness of the locales it has so-far |
| 100 | * looked up. This is in the form of a C string: */ |
| 101 | |
| 102 | #define UTF8NESS_SEP "\v" |
| 103 | #define UTF8NESS_PREFIX "\f" |
| 104 | |
| 105 | /* So, the string looks like: |
| 106 | * |
| 107 | * \vC\a0\vPOSIX\a0\vam_ET\a0\vaf_ZA.utf8\a1\ven_US.UTF-8\a1\0 |
| 108 | * |
| 109 | * where the digit 0 after the \a indicates that the locale starting just |
| 110 | * after the preceding \v is not UTF-8, and the digit 1 mean it is. */ |
| 111 | |
| 112 | STATIC_ASSERT_DECL(STRLENs(UTF8NESS_SEP) == 1); |
| 113 | STATIC_ASSERT_DECL(STRLENs(UTF8NESS_PREFIX) == 1); |
| 114 | |
| 115 | #define C_and_POSIX_utf8ness UTF8NESS_SEP "C" UTF8NESS_PREFIX "0" \ |
| 116 | UTF8NESS_SEP "POSIX" UTF8NESS_PREFIX "0" |
| 117 | |
| 118 | /* The cache is initialized to C_and_POSIX_utf8ness at start up. These are |
| 119 | * kept there always. The remining portion of the cache is LRU, with the |
| 120 | * oldest looked-up locale at the tail end */ |
| 121 | |
| 122 | STATIC char * |
| 123 | S_stdize_locale(pTHX_ char *locs) |
| 124 | { |
| 125 | /* Standardize the locale name from a string returned by 'setlocale', |
| 126 | * possibly modifying that string. |
| 127 | * |
| 128 | * The typical return value of setlocale() is either |
| 129 | * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL |
| 130 | * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL |
| 131 | * (the space-separated values represent the various sublocales, |
| 132 | * in some unspecified order). This is not handled by this function. |
| 133 | * |
| 134 | * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n", |
| 135 | * which is harmful for further use of the string in setlocale(). This |
| 136 | * function removes the trailing new line and everything up through the '=' |
| 137 | * */ |
| 138 | |
| 139 | const char * const s = strchr(locs, '='); |
| 140 | bool okay = TRUE; |
| 141 | |
| 142 | PERL_ARGS_ASSERT_STDIZE_LOCALE; |
| 143 | |
| 144 | if (s) { |
| 145 | const char * const t = strchr(s, '.'); |
| 146 | okay = FALSE; |
| 147 | if (t) { |
| 148 | const char * const u = strchr(t, '\n'); |
| 149 | if (u && (u[1] == 0)) { |
| 150 | const STRLEN len = u - s; |
| 151 | Move(s + 1, locs, len, char); |
| 152 | locs[len] = 0; |
| 153 | okay = TRUE; |
| 154 | } |
| 155 | } |
| 156 | } |
| 157 | |
| 158 | if (!okay) |
| 159 | Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs); |
| 160 | |
| 161 | return locs; |
| 162 | } |
| 163 | |
| 164 | /* Two parallel arrays; first the locale categories Perl uses on this system; |
| 165 | * the second array is their names. These arrays are in mostly arbitrary |
| 166 | * order. */ |
| 167 | |
| 168 | const int categories[] = { |
| 169 | |
| 170 | # ifdef USE_LOCALE_NUMERIC |
| 171 | LC_NUMERIC, |
| 172 | # endif |
| 173 | # ifdef USE_LOCALE_CTYPE |
| 174 | LC_CTYPE, |
| 175 | # endif |
| 176 | # ifdef USE_LOCALE_COLLATE |
| 177 | LC_COLLATE, |
| 178 | # endif |
| 179 | # ifdef USE_LOCALE_TIME |
| 180 | LC_TIME, |
| 181 | # endif |
| 182 | # ifdef USE_LOCALE_MESSAGES |
| 183 | LC_MESSAGES, |
| 184 | # endif |
| 185 | # ifdef USE_LOCALE_MONETARY |
| 186 | LC_MONETARY, |
| 187 | # endif |
| 188 | # ifdef USE_LOCALE_ADDRESS |
| 189 | LC_ADDRESS, |
| 190 | # endif |
| 191 | # ifdef USE_LOCALE_IDENTIFICATION |
| 192 | LC_IDENTIFICATION, |
| 193 | # endif |
| 194 | # ifdef USE_LOCALE_MEASUREMENT |
| 195 | LC_MEASUREMENT, |
| 196 | # endif |
| 197 | # ifdef USE_LOCALE_PAPER |
| 198 | LC_PAPER, |
| 199 | # endif |
| 200 | # ifdef USE_LOCALE_TELEPHONE |
| 201 | LC_TELEPHONE, |
| 202 | # endif |
| 203 | # ifdef LC_ALL |
| 204 | LC_ALL, |
| 205 | # endif |
| 206 | -1 /* Placeholder because C doesn't allow a |
| 207 | trailing comma, and it would get complicated |
| 208 | with all the #ifdef's */ |
| 209 | }; |
| 210 | |
| 211 | /* The top-most real element is LC_ALL */ |
| 212 | |
| 213 | const char * const category_names[] = { |
| 214 | |
| 215 | # ifdef USE_LOCALE_NUMERIC |
| 216 | "LC_NUMERIC", |
| 217 | # endif |
| 218 | # ifdef USE_LOCALE_CTYPE |
| 219 | "LC_CTYPE", |
| 220 | # endif |
| 221 | # ifdef USE_LOCALE_COLLATE |
| 222 | "LC_COLLATE", |
| 223 | # endif |
| 224 | # ifdef USE_LOCALE_TIME |
| 225 | "LC_TIME", |
| 226 | # endif |
| 227 | # ifdef USE_LOCALE_MESSAGES |
| 228 | "LC_MESSAGES", |
| 229 | # endif |
| 230 | # ifdef USE_LOCALE_MONETARY |
| 231 | "LC_MONETARY", |
| 232 | # endif |
| 233 | # ifdef USE_LOCALE_ADDRESS |
| 234 | "LC_ADDRESS", |
| 235 | # endif |
| 236 | # ifdef USE_LOCALE_IDENTIFICATION |
| 237 | "LC_IDENTIFICATION", |
| 238 | # endif |
| 239 | # ifdef USE_LOCALE_MEASUREMENT |
| 240 | "LC_MEASUREMENT", |
| 241 | # endif |
| 242 | # ifdef USE_LOCALE_PAPER |
| 243 | "LC_PAPER", |
| 244 | # endif |
| 245 | # ifdef USE_LOCALE_TELEPHONE |
| 246 | "LC_TELEPHONE", |
| 247 | # endif |
| 248 | # ifdef LC_ALL |
| 249 | "LC_ALL", |
| 250 | # endif |
| 251 | NULL /* Placeholder */ |
| 252 | }; |
| 253 | |
| 254 | # ifdef LC_ALL |
| 255 | |
| 256 | /* On systems with LC_ALL, it is kept in the highest index position. (-2 |
| 257 | * to account for the final unused placeholder element.) */ |
| 258 | # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 2) |
| 259 | |
| 260 | # else |
| 261 | |
| 262 | /* On systems without LC_ALL, we pretend it is there, one beyond the real |
| 263 | * top element, hence in the unused placeholder element. */ |
| 264 | # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 1) |
| 265 | |
| 266 | # endif |
| 267 | |
| 268 | /* Pretending there is an LC_ALL element just above allows us to avoid most |
| 269 | * special cases. Most loops through these arrays in the code below are |
| 270 | * written like 'for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++)'. They will work |
| 271 | * on either type of system. But the code must be written to not access the |
| 272 | * element at 'LC_ALL_INDEX' except on platforms that have it. This can be |
| 273 | * checked for at compile time by using the #define LC_ALL_INDEX which is only |
| 274 | * defined if we do have LC_ALL. */ |
| 275 | |
| 276 | STATIC const char * |
| 277 | S_category_name(const int category) |
| 278 | { |
| 279 | unsigned int i; |
| 280 | |
| 281 | #ifdef LC_ALL |
| 282 | |
| 283 | if (category == LC_ALL) { |
| 284 | return "LC_ALL"; |
| 285 | } |
| 286 | |
| 287 | #endif |
| 288 | |
| 289 | for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) { |
| 290 | if (category == categories[i]) { |
| 291 | return category_names[i]; |
| 292 | } |
| 293 | } |
| 294 | |
| 295 | { |
| 296 | const char suffix[] = " (unknown)"; |
| 297 | int temp = category; |
| 298 | Size_t length = sizeof(suffix) + 1; |
| 299 | char * unknown; |
| 300 | dTHX; |
| 301 | |
| 302 | if (temp < 0) { |
| 303 | length++; |
| 304 | temp = - temp; |
| 305 | } |
| 306 | |
| 307 | /* Calculate the number of digits */ |
| 308 | while (temp >= 10) { |
| 309 | temp /= 10; |
| 310 | length++; |
| 311 | } |
| 312 | |
| 313 | Newx(unknown, length, char); |
| 314 | my_snprintf(unknown, length, "%d%s", category, suffix); |
| 315 | SAVEFREEPV(unknown); |
| 316 | return unknown; |
| 317 | } |
| 318 | } |
| 319 | |
| 320 | /* Now create LC_foo_INDEX #defines for just those categories on this system */ |
| 321 | # ifdef USE_LOCALE_NUMERIC |
| 322 | # define LC_NUMERIC_INDEX 0 |
| 323 | # define _DUMMY_NUMERIC LC_NUMERIC_INDEX |
| 324 | # else |
| 325 | # define _DUMMY_NUMERIC -1 |
| 326 | # endif |
| 327 | # ifdef USE_LOCALE_CTYPE |
| 328 | # define LC_CTYPE_INDEX _DUMMY_NUMERIC + 1 |
| 329 | # define _DUMMY_CTYPE LC_CTYPE_INDEX |
| 330 | # else |
| 331 | # define _DUMMY_CTYPE _DUMMY_NUMERIC |
| 332 | # endif |
| 333 | # ifdef USE_LOCALE_COLLATE |
| 334 | # define LC_COLLATE_INDEX _DUMMY_CTYPE + 1 |
| 335 | # define _DUMMY_COLLATE LC_COLLATE_INDEX |
| 336 | # else |
| 337 | # define _DUMMY_COLLATE _DUMMY_CTYPE |
| 338 | # endif |
| 339 | # ifdef USE_LOCALE_TIME |
| 340 | # define LC_TIME_INDEX _DUMMY_COLLATE + 1 |
| 341 | # define _DUMMY_TIME LC_TIME_INDEX |
| 342 | # else |
| 343 | # define _DUMMY_TIME _DUMMY_COLLATE |
| 344 | # endif |
| 345 | # ifdef USE_LOCALE_MESSAGES |
| 346 | # define LC_MESSAGES_INDEX _DUMMY_TIME + 1 |
| 347 | # define _DUMMY_MESSAGES LC_MESSAGES_INDEX |
| 348 | # else |
| 349 | # define _DUMMY_MESSAGES _DUMMY_TIME |
| 350 | # endif |
| 351 | # ifdef USE_LOCALE_MONETARY |
| 352 | # define LC_MONETARY_INDEX _DUMMY_MESSAGES + 1 |
| 353 | # define _DUMMY_MONETARY LC_MONETARY_INDEX |
| 354 | # else |
| 355 | # define _DUMMY_MONETARY _DUMMY_MESSAGES |
| 356 | # endif |
| 357 | # ifdef USE_LOCALE_ADDRESS |
| 358 | # define LC_ADDRESS_INDEX _DUMMY_MONETARY + 1 |
| 359 | # define _DUMMY_ADDRESS LC_ADDRESS_INDEX |
| 360 | # else |
| 361 | # define _DUMMY_ADDRESS _DUMMY_MONETARY |
| 362 | # endif |
| 363 | # ifdef USE_LOCALE_IDENTIFICATION |
| 364 | # define LC_IDENTIFICATION_INDEX _DUMMY_ADDRESS + 1 |
| 365 | # define _DUMMY_IDENTIFICATION LC_IDENTIFICATION_INDEX |
| 366 | # else |
| 367 | # define _DUMMY_IDENTIFICATION _DUMMY_ADDRESS |
| 368 | # endif |
| 369 | # ifdef USE_LOCALE_MEASUREMENT |
| 370 | # define LC_MEASUREMENT_INDEX _DUMMY_IDENTIFICATION + 1 |
| 371 | # define _DUMMY_MEASUREMENT LC_MEASUREMENT_INDEX |
| 372 | # else |
| 373 | # define _DUMMY_MEASUREMENT _DUMMY_IDENTIFICATION |
| 374 | # endif |
| 375 | # ifdef USE_LOCALE_PAPER |
| 376 | # define LC_PAPER_INDEX _DUMMY_MEASUREMENT + 1 |
| 377 | # define _DUMMY_PAPER LC_PAPER_INDEX |
| 378 | # else |
| 379 | # define _DUMMY_PAPER _DUMMY_MEASUREMENT |
| 380 | # endif |
| 381 | # ifdef USE_LOCALE_TELEPHONE |
| 382 | # define LC_TELEPHONE_INDEX _DUMMY_PAPER + 1 |
| 383 | # define _DUMMY_TELEPHONE LC_TELEPHONE_INDEX |
| 384 | # else |
| 385 | # define _DUMMY_TELEPHONE _DUMMY_PAPER |
| 386 | # endif |
| 387 | # ifdef LC_ALL |
| 388 | # define LC_ALL_INDEX _DUMMY_TELEPHONE + 1 |
| 389 | # endif |
| 390 | #endif /* ifdef USE_LOCALE */ |
| 391 | |
| 392 | /* Windows requres a customized base-level setlocale() */ |
| 393 | #ifdef WIN32 |
| 394 | # define my_setlocale(cat, locale) win32_setlocale(cat, locale) |
| 395 | #else |
| 396 | # define my_setlocale(cat, locale) setlocale(cat, locale) |
| 397 | #endif |
| 398 | |
| 399 | #ifndef USE_POSIX_2008_LOCALE |
| 400 | |
| 401 | /* "do_setlocale_c" is intended to be called when the category is a constant |
| 402 | * known at compile time; "do_setlocale_r", not known until run time */ |
| 403 | # define do_setlocale_c(cat, locale) my_setlocale(cat, locale) |
| 404 | # define do_setlocale_r(cat, locale) my_setlocale(cat, locale) |
| 405 | |
| 406 | #else /* Below uses POSIX 2008 */ |
| 407 | |
| 408 | /* We emulate setlocale with our own function. LC_foo is not valid for the |
| 409 | * POSIX 2008 functions. Instead LC_foo_MASK is used, which we use an array |
| 410 | * lookup to convert to. At compile time we have defined LC_foo_INDEX as the |
| 411 | * proper offset into the array 'category_masks[]'. At runtime, we have to |
| 412 | * search through the array (as the actual numbers may not be small contiguous |
| 413 | * positive integers which would lend themselves to array lookup). */ |
| 414 | # define do_setlocale_c(cat, locale) \ |
| 415 | emulate_setlocale(cat, locale, cat ## _INDEX, TRUE) |
| 416 | # define do_setlocale_r(cat, locale) emulate_setlocale(cat, locale, 0, FALSE) |
| 417 | |
| 418 | /* A third array, parallel to the ones above to map from category to its |
| 419 | * equivalent mask */ |
| 420 | const int category_masks[] = { |
| 421 | # ifdef USE_LOCALE_NUMERIC |
| 422 | LC_NUMERIC_MASK, |
| 423 | # endif |
| 424 | # ifdef USE_LOCALE_CTYPE |
| 425 | LC_CTYPE_MASK, |
| 426 | # endif |
| 427 | # ifdef USE_LOCALE_COLLATE |
| 428 | LC_COLLATE_MASK, |
| 429 | # endif |
| 430 | # ifdef USE_LOCALE_TIME |
| 431 | LC_TIME_MASK, |
| 432 | # endif |
| 433 | # ifdef USE_LOCALE_MESSAGES |
| 434 | LC_MESSAGES_MASK, |
| 435 | # endif |
| 436 | # ifdef USE_LOCALE_MONETARY |
| 437 | LC_MONETARY_MASK, |
| 438 | # endif |
| 439 | # ifdef USE_LOCALE_ADDRESS |
| 440 | LC_ADDRESS_MASK, |
| 441 | # endif |
| 442 | # ifdef USE_LOCALE_IDENTIFICATION |
| 443 | LC_IDENTIFICATION_MASK, |
| 444 | # endif |
| 445 | # ifdef USE_LOCALE_MEASUREMENT |
| 446 | LC_MEASUREMENT_MASK, |
| 447 | # endif |
| 448 | # ifdef USE_LOCALE_PAPER |
| 449 | LC_PAPER_MASK, |
| 450 | # endif |
| 451 | # ifdef USE_LOCALE_TELEPHONE |
| 452 | LC_TELEPHONE_MASK, |
| 453 | # endif |
| 454 | /* LC_ALL can't be turned off by a Configure |
| 455 | * option, and in Posix 2008, should always be |
| 456 | * here, so compile it in unconditionally. |
| 457 | * This could catch some glitches at compile |
| 458 | * time */ |
| 459 | LC_ALL_MASK |
| 460 | }; |
| 461 | |
| 462 | STATIC const char * |
| 463 | S_emulate_setlocale(const int category, |
| 464 | const char * locale, |
| 465 | unsigned int index, |
| 466 | const bool is_index_valid |
| 467 | ) |
| 468 | { |
| 469 | /* This function effectively performs a setlocale() on just the current |
| 470 | * thread; thus it is thread-safe. It does this by using the POSIX 2008 |
| 471 | * locale functions to emulate the behavior of setlocale(). Similar to |
| 472 | * regular setlocale(), the return from this function points to memory that |
| 473 | * can be overwritten by other system calls, so needs to be copied |
| 474 | * immediately if you need to retain it. The difference here is that |
| 475 | * system calls besides another setlocale() can overwrite it. |
| 476 | * |
| 477 | * By doing this, most locale-sensitive functions become thread-safe. The |
| 478 | * exceptions are mostly those that return a pointer to static memory. |
| 479 | * |
| 480 | * This function takes the same parameters, 'category' and 'locale', that |
| 481 | * the regular setlocale() function does, but it also takes two additional |
| 482 | * ones. This is because the 2008 functions don't use a category; instead |
| 483 | * they use a corresponding mask. Because this function operates in both |
| 484 | * worlds, it may need one or the other or both. This function can |
| 485 | * calculate the mask from the input category, but to avoid this |
| 486 | * calculation, if the caller knows at compile time what the mask is, it |
| 487 | * can pass it, setting 'is_index_valid' to TRUE; otherwise the mask |
| 488 | * parameter is ignored. |
| 489 | * |
| 490 | * POSIX 2008, for some sick reason, chose not to provide a method to find |
| 491 | * the category name of a locale. Some vendors have created a |
| 492 | * querylocale() function to do just that. This function is a lot simpler |
| 493 | * to implement on systems that have this. Otherwise, we have to keep |
| 494 | * track of what the locale has been set to, so that we can return its |
| 495 | * name to emulate setlocale(). It's also possible for C code in some |
| 496 | * library to change the locale without us knowing it, though as of |
| 497 | * September 2017, there are no occurrences in CPAN of uselocale(). Some |
| 498 | * libraries do use setlocale(), but that changes the global locale, and |
| 499 | * threads using per-thread locales will just ignore those changes. |
| 500 | * Another problem is that without querylocale(), we have to guess at what |
| 501 | * was meant by setting a locale of "". We handle this by not actually |
| 502 | * ever setting to "" (unless querylocale exists), but to emulate what we |
| 503 | * think should happen for "". |
| 504 | */ |
| 505 | |
| 506 | int mask; |
| 507 | locale_t old_obj; |
| 508 | locale_t new_obj; |
| 509 | dTHX; |
| 510 | |
| 511 | # ifdef DEBUGGING |
| 512 | |
| 513 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 514 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale input=%d (%s), \"%s\", %d, %d\n", __FILE__, __LINE__, category, category_name(category), locale, index, is_index_valid); |
| 515 | } |
| 516 | |
| 517 | # endif |
| 518 | |
| 519 | /* If the input mask might be incorrect, calculate the correct one */ |
| 520 | if (! is_index_valid) { |
| 521 | unsigned int i; |
| 522 | |
| 523 | # ifdef DEBUGGING |
| 524 | |
| 525 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 526 | PerlIO_printf(Perl_debug_log, "%s:%d: finding index of category %d (%s)\n", __FILE__, __LINE__, category, category_name(category)); |
| 527 | } |
| 528 | |
| 529 | # endif |
| 530 | |
| 531 | for (i = 0; i <= LC_ALL_INDEX; i++) { |
| 532 | if (category == categories[i]) { |
| 533 | index = i; |
| 534 | goto found_index; |
| 535 | } |
| 536 | } |
| 537 | |
| 538 | /* Here, we don't know about this category, so can't handle it. |
| 539 | * Fallback to the early POSIX usages */ |
| 540 | Perl_warner(aTHX_ packWARN(WARN_LOCALE), |
| 541 | "Unknown locale category %d; can't set it to %s\n", |
| 542 | category, locale); |
| 543 | return NULL; |
| 544 | |
| 545 | found_index: ; |
| 546 | |
| 547 | # ifdef DEBUGGING |
| 548 | |
| 549 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 550 | PerlIO_printf(Perl_debug_log, "%s:%d: index is %d for %s\n", __FILE__, __LINE__, index, category_name(category)); |
| 551 | } |
| 552 | |
| 553 | # endif |
| 554 | |
| 555 | } |
| 556 | |
| 557 | mask = category_masks[index]; |
| 558 | |
| 559 | # ifdef DEBUGGING |
| 560 | |
| 561 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 562 | PerlIO_printf(Perl_debug_log, "%s:%d: category name is %s; mask is 0x%x\n", __FILE__, __LINE__, category_names[index], mask); |
| 563 | } |
| 564 | |
| 565 | # endif |
| 566 | |
| 567 | /* If just querying what the existing locale is ... */ |
| 568 | if (locale == NULL) { |
| 569 | locale_t cur_obj = uselocale((locale_t) 0); |
| 570 | |
| 571 | # ifdef DEBUGGING |
| 572 | |
| 573 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 574 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale querying %p\n", __FILE__, __LINE__, cur_obj); |
| 575 | } |
| 576 | |
| 577 | # endif |
| 578 | |
| 579 | if (cur_obj == LC_GLOBAL_LOCALE) { |
| 580 | return my_setlocale(category, NULL); |
| 581 | } |
| 582 | |
| 583 | # ifdef HAS_QUERYLOCALE |
| 584 | |
| 585 | return (char *) querylocale(mask, cur_obj); |
| 586 | |
| 587 | # else |
| 588 | |
| 589 | /* If this assert fails, adjust the size of curlocales in intrpvar.h */ |
| 590 | STATIC_ASSERT_STMT(C_ARRAY_LENGTH(PL_curlocales) > LC_ALL_INDEX); |
| 591 | |
| 592 | # if defined(_NL_LOCALE_NAME) \ |
| 593 | && defined(DEBUGGING) \ |
| 594 | && ! defined(SETLOCALE_ACCEPTS_ANY_LOCALE_NAME) |
| 595 | /* On systems that accept any locale name, the real underlying locale |
| 596 | * is often returned by this internal function, so we can't use it */ |
| 597 | { |
| 598 | /* Internal glibc for querylocale(), but doesn't handle |
| 599 | * empty-string ("") locale properly; who knows what other |
| 600 | * glitches. Check for it now, under debug. */ |
| 601 | |
| 602 | char * temp_name = nl_langinfo_l(_NL_LOCALE_NAME(category), |
| 603 | uselocale((locale_t) 0)); |
| 604 | /* |
| 605 | PerlIO_printf(Perl_debug_log, "%s:%d: temp_name=%s\n", __FILE__, __LINE__, temp_name ? temp_name : "NULL"); |
| 606 | PerlIO_printf(Perl_debug_log, "%s:%d: index=%d\n", __FILE__, __LINE__, index); |
| 607 | PerlIO_printf(Perl_debug_log, "%s:%d: PL_curlocales[index]=%s\n", __FILE__, __LINE__, PL_curlocales[index]); |
| 608 | */ |
| 609 | if (temp_name && PL_curlocales[index] && strNE(temp_name, "")) { |
| 610 | if ( strNE(PL_curlocales[index], temp_name) |
| 611 | && ! ( isNAME_C_OR_POSIX(temp_name) |
| 612 | && isNAME_C_OR_POSIX(PL_curlocales[index]))) { |
| 613 | |
| 614 | # ifdef USE_C_BACKTRACE |
| 615 | |
| 616 | dump_c_backtrace(Perl_debug_log, 20, 1); |
| 617 | |
| 618 | # endif |
| 619 | |
| 620 | Perl_croak(aTHX_ "panic: Mismatch between what Perl thinks %s is" |
| 621 | " (%s) and what internal glibc thinks" |
| 622 | " (%s)\n", category_names[index], |
| 623 | PL_curlocales[index], temp_name); |
| 624 | } |
| 625 | |
| 626 | return temp_name; |
| 627 | } |
| 628 | } |
| 629 | |
| 630 | # endif |
| 631 | |
| 632 | /* Without querylocale(), we have to use our record-keeping we've |
| 633 | * done. */ |
| 634 | |
| 635 | if (category != LC_ALL) { |
| 636 | |
| 637 | # ifdef DEBUGGING |
| 638 | |
| 639 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 640 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, PL_curlocales[index]); |
| 641 | } |
| 642 | |
| 643 | # endif |
| 644 | |
| 645 | return PL_curlocales[index]; |
| 646 | } |
| 647 | else { /* For LC_ALL */ |
| 648 | unsigned int i; |
| 649 | Size_t names_len = 0; |
| 650 | char * all_string; |
| 651 | bool are_all_categories_the_same_locale = TRUE; |
| 652 | |
| 653 | /* If we have a valid LC_ALL value, just return it */ |
| 654 | if (PL_curlocales[LC_ALL_INDEX]) { |
| 655 | |
| 656 | # ifdef DEBUGGING |
| 657 | |
| 658 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 659 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, PL_curlocales[LC_ALL_INDEX]); |
| 660 | } |
| 661 | |
| 662 | # endif |
| 663 | |
| 664 | return PL_curlocales[LC_ALL_INDEX]; |
| 665 | } |
| 666 | |
| 667 | /* Otherwise, we need to construct a string of name=value pairs. |
| 668 | * We use the glibc syntax, like |
| 669 | * LC_NUMERIC=C;LC_TIME=en_US.UTF-8;... |
| 670 | * First calculate the needed size. Along the way, check if all |
| 671 | * the locale names are the same */ |
| 672 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 673 | |
| 674 | # ifdef DEBUGGING |
| 675 | |
| 676 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 677 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale i=%d, name=%s, locale=%s\n", __FILE__, __LINE__, i, category_names[i], PL_curlocales[i]); |
| 678 | } |
| 679 | |
| 680 | # endif |
| 681 | |
| 682 | names_len += strlen(category_names[i]) |
| 683 | + 1 /* '=' */ |
| 684 | + strlen(PL_curlocales[i]) |
| 685 | + 1; /* ';' */ |
| 686 | |
| 687 | if (i > 0 && strNE(PL_curlocales[i], PL_curlocales[i-1])) { |
| 688 | are_all_categories_the_same_locale = FALSE; |
| 689 | } |
| 690 | } |
| 691 | |
| 692 | /* If they are the same, we don't actually have to construct the |
| 693 | * string; we just make the entry in LC_ALL_INDEX valid, and be |
| 694 | * that single name */ |
| 695 | if (are_all_categories_the_same_locale) { |
| 696 | PL_curlocales[LC_ALL_INDEX] = savepv(PL_curlocales[0]); |
| 697 | return PL_curlocales[LC_ALL_INDEX]; |
| 698 | } |
| 699 | |
| 700 | names_len++; /* Trailing '\0' */ |
| 701 | SAVEFREEPV(Newx(all_string, names_len, char)); |
| 702 | *all_string = '\0'; |
| 703 | |
| 704 | /* Then fill in the string */ |
| 705 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 706 | |
| 707 | # ifdef DEBUGGING |
| 708 | |
| 709 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 710 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale i=%d, name=%s, locale=%s\n", __FILE__, __LINE__, i, category_names[i], PL_curlocales[i]); |
| 711 | } |
| 712 | |
| 713 | # endif |
| 714 | |
| 715 | my_strlcat(all_string, category_names[i], names_len); |
| 716 | my_strlcat(all_string, "=", names_len); |
| 717 | my_strlcat(all_string, PL_curlocales[i], names_len); |
| 718 | my_strlcat(all_string, ";", names_len); |
| 719 | } |
| 720 | |
| 721 | # ifdef DEBUGGING |
| 722 | |
| 723 | if (DEBUG_L_TEST || debug_initialization) { |
| 724 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, all_string); |
| 725 | } |
| 726 | |
| 727 | #endif |
| 728 | |
| 729 | return all_string; |
| 730 | } |
| 731 | |
| 732 | # ifdef EINVAL |
| 733 | |
| 734 | SETERRNO(EINVAL, LIB_INVARG); |
| 735 | |
| 736 | # endif |
| 737 | |
| 738 | return NULL; |
| 739 | |
| 740 | # endif |
| 741 | |
| 742 | } /* End of this being setlocale(LC_foo, NULL) */ |
| 743 | |
| 744 | /* Here, we are switching locales. */ |
| 745 | |
| 746 | # ifndef HAS_QUERYLOCALE |
| 747 | |
| 748 | if (strEQ(locale, "")) { |
| 749 | |
| 750 | /* For non-querylocale() systems, we do the setting of "" ourselves to |
| 751 | * be sure that we really know what's going on. We follow the Linux |
| 752 | * documented behavior (but if that differs from the actual behavior, |
| 753 | * this won't work exactly as the OS implements). We go out and |
| 754 | * examine the environment based on our understanding of how the system |
| 755 | * works, and use that to figure things out */ |
| 756 | |
| 757 | const char * const lc_all = PerlEnv_getenv("LC_ALL"); |
| 758 | |
| 759 | /* Use any "LC_ALL" environment variable, as it overrides everything |
| 760 | * else. */ |
| 761 | if (lc_all && strNE(lc_all, "")) { |
| 762 | locale = lc_all; |
| 763 | } |
| 764 | else { |
| 765 | |
| 766 | /* Otherwise, we need to dig deeper. Unless overridden, the |
| 767 | * default is the LANG environment variable; if it doesn't exist, |
| 768 | * then "C" */ |
| 769 | |
| 770 | const char * default_name; |
| 771 | |
| 772 | default_name = PerlEnv_getenv("LANG"); |
| 773 | |
| 774 | if (! default_name || strEQ(default_name, "")) { |
| 775 | default_name = "C"; |
| 776 | } |
| 777 | else if (PL_scopestack_ix != 0) { |
| 778 | /* To minimize other threads messing with the environment, |
| 779 | * we copy the variable, making it a temporary. But this |
| 780 | * doesn't work upon program initialization before any |
| 781 | * scopes are created, and at this time, there's nothing |
| 782 | * else going on that would interfere. So skip the copy |
| 783 | * in that case */ |
| 784 | default_name = savepv(default_name); |
| 785 | SAVEFREEPV(default_name); |
| 786 | } |
| 787 | |
| 788 | if (category != LC_ALL) { |
| 789 | const char * const name = PerlEnv_getenv(category_names[index]); |
| 790 | |
| 791 | /* Here we are setting a single category. Assume will have the |
| 792 | * default name */ |
| 793 | locale = default_name; |
| 794 | |
| 795 | /* But then look for an overriding environment variable */ |
| 796 | if (name && strNE(name, "")) { |
| 797 | locale = name; |
| 798 | } |
| 799 | } |
| 800 | else { |
| 801 | bool did_override = FALSE; |
| 802 | unsigned int i; |
| 803 | |
| 804 | /* Here, we are getting LC_ALL. Any categories that don't have |
| 805 | * a corresponding environment variable set should be set to |
| 806 | * LANG, or to "C" if there is no LANG. If no individual |
| 807 | * categories differ from this, we can just set LC_ALL. This |
| 808 | * is buggy on systems that have extra categories that we don't |
| 809 | * know about. If there is an environment variable that sets |
| 810 | * that category, we won't know to look for it, and so our use |
| 811 | * of LANG or "C" improperly overrides it. On the other hand, |
| 812 | * if we don't do what is done here, and there is no |
| 813 | * environment variable, the category's locale should be set to |
| 814 | * LANG or "C". So there is no good solution. khw thinks the |
| 815 | * best is to look at systems to see what categories they have, |
| 816 | * and include them, and then to assume that we know the |
| 817 | * complete set */ |
| 818 | |
| 819 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 820 | const char * const env_override |
| 821 | = savepv(PerlEnv_getenv(category_names[i])); |
| 822 | const char * this_locale = ( env_override |
| 823 | && strNE(env_override, "")) |
| 824 | ? env_override |
| 825 | : default_name; |
| 826 | if (! emulate_setlocale(categories[i], this_locale, i, TRUE)) |
| 827 | { |
| 828 | Safefree(env_override); |
| 829 | return NULL; |
| 830 | } |
| 831 | |
| 832 | if (strNE(this_locale, default_name)) { |
| 833 | did_override = TRUE; |
| 834 | } |
| 835 | |
| 836 | Safefree(env_override); |
| 837 | } |
| 838 | |
| 839 | /* If all the categories are the same, we can set LC_ALL to |
| 840 | * that */ |
| 841 | if (! did_override) { |
| 842 | locale = default_name; |
| 843 | } |
| 844 | else { |
| 845 | |
| 846 | /* Here, LC_ALL is no longer valid, as some individual |
| 847 | * categories don't match it. We call ourselves |
| 848 | * recursively, as that will execute the code that |
| 849 | * generates the proper locale string for this situation. |
| 850 | * We don't do the remainder of this function, as that is |
| 851 | * to update our records, and we've just done that for the |
| 852 | * individual categories in the loop above, and doing so |
| 853 | * would cause LC_ALL to be done as well */ |
| 854 | return emulate_setlocale(LC_ALL, NULL, LC_ALL_INDEX, TRUE); |
| 855 | } |
| 856 | } |
| 857 | } |
| 858 | } /* End of this being setlocale(LC_foo, "") */ |
| 859 | else if (strchr(locale, ';')) { |
| 860 | |
| 861 | /* LC_ALL may actually incude a conglomeration of various categories. |
| 862 | * Without querylocale, this code uses the glibc (as of this writing) |
| 863 | * syntax for representing that, but that is not a stable API, and |
| 864 | * other platforms do it differently, so we have to handle all cases |
| 865 | * ourselves */ |
| 866 | |
| 867 | unsigned int i; |
| 868 | const char * s = locale; |
| 869 | const char * e = locale + strlen(locale); |
| 870 | const char * p = s; |
| 871 | const char * category_end; |
| 872 | const char * name_start; |
| 873 | const char * name_end; |
| 874 | |
| 875 | /* If the string that gives what to set doesn't include all categories, |
| 876 | * the omitted ones get set to "C". To get this behavior, first set |
| 877 | * all the individual categories to "C", and override the furnished |
| 878 | * ones below */ |
| 879 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 880 | if (! emulate_setlocale(categories[i], "C", i, TRUE)) { |
| 881 | return NULL; |
| 882 | } |
| 883 | } |
| 884 | |
| 885 | while (s < e) { |
| 886 | |
| 887 | /* Parse through the category */ |
| 888 | while (isWORDCHAR(*p)) { |
| 889 | p++; |
| 890 | } |
| 891 | category_end = p; |
| 892 | |
| 893 | if (*p++ != '=') { |
| 894 | Perl_croak(aTHX_ |
| 895 | "panic: %s: %d: Unexpected character in locale name '%02X", |
| 896 | __FILE__, __LINE__, *(p-1)); |
| 897 | } |
| 898 | |
| 899 | /* Parse through the locale name */ |
| 900 | name_start = p; |
| 901 | while (p < e && *p != ';') { |
| 902 | if (! isGRAPH(*p)) { |
| 903 | Perl_croak(aTHX_ |
| 904 | "panic: %s: %d: Unexpected character in locale name '%02X", |
| 905 | __FILE__, __LINE__, *(p-1)); |
| 906 | } |
| 907 | p++; |
| 908 | } |
| 909 | name_end = p; |
| 910 | |
| 911 | /* Space past the semi-colon */ |
| 912 | if (p < e) { |
| 913 | p++; |
| 914 | } |
| 915 | |
| 916 | /* Find the index of the category name in our lists */ |
| 917 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 918 | char * individ_locale; |
| 919 | |
| 920 | /* Keep going if this isn't the index. The strnNE() avoids a |
| 921 | * Perl_form(), but would fail if ever a category name could be |
| 922 | * a substring of another one, like if there were a |
| 923 | * "LC_TIME_DATE" */ |
| 924 | if strnNE(s, category_names[i], category_end - s) { |
| 925 | continue; |
| 926 | } |
| 927 | |
| 928 | /* If this index is for the single category we're changing, we |
| 929 | * have found the locale to set it to. */ |
| 930 | if (category == categories[i]) { |
| 931 | locale = Perl_form(aTHX_ "%.*s", |
| 932 | (int) (name_end - name_start), |
| 933 | name_start); |
| 934 | goto ready_to_set; |
| 935 | } |
| 936 | |
| 937 | assert(category == LC_ALL); |
| 938 | individ_locale = Perl_form(aTHX_ "%.*s", |
| 939 | (int) (name_end - name_start), name_start); |
| 940 | if (! emulate_setlocale(categories[i], individ_locale, i, TRUE)) |
| 941 | { |
| 942 | return NULL; |
| 943 | } |
| 944 | } |
| 945 | |
| 946 | s = p; |
| 947 | } |
| 948 | |
| 949 | /* Here we have set all the individual categories by recursive calls. |
| 950 | * These collectively should have fixed up LC_ALL, so can just query |
| 951 | * what that now is */ |
| 952 | assert(category == LC_ALL); |
| 953 | |
| 954 | return do_setlocale_c(LC_ALL, NULL); |
| 955 | } /* End of this being setlocale(LC_ALL, |
| 956 | "LC_CTYPE=foo;LC_NUMERIC=bar;...") */ |
| 957 | |
| 958 | ready_to_set: ; |
| 959 | |
| 960 | /* Here at the end of having to deal with the absence of querylocale(). |
| 961 | * Some cases have already been fully handled by recursive calls to this |
| 962 | * function. But at this point, we haven't dealt with those, but are now |
| 963 | * prepared to, knowing what the locale name to set this category to is. |
| 964 | * This would have come for free if this system had had querylocale() */ |
| 965 | |
| 966 | # endif /* end of ! querylocale */ |
| 967 | |
| 968 | assert(PL_C_locale_obj); |
| 969 | |
| 970 | /* Switching locales generally entails freeing the current one's space (at |
| 971 | * the C library's discretion). We need to stop using that locale before |
| 972 | * the switch. So switch to a known locale object that we don't otherwise |
| 973 | * mess with. This returns the locale object in effect at the time of the |
| 974 | * switch. */ |
| 975 | old_obj = uselocale(PL_C_locale_obj); |
| 976 | |
| 977 | # ifdef DEBUGGING |
| 978 | |
| 979 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 980 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale was using %p\n", __FILE__, __LINE__, old_obj); |
| 981 | } |
| 982 | |
| 983 | # endif |
| 984 | |
| 985 | if (! old_obj) { |
| 986 | |
| 987 | # ifdef DEBUGGING |
| 988 | |
| 989 | if (DEBUG_L_TEST || debug_initialization) { |
| 990 | dSAVE_ERRNO; |
| 991 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale switching to C failed: %d\n", __FILE__, __LINE__, GET_ERRNO); |
| 992 | RESTORE_ERRNO; |
| 993 | } |
| 994 | |
| 995 | # endif |
| 996 | |
| 997 | return NULL; |
| 998 | } |
| 999 | |
| 1000 | # ifdef DEBUGGING |
| 1001 | |
| 1002 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1003 | PerlIO_printf(Perl_debug_log, |
| 1004 | "%s:%d: emulate_setlocale now using %p\n", |
| 1005 | __FILE__, __LINE__, PL_C_locale_obj); |
| 1006 | } |
| 1007 | |
| 1008 | # endif |
| 1009 | |
| 1010 | /* If this call is to switch to the LC_ALL C locale, it already exists, and |
| 1011 | * in fact, we already have switched to it (in preparation for what |
| 1012 | * normally is to come). But since we're already there, continue to use |
| 1013 | * it instead of trying to create a new locale */ |
| 1014 | if (mask == LC_ALL_MASK && isNAME_C_OR_POSIX(locale)) { |
| 1015 | |
| 1016 | # ifdef DEBUGGING |
| 1017 | |
| 1018 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1019 | PerlIO_printf(Perl_debug_log, |
| 1020 | "%s:%d: will stay in C object\n", __FILE__, __LINE__); |
| 1021 | } |
| 1022 | |
| 1023 | # endif |
| 1024 | |
| 1025 | new_obj = PL_C_locale_obj; |
| 1026 | |
| 1027 | /* We already had switched to the C locale in preparation for freeing |
| 1028 | * 'old_obj' */ |
| 1029 | if (old_obj != LC_GLOBAL_LOCALE && old_obj != PL_C_locale_obj) { |
| 1030 | freelocale(old_obj); |
| 1031 | } |
| 1032 | } |
| 1033 | else { |
| 1034 | /* If we weren't in a thread safe locale, set so that newlocale() below |
| 1035 | * which uses 'old_obj', uses an empty one. Same for our reserved C |
| 1036 | * object. The latter is defensive coding, so that, even if there is |
| 1037 | * some bug, we will never end up trying to modify either of these, as |
| 1038 | * if passed to newlocale(), they can be. */ |
| 1039 | if (old_obj == LC_GLOBAL_LOCALE || old_obj == PL_C_locale_obj) { |
| 1040 | old_obj = (locale_t) 0; |
| 1041 | } |
| 1042 | |
| 1043 | /* Ready to create a new locale by modification of the exising one */ |
| 1044 | new_obj = newlocale(mask, locale, old_obj); |
| 1045 | |
| 1046 | if (! new_obj) { |
| 1047 | dSAVE_ERRNO; |
| 1048 | |
| 1049 | # ifdef DEBUGGING |
| 1050 | |
| 1051 | if (DEBUG_L_TEST || debug_initialization) { |
| 1052 | PerlIO_printf(Perl_debug_log, |
| 1053 | "%s:%d: emulate_setlocale creating new object" |
| 1054 | " failed: %d\n", __FILE__, __LINE__, GET_ERRNO); |
| 1055 | } |
| 1056 | |
| 1057 | # endif |
| 1058 | |
| 1059 | if (! uselocale(old_obj)) { |
| 1060 | |
| 1061 | # ifdef DEBUGGING |
| 1062 | |
| 1063 | if (DEBUG_L_TEST || debug_initialization) { |
| 1064 | PerlIO_printf(Perl_debug_log, |
| 1065 | "%s:%d: switching back failed: %d\n", |
| 1066 | __FILE__, __LINE__, GET_ERRNO); |
| 1067 | } |
| 1068 | |
| 1069 | # endif |
| 1070 | |
| 1071 | } |
| 1072 | RESTORE_ERRNO; |
| 1073 | return NULL; |
| 1074 | } |
| 1075 | |
| 1076 | # ifdef DEBUGGING |
| 1077 | |
| 1078 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1079 | PerlIO_printf(Perl_debug_log, |
| 1080 | "%s:%d: emulate_setlocale created %p", |
| 1081 | __FILE__, __LINE__, new_obj); |
| 1082 | if (old_obj) { |
| 1083 | PerlIO_printf(Perl_debug_log, |
| 1084 | "; should have freed %p", old_obj); |
| 1085 | } |
| 1086 | PerlIO_printf(Perl_debug_log, "\n"); |
| 1087 | } |
| 1088 | |
| 1089 | # endif |
| 1090 | |
| 1091 | /* And switch into it */ |
| 1092 | if (! uselocale(new_obj)) { |
| 1093 | dSAVE_ERRNO; |
| 1094 | |
| 1095 | # ifdef DEBUGGING |
| 1096 | |
| 1097 | if (DEBUG_L_TEST || debug_initialization) { |
| 1098 | PerlIO_printf(Perl_debug_log, |
| 1099 | "%s:%d: emulate_setlocale switching to new object" |
| 1100 | " failed\n", __FILE__, __LINE__); |
| 1101 | } |
| 1102 | |
| 1103 | # endif |
| 1104 | |
| 1105 | if (! uselocale(old_obj)) { |
| 1106 | |
| 1107 | # ifdef DEBUGGING |
| 1108 | |
| 1109 | if (DEBUG_L_TEST || debug_initialization) { |
| 1110 | PerlIO_printf(Perl_debug_log, |
| 1111 | "%s:%d: switching back failed: %d\n", |
| 1112 | __FILE__, __LINE__, GET_ERRNO); |
| 1113 | } |
| 1114 | |
| 1115 | # endif |
| 1116 | |
| 1117 | } |
| 1118 | freelocale(new_obj); |
| 1119 | RESTORE_ERRNO; |
| 1120 | return NULL; |
| 1121 | } |
| 1122 | } |
| 1123 | |
| 1124 | # ifdef DEBUGGING |
| 1125 | |
| 1126 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1127 | PerlIO_printf(Perl_debug_log, |
| 1128 | "%s:%d: emulate_setlocale now using %p\n", |
| 1129 | __FILE__, __LINE__, new_obj); |
| 1130 | } |
| 1131 | |
| 1132 | # endif |
| 1133 | |
| 1134 | /* We are done, except for updating our records (if the system doesn't keep |
| 1135 | * them) and in the case of locale "", we don't actually know what the |
| 1136 | * locale that got switched to is, as it came from the environment. So |
| 1137 | * have to find it */ |
| 1138 | |
| 1139 | # ifdef HAS_QUERYLOCALE |
| 1140 | |
| 1141 | if (strEQ(locale, "")) { |
| 1142 | locale = querylocale(mask, new_obj); |
| 1143 | } |
| 1144 | |
| 1145 | # else |
| 1146 | |
| 1147 | /* Here, 'locale' is the return value */ |
| 1148 | |
| 1149 | /* Without querylocale(), we have to update our records */ |
| 1150 | |
| 1151 | if (category == LC_ALL) { |
| 1152 | unsigned int i; |
| 1153 | |
| 1154 | /* For LC_ALL, we change all individual categories to correspond */ |
| 1155 | /* PL_curlocales is a parallel array, so has same |
| 1156 | * length as 'categories' */ |
| 1157 | for (i = 0; i <= LC_ALL_INDEX; i++) { |
| 1158 | Safefree(PL_curlocales[i]); |
| 1159 | PL_curlocales[i] = savepv(locale); |
| 1160 | } |
| 1161 | } |
| 1162 | else { |
| 1163 | |
| 1164 | /* For a single category, if it's not the same as the one in LC_ALL, we |
| 1165 | * nullify LC_ALL */ |
| 1166 | |
| 1167 | if (PL_curlocales[LC_ALL_INDEX] && strNE(PL_curlocales[LC_ALL_INDEX], locale)) { |
| 1168 | Safefree(PL_curlocales[LC_ALL_INDEX]); |
| 1169 | PL_curlocales[LC_ALL_INDEX] = NULL; |
| 1170 | } |
| 1171 | |
| 1172 | /* Then update the category's record */ |
| 1173 | Safefree(PL_curlocales[index]); |
| 1174 | PL_curlocales[index] = savepv(locale); |
| 1175 | } |
| 1176 | |
| 1177 | # endif |
| 1178 | |
| 1179 | return locale; |
| 1180 | } |
| 1181 | |
| 1182 | #endif /* USE_POSIX_2008_LOCALE */ |
| 1183 | |
| 1184 | #if 0 /* Code that was to emulate thread-safe locales on platforms that |
| 1185 | didn't natively support them */ |
| 1186 | |
| 1187 | /* The way this would work is that we would keep a per-thread list of the |
| 1188 | * correct locale for that thread. Any operation that was locale-sensitive |
| 1189 | * would have to be changed so that it would look like this: |
| 1190 | * |
| 1191 | * LOCALE_LOCK; |
| 1192 | * setlocale to the correct locale for this operation |
| 1193 | * do operation |
| 1194 | * LOCALE_UNLOCK |
| 1195 | * |
| 1196 | * This leaves the global locale in the most recently used operation's, but it |
| 1197 | * was locked long enough to get the result. If that result is static, it |
| 1198 | * needs to be copied before the unlock. |
| 1199 | * |
| 1200 | * Macros could be written like SETUP_LOCALE_DEPENDENT_OP(category) that did |
| 1201 | * the setup, but are no-ops when not needed, and similarly, |
| 1202 | * END_LOCALE_DEPENDENT_OP for the tear-down |
| 1203 | * |
| 1204 | * But every call to a locale-sensitive function would have to be changed, and |
| 1205 | * if a module didn't cooperate by using the mutex, things would break. |
| 1206 | * |
| 1207 | * This code was abandoned before being completed or tested, and is left as-is |
| 1208 | */ |
| 1209 | |
| 1210 | # define do_setlocale_c(cat, locale) locking_setlocale(cat, locale, cat ## _INDEX, TRUE) |
| 1211 | # define do_setlocale_r(cat, locale) locking_setlocale(cat, locale, 0, FALSE) |
| 1212 | |
| 1213 | STATIC char * |
| 1214 | S_locking_setlocale(pTHX_ |
| 1215 | const int category, |
| 1216 | const char * locale, |
| 1217 | int index, |
| 1218 | const bool is_index_valid |
| 1219 | ) |
| 1220 | { |
| 1221 | /* This function kind of performs a setlocale() on just the current thread; |
| 1222 | * thus it is kind of thread-safe. It does this by keeping a thread-level |
| 1223 | * array of the current locales for each category. Every time a locale is |
| 1224 | * switched to, it does the switch globally, but updates the thread's |
| 1225 | * array. A query as to what the current locale is just returns the |
| 1226 | * appropriate element from the array, and doesn't actually call the system |
| 1227 | * setlocale(). The saving into the array is done in an uninterruptible |
| 1228 | * section of code, so is unaffected by whatever any other threads might be |
| 1229 | * doing. |
| 1230 | * |
| 1231 | * All locale-sensitive operations must work by first starting a critical |
| 1232 | * section, then switching to the thread's locale as kept by this function, |
| 1233 | * and then doing the operation, then ending the critical section. Thus, |
| 1234 | * each gets done in the appropriate locale. simulating thread-safety. |
| 1235 | * |
| 1236 | * This function takes the same parameters, 'category' and 'locale', that |
| 1237 | * the regular setlocale() function does, but it also takes two additional |
| 1238 | * ones. This is because as described earlier. If we know on input the |
| 1239 | * index corresponding to the category into the array where we store the |
| 1240 | * current locales, we don't have to calculate it. If the caller knows at |
| 1241 | * compile time what the index is, it it can pass it, setting |
| 1242 | * 'is_index_valid' to TRUE; otherwise the index parameter is ignored. |
| 1243 | * |
| 1244 | */ |
| 1245 | |
| 1246 | /* If the input index might be incorrect, calculate the correct one */ |
| 1247 | if (! is_index_valid) { |
| 1248 | unsigned int i; |
| 1249 | |
| 1250 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1251 | PerlIO_printf(Perl_debug_log, "%s:%d: converting category %d to index\n", __FILE__, __LINE__, category); |
| 1252 | } |
| 1253 | |
| 1254 | for (i = 0; i <= LC_ALL_INDEX; i++) { |
| 1255 | if (category == categories[i]) { |
| 1256 | index = i; |
| 1257 | goto found_index; |
| 1258 | } |
| 1259 | } |
| 1260 | |
| 1261 | /* Here, we don't know about this category, so can't handle it. |
| 1262 | * XXX best we can do is to unsafely set this |
| 1263 | * XXX warning */ |
| 1264 | |
| 1265 | return my_setlocale(category, locale); |
| 1266 | |
| 1267 | found_index: ; |
| 1268 | |
| 1269 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 1270 | PerlIO_printf(Perl_debug_log, "%s:%d: index is 0x%x\n", __FILE__, __LINE__, index); |
| 1271 | } |
| 1272 | } |
| 1273 | |
| 1274 | /* For a query, just return what's in our records */ |
| 1275 | if (new_locale == NULL) { |
| 1276 | return curlocales[index]; |
| 1277 | } |
| 1278 | |
| 1279 | |
| 1280 | /* Otherwise, we need to do the switch, and save the result, all in a |
| 1281 | * critical section */ |
| 1282 | |
| 1283 | Safefree(curlocales[[index]]); |
| 1284 | |
| 1285 | /* It might be that this is called from an already-locked section of code. |
| 1286 | * We would have to detect and skip the LOCK/UNLOCK if so */ |
| 1287 | LOCALE_LOCK; |
| 1288 | |
| 1289 | curlocales[index] = savepv(my_setlocale(category, new_locale)); |
| 1290 | |
| 1291 | if (strEQ(new_locale, "")) { |
| 1292 | |
| 1293 | #ifdef LC_ALL |
| 1294 | |
| 1295 | /* The locale values come from the environment, and may not all be the |
| 1296 | * same, so for LC_ALL, we have to update all the others, while the |
| 1297 | * mutex is still locked */ |
| 1298 | |
| 1299 | if (category == LC_ALL) { |
| 1300 | unsigned int i; |
| 1301 | for (i = 0; i < LC_ALL_INDEX) { |
| 1302 | curlocales[i] = my_setlocale(categories[i], NULL); |
| 1303 | } |
| 1304 | } |
| 1305 | } |
| 1306 | |
| 1307 | #endif |
| 1308 | |
| 1309 | LOCALE_UNLOCK; |
| 1310 | |
| 1311 | return curlocales[index]; |
| 1312 | } |
| 1313 | |
| 1314 | #endif |
| 1315 | #ifdef USE_LOCALE |
| 1316 | |
| 1317 | STATIC void |
| 1318 | S_set_numeric_radix(pTHX_ const bool use_locale) |
| 1319 | { |
| 1320 | /* If 'use_locale' is FALSE, set to use a dot for the radix character. If |
| 1321 | * TRUE, use the radix character derived from the current locale */ |
| 1322 | |
| 1323 | #if defined(USE_LOCALE_NUMERIC) && ( defined(HAS_LOCALECONV) \ |
| 1324 | || defined(HAS_NL_LANGINFO)) |
| 1325 | |
| 1326 | const char * radix = (use_locale) |
| 1327 | ? my_nl_langinfo(RADIXCHAR, FALSE) |
| 1328 | /* FALSE => already in dest locale */ |
| 1329 | : "."; |
| 1330 | |
| 1331 | sv_setpv(PL_numeric_radix_sv, radix); |
| 1332 | |
| 1333 | /* If this is valid UTF-8 that isn't totally ASCII, and we are in |
| 1334 | * a UTF-8 locale, then mark the radix as being in UTF-8 */ |
| 1335 | if (is_utf8_non_invariant_string((U8 *) SvPVX(PL_numeric_radix_sv), |
| 1336 | SvCUR(PL_numeric_radix_sv)) |
| 1337 | && _is_cur_LC_category_utf8(LC_NUMERIC)) |
| 1338 | { |
| 1339 | SvUTF8_on(PL_numeric_radix_sv); |
| 1340 | } |
| 1341 | |
| 1342 | # ifdef DEBUGGING |
| 1343 | |
| 1344 | if (DEBUG_L_TEST || debug_initialization) { |
| 1345 | PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n", |
| 1346 | SvPVX(PL_numeric_radix_sv), |
| 1347 | cBOOL(SvUTF8(PL_numeric_radix_sv))); |
| 1348 | } |
| 1349 | |
| 1350 | # endif |
| 1351 | #else |
| 1352 | |
| 1353 | PERL_UNUSED_ARG(use_locale); |
| 1354 | |
| 1355 | #endif /* USE_LOCALE_NUMERIC and can find the radix char */ |
| 1356 | |
| 1357 | } |
| 1358 | |
| 1359 | STATIC void |
| 1360 | S_new_numeric(pTHX_ const char *newnum) |
| 1361 | { |
| 1362 | |
| 1363 | #ifndef USE_LOCALE_NUMERIC |
| 1364 | |
| 1365 | PERL_UNUSED_ARG(newnum); |
| 1366 | |
| 1367 | #else |
| 1368 | |
| 1369 | /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell |
| 1370 | * core Perl this and that 'newnum' is the name of the new locale. |
| 1371 | * It installs this locale as the current underlying default. |
| 1372 | * |
| 1373 | * The default locale and the C locale can be toggled between by use of the |
| 1374 | * set_numeric_underlying() and set_numeric_standard() functions, which |
| 1375 | * should probably not be called directly, but only via macros like |
| 1376 | * SET_NUMERIC_STANDARD() in perl.h. |
| 1377 | * |
| 1378 | * The toggling is necessary mainly so that a non-dot radix decimal point |
| 1379 | * character can be output, while allowing internal calculations to use a |
| 1380 | * dot. |
| 1381 | * |
| 1382 | * This sets several interpreter-level variables: |
| 1383 | * PL_numeric_name The underlying locale's name: a copy of 'newnum' |
| 1384 | * PL_numeric_underlying A boolean indicating if the toggled state is such |
| 1385 | * that the current locale is the program's underlying |
| 1386 | * locale |
| 1387 | * PL_numeric_standard An int indicating if the toggled state is such |
| 1388 | * that the current locale is the C locale or |
| 1389 | * indistinguishable from the C locale. If non-zero, it |
| 1390 | * is in C; if > 1, it means it may not be toggled away |
| 1391 | * from C. |
| 1392 | * PL_numeric_underlying_is_standard A bool kept by this function |
| 1393 | * indicating that the underlying locale and the standard |
| 1394 | * C locale are indistinguishable for the purposes of |
| 1395 | * LC_NUMERIC. This happens when both of the above two |
| 1396 | * variables are true at the same time. (Toggling is a |
| 1397 | * no-op under these circumstances.) This variable is |
| 1398 | * used to avoid having to recalculate. |
| 1399 | */ |
| 1400 | |
| 1401 | char *save_newnum; |
| 1402 | |
| 1403 | if (! newnum) { |
| 1404 | Safefree(PL_numeric_name); |
| 1405 | PL_numeric_name = NULL; |
| 1406 | PL_numeric_standard = TRUE; |
| 1407 | PL_numeric_underlying = TRUE; |
| 1408 | PL_numeric_underlying_is_standard = TRUE; |
| 1409 | return; |
| 1410 | } |
| 1411 | |
| 1412 | save_newnum = stdize_locale(savepv(newnum)); |
| 1413 | PL_numeric_underlying = TRUE; |
| 1414 | PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum); |
| 1415 | |
| 1416 | #ifndef TS_W32_BROKEN_LOCALECONV |
| 1417 | |
| 1418 | /* If its name isn't C nor POSIX, it could still be indistinguishable from |
| 1419 | * them. But on broken Windows systems calling my_nl_langinfo() for |
| 1420 | * THOUSEP can currently (but rarely) cause a race, so avoid doing that, |
| 1421 | * and just always change the locale if not C nor POSIX on those systems */ |
| 1422 | if (! PL_numeric_standard) { |
| 1423 | PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(RADIXCHAR, |
| 1424 | FALSE /* Don't toggle locale */ )) |
| 1425 | && strEQ("", my_nl_langinfo(THOUSEP, FALSE))); |
| 1426 | } |
| 1427 | |
| 1428 | #endif |
| 1429 | |
| 1430 | /* Save the new name if it isn't the same as the previous one, if any */ |
| 1431 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { |
| 1432 | Safefree(PL_numeric_name); |
| 1433 | PL_numeric_name = save_newnum; |
| 1434 | } |
| 1435 | else { |
| 1436 | Safefree(save_newnum); |
| 1437 | } |
| 1438 | |
| 1439 | PL_numeric_underlying_is_standard = PL_numeric_standard; |
| 1440 | |
| 1441 | # ifdef HAS_POSIX_2008_LOCALE |
| 1442 | |
| 1443 | PL_underlying_numeric_obj = newlocale(LC_NUMERIC_MASK, |
| 1444 | PL_numeric_name, |
| 1445 | PL_underlying_numeric_obj); |
| 1446 | |
| 1447 | #endif |
| 1448 | |
| 1449 | if (DEBUG_L_TEST || debug_initialization) { |
| 1450 | PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name); |
| 1451 | } |
| 1452 | |
| 1453 | /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't |
| 1454 | * have to worry about the radix being a non-dot. (Core operations that |
| 1455 | * need the underlying locale change to it temporarily). */ |
| 1456 | if (PL_numeric_standard) { |
| 1457 | set_numeric_radix(0); |
| 1458 | } |
| 1459 | else { |
| 1460 | set_numeric_standard(); |
| 1461 | } |
| 1462 | |
| 1463 | #endif /* USE_LOCALE_NUMERIC */ |
| 1464 | |
| 1465 | } |
| 1466 | |
| 1467 | void |
| 1468 | Perl_set_numeric_standard(pTHX) |
| 1469 | { |
| 1470 | |
| 1471 | #ifdef USE_LOCALE_NUMERIC |
| 1472 | |
| 1473 | /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like |
| 1474 | * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The |
| 1475 | * macro avoids calling this routine if toggling isn't necessary according |
| 1476 | * to our records (which could be wrong if some XS code has changed the |
| 1477 | * locale behind our back) */ |
| 1478 | |
| 1479 | # ifdef DEBUGGING |
| 1480 | |
| 1481 | if (DEBUG_L_TEST || debug_initialization) { |
| 1482 | PerlIO_printf(Perl_debug_log, |
| 1483 | "Setting LC_NUMERIC locale to standard C\n"); |
| 1484 | } |
| 1485 | |
| 1486 | # endif |
| 1487 | |
| 1488 | do_setlocale_c(LC_NUMERIC, "C"); |
| 1489 | PL_numeric_standard = TRUE; |
| 1490 | PL_numeric_underlying = PL_numeric_underlying_is_standard; |
| 1491 | set_numeric_radix(0); |
| 1492 | |
| 1493 | #endif /* USE_LOCALE_NUMERIC */ |
| 1494 | |
| 1495 | } |
| 1496 | |
| 1497 | void |
| 1498 | Perl_set_numeric_underlying(pTHX) |
| 1499 | { |
| 1500 | |
| 1501 | #ifdef USE_LOCALE_NUMERIC |
| 1502 | |
| 1503 | /* Toggle the LC_NUMERIC locale to the current underlying default. Most |
| 1504 | * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h |
| 1505 | * instead of calling this directly. The macro avoids calling this routine |
| 1506 | * if toggling isn't necessary according to our records (which could be |
| 1507 | * wrong if some XS code has changed the locale behind our back) */ |
| 1508 | |
| 1509 | # ifdef DEBUGGING |
| 1510 | |
| 1511 | if (DEBUG_L_TEST || debug_initialization) { |
| 1512 | PerlIO_printf(Perl_debug_log, |
| 1513 | "Setting LC_NUMERIC locale to %s\n", |
| 1514 | PL_numeric_name); |
| 1515 | } |
| 1516 | |
| 1517 | # endif |
| 1518 | |
| 1519 | do_setlocale_c(LC_NUMERIC, PL_numeric_name); |
| 1520 | PL_numeric_standard = PL_numeric_underlying_is_standard; |
| 1521 | PL_numeric_underlying = TRUE; |
| 1522 | set_numeric_radix(! PL_numeric_standard); |
| 1523 | |
| 1524 | #endif /* USE_LOCALE_NUMERIC */ |
| 1525 | |
| 1526 | } |
| 1527 | |
| 1528 | /* |
| 1529 | * Set up for a new ctype locale. |
| 1530 | */ |
| 1531 | STATIC void |
| 1532 | S_new_ctype(pTHX_ const char *newctype) |
| 1533 | { |
| 1534 | |
| 1535 | #ifndef USE_LOCALE_CTYPE |
| 1536 | |
| 1537 | PERL_UNUSED_ARG(newctype); |
| 1538 | PERL_UNUSED_CONTEXT; |
| 1539 | |
| 1540 | #else |
| 1541 | |
| 1542 | /* Called after each libc setlocale() call affecting LC_CTYPE, to tell |
| 1543 | * core Perl this and that 'newctype' is the name of the new locale. |
| 1544 | * |
| 1545 | * This function sets up the folding arrays for all 256 bytes, assuming |
| 1546 | * that tofold() is tolc() since fold case is not a concept in POSIX, |
| 1547 | * |
| 1548 | * Any code changing the locale (outside this file) should use |
| 1549 | * Perl_setlocale or POSIX::setlocale, which call this function. Therefore |
| 1550 | * this function should be called directly only from this file and from |
| 1551 | * POSIX::setlocale() */ |
| 1552 | |
| 1553 | dVAR; |
| 1554 | unsigned int i; |
| 1555 | |
| 1556 | /* Don't check for problems if we are suppressing the warnings */ |
| 1557 | bool check_for_problems = ckWARN_d(WARN_LOCALE) || UNLIKELY(DEBUG_L_TEST); |
| 1558 | bool maybe_utf8_turkic = FALSE; |
| 1559 | |
| 1560 | PERL_ARGS_ASSERT_NEW_CTYPE; |
| 1561 | |
| 1562 | /* We will replace any bad locale warning with 1) nothing if the new one is |
| 1563 | * ok; or 2) a new warning for the bad new locale */ |
| 1564 | if (PL_warn_locale) { |
| 1565 | SvREFCNT_dec_NN(PL_warn_locale); |
| 1566 | PL_warn_locale = NULL; |
| 1567 | } |
| 1568 | |
| 1569 | PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE); |
| 1570 | |
| 1571 | /* A UTF-8 locale gets standard rules. But note that code still has to |
| 1572 | * handle this specially because of the three problematic code points */ |
| 1573 | if (PL_in_utf8_CTYPE_locale) { |
| 1574 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); |
| 1575 | |
| 1576 | /* UTF-8 locales can have special handling for 'I' and 'i' if they are |
| 1577 | * Turkic. Make sure these two are the only anomalies. (We don't use |
| 1578 | * towupper and towlower because they aren't in C89.) */ |
| 1579 | |
| 1580 | #if defined(HAS_TOWUPPER) && defined (HAS_TOWLOWER) |
| 1581 | |
| 1582 | if (towupper('i') == 0x130 && towlower('I') == 0x131) { |
| 1583 | |
| 1584 | #else |
| 1585 | |
| 1586 | if (toupper('i') == 'i' && tolower('I') == 'I') { |
| 1587 | |
| 1588 | #endif |
| 1589 | check_for_problems = TRUE; |
| 1590 | maybe_utf8_turkic = TRUE; |
| 1591 | } |
| 1592 | } |
| 1593 | |
| 1594 | /* We don't populate the other lists if a UTF-8 locale, but do check that |
| 1595 | * everything works as expected, unless checking turned off */ |
| 1596 | if (check_for_problems || ! PL_in_utf8_CTYPE_locale) { |
| 1597 | /* Assume enough space for every character being bad. 4 spaces each |
| 1598 | * for the 94 printable characters that are output like "'x' "; and 5 |
| 1599 | * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating |
| 1600 | * NUL */ |
| 1601 | char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ] = { '\0' }; |
| 1602 | bool multi_byte_locale = FALSE; /* Assume is a single-byte locale |
| 1603 | to start */ |
| 1604 | unsigned int bad_count = 0; /* Count of bad characters */ |
| 1605 | |
| 1606 | for (i = 0; i < 256; i++) { |
| 1607 | if (! PL_in_utf8_CTYPE_locale) { |
| 1608 | if (isupper(i)) |
| 1609 | PL_fold_locale[i] = (U8) tolower(i); |
| 1610 | else if (islower(i)) |
| 1611 | PL_fold_locale[i] = (U8) toupper(i); |
| 1612 | else |
| 1613 | PL_fold_locale[i] = (U8) i; |
| 1614 | } |
| 1615 | |
| 1616 | /* If checking for locale problems, see if the native ASCII-range |
| 1617 | * printables plus \n and \t are in their expected categories in |
| 1618 | * the new locale. If not, this could mean big trouble, upending |
| 1619 | * Perl's and most programs' assumptions, like having a |
| 1620 | * metacharacter with special meaning become a \w. Fortunately, |
| 1621 | * it's very rare to find locales that aren't supersets of ASCII |
| 1622 | * nowadays. It isn't a problem for most controls to be changed |
| 1623 | * into something else; we check only \n and \t, though perhaps \r |
| 1624 | * could be an issue as well. */ |
| 1625 | if ( check_for_problems |
| 1626 | && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n')) |
| 1627 | { |
| 1628 | bool is_bad = FALSE; |
| 1629 | char name[4] = { '\0' }; |
| 1630 | |
| 1631 | /* Convert the name into a string */ |
| 1632 | if (isGRAPH_A(i)) { |
| 1633 | name[0] = i; |
| 1634 | name[1] = '\0'; |
| 1635 | } |
| 1636 | else if (i == '\n') { |
| 1637 | my_strlcpy(name, "\\n", sizeof(name)); |
| 1638 | } |
| 1639 | else if (i == '\t') { |
| 1640 | my_strlcpy(name, "\\t", sizeof(name)); |
| 1641 | } |
| 1642 | else { |
| 1643 | assert(i == ' '); |
| 1644 | my_strlcpy(name, "' '", sizeof(name)); |
| 1645 | } |
| 1646 | |
| 1647 | /* Check each possibe class */ |
| 1648 | if (UNLIKELY(cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC_A(i)))) { |
| 1649 | is_bad = TRUE; |
| 1650 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1651 | "isalnum('%s') unexpectedly is %d\n", |
| 1652 | name, cBOOL(isalnum(i)))); |
| 1653 | } |
| 1654 | if (UNLIKELY(cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i)))) { |
| 1655 | is_bad = TRUE; |
| 1656 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1657 | "isalpha('%s') unexpectedly is %d\n", |
| 1658 | name, cBOOL(isalpha(i)))); |
| 1659 | } |
| 1660 | if (UNLIKELY(cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i)))) { |
| 1661 | is_bad = TRUE; |
| 1662 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1663 | "isdigit('%s') unexpectedly is %d\n", |
| 1664 | name, cBOOL(isdigit(i)))); |
| 1665 | } |
| 1666 | if (UNLIKELY(cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i)))) { |
| 1667 | is_bad = TRUE; |
| 1668 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1669 | "isgraph('%s') unexpectedly is %d\n", |
| 1670 | name, cBOOL(isgraph(i)))); |
| 1671 | } |
| 1672 | if (UNLIKELY(cBOOL(islower(i)) != cBOOL(isLOWER_A(i)))) { |
| 1673 | is_bad = TRUE; |
| 1674 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1675 | "islower('%s') unexpectedly is %d\n", |
| 1676 | name, cBOOL(islower(i)))); |
| 1677 | } |
| 1678 | if (UNLIKELY(cBOOL(isprint(i)) != cBOOL(isPRINT_A(i)))) { |
| 1679 | is_bad = TRUE; |
| 1680 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1681 | "isprint('%s') unexpectedly is %d\n", |
| 1682 | name, cBOOL(isprint(i)))); |
| 1683 | } |
| 1684 | if (UNLIKELY(cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i)))) { |
| 1685 | is_bad = TRUE; |
| 1686 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1687 | "ispunct('%s') unexpectedly is %d\n", |
| 1688 | name, cBOOL(ispunct(i)))); |
| 1689 | } |
| 1690 | if (UNLIKELY(cBOOL(isspace(i)) != cBOOL(isSPACE_A(i)))) { |
| 1691 | is_bad = TRUE; |
| 1692 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1693 | "isspace('%s') unexpectedly is %d\n", |
| 1694 | name, cBOOL(isspace(i)))); |
| 1695 | } |
| 1696 | if (UNLIKELY(cBOOL(isupper(i)) != cBOOL(isUPPER_A(i)))) { |
| 1697 | is_bad = TRUE; |
| 1698 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1699 | "isupper('%s') unexpectedly is %d\n", |
| 1700 | name, cBOOL(isupper(i)))); |
| 1701 | } |
| 1702 | if (UNLIKELY(cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i)))) { |
| 1703 | is_bad = TRUE; |
| 1704 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1705 | "isxdigit('%s') unexpectedly is %d\n", |
| 1706 | name, cBOOL(isxdigit(i)))); |
| 1707 | } |
| 1708 | if (UNLIKELY(tolower(i) != (int) toLOWER_A(i))) { |
| 1709 | is_bad = TRUE; |
| 1710 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1711 | "tolower('%s')=0x%x instead of the expected 0x%x\n", |
| 1712 | name, tolower(i), (int) toLOWER_A(i))); |
| 1713 | } |
| 1714 | if (UNLIKELY(toupper(i) != (int) toUPPER_A(i))) { |
| 1715 | is_bad = TRUE; |
| 1716 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1717 | "toupper('%s')=0x%x instead of the expected 0x%x\n", |
| 1718 | name, toupper(i), (int) toUPPER_A(i))); |
| 1719 | } |
| 1720 | if (UNLIKELY((i == '\n' && ! isCNTRL_LC(i)))) { |
| 1721 | is_bad = TRUE; |
| 1722 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 1723 | "'\\n' (=%02X) is not a control\n", (int) i)); |
| 1724 | } |
| 1725 | |
| 1726 | /* Add to the list; Separate multiple entries with a blank */ |
| 1727 | if (is_bad) { |
| 1728 | if (bad_count) { |
| 1729 | my_strlcat(bad_chars_list, " ", sizeof(bad_chars_list)); |
| 1730 | } |
| 1731 | my_strlcat(bad_chars_list, name, sizeof(bad_chars_list)); |
| 1732 | bad_count++; |
| 1733 | } |
| 1734 | } |
| 1735 | } |
| 1736 | |
| 1737 | if (bad_count == 2 && maybe_utf8_turkic) { |
| 1738 | bad_count = 0; |
| 1739 | *bad_chars_list = '\0'; |
| 1740 | PL_fold_locale['I'] = 'I'; |
| 1741 | PL_fold_locale['i'] = 'i'; |
| 1742 | PL_in_utf8_turkic_locale = TRUE; |
| 1743 | DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s is turkic\n", |
| 1744 | __FILE__, __LINE__, newctype)); |
| 1745 | } |
| 1746 | else { |
| 1747 | PL_in_utf8_turkic_locale = FALSE; |
| 1748 | } |
| 1749 | |
| 1750 | # ifdef MB_CUR_MAX |
| 1751 | |
| 1752 | /* We only handle single-byte locales (outside of UTF-8 ones; so if |
| 1753 | * this locale requires more than one byte, there are going to be |
| 1754 | * problems. */ |
| 1755 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 1756 | "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n", |
| 1757 | __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX)); |
| 1758 | |
| 1759 | if ( check_for_problems && MB_CUR_MAX > 1 |
| 1760 | && ! PL_in_utf8_CTYPE_locale |
| 1761 | |
| 1762 | /* Some platforms return MB_CUR_MAX > 1 for even the "C" |
| 1763 | * locale. Just assume that the implementation for them (plus |
| 1764 | * for POSIX) is correct and the > 1 value is spurious. (Since |
| 1765 | * these are specially handled to never be considered UTF-8 |
| 1766 | * locales, as long as this is the only problem, everything |
| 1767 | * should work fine */ |
| 1768 | && strNE(newctype, "C") && strNE(newctype, "POSIX")) |
| 1769 | { |
| 1770 | multi_byte_locale = TRUE; |
| 1771 | } |
| 1772 | |
| 1773 | # endif |
| 1774 | |
| 1775 | /* If we found problems and we want them output, do so */ |
| 1776 | if ( (UNLIKELY(bad_count) || UNLIKELY(multi_byte_locale)) |
| 1777 | && (LIKELY(ckWARN_d(WARN_LOCALE)) || UNLIKELY(DEBUG_L_TEST))) |
| 1778 | { |
| 1779 | if (UNLIKELY(bad_count) && PL_in_utf8_CTYPE_locale) { |
| 1780 | PL_warn_locale = Perl_newSVpvf(aTHX_ |
| 1781 | "Locale '%s' contains (at least) the following characters" |
| 1782 | " which have\nunexpected meanings: %s\nThe Perl program" |
| 1783 | " will use the expected meanings", |
| 1784 | newctype, bad_chars_list); |
| 1785 | } |
| 1786 | else { |
| 1787 | PL_warn_locale = Perl_newSVpvf(aTHX_ |
| 1788 | "Locale '%s' may not work well.%s%s%s\n", |
| 1789 | newctype, |
| 1790 | (multi_byte_locale) |
| 1791 | ? " Some characters in it are not recognized by" |
| 1792 | " Perl." |
| 1793 | : "", |
| 1794 | (bad_count) |
| 1795 | ? "\nThe following characters (and maybe others)" |
| 1796 | " may not have the same meaning as the Perl" |
| 1797 | " program expects:\n" |
| 1798 | : "", |
| 1799 | (bad_count) |
| 1800 | ? bad_chars_list |
| 1801 | : "" |
| 1802 | ); |
| 1803 | } |
| 1804 | |
| 1805 | # ifdef HAS_NL_LANGINFO |
| 1806 | |
| 1807 | Perl_sv_catpvf(aTHX_ PL_warn_locale, "; codeset=%s", |
| 1808 | /* parameter FALSE is a don't care here */ |
| 1809 | my_nl_langinfo(CODESET, FALSE)); |
| 1810 | |
| 1811 | # endif |
| 1812 | |
| 1813 | Perl_sv_catpvf(aTHX_ PL_warn_locale, "\n"); |
| 1814 | |
| 1815 | /* If we are actually in the scope of the locale or are debugging, |
| 1816 | * output the message now. If not in that scope, we save the |
| 1817 | * message to be output at the first operation using this locale, |
| 1818 | * if that actually happens. Most programs don't use locales, so |
| 1819 | * they are immune to bad ones. */ |
| 1820 | if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) { |
| 1821 | |
| 1822 | /* The '0' below suppresses a bogus gcc compiler warning */ |
| 1823 | Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0); |
| 1824 | |
| 1825 | if (IN_LC(LC_CTYPE)) { |
| 1826 | SvREFCNT_dec_NN(PL_warn_locale); |
| 1827 | PL_warn_locale = NULL; |
| 1828 | } |
| 1829 | } |
| 1830 | } |
| 1831 | } |
| 1832 | |
| 1833 | #endif /* USE_LOCALE_CTYPE */ |
| 1834 | |
| 1835 | } |
| 1836 | |
| 1837 | void |
| 1838 | Perl__warn_problematic_locale() |
| 1839 | { |
| 1840 | |
| 1841 | #ifdef USE_LOCALE_CTYPE |
| 1842 | |
| 1843 | dTHX; |
| 1844 | |
| 1845 | /* Internal-to-core function that outputs the message in PL_warn_locale, |
| 1846 | * and then NULLS it. Should be called only through the macro |
| 1847 | * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */ |
| 1848 | |
| 1849 | if (PL_warn_locale) { |
| 1850 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), |
| 1851 | SvPVX(PL_warn_locale), |
| 1852 | 0 /* dummy to avoid compiler warning */ ); |
| 1853 | SvREFCNT_dec_NN(PL_warn_locale); |
| 1854 | PL_warn_locale = NULL; |
| 1855 | } |
| 1856 | |
| 1857 | #endif |
| 1858 | |
| 1859 | } |
| 1860 | |
| 1861 | STATIC void |
| 1862 | S_new_collate(pTHX_ const char *newcoll) |
| 1863 | { |
| 1864 | |
| 1865 | #ifndef USE_LOCALE_COLLATE |
| 1866 | |
| 1867 | PERL_UNUSED_ARG(newcoll); |
| 1868 | PERL_UNUSED_CONTEXT; |
| 1869 | |
| 1870 | #else |
| 1871 | |
| 1872 | /* Called after each libc setlocale() call affecting LC_COLLATE, to tell |
| 1873 | * core Perl this and that 'newcoll' is the name of the new locale. |
| 1874 | * |
| 1875 | * The design of locale collation is that every locale change is given an |
| 1876 | * index 'PL_collation_ix'. The first time a string particpates in an |
| 1877 | * operation that requires collation while locale collation is active, it |
| 1878 | * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That |
| 1879 | * magic includes the collation index, and the transformation of the string |
| 1880 | * by strxfrm(), q.v. That transformation is used when doing comparisons, |
| 1881 | * instead of the string itself. If a string changes, the magic is |
| 1882 | * cleared. The next time the locale changes, the index is incremented, |
| 1883 | * and so we know during a comparison that the transformation is not |
| 1884 | * necessarily still valid, and so is recomputed. Note that if the locale |
| 1885 | * changes enough times, the index could wrap (a U32), and it is possible |
| 1886 | * that a transformation would improperly be considered valid, leading to |
| 1887 | * an unlikely bug */ |
| 1888 | |
| 1889 | if (! newcoll) { |
| 1890 | if (PL_collation_name) { |
| 1891 | ++PL_collation_ix; |
| 1892 | Safefree(PL_collation_name); |
| 1893 | PL_collation_name = NULL; |
| 1894 | } |
| 1895 | PL_collation_standard = TRUE; |
| 1896 | is_standard_collation: |
| 1897 | PL_collxfrm_base = 0; |
| 1898 | PL_collxfrm_mult = 2; |
| 1899 | PL_in_utf8_COLLATE_locale = FALSE; |
| 1900 | PL_strxfrm_NUL_replacement = '\0'; |
| 1901 | PL_strxfrm_max_cp = 0; |
| 1902 | return; |
| 1903 | } |
| 1904 | |
| 1905 | /* If this is not the same locale as currently, set the new one up */ |
| 1906 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { |
| 1907 | ++PL_collation_ix; |
| 1908 | Safefree(PL_collation_name); |
| 1909 | PL_collation_name = stdize_locale(savepv(newcoll)); |
| 1910 | PL_collation_standard = isNAME_C_OR_POSIX(newcoll); |
| 1911 | if (PL_collation_standard) { |
| 1912 | goto is_standard_collation; |
| 1913 | } |
| 1914 | |
| 1915 | PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE); |
| 1916 | PL_strxfrm_NUL_replacement = '\0'; |
| 1917 | PL_strxfrm_max_cp = 0; |
| 1918 | |
| 1919 | /* A locale collation definition includes primary, secondary, tertiary, |
| 1920 | * etc. weights for each character. To sort, the primary weights are |
| 1921 | * used, and only if they compare equal, then the secondary weights are |
| 1922 | * used, and only if they compare equal, then the tertiary, etc. |
| 1923 | * |
| 1924 | * strxfrm() works by taking the input string, say ABC, and creating an |
| 1925 | * output transformed string consisting of first the primary weights, |
| 1926 | * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the |
| 1927 | * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters |
| 1928 | * may not have weights at every level. In our example, let's say B |
| 1929 | * doesn't have a tertiary weight, and A doesn't have a secondary |
| 1930 | * weight. The constructed string is then going to be |
| 1931 | * A¹B¹C¹ B²C² A³C³ .... |
| 1932 | * This has the desired effect that strcmp() will look at the secondary |
| 1933 | * or tertiary weights only if the strings compare equal at all higher |
| 1934 | * priority weights. The spaces shown here, like in |
| 1935 | * "A¹B¹C¹ A²B²C² " |
| 1936 | * are not just for readability. In the general case, these must |
| 1937 | * actually be bytes, which we will call here 'separator weights'; and |
| 1938 | * they must be smaller than any other weight value, but since these |
| 1939 | * are C strings, only the terminating one can be a NUL (some |
| 1940 | * implementations may include a non-NUL separator weight just before |
| 1941 | * the NUL). Implementations tend to reserve 01 for the separator |
| 1942 | * weights. They are needed so that a shorter string's secondary |
| 1943 | * weights won't be misconstrued as primary weights of a longer string, |
| 1944 | * etc. By making them smaller than any other weight, the shorter |
| 1945 | * string will sort first. (Actually, if all secondary weights are |
| 1946 | * smaller than all primary ones, there is no need for a separator |
| 1947 | * weight between those two levels, etc.) |
| 1948 | * |
| 1949 | * The length of the transformed string is roughly a linear function of |
| 1950 | * the input string. It's not exactly linear because some characters |
| 1951 | * don't have weights at all levels. When we call strxfrm() we have to |
| 1952 | * allocate some memory to hold the transformed string. The |
| 1953 | * calculations below try to find coefficients 'm' and 'b' for this |
| 1954 | * locale so that m*x + b equals how much space we need, given the size |
| 1955 | * of the input string in 'x'. If we calculate too small, we increase |
| 1956 | * the size as needed, and call strxfrm() again, but it is better to |
| 1957 | * get it right the first time to avoid wasted expensive string |
| 1958 | * transformations. */ |
| 1959 | |
| 1960 | { |
| 1961 | /* We use the string below to find how long the tranformation of it |
| 1962 | * is. Almost all locales are supersets of ASCII, or at least the |
| 1963 | * ASCII letters. We use all of them, half upper half lower, |
| 1964 | * because if we used fewer, we might hit just the ones that are |
| 1965 | * outliers in a particular locale. Most of the strings being |
| 1966 | * collated will contain a preponderance of letters, and even if |
| 1967 | * they are above-ASCII, they are likely to have the same number of |
| 1968 | * weight levels as the ASCII ones. It turns out that digits tend |
| 1969 | * to have fewer levels, and some punctuation has more, but those |
| 1970 | * are relatively sparse in text, and khw believes this gives a |
| 1971 | * reasonable result, but it could be changed if experience so |
| 1972 | * dictates. */ |
| 1973 | const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz"; |
| 1974 | char * x_longer; /* Transformed 'longer' */ |
| 1975 | Size_t x_len_longer; /* Length of 'x_longer' */ |
| 1976 | |
| 1977 | char * x_shorter; /* We also transform a substring of 'longer' */ |
| 1978 | Size_t x_len_shorter; |
| 1979 | |
| 1980 | /* _mem_collxfrm() is used get the transformation (though here we |
| 1981 | * are interested only in its length). It is used because it has |
| 1982 | * the intelligence to handle all cases, but to work, it needs some |
| 1983 | * values of 'm' and 'b' to get it started. For the purposes of |
| 1984 | * this calculation we use a very conservative estimate of 'm' and |
| 1985 | * 'b'. This assumes a weight can be multiple bytes, enough to |
| 1986 | * hold any UV on the platform, and there are 5 levels, 4 weight |
| 1987 | * bytes, and a trailing NUL. */ |
| 1988 | PL_collxfrm_base = 5; |
| 1989 | PL_collxfrm_mult = 5 * sizeof(UV); |
| 1990 | |
| 1991 | /* Find out how long the transformation really is */ |
| 1992 | x_longer = _mem_collxfrm(longer, |
| 1993 | sizeof(longer) - 1, |
| 1994 | &x_len_longer, |
| 1995 | |
| 1996 | /* We avoid converting to UTF-8 in the |
| 1997 | * called function by telling it the |
| 1998 | * string is in UTF-8 if the locale is a |
| 1999 | * UTF-8 one. Since the string passed |
| 2000 | * here is invariant under UTF-8, we can |
| 2001 | * claim it's UTF-8 even though it isn't. |
| 2002 | * */ |
| 2003 | PL_in_utf8_COLLATE_locale); |
| 2004 | Safefree(x_longer); |
| 2005 | |
| 2006 | /* Find out how long the transformation of a substring of 'longer' |
| 2007 | * is. Together the lengths of these transformations are |
| 2008 | * sufficient to calculate 'm' and 'b'. The substring is all of |
| 2009 | * 'longer' except the first character. This minimizes the chances |
| 2010 | * of being swayed by outliers */ |
| 2011 | x_shorter = _mem_collxfrm(longer + 1, |
| 2012 | sizeof(longer) - 2, |
| 2013 | &x_len_shorter, |
| 2014 | PL_in_utf8_COLLATE_locale); |
| 2015 | Safefree(x_shorter); |
| 2016 | |
| 2017 | /* If the results are nonsensical for this simple test, the whole |
| 2018 | * locale definition is suspect. Mark it so that locale collation |
| 2019 | * is not active at all for it. XXX Should we warn? */ |
| 2020 | if ( x_len_shorter == 0 |
| 2021 | || x_len_longer == 0 |
| 2022 | || x_len_shorter >= x_len_longer) |
| 2023 | { |
| 2024 | PL_collxfrm_mult = 0; |
| 2025 | PL_collxfrm_base = 0; |
| 2026 | } |
| 2027 | else { |
| 2028 | SSize_t base; /* Temporary */ |
| 2029 | |
| 2030 | /* We have both: m * strlen(longer) + b = x_len_longer |
| 2031 | * m * strlen(shorter) + b = x_len_shorter; |
| 2032 | * subtracting yields: |
| 2033 | * m * (strlen(longer) - strlen(shorter)) |
| 2034 | * = x_len_longer - x_len_shorter |
| 2035 | * But we have set things up so that 'shorter' is 1 byte smaller |
| 2036 | * than 'longer'. Hence: |
| 2037 | * m = x_len_longer - x_len_shorter |
| 2038 | * |
| 2039 | * But if something went wrong, make sure the multiplier is at |
| 2040 | * least 1. |
| 2041 | */ |
| 2042 | if (x_len_longer > x_len_shorter) { |
| 2043 | PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter; |
| 2044 | } |
| 2045 | else { |
| 2046 | PL_collxfrm_mult = 1; |
| 2047 | } |
| 2048 | |
| 2049 | /* mx + b = len |
| 2050 | * so: b = len - mx |
| 2051 | * but in case something has gone wrong, make sure it is |
| 2052 | * non-negative */ |
| 2053 | base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1); |
| 2054 | if (base < 0) { |
| 2055 | base = 0; |
| 2056 | } |
| 2057 | |
| 2058 | /* Add 1 for the trailing NUL */ |
| 2059 | PL_collxfrm_base = base + 1; |
| 2060 | } |
| 2061 | |
| 2062 | # ifdef DEBUGGING |
| 2063 | |
| 2064 | if (DEBUG_L_TEST || debug_initialization) { |
| 2065 | PerlIO_printf(Perl_debug_log, |
| 2066 | "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, " |
| 2067 | "x_len_longer=%zu," |
| 2068 | " collate multipler=%zu, collate base=%zu\n", |
| 2069 | __FILE__, __LINE__, |
| 2070 | PL_in_utf8_COLLATE_locale, |
| 2071 | x_len_shorter, x_len_longer, |
| 2072 | PL_collxfrm_mult, PL_collxfrm_base); |
| 2073 | } |
| 2074 | # endif |
| 2075 | |
| 2076 | } |
| 2077 | } |
| 2078 | |
| 2079 | #endif /* USE_LOCALE_COLLATE */ |
| 2080 | |
| 2081 | } |
| 2082 | |
| 2083 | #endif |
| 2084 | |
| 2085 | #ifdef WIN32 |
| 2086 | |
| 2087 | STATIC char * |
| 2088 | S_win32_setlocale(pTHX_ int category, const char* locale) |
| 2089 | { |
| 2090 | /* This, for Windows, emulates POSIX setlocale() behavior. There is no |
| 2091 | * difference between the two unless the input locale is "", which normally |
| 2092 | * means on Windows to get the machine default, which is set via the |
| 2093 | * computer's "Regional and Language Options" (or its current equivalent). |
| 2094 | * In POSIX, it instead means to find the locale from the user's |
| 2095 | * environment. This routine changes the Windows behavior to first look in |
| 2096 | * the environment, and, if anything is found, use that instead of going to |
| 2097 | * the machine default. If there is no environment override, the machine |
| 2098 | * default is used, by calling the real setlocale() with "". |
| 2099 | * |
| 2100 | * The POSIX behavior is to use the LC_ALL variable if set; otherwise to |
| 2101 | * use the particular category's variable if set; otherwise to use the LANG |
| 2102 | * variable. */ |
| 2103 | |
| 2104 | bool override_LC_ALL = FALSE; |
| 2105 | char * result; |
| 2106 | unsigned int i; |
| 2107 | |
| 2108 | if (locale && strEQ(locale, "")) { |
| 2109 | |
| 2110 | # ifdef LC_ALL |
| 2111 | |
| 2112 | locale = PerlEnv_getenv("LC_ALL"); |
| 2113 | if (! locale) { |
| 2114 | if (category == LC_ALL) { |
| 2115 | override_LC_ALL = TRUE; |
| 2116 | } |
| 2117 | else { |
| 2118 | |
| 2119 | # endif |
| 2120 | |
| 2121 | for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) { |
| 2122 | if (category == categories[i]) { |
| 2123 | locale = PerlEnv_getenv(category_names[i]); |
| 2124 | goto found_locale; |
| 2125 | } |
| 2126 | } |
| 2127 | |
| 2128 | locale = PerlEnv_getenv("LANG"); |
| 2129 | if (! locale) { |
| 2130 | locale = ""; |
| 2131 | } |
| 2132 | |
| 2133 | found_locale: ; |
| 2134 | |
| 2135 | # ifdef LC_ALL |
| 2136 | |
| 2137 | } |
| 2138 | } |
| 2139 | |
| 2140 | # endif |
| 2141 | |
| 2142 | } |
| 2143 | |
| 2144 | result = setlocale(category, locale); |
| 2145 | DEBUG_L(STMT_START { |
| 2146 | dSAVE_ERRNO; |
| 2147 | PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__, |
| 2148 | setlocale_debug_string(category, locale, result)); |
| 2149 | RESTORE_ERRNO; |
| 2150 | } STMT_END); |
| 2151 | |
| 2152 | if (! override_LC_ALL) { |
| 2153 | return result; |
| 2154 | } |
| 2155 | |
| 2156 | /* Here the input category was LC_ALL, and we have set it to what is in the |
| 2157 | * LANG variable or the system default if there is no LANG. But these have |
| 2158 | * lower priority than the other LC_foo variables, so override it for each |
| 2159 | * one that is set. (If they are set to "", it means to use the same thing |
| 2160 | * we just set LC_ALL to, so can skip) */ |
| 2161 | |
| 2162 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 2163 | result = PerlEnv_getenv(category_names[i]); |
| 2164 | if (result && strNE(result, "")) { |
| 2165 | setlocale(categories[i], result); |
| 2166 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", |
| 2167 | __FILE__, __LINE__, |
| 2168 | setlocale_debug_string(categories[i], result, "not captured"))); |
| 2169 | } |
| 2170 | } |
| 2171 | |
| 2172 | result = setlocale(LC_ALL, NULL); |
| 2173 | DEBUG_L(STMT_START { |
| 2174 | dSAVE_ERRNO; |
| 2175 | PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", |
| 2176 | __FILE__, __LINE__, |
| 2177 | setlocale_debug_string(LC_ALL, NULL, result)); |
| 2178 | RESTORE_ERRNO; |
| 2179 | } STMT_END); |
| 2180 | |
| 2181 | return result; |
| 2182 | } |
| 2183 | |
| 2184 | #endif |
| 2185 | |
| 2186 | /* |
| 2187 | |
| 2188 | =head1 Locale-related functions and macros |
| 2189 | |
| 2190 | =for apidoc Perl_setlocale |
| 2191 | |
| 2192 | This is an (almost) drop-in replacement for the system L<C<setlocale(3)>>, |
| 2193 | taking the same parameters, and returning the same information, except that it |
| 2194 | returns the correct underlying C<LC_NUMERIC> locale. Regular C<setlocale> will |
| 2195 | instead return C<C> if the underlying locale has a non-dot decimal point |
| 2196 | character, or a non-empty thousands separator for displaying floating point |
| 2197 | numbers. This is because perl keeps that locale category such that it has a |
| 2198 | dot and empty separator, changing the locale briefly during the operations |
| 2199 | where the underlying one is required. C<Perl_setlocale> knows about this, and |
| 2200 | compensates; regular C<setlocale> doesn't. |
| 2201 | |
| 2202 | Another reason it isn't completely a drop-in replacement is that it is |
| 2203 | declared to return S<C<const char *>>, whereas the system setlocale omits the |
| 2204 | C<const> (presumably because its API was specified long ago, and can't be |
| 2205 | updated; it is illegal to change the information C<setlocale> returns; doing |
| 2206 | so leads to segfaults.) |
| 2207 | |
| 2208 | Finally, C<Perl_setlocale> works under all circumstances, whereas plain |
| 2209 | C<setlocale> can be completely ineffective on some platforms under some |
| 2210 | configurations. |
| 2211 | |
| 2212 | C<Perl_setlocale> should not be used to change the locale except on systems |
| 2213 | where the predefined variable C<${^SAFE_LOCALES}> is 1. On some such systems, |
| 2214 | the system C<setlocale()> is ineffective, returning the wrong information, and |
| 2215 | failing to actually change the locale. C<Perl_setlocale>, however works |
| 2216 | properly in all circumstances. |
| 2217 | |
| 2218 | The return points to a per-thread static buffer, which is overwritten the next |
| 2219 | time C<Perl_setlocale> is called from the same thread. |
| 2220 | |
| 2221 | =cut |
| 2222 | |
| 2223 | */ |
| 2224 | |
| 2225 | const char * |
| 2226 | Perl_setlocale(const int category, const char * locale) |
| 2227 | { |
| 2228 | /* This wraps POSIX::setlocale() */ |
| 2229 | |
| 2230 | #ifndef USE_LOCALE |
| 2231 | |
| 2232 | PERL_UNUSED_ARG(category); |
| 2233 | PERL_UNUSED_ARG(locale); |
| 2234 | |
| 2235 | return "C"; |
| 2236 | |
| 2237 | #else |
| 2238 | |
| 2239 | const char * retval; |
| 2240 | const char * newlocale; |
| 2241 | dSAVEDERRNO; |
| 2242 | dTHX; |
| 2243 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; |
| 2244 | |
| 2245 | #ifdef USE_LOCALE_NUMERIC |
| 2246 | |
| 2247 | /* A NULL locale means only query what the current one is. We have the |
| 2248 | * LC_NUMERIC name saved, because we are normally switched into the C |
| 2249 | * (or equivalent) locale for it. For an LC_ALL query, switch back to get |
| 2250 | * the correct results. All other categories don't require special |
| 2251 | * handling */ |
| 2252 | if (locale == NULL) { |
| 2253 | if (category == LC_NUMERIC) { |
| 2254 | |
| 2255 | /* We don't have to copy this return value, as it is a per-thread |
| 2256 | * variable, and won't change until a future setlocale */ |
| 2257 | return PL_numeric_name; |
| 2258 | } |
| 2259 | |
| 2260 | # ifdef LC_ALL |
| 2261 | |
| 2262 | else if (category == LC_ALL) { |
| 2263 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
| 2264 | } |
| 2265 | |
| 2266 | # endif |
| 2267 | |
| 2268 | } |
| 2269 | |
| 2270 | #endif |
| 2271 | |
| 2272 | retval = save_to_buffer(do_setlocale_r(category, locale), |
| 2273 | &PL_setlocale_buf, &PL_setlocale_bufsize, 0); |
| 2274 | SAVE_ERRNO; |
| 2275 | |
| 2276 | #if defined(USE_LOCALE_NUMERIC) && defined(LC_ALL) |
| 2277 | |
| 2278 | if (locale == NULL && category == LC_ALL) { |
| 2279 | RESTORE_LC_NUMERIC(); |
| 2280 | } |
| 2281 | |
| 2282 | #endif |
| 2283 | |
| 2284 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 2285 | "%s:%d: %s\n", __FILE__, __LINE__, |
| 2286 | setlocale_debug_string(category, locale, retval))); |
| 2287 | |
| 2288 | RESTORE_ERRNO; |
| 2289 | |
| 2290 | if (! retval) { |
| 2291 | return NULL; |
| 2292 | } |
| 2293 | |
| 2294 | /* If locale == NULL, we are just querying the state */ |
| 2295 | if (locale == NULL) { |
| 2296 | return retval; |
| 2297 | } |
| 2298 | |
| 2299 | /* Now that have switched locales, we have to update our records to |
| 2300 | * correspond. */ |
| 2301 | |
| 2302 | switch (category) { |
| 2303 | |
| 2304 | #ifdef USE_LOCALE_CTYPE |
| 2305 | |
| 2306 | case LC_CTYPE: |
| 2307 | new_ctype(retval); |
| 2308 | break; |
| 2309 | |
| 2310 | #endif |
| 2311 | #ifdef USE_LOCALE_COLLATE |
| 2312 | |
| 2313 | case LC_COLLATE: |
| 2314 | new_collate(retval); |
| 2315 | break; |
| 2316 | |
| 2317 | #endif |
| 2318 | #ifdef USE_LOCALE_NUMERIC |
| 2319 | |
| 2320 | case LC_NUMERIC: |
| 2321 | new_numeric(retval); |
| 2322 | break; |
| 2323 | |
| 2324 | #endif |
| 2325 | #ifdef LC_ALL |
| 2326 | |
| 2327 | case LC_ALL: |
| 2328 | |
| 2329 | /* LC_ALL updates all the things we care about. The values may not |
| 2330 | * be the same as 'retval', as the locale "" may have set things |
| 2331 | * individually */ |
| 2332 | |
| 2333 | # ifdef USE_LOCALE_CTYPE |
| 2334 | |
| 2335 | newlocale = savepv(do_setlocale_c(LC_CTYPE, NULL)); |
| 2336 | new_ctype(newlocale); |
| 2337 | Safefree(newlocale); |
| 2338 | |
| 2339 | # endif /* USE_LOCALE_CTYPE */ |
| 2340 | # ifdef USE_LOCALE_COLLATE |
| 2341 | |
| 2342 | newlocale = savepv(do_setlocale_c(LC_COLLATE, NULL)); |
| 2343 | new_collate(newlocale); |
| 2344 | Safefree(newlocale); |
| 2345 | |
| 2346 | # endif |
| 2347 | # ifdef USE_LOCALE_NUMERIC |
| 2348 | |
| 2349 | newlocale = savepv(do_setlocale_c(LC_NUMERIC, NULL)); |
| 2350 | new_numeric(newlocale); |
| 2351 | Safefree(newlocale); |
| 2352 | |
| 2353 | # endif /* USE_LOCALE_NUMERIC */ |
| 2354 | #endif /* LC_ALL */ |
| 2355 | |
| 2356 | default: |
| 2357 | break; |
| 2358 | } |
| 2359 | |
| 2360 | return retval; |
| 2361 | |
| 2362 | #endif |
| 2363 | |
| 2364 | } |
| 2365 | |
| 2366 | PERL_STATIC_INLINE const char * |
| 2367 | S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset) |
| 2368 | { |
| 2369 | /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size', |
| 2370 | * growing it if necessary */ |
| 2371 | |
| 2372 | Size_t string_size; |
| 2373 | |
| 2374 | PERL_ARGS_ASSERT_SAVE_TO_BUFFER; |
| 2375 | |
| 2376 | if (! string) { |
| 2377 | return NULL; |
| 2378 | } |
| 2379 | |
| 2380 | string_size = strlen(string) + offset + 1; |
| 2381 | |
| 2382 | if (*buf_size == 0) { |
| 2383 | Newx(*buf, string_size, char); |
| 2384 | *buf_size = string_size; |
| 2385 | } |
| 2386 | else if (string_size > *buf_size) { |
| 2387 | Renew(*buf, string_size, char); |
| 2388 | *buf_size = string_size; |
| 2389 | } |
| 2390 | |
| 2391 | Copy(string, *buf + offset, string_size - offset, char); |
| 2392 | return *buf; |
| 2393 | } |
| 2394 | |
| 2395 | /* |
| 2396 | |
| 2397 | =for apidoc Perl_langinfo |
| 2398 | |
| 2399 | This is an (almost) drop-in replacement for the system C<L<nl_langinfo(3)>>, |
| 2400 | taking the same C<item> parameter values, and returning the same information. |
| 2401 | But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks |
| 2402 | of Perl's locale handling from your code, and can be used on systems that lack |
| 2403 | a native C<nl_langinfo>. |
| 2404 | |
| 2405 | Expanding on these: |
| 2406 | |
| 2407 | =over |
| 2408 | |
| 2409 | =item * |
| 2410 | |
| 2411 | The reason it isn't quite a drop-in replacement is actually an advantage. The |
| 2412 | only difference is that it returns S<C<const char *>>, whereas plain |
| 2413 | C<nl_langinfo()> returns S<C<char *>>, but you are (only by documentation) |
| 2414 | forbidden to write into the buffer. By declaring this C<const>, the compiler |
| 2415 | enforces this restriction, so if it is violated, you know at compilation time, |
| 2416 | rather than getting segfaults at runtime. |
| 2417 | |
| 2418 | =item * |
| 2419 | |
| 2420 | It delivers the correct results for the C<RADIXCHAR> and C<THOUSEP> items, |
| 2421 | without you having to write extra code. The reason for the extra code would be |
| 2422 | because these are from the C<LC_NUMERIC> locale category, which is normally |
| 2423 | kept set by Perl so that the radix is a dot, and the separator is the empty |
| 2424 | string, no matter what the underlying locale is supposed to be, and so to get |
| 2425 | the expected results, you have to temporarily toggle into the underlying |
| 2426 | locale, and later toggle back. (You could use plain C<nl_langinfo> and |
| 2427 | C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this but then you wouldn't get |
| 2428 | the other advantages of C<Perl_langinfo()>; not keeping C<LC_NUMERIC> in the C |
| 2429 | (or equivalent) locale would break a lot of CPAN, which is expecting the radix |
| 2430 | (decimal point) character to be a dot.) |
| 2431 | |
| 2432 | =item * |
| 2433 | |
| 2434 | The system function it replaces can have its static return buffer trashed, |
| 2435 | not only by a subesequent call to that function, but by a C<freelocale>, |
| 2436 | C<setlocale>, or other locale change. The returned buffer of this function is |
| 2437 | not changed until the next call to it, so the buffer is never in a trashed |
| 2438 | state. |
| 2439 | |
| 2440 | =item * |
| 2441 | |
| 2442 | Its return buffer is per-thread, so it also is never overwritten by a call to |
| 2443 | this function from another thread; unlike the function it replaces. |
| 2444 | |
| 2445 | =item * |
| 2446 | |
| 2447 | But most importantly, it works on systems that don't have C<nl_langinfo>, such |
| 2448 | as Windows, hence makes your code more portable. Of the fifty-some possible |
| 2449 | items specified by the POSIX 2008 standard, |
| 2450 | L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>, |
| 2451 | only one is completely unimplemented, though on non-Windows platforms, another |
| 2452 | significant one is also not implemented). It uses various techniques to |
| 2453 | recover the other items, including calling C<L<localeconv(3)>>, and |
| 2454 | C<L<strftime(3)>>, both of which are specified in C89, so should be always be |
| 2455 | available. Later C<strftime()> versions have additional capabilities; C<""> is |
| 2456 | returned for those not available on your system. |
| 2457 | |
| 2458 | It is important to note that when called with an item that is recovered by |
| 2459 | using C<localeconv>, the buffer from any previous explicit call to |
| 2460 | C<localeconv> will be overwritten. This means you must save that buffer's |
| 2461 | contents if you need to access them after a call to this function. (But note |
| 2462 | that you might not want to be using C<localeconv()> directly anyway, because of |
| 2463 | issues like the ones listed in the second item of this list (above) for |
| 2464 | C<RADIXCHAR> and C<THOUSEP>. You can use the methods given in L<perlcall> to |
| 2465 | call L<POSIX/localeconv> and avoid all the issues, but then you have a hash to |
| 2466 | unpack). |
| 2467 | |
| 2468 | The details for those items which may deviate from what this emulation returns |
| 2469 | and what a native C<nl_langinfo()> would return are specified in |
| 2470 | L<I18N::Langinfo>. |
| 2471 | |
| 2472 | =back |
| 2473 | |
| 2474 | When using C<Perl_langinfo> on systems that don't have a native |
| 2475 | C<nl_langinfo()>, you must |
| 2476 | |
| 2477 | #include "perl_langinfo.h" |
| 2478 | |
| 2479 | before the C<perl.h> C<#include>. You can replace your C<langinfo.h> |
| 2480 | C<#include> with this one. (Doing it this way keeps out the symbols that plain |
| 2481 | C<langinfo.h> would try to import into the namespace for code that doesn't need |
| 2482 | it.) |
| 2483 | |
| 2484 | The original impetus for C<Perl_langinfo()> was so that code that needs to |
| 2485 | find out the current currency symbol, floating point radix character, or digit |
| 2486 | grouping separator can use, on all systems, the simpler and more |
| 2487 | thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a |
| 2488 | pain to make thread-friendly. For other fields returned by C<localeconv>, it |
| 2489 | is better to use the methods given in L<perlcall> to call |
| 2490 | L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly. |
| 2491 | |
| 2492 | =cut |
| 2493 | |
| 2494 | */ |
| 2495 | |
| 2496 | const char * |
| 2497 | #ifdef HAS_NL_LANGINFO |
| 2498 | Perl_langinfo(const nl_item item) |
| 2499 | #else |
| 2500 | Perl_langinfo(const int item) |
| 2501 | #endif |
| 2502 | { |
| 2503 | return my_nl_langinfo(item, TRUE); |
| 2504 | } |
| 2505 | |
| 2506 | STATIC const char * |
| 2507 | #ifdef HAS_NL_LANGINFO |
| 2508 | S_my_nl_langinfo(const nl_item item, bool toggle) |
| 2509 | #else |
| 2510 | S_my_nl_langinfo(const int item, bool toggle) |
| 2511 | #endif |
| 2512 | { |
| 2513 | dTHX; |
| 2514 | const char * retval; |
| 2515 | |
| 2516 | #ifdef USE_LOCALE_NUMERIC |
| 2517 | |
| 2518 | /* We only need to toggle into the underlying LC_NUMERIC locale for these |
| 2519 | * two items, and only if not already there */ |
| 2520 | if (toggle && (( item != RADIXCHAR && item != THOUSEP) |
| 2521 | || PL_numeric_underlying)) |
| 2522 | |
| 2523 | #endif /* No toggling needed if not using LC_NUMERIC */ |
| 2524 | |
| 2525 | toggle = FALSE; |
| 2526 | |
| 2527 | #if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */ |
| 2528 | # if ! defined(HAS_THREAD_SAFE_NL_LANGINFO_L) \ |
| 2529 | || ! defined(HAS_POSIX_2008_LOCALE) \ |
| 2530 | || ! defined(DUPLOCALE) |
| 2531 | |
| 2532 | /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC |
| 2533 | * for those items dependent on it. This must be copied to a buffer before |
| 2534 | * switching back, as some systems destroy the buffer when setlocale() is |
| 2535 | * called */ |
| 2536 | |
| 2537 | { |
| 2538 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; |
| 2539 | |
| 2540 | if (toggle) { |
| 2541 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
| 2542 | } |
| 2543 | |
| 2544 | LOCALE_LOCK; /* Prevent interference from another thread executing |
| 2545 | this code section (the only call to nl_langinfo in |
| 2546 | the core) */ |
| 2547 | |
| 2548 | |
| 2549 | /* Copy to a per-thread buffer, which is also one that won't be |
| 2550 | * destroyed by a subsequent setlocale(), such as the |
| 2551 | * RESTORE_LC_NUMERIC may do just below. */ |
| 2552 | retval = save_to_buffer(nl_langinfo(item), |
| 2553 | &PL_langinfo_buf, &PL_langinfo_bufsize, 0); |
| 2554 | |
| 2555 | LOCALE_UNLOCK; |
| 2556 | |
| 2557 | if (toggle) { |
| 2558 | RESTORE_LC_NUMERIC(); |
| 2559 | } |
| 2560 | } |
| 2561 | |
| 2562 | # else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */ |
| 2563 | |
| 2564 | { |
| 2565 | bool do_free = FALSE; |
| 2566 | locale_t cur = uselocale((locale_t) 0); |
| 2567 | |
| 2568 | if (cur == LC_GLOBAL_LOCALE) { |
| 2569 | cur = duplocale(LC_GLOBAL_LOCALE); |
| 2570 | do_free = TRUE; |
| 2571 | } |
| 2572 | |
| 2573 | # ifdef USE_LOCALE_NUMERIC |
| 2574 | |
| 2575 | if (toggle) { |
| 2576 | if (PL_underlying_numeric_obj) { |
| 2577 | cur = PL_underlying_numeric_obj; |
| 2578 | } |
| 2579 | else { |
| 2580 | cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur); |
| 2581 | do_free = TRUE; |
| 2582 | } |
| 2583 | } |
| 2584 | |
| 2585 | # endif |
| 2586 | |
| 2587 | /* We have to save it to a buffer, because the freelocale() just below |
| 2588 | * can invalidate the internal one */ |
| 2589 | retval = save_to_buffer(nl_langinfo_l(item, cur), |
| 2590 | &PL_langinfo_buf, &PL_langinfo_bufsize, 0); |
| 2591 | |
| 2592 | if (do_free) { |
| 2593 | freelocale(cur); |
| 2594 | } |
| 2595 | } |
| 2596 | |
| 2597 | # endif |
| 2598 | |
| 2599 | if (strEQ(retval, "")) { |
| 2600 | if (item == YESSTR) { |
| 2601 | return "yes"; |
| 2602 | } |
| 2603 | if (item == NOSTR) { |
| 2604 | return "no"; |
| 2605 | } |
| 2606 | } |
| 2607 | |
| 2608 | return retval; |
| 2609 | |
| 2610 | #else /* Below, emulate nl_langinfo as best we can */ |
| 2611 | |
| 2612 | { |
| 2613 | |
| 2614 | # ifdef HAS_LOCALECONV |
| 2615 | |
| 2616 | const struct lconv* lc; |
| 2617 | const char * temp; |
| 2618 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; |
| 2619 | |
| 2620 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2621 | |
| 2622 | const char * save_global; |
| 2623 | const char * save_thread; |
| 2624 | int needed_size; |
| 2625 | char * ptr; |
| 2626 | char * e; |
| 2627 | char * item_start; |
| 2628 | |
| 2629 | # endif |
| 2630 | # endif |
| 2631 | # ifdef HAS_STRFTIME |
| 2632 | |
| 2633 | struct tm tm; |
| 2634 | bool return_format = FALSE; /* Return the %format, not the value */ |
| 2635 | const char * format; |
| 2636 | |
| 2637 | # endif |
| 2638 | |
| 2639 | /* We copy the results to a per-thread buffer, even if not |
| 2640 | * multi-threaded. This is in part to simplify this code, and partly |
| 2641 | * because we need a buffer anyway for strftime(), and partly because a |
| 2642 | * call of localeconv() could otherwise wipe out the buffer, and the |
| 2643 | * programmer would not be expecting this, as this is a nl_langinfo() |
| 2644 | * substitute after all, so s/he might be thinking their localeconv() |
| 2645 | * is safe until another localeconv() call. */ |
| 2646 | |
| 2647 | switch (item) { |
| 2648 | Size_t len; |
| 2649 | |
| 2650 | /* This is unimplemented */ |
| 2651 | case ERA: /* For use with strftime() %E modifier */ |
| 2652 | |
| 2653 | default: |
| 2654 | return ""; |
| 2655 | |
| 2656 | /* We use only an English set, since we don't know any more */ |
| 2657 | case YESEXPR: return "^[+1yY]"; |
| 2658 | case YESSTR: return "yes"; |
| 2659 | case NOEXPR: return "^[-0nN]"; |
| 2660 | case NOSTR: return "no"; |
| 2661 | |
| 2662 | case CODESET: |
| 2663 | |
| 2664 | # ifndef WIN32 |
| 2665 | |
| 2666 | /* On non-windows, this is unimplemented, in part because of |
| 2667 | * inconsistencies between vendors. The Darwin native |
| 2668 | * nl_langinfo() implementation simply looks at everything past |
| 2669 | * any dot in the name, but that doesn't work for other |
| 2670 | * vendors. Many Linux locales that don't have UTF-8 in their |
| 2671 | * names really are UTF-8, for example; z/OS locales that do |
| 2672 | * have UTF-8 in their names, aren't really UTF-8 */ |
| 2673 | return ""; |
| 2674 | |
| 2675 | # else |
| 2676 | |
| 2677 | { /* But on Windows, the name does seem to be consistent, so |
| 2678 | use that. */ |
| 2679 | const char * p; |
| 2680 | const char * first; |
| 2681 | Size_t offset = 0; |
| 2682 | const char * name = my_setlocale(LC_CTYPE, NULL); |
| 2683 | |
| 2684 | if (isNAME_C_OR_POSIX(name)) { |
| 2685 | return "ANSI_X3.4-1968"; |
| 2686 | } |
| 2687 | |
| 2688 | /* Find the dot in the locale name */ |
| 2689 | first = (const char *) strchr(name, '.'); |
| 2690 | if (! first) { |
| 2691 | first = name; |
| 2692 | goto has_nondigit; |
| 2693 | } |
| 2694 | |
| 2695 | /* Look at everything past the dot */ |
| 2696 | first++; |
| 2697 | p = first; |
| 2698 | |
| 2699 | while (*p) { |
| 2700 | if (! isDIGIT(*p)) { |
| 2701 | goto has_nondigit; |
| 2702 | } |
| 2703 | |
| 2704 | p++; |
| 2705 | } |
| 2706 | |
| 2707 | /* Here everything past the dot is a digit. Treat it as a |
| 2708 | * code page */ |
| 2709 | retval = save_to_buffer("CP", &PL_langinfo_buf, |
| 2710 | &PL_langinfo_bufsize, 0); |
| 2711 | offset = STRLENs("CP"); |
| 2712 | |
| 2713 | has_nondigit: |
| 2714 | |
| 2715 | retval = save_to_buffer(first, &PL_langinfo_buf, |
| 2716 | &PL_langinfo_bufsize, offset); |
| 2717 | } |
| 2718 | |
| 2719 | break; |
| 2720 | |
| 2721 | # endif |
| 2722 | # ifdef HAS_LOCALECONV |
| 2723 | |
| 2724 | case CRNCYSTR: |
| 2725 | |
| 2726 | /* We don't bother with localeconv_l() because any system that |
| 2727 | * has it is likely to also have nl_langinfo() */ |
| 2728 | |
| 2729 | LOCALE_LOCK_V; /* Prevent interference with other threads |
| 2730 | using localeconv() */ |
| 2731 | |
| 2732 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2733 | |
| 2734 | /* This is a workaround for a Windows bug prior to VS 15. |
| 2735 | * What we do here is, while locked, switch to the global |
| 2736 | * locale so localeconv() works; then switch back just before |
| 2737 | * the unlock. This can screw things up if some thread is |
| 2738 | * already using the global locale while assuming no other is. |
| 2739 | * A different workaround would be to call GetCurrencyFormat on |
| 2740 | * a known value, and parse it; patches welcome |
| 2741 | * |
| 2742 | * We have to use LC_ALL instead of LC_MONETARY because of |
| 2743 | * another bug in Windows */ |
| 2744 | |
| 2745 | save_thread = savepv(my_setlocale(LC_ALL, NULL)); |
| 2746 | _configthreadlocale(_DISABLE_PER_THREAD_LOCALE); |
| 2747 | save_global= savepv(my_setlocale(LC_ALL, NULL)); |
| 2748 | my_setlocale(LC_ALL, save_thread); |
| 2749 | |
| 2750 | # endif |
| 2751 | |
| 2752 | lc = localeconv(); |
| 2753 | if ( ! lc |
| 2754 | || ! lc->currency_symbol |
| 2755 | || strEQ("", lc->currency_symbol)) |
| 2756 | { |
| 2757 | LOCALE_UNLOCK_V; |
| 2758 | return ""; |
| 2759 | } |
| 2760 | |
| 2761 | /* Leave the first spot empty to be filled in below */ |
| 2762 | retval = save_to_buffer(lc->currency_symbol, &PL_langinfo_buf, |
| 2763 | &PL_langinfo_bufsize, 1); |
| 2764 | if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, "")) |
| 2765 | { /* khw couldn't figure out how the localedef specifications |
| 2766 | would show that the $ should replace the radix; this is |
| 2767 | just a guess as to how it might work.*/ |
| 2768 | PL_langinfo_buf[0] = '.'; |
| 2769 | } |
| 2770 | else if (lc->p_cs_precedes) { |
| 2771 | PL_langinfo_buf[0] = '-'; |
| 2772 | } |
| 2773 | else { |
| 2774 | PL_langinfo_buf[0] = '+'; |
| 2775 | } |
| 2776 | |
| 2777 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2778 | |
| 2779 | my_setlocale(LC_ALL, save_global); |
| 2780 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); |
| 2781 | my_setlocale(LC_ALL, save_thread); |
| 2782 | Safefree(save_global); |
| 2783 | Safefree(save_thread); |
| 2784 | |
| 2785 | # endif |
| 2786 | |
| 2787 | LOCALE_UNLOCK_V; |
| 2788 | break; |
| 2789 | |
| 2790 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2791 | |
| 2792 | case RADIXCHAR: |
| 2793 | |
| 2794 | /* For this, we output a known simple floating point number to |
| 2795 | * a buffer, and parse it, looking for the radix */ |
| 2796 | |
| 2797 | if (toggle) { |
| 2798 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
| 2799 | } |
| 2800 | |
| 2801 | if (PL_langinfo_bufsize < 10) { |
| 2802 | PL_langinfo_bufsize = 10; |
| 2803 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); |
| 2804 | } |
| 2805 | |
| 2806 | needed_size = my_snprintf(PL_langinfo_buf, PL_langinfo_bufsize, |
| 2807 | "%.1f", 1.5); |
| 2808 | if (needed_size >= (int) PL_langinfo_bufsize) { |
| 2809 | PL_langinfo_bufsize = needed_size + 1; |
| 2810 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); |
| 2811 | needed_size = my_snprintf(PL_langinfo_buf, PL_langinfo_bufsize, |
| 2812 | "%.1f", 1.5); |
| 2813 | assert(needed_size < (int) PL_langinfo_bufsize); |
| 2814 | } |
| 2815 | |
| 2816 | ptr = PL_langinfo_buf; |
| 2817 | e = PL_langinfo_buf + PL_langinfo_bufsize; |
| 2818 | |
| 2819 | /* Find the '1' */ |
| 2820 | while (ptr < e && *ptr != '1') { |
| 2821 | ptr++; |
| 2822 | } |
| 2823 | ptr++; |
| 2824 | |
| 2825 | /* Find the '5' */ |
| 2826 | item_start = ptr; |
| 2827 | while (ptr < e && *ptr != '5') { |
| 2828 | ptr++; |
| 2829 | } |
| 2830 | |
| 2831 | /* Everything in between is the radix string */ |
| 2832 | if (ptr >= e) { |
| 2833 | PL_langinfo_buf[0] = '?'; |
| 2834 | PL_langinfo_buf[1] = '\0'; |
| 2835 | } |
| 2836 | else { |
| 2837 | *ptr = '\0'; |
| 2838 | Move(item_start, PL_langinfo_buf, ptr - PL_langinfo_buf, char); |
| 2839 | } |
| 2840 | |
| 2841 | if (toggle) { |
| 2842 | RESTORE_LC_NUMERIC(); |
| 2843 | } |
| 2844 | |
| 2845 | retval = PL_langinfo_buf; |
| 2846 | break; |
| 2847 | |
| 2848 | # else |
| 2849 | |
| 2850 | case RADIXCHAR: /* No special handling needed */ |
| 2851 | |
| 2852 | # endif |
| 2853 | |
| 2854 | case THOUSEP: |
| 2855 | |
| 2856 | if (toggle) { |
| 2857 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
| 2858 | } |
| 2859 | |
| 2860 | LOCALE_LOCK_V; /* Prevent interference with other threads |
| 2861 | using localeconv() */ |
| 2862 | |
| 2863 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2864 | |
| 2865 | /* This should only be for the thousands separator. A |
| 2866 | * different work around would be to use GetNumberFormat on a |
| 2867 | * known value and parse the result to find the separator */ |
| 2868 | save_thread = savepv(my_setlocale(LC_ALL, NULL)); |
| 2869 | _configthreadlocale(_DISABLE_PER_THREAD_LOCALE); |
| 2870 | save_global = savepv(my_setlocale(LC_ALL, NULL)); |
| 2871 | my_setlocale(LC_ALL, save_thread); |
| 2872 | # if 0 |
| 2873 | /* This is the start of code that for broken Windows replaces |
| 2874 | * the above and below code, and instead calls |
| 2875 | * GetNumberFormat() and then would parse that to find the |
| 2876 | * thousands separator. It needs to handle UTF-16 vs -8 |
| 2877 | * issues. */ |
| 2878 | |
| 2879 | needed_size = GetNumberFormatEx(PL_numeric_name, 0, "1234.5", NULL, PL_langinfo_buf, PL_langinfo_bufsize); |
| 2880 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 2881 | "%s: %d: return from GetNumber, count=%d, val=%s\n", |
| 2882 | __FILE__, __LINE__, needed_size, PL_langinfo_buf)); |
| 2883 | |
| 2884 | # endif |
| 2885 | # endif |
| 2886 | |
| 2887 | lc = localeconv(); |
| 2888 | if (! lc) { |
| 2889 | temp = ""; |
| 2890 | } |
| 2891 | else { |
| 2892 | temp = (item == RADIXCHAR) |
| 2893 | ? lc->decimal_point |
| 2894 | : lc->thousands_sep; |
| 2895 | if (! temp) { |
| 2896 | temp = ""; |
| 2897 | } |
| 2898 | } |
| 2899 | |
| 2900 | retval = save_to_buffer(temp, &PL_langinfo_buf, |
| 2901 | &PL_langinfo_bufsize, 0); |
| 2902 | |
| 2903 | # ifdef TS_W32_BROKEN_LOCALECONV |
| 2904 | |
| 2905 | my_setlocale(LC_ALL, save_global); |
| 2906 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); |
| 2907 | my_setlocale(LC_ALL, save_thread); |
| 2908 | Safefree(save_global); |
| 2909 | Safefree(save_thread); |
| 2910 | |
| 2911 | # endif |
| 2912 | |
| 2913 | LOCALE_UNLOCK_V; |
| 2914 | |
| 2915 | if (toggle) { |
| 2916 | RESTORE_LC_NUMERIC(); |
| 2917 | } |
| 2918 | |
| 2919 | break; |
| 2920 | |
| 2921 | # endif |
| 2922 | # ifdef HAS_STRFTIME |
| 2923 | |
| 2924 | /* These are defined by C89, so we assume that strftime supports |
| 2925 | * them, and so are returned unconditionally; they may not be what |
| 2926 | * the locale actually says, but should give good enough results |
| 2927 | * for someone using them as formats (as opposed to trying to parse |
| 2928 | * them to figure out what the locale says). The other format |
| 2929 | * items are actually tested to verify they work on the platform */ |
| 2930 | case D_FMT: return "%x"; |
| 2931 | case T_FMT: return "%X"; |
| 2932 | case D_T_FMT: return "%c"; |
| 2933 | |
| 2934 | /* These formats are only available in later strfmtime's */ |
| 2935 | case ERA_D_FMT: case ERA_T_FMT: case ERA_D_T_FMT: case T_FMT_AMPM: |
| 2936 | |
| 2937 | /* The rest can be gotten from most versions of strftime(). */ |
| 2938 | case ABDAY_1: case ABDAY_2: case ABDAY_3: |
| 2939 | case ABDAY_4: case ABDAY_5: case ABDAY_6: case ABDAY_7: |
| 2940 | case ALT_DIGITS: |
| 2941 | case AM_STR: case PM_STR: |
| 2942 | case ABMON_1: case ABMON_2: case ABMON_3: case ABMON_4: |
| 2943 | case ABMON_5: case ABMON_6: case ABMON_7: case ABMON_8: |
| 2944 | case ABMON_9: case ABMON_10: case ABMON_11: case ABMON_12: |
| 2945 | case DAY_1: case DAY_2: case DAY_3: case DAY_4: |
| 2946 | case DAY_5: case DAY_6: case DAY_7: |
| 2947 | case MON_1: case MON_2: case MON_3: case MON_4: |
| 2948 | case MON_5: case MON_6: case MON_7: case MON_8: |
| 2949 | case MON_9: case MON_10: case MON_11: case MON_12: |
| 2950 | |
| 2951 | LOCALE_LOCK; |
| 2952 | |
| 2953 | init_tm(&tm); /* Precaution against core dumps */ |
| 2954 | tm.tm_sec = 30; |
| 2955 | tm.tm_min = 30; |
| 2956 | tm.tm_hour = 6; |
| 2957 | tm.tm_year = 2017 - 1900; |
| 2958 | tm.tm_wday = 0; |
| 2959 | tm.tm_mon = 0; |
| 2960 | switch (item) { |
| 2961 | default: |
| 2962 | LOCALE_UNLOCK; |
| 2963 | Perl_croak(aTHX_ |
| 2964 | "panic: %s: %d: switch case: %d problem", |
| 2965 | __FILE__, __LINE__, item); |
| 2966 | NOT_REACHED; /* NOTREACHED */ |
| 2967 | |
| 2968 | case PM_STR: tm.tm_hour = 18; |
| 2969 | case AM_STR: |
| 2970 | format = "%p"; |
| 2971 | break; |
| 2972 | |
| 2973 | case ABDAY_7: tm.tm_wday++; |
| 2974 | case ABDAY_6: tm.tm_wday++; |
| 2975 | case ABDAY_5: tm.tm_wday++; |
| 2976 | case ABDAY_4: tm.tm_wday++; |
| 2977 | case ABDAY_3: tm.tm_wday++; |
| 2978 | case ABDAY_2: tm.tm_wday++; |
| 2979 | case ABDAY_1: |
| 2980 | format = "%a"; |
| 2981 | break; |
| 2982 | |
| 2983 | case DAY_7: tm.tm_wday++; |
| 2984 | case DAY_6: tm.tm_wday++; |
| 2985 | case DAY_5: tm.tm_wday++; |
| 2986 | case DAY_4: tm.tm_wday++; |
| 2987 | case DAY_3: tm.tm_wday++; |
| 2988 | case DAY_2: tm.tm_wday++; |
| 2989 | case DAY_1: |
| 2990 | format = "%A"; |
| 2991 | break; |
| 2992 | |
| 2993 | case ABMON_12: tm.tm_mon++; |
| 2994 | case ABMON_11: tm.tm_mon++; |
| 2995 | case ABMON_10: tm.tm_mon++; |
| 2996 | case ABMON_9: tm.tm_mon++; |
| 2997 | case ABMON_8: tm.tm_mon++; |
| 2998 | case ABMON_7: tm.tm_mon++; |
| 2999 | case ABMON_6: tm.tm_mon++; |
| 3000 | case ABMON_5: tm.tm_mon++; |
| 3001 | case ABMON_4: tm.tm_mon++; |
| 3002 | case ABMON_3: tm.tm_mon++; |
| 3003 | case ABMON_2: tm.tm_mon++; |
| 3004 | case ABMON_1: |
| 3005 | format = "%b"; |
| 3006 | break; |
| 3007 | |
| 3008 | case MON_12: tm.tm_mon++; |
| 3009 | case MON_11: tm.tm_mon++; |
| 3010 | case MON_10: tm.tm_mon++; |
| 3011 | case MON_9: tm.tm_mon++; |
| 3012 | case MON_8: tm.tm_mon++; |
| 3013 | case MON_7: tm.tm_mon++; |
| 3014 | case MON_6: tm.tm_mon++; |
| 3015 | case MON_5: tm.tm_mon++; |
| 3016 | case MON_4: tm.tm_mon++; |
| 3017 | case MON_3: tm.tm_mon++; |
| 3018 | case MON_2: tm.tm_mon++; |
| 3019 | case MON_1: |
| 3020 | format = "%B"; |
| 3021 | break; |
| 3022 | |
| 3023 | case T_FMT_AMPM: |
| 3024 | format = "%r"; |
| 3025 | return_format = TRUE; |
| 3026 | break; |
| 3027 | |
| 3028 | case ERA_D_FMT: |
| 3029 | format = "%Ex"; |
| 3030 | return_format = TRUE; |
| 3031 | break; |
| 3032 | |
| 3033 | case ERA_T_FMT: |
| 3034 | format = "%EX"; |
| 3035 | return_format = TRUE; |
| 3036 | break; |
| 3037 | |
| 3038 | case ERA_D_T_FMT: |
| 3039 | format = "%Ec"; |
| 3040 | return_format = TRUE; |
| 3041 | break; |
| 3042 | |
| 3043 | case ALT_DIGITS: |
| 3044 | tm.tm_wday = 0; |
| 3045 | format = "%Ow"; /* Find the alternate digit for 0 */ |
| 3046 | break; |
| 3047 | } |
| 3048 | |
| 3049 | /* We can't use my_strftime() because it doesn't look at |
| 3050 | * tm_wday */ |
| 3051 | while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize, |
| 3052 | format, &tm)) |
| 3053 | { |
| 3054 | /* A zero return means one of: |
| 3055 | * a) there wasn't enough space in PL_langinfo_buf |
| 3056 | * b) the format, like a plain %p, returns empty |
| 3057 | * c) it was an illegal format, though some |
| 3058 | * implementations of strftime will just return the |
| 3059 | * illegal format as a plain character sequence. |
| 3060 | * |
| 3061 | * To quickly test for case 'b)', try again but precede |
| 3062 | * the format with a plain character. If that result is |
| 3063 | * still empty, the problem is either 'a)' or 'c)' */ |
| 3064 | |
| 3065 | Size_t format_size = strlen(format) + 1; |
| 3066 | Size_t mod_size = format_size + 1; |
| 3067 | char * mod_format; |
| 3068 | char * temp_result; |
| 3069 | |
| 3070 | Newx(mod_format, mod_size, char); |
| 3071 | Newx(temp_result, PL_langinfo_bufsize, char); |
| 3072 | *mod_format = ' '; |
| 3073 | my_strlcpy(mod_format + 1, format, mod_size); |
| 3074 | len = strftime(temp_result, |
| 3075 | PL_langinfo_bufsize, |
| 3076 | mod_format, &tm); |
| 3077 | Safefree(mod_format); |
| 3078 | Safefree(temp_result); |
| 3079 | |
| 3080 | /* If 'len' is non-zero, it means that we had a case like |
| 3081 | * %p which means the current locale doesn't use a.m. or |
| 3082 | * p.m., and that is valid */ |
| 3083 | if (len == 0) { |
| 3084 | |
| 3085 | /* Here, still didn't work. If we get well beyond a |
| 3086 | * reasonable size, bail out to prevent an infinite |
| 3087 | * loop. */ |
| 3088 | |
| 3089 | if (PL_langinfo_bufsize > 100 * format_size) { |
| 3090 | *PL_langinfo_buf = '\0'; |
| 3091 | } |
| 3092 | else { |
| 3093 | /* Double the buffer size to retry; Add 1 in case |
| 3094 | * original was 0, so we aren't stuck at 0. */ |
| 3095 | PL_langinfo_bufsize *= 2; |
| 3096 | PL_langinfo_bufsize++; |
| 3097 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); |
| 3098 | continue; |
| 3099 | } |
| 3100 | } |
| 3101 | |
| 3102 | break; |
| 3103 | } |
| 3104 | |
| 3105 | /* Here, we got a result. |
| 3106 | * |
| 3107 | * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the |
| 3108 | * alternate format for wday 0. If the value is the same as |
| 3109 | * the normal 0, there isn't an alternate, so clear the buffer. |
| 3110 | * */ |
| 3111 | if ( item == ALT_DIGITS |
| 3112 | && strEQ(PL_langinfo_buf, "0")) |
| 3113 | { |
| 3114 | *PL_langinfo_buf = '\0'; |
| 3115 | } |
| 3116 | |
| 3117 | /* ALT_DIGITS is problematic. Experiments on it showed that |
| 3118 | * strftime() did not always work properly when going from |
| 3119 | * alt-9 to alt-10. Only a few locales have this item defined, |
| 3120 | * and in all of them on Linux that khw was able to find, |
| 3121 | * nl_langinfo() merely returned the alt-0 character, possibly |
| 3122 | * doubled. Most Unicode digits are in blocks of 10 |
| 3123 | * consecutive code points, so that is sufficient information |
| 3124 | * for those scripts, as we can infer alt-1, alt-2, .... But |
| 3125 | * for a Japanese locale, a CJK ideographic 0 is returned, and |
| 3126 | * the CJK digits are not in code point order, so you can't |
| 3127 | * really infer anything. The localedef for this locale did |
| 3128 | * specify the succeeding digits, so that strftime() works |
| 3129 | * properly on them, without needing to infer anything. But |
| 3130 | * the nl_langinfo() return did not give sufficient information |
| 3131 | * for the caller to understand what's going on. So until |
| 3132 | * there is evidence that it should work differently, this |
| 3133 | * returns the alt-0 string for ALT_DIGITS. |
| 3134 | * |
| 3135 | * wday was chosen because its range is all a single digit. |
| 3136 | * Things like tm_sec have two digits as the minimum: '00' */ |
| 3137 | |
| 3138 | LOCALE_UNLOCK; |
| 3139 | |
| 3140 | retval = PL_langinfo_buf; |
| 3141 | |
| 3142 | /* If to return the format, not the value, overwrite the buffer |
| 3143 | * with it. But some strftime()s will keep the original format |
| 3144 | * if illegal, so change those to "" */ |
| 3145 | if (return_format) { |
| 3146 | if (strEQ(PL_langinfo_buf, format)) { |
| 3147 | *PL_langinfo_buf = '\0'; |
| 3148 | } |
| 3149 | else { |
| 3150 | retval = save_to_buffer(format, &PL_langinfo_buf, |
| 3151 | &PL_langinfo_bufsize, 0); |
| 3152 | } |
| 3153 | } |
| 3154 | |
| 3155 | break; |
| 3156 | |
| 3157 | # endif |
| 3158 | |
| 3159 | } |
| 3160 | } |
| 3161 | |
| 3162 | return retval; |
| 3163 | |
| 3164 | #endif |
| 3165 | |
| 3166 | } |
| 3167 | |
| 3168 | /* |
| 3169 | * Initialize locale awareness. |
| 3170 | */ |
| 3171 | int |
| 3172 | Perl_init_i18nl10n(pTHX_ int printwarn) |
| 3173 | { |
| 3174 | /* printwarn is |
| 3175 | * |
| 3176 | * 0 if not to output warning when setup locale is bad |
| 3177 | * 1 if to output warning based on value of PERL_BADLANG |
| 3178 | * >1 if to output regardless of PERL_BADLANG |
| 3179 | * |
| 3180 | * returns |
| 3181 | * 1 = set ok or not applicable, |
| 3182 | * 0 = fallback to a locale of lower priority |
| 3183 | * -1 = fallback to all locales failed, not even to the C locale |
| 3184 | * |
| 3185 | * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is |
| 3186 | * set, debugging information is output. |
| 3187 | * |
| 3188 | * This looks more complicated than it is, mainly due to the #ifdefs. |
| 3189 | * |
| 3190 | * We try to set LC_ALL to the value determined by the environment. If |
| 3191 | * there is no LC_ALL on this platform, we try the individual categories we |
| 3192 | * know about. If this works, we are done. |
| 3193 | * |
| 3194 | * But if it doesn't work, we have to do something else. We search the |
| 3195 | * environment variables ourselves instead of relying on the system to do |
| 3196 | * it. We look at, in order, LC_ALL, LANG, a system default locale (if we |
| 3197 | * think there is one), and the ultimate fallback "C". This is all done in |
| 3198 | * the same loop as above to avoid duplicating code, but it makes things |
| 3199 | * more complex. The 'trial_locales' array is initialized with just one |
| 3200 | * element; it causes the behavior described in the paragraph above this to |
| 3201 | * happen. If that fails, we add elements to 'trial_locales', and do extra |
| 3202 | * loop iterations to cause the behavior described in this paragraph. |
| 3203 | * |
| 3204 | * On Ultrix, the locale MUST come from the environment, so there is |
| 3205 | * preliminary code to set it. I (khw) am not sure that it is necessary, |
| 3206 | * and that this couldn't be folded into the loop, but barring any real |
| 3207 | * platforms to test on, it's staying as-is |
| 3208 | * |
| 3209 | * A slight complication is that in embedded Perls, the locale may already |
| 3210 | * be set-up, and we don't want to get it from the normal environment |
| 3211 | * variables. This is handled by having a special environment variable |
| 3212 | * indicate we're in this situation. We simply set setlocale's 2nd |
| 3213 | * parameter to be a NULL instead of "". That indicates to setlocale that |
| 3214 | * it is not to change anything, but to return the current value, |
| 3215 | * effectively initializing perl's db to what the locale already is. |
| 3216 | * |
| 3217 | * We play the same trick with NULL if a LC_ALL succeeds. We call |
| 3218 | * setlocale() on the individual categores with NULL to get their existing |
| 3219 | * values for our db, instead of trying to change them. |
| 3220 | * */ |
| 3221 | |
| 3222 | dVAR; |
| 3223 | |
| 3224 | int ok = 1; |
| 3225 | |
| 3226 | #ifndef USE_LOCALE |
| 3227 | |
| 3228 | PERL_UNUSED_ARG(printwarn); |
| 3229 | |
| 3230 | #else /* USE_LOCALE */ |
| 3231 | # ifdef __GLIBC__ |
| 3232 | |
| 3233 | const char * const language = savepv(PerlEnv_getenv("LANGUAGE")); |
| 3234 | |
| 3235 | # endif |
| 3236 | |
| 3237 | /* NULL uses the existing already set up locale */ |
| 3238 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) |
| 3239 | ? NULL |
| 3240 | : ""; |
| 3241 | const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */ |
| 3242 | unsigned int trial_locales_count; |
| 3243 | const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL")); |
| 3244 | const char * const lang = savepv(PerlEnv_getenv("LANG")); |
| 3245 | bool setlocale_failure = FALSE; |
| 3246 | unsigned int i; |
| 3247 | |
| 3248 | /* A later getenv() could zap this, so only use here */ |
| 3249 | const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG"); |
| 3250 | |
| 3251 | const bool locwarn = (printwarn > 1 |
| 3252 | || ( printwarn |
| 3253 | && ( ! bad_lang_use_once |
| 3254 | || ( |
| 3255 | /* disallow with "" or "0" */ |
| 3256 | *bad_lang_use_once |
| 3257 | && strNE("0", bad_lang_use_once))))); |
| 3258 | |
| 3259 | /* setlocale() return vals; not copied so must be looked at immediately */ |
| 3260 | const char * sl_result[NOMINAL_LC_ALL_INDEX + 1]; |
| 3261 | |
| 3262 | /* current locale for given category; should have been copied so aren't |
| 3263 | * volatile */ |
| 3264 | const char * curlocales[NOMINAL_LC_ALL_INDEX + 1]; |
| 3265 | |
| 3266 | # ifdef WIN32 |
| 3267 | |
| 3268 | /* In some systems you can find out the system default locale |
| 3269 | * and use that as the fallback locale. */ |
| 3270 | # define SYSTEM_DEFAULT_LOCALE |
| 3271 | # endif |
| 3272 | # ifdef SYSTEM_DEFAULT_LOCALE |
| 3273 | |
| 3274 | const char *system_default_locale = NULL; |
| 3275 | |
| 3276 | # endif |
| 3277 | |
| 3278 | # ifndef DEBUGGING |
| 3279 | # define DEBUG_LOCALE_INIT(a,b,c) |
| 3280 | # else |
| 3281 | |
| 3282 | DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))); |
| 3283 | |
| 3284 | # define DEBUG_LOCALE_INIT(category, locale, result) \ |
| 3285 | STMT_START { \ |
| 3286 | if (debug_initialization) { \ |
| 3287 | PerlIO_printf(Perl_debug_log, \ |
| 3288 | "%s:%d: %s\n", \ |
| 3289 | __FILE__, __LINE__, \ |
| 3290 | setlocale_debug_string(category, \ |
| 3291 | locale, \ |
| 3292 | result)); \ |
| 3293 | } \ |
| 3294 | } STMT_END |
| 3295 | |
| 3296 | /* Make sure the parallel arrays are properly set up */ |
| 3297 | # ifdef USE_LOCALE_NUMERIC |
| 3298 | assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC); |
| 3299 | assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC")); |
| 3300 | # ifdef USE_POSIX_2008_LOCALE |
| 3301 | assert(category_masks[LC_NUMERIC_INDEX] == LC_NUMERIC_MASK); |
| 3302 | # endif |
| 3303 | # endif |
| 3304 | # ifdef USE_LOCALE_CTYPE |
| 3305 | assert(categories[LC_CTYPE_INDEX] == LC_CTYPE); |
| 3306 | assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE")); |
| 3307 | # ifdef USE_POSIX_2008_LOCALE |
| 3308 | assert(category_masks[LC_CTYPE_INDEX] == LC_CTYPE_MASK); |
| 3309 | # endif |
| 3310 | # endif |
| 3311 | # ifdef USE_LOCALE_COLLATE |
| 3312 | assert(categories[LC_COLLATE_INDEX] == LC_COLLATE); |
| 3313 | assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE")); |
| 3314 | # ifdef USE_POSIX_2008_LOCALE |
| 3315 | assert(category_masks[LC_COLLATE_INDEX] == LC_COLLATE_MASK); |
| 3316 | # endif |
| 3317 | # endif |
| 3318 | # ifdef USE_LOCALE_TIME |
| 3319 | assert(categories[LC_TIME_INDEX] == LC_TIME); |
| 3320 | assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME")); |
| 3321 | # ifdef USE_POSIX_2008_LOCALE |
| 3322 | assert(category_masks[LC_TIME_INDEX] == LC_TIME_MASK); |
| 3323 | # endif |
| 3324 | # endif |
| 3325 | # ifdef USE_LOCALE_MESSAGES |
| 3326 | assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES); |
| 3327 | assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES")); |
| 3328 | # ifdef USE_POSIX_2008_LOCALE |
| 3329 | assert(category_masks[LC_MESSAGES_INDEX] == LC_MESSAGES_MASK); |
| 3330 | # endif |
| 3331 | # endif |
| 3332 | # ifdef USE_LOCALE_MONETARY |
| 3333 | assert(categories[LC_MONETARY_INDEX] == LC_MONETARY); |
| 3334 | assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY")); |
| 3335 | # ifdef USE_POSIX_2008_LOCALE |
| 3336 | assert(category_masks[LC_MONETARY_INDEX] == LC_MONETARY_MASK); |
| 3337 | # endif |
| 3338 | # endif |
| 3339 | # ifdef USE_LOCALE_ADDRESS |
| 3340 | assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS); |
| 3341 | assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS")); |
| 3342 | # ifdef USE_POSIX_2008_LOCALE |
| 3343 | assert(category_masks[LC_ADDRESS_INDEX] == LC_ADDRESS_MASK); |
| 3344 | # endif |
| 3345 | # endif |
| 3346 | # ifdef USE_LOCALE_IDENTIFICATION |
| 3347 | assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION); |
| 3348 | assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION")); |
| 3349 | # ifdef USE_POSIX_2008_LOCALE |
| 3350 | assert(category_masks[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION_MASK); |
| 3351 | # endif |
| 3352 | # endif |
| 3353 | # ifdef USE_LOCALE_MEASUREMENT |
| 3354 | assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT); |
| 3355 | assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT")); |
| 3356 | # ifdef USE_POSIX_2008_LOCALE |
| 3357 | assert(category_masks[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT_MASK); |
| 3358 | # endif |
| 3359 | # endif |
| 3360 | # ifdef USE_LOCALE_PAPER |
| 3361 | assert(categories[LC_PAPER_INDEX] == LC_PAPER); |
| 3362 | assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER")); |
| 3363 | # ifdef USE_POSIX_2008_LOCALE |
| 3364 | assert(category_masks[LC_PAPER_INDEX] == LC_PAPER_MASK); |
| 3365 | # endif |
| 3366 | # endif |
| 3367 | # ifdef USE_LOCALE_TELEPHONE |
| 3368 | assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE); |
| 3369 | assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE")); |
| 3370 | # ifdef USE_POSIX_2008_LOCALE |
| 3371 | assert(category_masks[LC_TELEPHONE_INDEX] == LC_TELEPHONE_MASK); |
| 3372 | # endif |
| 3373 | # endif |
| 3374 | # ifdef LC_ALL |
| 3375 | assert(categories[LC_ALL_INDEX] == LC_ALL); |
| 3376 | assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL")); |
| 3377 | assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX); |
| 3378 | # ifdef USE_POSIX_2008_LOCALE |
| 3379 | assert(category_masks[LC_ALL_INDEX] == LC_ALL_MASK); |
| 3380 | # endif |
| 3381 | # endif |
| 3382 | # endif /* DEBUGGING */ |
| 3383 | |
| 3384 | /* Initialize the cache of the program's UTF-8ness for the always known |
| 3385 | * locales C and POSIX */ |
| 3386 | my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness, |
| 3387 | sizeof(PL_locale_utf8ness)); |
| 3388 | |
| 3389 | # ifdef USE_THREAD_SAFE_LOCALE |
| 3390 | # ifdef WIN32 |
| 3391 | |
| 3392 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); |
| 3393 | |
| 3394 | # endif |
| 3395 | # endif |
| 3396 | # ifdef USE_POSIX_2008_LOCALE |
| 3397 | |
| 3398 | PL_C_locale_obj = newlocale(LC_ALL_MASK, "C", (locale_t) 0); |
| 3399 | if (! PL_C_locale_obj) { |
| 3400 | Perl_croak_nocontext( |
| 3401 | "panic: Cannot create POSIX 2008 C locale object; errno=%d", errno); |
| 3402 | } |
| 3403 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 3404 | PerlIO_printf(Perl_debug_log, "%s:%d: created C object %p\n", __FILE__, __LINE__, PL_C_locale_obj); |
| 3405 | } |
| 3406 | |
| 3407 | # endif |
| 3408 | |
| 3409 | # ifdef USE_LOCALE_NUMERIC |
| 3410 | |
| 3411 | PL_numeric_radix_sv = newSVpvs("."); |
| 3412 | |
| 3413 | # endif |
| 3414 | |
| 3415 | # if defined(USE_POSIX_2008_LOCALE) && ! defined(HAS_QUERYLOCALE) |
| 3416 | |
| 3417 | /* Initialize our records. If we have POSIX 2008, we have LC_ALL */ |
| 3418 | do_setlocale_c(LC_ALL, my_setlocale(LC_ALL, NULL)); |
| 3419 | |
| 3420 | # endif |
| 3421 | # ifdef LOCALE_ENVIRON_REQUIRED |
| 3422 | |
| 3423 | /* |
| 3424 | * Ultrix setlocale(..., "") fails if there are no environment |
| 3425 | * variables from which to get a locale name. |
| 3426 | */ |
| 3427 | |
| 3428 | # ifndef LC_ALL |
| 3429 | # error Ultrix without LC_ALL not implemented |
| 3430 | # else |
| 3431 | |
| 3432 | { |
| 3433 | bool done = FALSE; |
| 3434 | if (lang) { |
| 3435 | sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init); |
| 3436 | DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]); |
| 3437 | if (sl_result[LC_ALL_INDEX]) |
| 3438 | done = TRUE; |
| 3439 | else |
| 3440 | setlocale_failure = TRUE; |
| 3441 | } |
| 3442 | if (! setlocale_failure) { |
| 3443 | const char * locale_param; |
| 3444 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 3445 | locale_param = (! done && (lang || PerlEnv_getenv(category_names[i]))) |
| 3446 | ? setlocale_init |
| 3447 | : NULL; |
| 3448 | sl_result[i] = do_setlocale_r(categories[i], locale_param); |
| 3449 | if (! sl_result[i]) { |
| 3450 | setlocale_failure = TRUE; |
| 3451 | } |
| 3452 | DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]); |
| 3453 | } |
| 3454 | } |
| 3455 | } |
| 3456 | |
| 3457 | # endif /* LC_ALL */ |
| 3458 | # endif /* LOCALE_ENVIRON_REQUIRED */ |
| 3459 | |
| 3460 | /* We try each locale in the list until we get one that works, or exhaust |
| 3461 | * the list. Normally the loop is executed just once. But if setting the |
| 3462 | * locale fails, inside the loop we add fallback trials to the array and so |
| 3463 | * will execute the loop multiple times */ |
| 3464 | trial_locales[0] = setlocale_init; |
| 3465 | trial_locales_count = 1; |
| 3466 | |
| 3467 | for (i= 0; i < trial_locales_count; i++) { |
| 3468 | const char * trial_locale = trial_locales[i]; |
| 3469 | |
| 3470 | if (i > 0) { |
| 3471 | |
| 3472 | /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED |
| 3473 | * when i==0, but I (khw) don't think that behavior makes much |
| 3474 | * sense */ |
| 3475 | setlocale_failure = FALSE; |
| 3476 | |
| 3477 | # ifdef SYSTEM_DEFAULT_LOCALE |
| 3478 | # ifdef WIN32 /* Note that assumes Win32 has LC_ALL */ |
| 3479 | |
| 3480 | /* On Windows machines, an entry of "" after the 0th means to use |
| 3481 | * the system default locale, which we now proceed to get. */ |
| 3482 | if (strEQ(trial_locale, "")) { |
| 3483 | unsigned int j; |
| 3484 | |
| 3485 | /* Note that this may change the locale, but we are going to do |
| 3486 | * that anyway just below */ |
| 3487 | system_default_locale = do_setlocale_c(LC_ALL, ""); |
| 3488 | DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale); |
| 3489 | |
| 3490 | /* Skip if invalid or if it's already on the list of locales to |
| 3491 | * try */ |
| 3492 | if (! system_default_locale) { |
| 3493 | goto next_iteration; |
| 3494 | } |
| 3495 | for (j = 0; j < trial_locales_count; j++) { |
| 3496 | if (strEQ(system_default_locale, trial_locales[j])) { |
| 3497 | goto next_iteration; |
| 3498 | } |
| 3499 | } |
| 3500 | |
| 3501 | trial_locale = system_default_locale; |
| 3502 | } |
| 3503 | # else |
| 3504 | # error SYSTEM_DEFAULT_LOCALE only implemented for Win32 |
| 3505 | # endif |
| 3506 | # endif /* SYSTEM_DEFAULT_LOCALE */ |
| 3507 | |
| 3508 | } /* For i > 0 */ |
| 3509 | |
| 3510 | # ifdef LC_ALL |
| 3511 | |
| 3512 | sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale); |
| 3513 | DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]); |
| 3514 | if (! sl_result[LC_ALL_INDEX]) { |
| 3515 | setlocale_failure = TRUE; |
| 3516 | } |
| 3517 | else { |
| 3518 | /* Since LC_ALL succeeded, it should have changed all the other |
| 3519 | * categories it can to its value; so we massage things so that the |
| 3520 | * setlocales below just return their category's current values. |
| 3521 | * This adequately handles the case in NetBSD where LC_COLLATE may |
| 3522 | * not be defined for a locale, and setting it individually will |
| 3523 | * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to |
| 3524 | * the POSIX locale. */ |
| 3525 | trial_locale = NULL; |
| 3526 | } |
| 3527 | |
| 3528 | # endif /* LC_ALL */ |
| 3529 | |
| 3530 | if (! setlocale_failure) { |
| 3531 | unsigned int j; |
| 3532 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { |
| 3533 | curlocales[j] |
| 3534 | = savepv(do_setlocale_r(categories[j], trial_locale)); |
| 3535 | if (! curlocales[j]) { |
| 3536 | setlocale_failure = TRUE; |
| 3537 | } |
| 3538 | DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]); |
| 3539 | } |
| 3540 | |
| 3541 | if (! setlocale_failure) { /* All succeeded */ |
| 3542 | break; /* Exit trial_locales loop */ |
| 3543 | } |
| 3544 | } |
| 3545 | |
| 3546 | /* Here, something failed; will need to try a fallback. */ |
| 3547 | ok = 0; |
| 3548 | |
| 3549 | if (i == 0) { |
| 3550 | unsigned int j; |
| 3551 | |
| 3552 | if (locwarn) { /* Output failure info only on the first one */ |
| 3553 | |
| 3554 | # ifdef LC_ALL |
| 3555 | |
| 3556 | PerlIO_printf(Perl_error_log, |
| 3557 | "perl: warning: Setting locale failed.\n"); |
| 3558 | |
| 3559 | # else /* !LC_ALL */ |
| 3560 | |
| 3561 | PerlIO_printf(Perl_error_log, |
| 3562 | "perl: warning: Setting locale failed for the categories:\n\t"); |
| 3563 | |
| 3564 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { |
| 3565 | if (! curlocales[j]) { |
| 3566 | PerlIO_printf(Perl_error_log, category_names[j]); |
| 3567 | } |
| 3568 | else { |
| 3569 | Safefree(curlocales[j]); |
| 3570 | } |
| 3571 | } |
| 3572 | |
| 3573 | # endif /* LC_ALL */ |
| 3574 | |
| 3575 | PerlIO_printf(Perl_error_log, |
| 3576 | "perl: warning: Please check that your locale settings:\n"); |
| 3577 | |
| 3578 | # ifdef __GLIBC__ |
| 3579 | |
| 3580 | PerlIO_printf(Perl_error_log, |
| 3581 | "\tLANGUAGE = %c%s%c,\n", |
| 3582 | language ? '"' : '(', |
| 3583 | language ? language : "unset", |
| 3584 | language ? '"' : ')'); |
| 3585 | # endif |
| 3586 | |
| 3587 | PerlIO_printf(Perl_error_log, |
| 3588 | "\tLC_ALL = %c%s%c,\n", |
| 3589 | lc_all ? '"' : '(', |
| 3590 | lc_all ? lc_all : "unset", |
| 3591 | lc_all ? '"' : ')'); |
| 3592 | |
| 3593 | # if defined(USE_ENVIRON_ARRAY) |
| 3594 | |
| 3595 | { |
| 3596 | char **e; |
| 3597 | |
| 3598 | /* Look through the environment for any variables of the |
| 3599 | * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was |
| 3600 | * already handled above. These are assumed to be locale |
| 3601 | * settings. Output them and their values. */ |
| 3602 | for (e = environ; *e; e++) { |
| 3603 | const STRLEN prefix_len = sizeof("LC_") - 1; |
| 3604 | STRLEN uppers_len; |
| 3605 | |
| 3606 | if ( strBEGINs(*e, "LC_") |
| 3607 | && ! strBEGINs(*e, "LC_ALL=") |
| 3608 | && (uppers_len = strspn(*e + prefix_len, |
| 3609 | "ABCDEFGHIJKLMNOPQRSTUVWXYZ")) |
| 3610 | && ((*e)[prefix_len + uppers_len] == '=')) |
| 3611 | { |
| 3612 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", |
| 3613 | (int) (prefix_len + uppers_len), *e, |
| 3614 | *e + prefix_len + uppers_len + 1); |
| 3615 | } |
| 3616 | } |
| 3617 | } |
| 3618 | |
| 3619 | # else |
| 3620 | |
| 3621 | PerlIO_printf(Perl_error_log, |
| 3622 | "\t(possibly more locale environment variables)\n"); |
| 3623 | |
| 3624 | # endif |
| 3625 | |
| 3626 | PerlIO_printf(Perl_error_log, |
| 3627 | "\tLANG = %c%s%c\n", |
| 3628 | lang ? '"' : '(', |
| 3629 | lang ? lang : "unset", |
| 3630 | lang ? '"' : ')'); |
| 3631 | |
| 3632 | PerlIO_printf(Perl_error_log, |
| 3633 | " are supported and installed on your system.\n"); |
| 3634 | } |
| 3635 | |
| 3636 | /* Calculate what fallback locales to try. We have avoided this |
| 3637 | * until we have to, because failure is quite unlikely. This will |
| 3638 | * usually change the upper bound of the loop we are in. |
| 3639 | * |
| 3640 | * Since the system's default way of setting the locale has not |
| 3641 | * found one that works, We use Perl's defined ordering: LC_ALL, |
| 3642 | * LANG, and the C locale. We don't try the same locale twice, so |
| 3643 | * don't add to the list if already there. (On POSIX systems, the |
| 3644 | * LC_ALL element will likely be a repeat of the 0th element "", |
| 3645 | * but there's no harm done by doing it explicitly. |
| 3646 | * |
| 3647 | * Note that this tries the LC_ALL environment variable even on |
| 3648 | * systems which have no LC_ALL locale setting. This may or may |
| 3649 | * not have been originally intentional, but there's no real need |
| 3650 | * to change the behavior. */ |
| 3651 | if (lc_all) { |
| 3652 | for (j = 0; j < trial_locales_count; j++) { |
| 3653 | if (strEQ(lc_all, trial_locales[j])) { |
| 3654 | goto done_lc_all; |
| 3655 | } |
| 3656 | } |
| 3657 | trial_locales[trial_locales_count++] = lc_all; |
| 3658 | } |
| 3659 | done_lc_all: |
| 3660 | |
| 3661 | if (lang) { |
| 3662 | for (j = 0; j < trial_locales_count; j++) { |
| 3663 | if (strEQ(lang, trial_locales[j])) { |
| 3664 | goto done_lang; |
| 3665 | } |
| 3666 | } |
| 3667 | trial_locales[trial_locales_count++] = lang; |
| 3668 | } |
| 3669 | done_lang: |
| 3670 | |
| 3671 | # if defined(WIN32) && defined(LC_ALL) |
| 3672 | |
| 3673 | /* For Windows, we also try the system default locale before "C". |
| 3674 | * (If there exists a Windows without LC_ALL we skip this because |
| 3675 | * it gets too complicated. For those, the "C" is the next |
| 3676 | * fallback possibility). The "" is the same as the 0th element of |
| 3677 | * the array, but the code at the loop above knows to treat it |
| 3678 | * differently when not the 0th */ |
| 3679 | trial_locales[trial_locales_count++] = ""; |
| 3680 | |
| 3681 | # endif |
| 3682 | |
| 3683 | for (j = 0; j < trial_locales_count; j++) { |
| 3684 | if (strEQ("C", trial_locales[j])) { |
| 3685 | goto done_C; |
| 3686 | } |
| 3687 | } |
| 3688 | trial_locales[trial_locales_count++] = "C"; |
| 3689 | |
| 3690 | done_C: ; |
| 3691 | } /* end of first time through the loop */ |
| 3692 | |
| 3693 | # ifdef WIN32 |
| 3694 | |
| 3695 | next_iteration: ; |
| 3696 | |
| 3697 | # endif |
| 3698 | |
| 3699 | } /* end of looping through the trial locales */ |
| 3700 | |
| 3701 | if (ok < 1) { /* If we tried to fallback */ |
| 3702 | const char* msg; |
| 3703 | if (! setlocale_failure) { /* fallback succeeded */ |
| 3704 | msg = "Falling back to"; |
| 3705 | } |
| 3706 | else { /* fallback failed */ |
| 3707 | unsigned int j; |
| 3708 | |
| 3709 | /* We dropped off the end of the loop, so have to decrement i to |
| 3710 | * get back to the value the last time through */ |
| 3711 | i--; |
| 3712 | |
| 3713 | ok = -1; |
| 3714 | msg = "Failed to fall back to"; |
| 3715 | |
| 3716 | /* To continue, we should use whatever values we've got */ |
| 3717 | |
| 3718 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { |
| 3719 | Safefree(curlocales[j]); |
| 3720 | curlocales[j] = savepv(do_setlocale_r(categories[j], NULL)); |
| 3721 | DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]); |
| 3722 | } |
| 3723 | } |
| 3724 | |
| 3725 | if (locwarn) { |
| 3726 | const char * description; |
| 3727 | const char * name = ""; |
| 3728 | if (strEQ(trial_locales[i], "C")) { |
| 3729 | description = "the standard locale"; |
| 3730 | name = "C"; |
| 3731 | } |
| 3732 | |
| 3733 | # ifdef SYSTEM_DEFAULT_LOCALE |
| 3734 | |
| 3735 | else if (strEQ(trial_locales[i], "")) { |
| 3736 | description = "the system default locale"; |
| 3737 | if (system_default_locale) { |
| 3738 | name = system_default_locale; |
| 3739 | } |
| 3740 | } |
| 3741 | |
| 3742 | # endif /* SYSTEM_DEFAULT_LOCALE */ |
| 3743 | |
| 3744 | else { |
| 3745 | description = "a fallback locale"; |
| 3746 | name = trial_locales[i]; |
| 3747 | } |
| 3748 | if (name && strNE(name, "")) { |
| 3749 | PerlIO_printf(Perl_error_log, |
| 3750 | "perl: warning: %s %s (\"%s\").\n", msg, description, name); |
| 3751 | } |
| 3752 | else { |
| 3753 | PerlIO_printf(Perl_error_log, |
| 3754 | "perl: warning: %s %s.\n", msg, description); |
| 3755 | } |
| 3756 | } |
| 3757 | } /* End of tried to fallback */ |
| 3758 | |
| 3759 | /* Done with finding the locales; update our records */ |
| 3760 | |
| 3761 | # ifdef USE_LOCALE_CTYPE |
| 3762 | |
| 3763 | new_ctype(curlocales[LC_CTYPE_INDEX]); |
| 3764 | |
| 3765 | # endif |
| 3766 | # ifdef USE_LOCALE_COLLATE |
| 3767 | |
| 3768 | new_collate(curlocales[LC_COLLATE_INDEX]); |
| 3769 | |
| 3770 | # endif |
| 3771 | # ifdef USE_LOCALE_NUMERIC |
| 3772 | |
| 3773 | new_numeric(curlocales[LC_NUMERIC_INDEX]); |
| 3774 | |
| 3775 | # endif |
| 3776 | |
| 3777 | for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) { |
| 3778 | |
| 3779 | # if defined(USE_ITHREADS) && ! defined(USE_THREAD_SAFE_LOCALE) |
| 3780 | |
| 3781 | /* This caches whether each category's locale is UTF-8 or not. This |
| 3782 | * may involve changing the locale. It is ok to do this at |
| 3783 | * initialization time before any threads have started, but not later |
| 3784 | * unless thread-safe operations are used. |
| 3785 | * Caching means that if the program heeds our dictate not to change |
| 3786 | * locales in threaded applications, this data will remain valid, and |
| 3787 | * it may get queried without having to change locales. If the |
| 3788 | * environment is such that all categories have the same locale, this |
| 3789 | * isn't needed, as the code will not change the locale; but this |
| 3790 | * handles the uncommon case where the environment has disparate |
| 3791 | * locales for the categories */ |
| 3792 | (void) _is_cur_LC_category_utf8(categories[i]); |
| 3793 | |
| 3794 | # endif |
| 3795 | |
| 3796 | Safefree(curlocales[i]); |
| 3797 | } |
| 3798 | |
| 3799 | # if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE) |
| 3800 | |
| 3801 | /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE |
| 3802 | * locale is UTF-8. The call to new_ctype() just above has already |
| 3803 | * calculated the latter value and saved it in PL_in_utf8_CTYPE_locale. If |
| 3804 | * both PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE}) |
| 3805 | * are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer on |
| 3806 | * STDIN, STDOUT, STDERR, _and_ the default open discipline. */ |
| 3807 | PL_utf8locale = PL_in_utf8_CTYPE_locale; |
| 3808 | |
| 3809 | /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO. |
| 3810 | This is an alternative to using the -C command line switch |
| 3811 | (the -C if present will override this). */ |
| 3812 | { |
| 3813 | const char *p = PerlEnv_getenv("PERL_UNICODE"); |
| 3814 | PL_unicode = p ? parse_unicode_opts(&p) : 0; |
| 3815 | if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG) |
| 3816 | PL_utf8cache = -1; |
| 3817 | } |
| 3818 | |
| 3819 | # endif |
| 3820 | # ifdef __GLIBC__ |
| 3821 | |
| 3822 | Safefree(language); |
| 3823 | |
| 3824 | # endif |
| 3825 | |
| 3826 | Safefree(lc_all); |
| 3827 | Safefree(lang); |
| 3828 | |
| 3829 | #endif /* USE_LOCALE */ |
| 3830 | #ifdef DEBUGGING |
| 3831 | |
| 3832 | /* So won't continue to output stuff */ |
| 3833 | DEBUG_INITIALIZATION_set(FALSE); |
| 3834 | |
| 3835 | #endif |
| 3836 | |
| 3837 | return ok; |
| 3838 | } |
| 3839 | |
| 3840 | #ifdef USE_LOCALE_COLLATE |
| 3841 | |
| 3842 | char * |
| 3843 | Perl__mem_collxfrm(pTHX_ const char *input_string, |
| 3844 | STRLEN len, /* Length of 'input_string' */ |
| 3845 | STRLEN *xlen, /* Set to length of returned string |
| 3846 | (not including the collation index |
| 3847 | prefix) */ |
| 3848 | bool utf8 /* Is the input in UTF-8? */ |
| 3849 | ) |
| 3850 | { |
| 3851 | |
| 3852 | /* _mem_collxfrm() is a bit like strxfrm() but with two important |
| 3853 | * differences. First, it handles embedded NULs. Second, it allocates a bit |
| 3854 | * more memory than needed for the transformed data itself. The real |
| 3855 | * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to |
| 3856 | * the length of that, and doesn't include the collation index size. |
| 3857 | * Please see sv_collxfrm() to see how this is used. */ |
| 3858 | |
| 3859 | #define COLLXFRM_HDR_LEN sizeof(PL_collation_ix) |
| 3860 | |
| 3861 | char * s = (char *) input_string; |
| 3862 | STRLEN s_strlen = strlen(input_string); |
| 3863 | char *xbuf = NULL; |
| 3864 | STRLEN xAlloc; /* xalloc is a reserved word in VC */ |
| 3865 | STRLEN length_in_chars; |
| 3866 | bool first_time = TRUE; /* Cleared after first loop iteration */ |
| 3867 | |
| 3868 | PERL_ARGS_ASSERT__MEM_COLLXFRM; |
| 3869 | |
| 3870 | /* Must be NUL-terminated */ |
| 3871 | assert(*(input_string + len) == '\0'); |
| 3872 | |
| 3873 | /* If this locale has defective collation, skip */ |
| 3874 | if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) { |
| 3875 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 3876 | "_mem_collxfrm: locale's collation is defective\n")); |
| 3877 | goto bad; |
| 3878 | } |
| 3879 | |
| 3880 | /* Replace any embedded NULs with the control that sorts before any others. |
| 3881 | * This will give as good as possible results on strings that don't |
| 3882 | * otherwise contain that character, but otherwise there may be |
| 3883 | * less-than-perfect results with that character and NUL. This is |
| 3884 | * unavoidable unless we replace strxfrm with our own implementation. */ |
| 3885 | if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded |
| 3886 | NUL */ |
| 3887 | char * e = s + len; |
| 3888 | char * sans_nuls; |
| 3889 | STRLEN sans_nuls_len; |
| 3890 | int try_non_controls; |
| 3891 | char this_replacement_char[] = "?\0"; /* Room for a two-byte string, |
| 3892 | making sure 2nd byte is NUL. |
| 3893 | */ |
| 3894 | STRLEN this_replacement_len; |
| 3895 | |
| 3896 | /* If we don't know what non-NUL control character sorts lowest for |
| 3897 | * this locale, find it */ |
| 3898 | if (PL_strxfrm_NUL_replacement == '\0') { |
| 3899 | int j; |
| 3900 | char * cur_min_x = NULL; /* The min_char's xfrm, (except it also |
| 3901 | includes the collation index |
| 3902 | prefixed. */ |
| 3903 | |
| 3904 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n")); |
| 3905 | |
| 3906 | /* Unlikely, but it may be that no control will work to replace |
| 3907 | * NUL, in which case we instead look for any character. Controls |
| 3908 | * are preferred because collation order is, in general, context |
| 3909 | * sensitive, with adjoining characters affecting the order, and |
| 3910 | * controls are less likely to have such interactions, allowing the |
| 3911 | * NUL-replacement to stand on its own. (Another way to look at it |
| 3912 | * is to imagine what would happen if the NUL were replaced by a |
| 3913 | * combining character; it wouldn't work out all that well.) */ |
| 3914 | for (try_non_controls = 0; |
| 3915 | try_non_controls < 2; |
| 3916 | try_non_controls++) |
| 3917 | { |
| 3918 | /* Look through all legal code points (NUL isn't) */ |
| 3919 | for (j = 1; j < 256; j++) { |
| 3920 | char * x; /* j's xfrm plus collation index */ |
| 3921 | STRLEN x_len; /* length of 'x' */ |
| 3922 | STRLEN trial_len = 1; |
| 3923 | char cur_source[] = { '\0', '\0' }; |
| 3924 | |
| 3925 | /* Skip non-controls the first time through the loop. The |
| 3926 | * controls in a UTF-8 locale are the L1 ones */ |
| 3927 | if (! try_non_controls && (PL_in_utf8_COLLATE_locale) |
| 3928 | ? ! isCNTRL_L1(j) |
| 3929 | : ! isCNTRL_LC(j)) |
| 3930 | { |
| 3931 | continue; |
| 3932 | } |
| 3933 | |
| 3934 | /* Create a 1-char string of the current code point */ |
| 3935 | cur_source[0] = (char) j; |
| 3936 | |
| 3937 | /* Then transform it */ |
| 3938 | x = _mem_collxfrm(cur_source, trial_len, &x_len, |
| 3939 | 0 /* The string is not in UTF-8 */); |
| 3940 | |
| 3941 | /* Ignore any character that didn't successfully transform. |
| 3942 | * */ |
| 3943 | if (! x) { |
| 3944 | continue; |
| 3945 | } |
| 3946 | |
| 3947 | /* If this character's transformation is lower than |
| 3948 | * the current lowest, this one becomes the lowest */ |
| 3949 | if ( cur_min_x == NULL |
| 3950 | || strLT(x + COLLXFRM_HDR_LEN, |
| 3951 | cur_min_x + COLLXFRM_HDR_LEN)) |
| 3952 | { |
| 3953 | PL_strxfrm_NUL_replacement = j; |
| 3954 | Safefree(cur_min_x); |
| 3955 | cur_min_x = x; |
| 3956 | } |
| 3957 | else { |
| 3958 | Safefree(x); |
| 3959 | } |
| 3960 | } /* end of loop through all 255 characters */ |
| 3961 | |
| 3962 | /* Stop looking if found */ |
| 3963 | if (cur_min_x) { |
| 3964 | break; |
| 3965 | } |
| 3966 | |
| 3967 | /* Unlikely, but possible, if there aren't any controls that |
| 3968 | * work in the locale, repeat the loop, looking for any |
| 3969 | * character that works */ |
| 3970 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 3971 | "_mem_collxfrm: No control worked. Trying non-controls\n")); |
| 3972 | } /* End of loop to try first the controls, then any char */ |
| 3973 | |
| 3974 | if (! cur_min_x) { |
| 3975 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 3976 | "_mem_collxfrm: Couldn't find any character to replace" |
| 3977 | " embedded NULs in locale %s with", PL_collation_name)); |
| 3978 | goto bad; |
| 3979 | } |
| 3980 | |
| 3981 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 3982 | "_mem_collxfrm: Replacing embedded NULs in locale %s with " |
| 3983 | "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement)); |
| 3984 | |
| 3985 | Safefree(cur_min_x); |
| 3986 | } /* End of determining the character that is to replace NULs */ |
| 3987 | |
| 3988 | /* If the replacement is variant under UTF-8, it must match the |
| 3989 | * UTF8-ness of the original */ |
| 3990 | if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) { |
| 3991 | this_replacement_char[0] = |
| 3992 | UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement); |
| 3993 | this_replacement_char[1] = |
| 3994 | UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement); |
| 3995 | this_replacement_len = 2; |
| 3996 | } |
| 3997 | else { |
| 3998 | this_replacement_char[0] = PL_strxfrm_NUL_replacement; |
| 3999 | /* this_replacement_char[1] = '\0' was done at initialization */ |
| 4000 | this_replacement_len = 1; |
| 4001 | } |
| 4002 | |
| 4003 | /* The worst case length for the replaced string would be if every |
| 4004 | * character in it is NUL. Multiply that by the length of each |
| 4005 | * replacement, and allow for a trailing NUL */ |
| 4006 | sans_nuls_len = (len * this_replacement_len) + 1; |
| 4007 | Newx(sans_nuls, sans_nuls_len, char); |
| 4008 | *sans_nuls = '\0'; |
| 4009 | |
| 4010 | /* Replace each NUL with the lowest collating control. Loop until have |
| 4011 | * exhausted all the NULs */ |
| 4012 | while (s + s_strlen < e) { |
| 4013 | my_strlcat(sans_nuls, s, sans_nuls_len); |
| 4014 | |
| 4015 | /* Do the actual replacement */ |
| 4016 | my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len); |
| 4017 | |
| 4018 | /* Move past the input NUL */ |
| 4019 | s += s_strlen + 1; |
| 4020 | s_strlen = strlen(s); |
| 4021 | } |
| 4022 | |
| 4023 | /* And add anything that trails the final NUL */ |
| 4024 | my_strlcat(sans_nuls, s, sans_nuls_len); |
| 4025 | |
| 4026 | /* Switch so below we transform this modified string */ |
| 4027 | s = sans_nuls; |
| 4028 | len = strlen(s); |
| 4029 | } /* End of replacing NULs */ |
| 4030 | |
| 4031 | /* Make sure the UTF8ness of the string and locale match */ |
| 4032 | if (utf8 != PL_in_utf8_COLLATE_locale) { |
| 4033 | /* XXX convert above Unicode to 10FFFF? */ |
| 4034 | const char * const t = s; /* Temporary so we can later find where the |
| 4035 | input was */ |
| 4036 | |
| 4037 | /* Here they don't match. Change the string's to be what the locale is |
| 4038 | * expecting */ |
| 4039 | |
| 4040 | if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */ |
| 4041 | s = (char *) bytes_to_utf8((const U8 *) s, &len); |
| 4042 | utf8 = TRUE; |
| 4043 | } |
| 4044 | else { /* locale is not UTF-8; but input is; downgrade the input */ |
| 4045 | |
| 4046 | s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8); |
| 4047 | |
| 4048 | /* If the downgrade was successful we are done, but if the input |
| 4049 | * contains things that require UTF-8 to represent, have to do |
| 4050 | * damage control ... */ |
| 4051 | if (UNLIKELY(utf8)) { |
| 4052 | |
| 4053 | /* What we do is construct a non-UTF-8 string with |
| 4054 | * 1) the characters representable by a single byte converted |
| 4055 | * to be so (if necessary); |
| 4056 | * 2) and the rest converted to collate the same as the |
| 4057 | * highest collating representable character. That makes |
| 4058 | * them collate at the end. This is similar to how we |
| 4059 | * handle embedded NULs, but we use the highest collating |
| 4060 | * code point instead of the smallest. Like the NUL case, |
| 4061 | * this isn't perfect, but is the best we can reasonably |
| 4062 | * do. Every above-255 code point will sort the same as |
| 4063 | * the highest-sorting 0-255 code point. If that code |
| 4064 | * point can combine in a sequence with some other code |
| 4065 | * points for weight calculations, us changing something to |
| 4066 | * be it can adversely affect the results. But in most |
| 4067 | * cases, it should work reasonably. And note that this is |
| 4068 | * really an illegal situation: using code points above 255 |
| 4069 | * on a locale where only 0-255 are valid. If two strings |
| 4070 | * sort entirely equal, then the sort order for the |
| 4071 | * above-255 code points will be in code point order. */ |
| 4072 | |
| 4073 | utf8 = FALSE; |
| 4074 | |
| 4075 | /* If we haven't calculated the code point with the maximum |
| 4076 | * collating order for this locale, do so now */ |
| 4077 | if (! PL_strxfrm_max_cp) { |
| 4078 | int j; |
| 4079 | |
| 4080 | /* The current transformed string that collates the |
| 4081 | * highest (except it also includes the prefixed collation |
| 4082 | * index. */ |
| 4083 | char * cur_max_x = NULL; |
| 4084 | |
| 4085 | /* Look through all legal code points (NUL isn't) */ |
| 4086 | for (j = 1; j < 256; j++) { |
| 4087 | char * x; |
| 4088 | STRLEN x_len; |
| 4089 | char cur_source[] = { '\0', '\0' }; |
| 4090 | |
| 4091 | /* Create a 1-char string of the current code point */ |
| 4092 | cur_source[0] = (char) j; |
| 4093 | |
| 4094 | /* Then transform it */ |
| 4095 | x = _mem_collxfrm(cur_source, 1, &x_len, FALSE); |
| 4096 | |
| 4097 | /* If something went wrong (which it shouldn't), just |
| 4098 | * ignore this code point */ |
| 4099 | if (! x) { |
| 4100 | continue; |
| 4101 | } |
| 4102 | |
| 4103 | /* If this character's transformation is higher than |
| 4104 | * the current highest, this one becomes the highest */ |
| 4105 | if ( cur_max_x == NULL |
| 4106 | || strGT(x + COLLXFRM_HDR_LEN, |
| 4107 | cur_max_x + COLLXFRM_HDR_LEN)) |
| 4108 | { |
| 4109 | PL_strxfrm_max_cp = j; |
| 4110 | Safefree(cur_max_x); |
| 4111 | cur_max_x = x; |
| 4112 | } |
| 4113 | else { |
| 4114 | Safefree(x); |
| 4115 | } |
| 4116 | } |
| 4117 | |
| 4118 | if (! cur_max_x) { |
| 4119 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4120 | "_mem_collxfrm: Couldn't find any character to" |
| 4121 | " replace above-Latin1 chars in locale %s with", |
| 4122 | PL_collation_name)); |
| 4123 | goto bad; |
| 4124 | } |
| 4125 | |
| 4126 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4127 | "_mem_collxfrm: highest 1-byte collating character" |
| 4128 | " in locale %s is 0x%02X\n", |
| 4129 | PL_collation_name, |
| 4130 | PL_strxfrm_max_cp)); |
| 4131 | |
| 4132 | Safefree(cur_max_x); |
| 4133 | } |
| 4134 | |
| 4135 | /* Here we know which legal code point collates the highest. |
| 4136 | * We are ready to construct the non-UTF-8 string. The length |
| 4137 | * will be at least 1 byte smaller than the input string |
| 4138 | * (because we changed at least one 2-byte character into a |
| 4139 | * single byte), but that is eaten up by the trailing NUL */ |
| 4140 | Newx(s, len, char); |
| 4141 | |
| 4142 | { |
| 4143 | STRLEN i; |
| 4144 | STRLEN d= 0; |
| 4145 | char * e = (char *) t + len; |
| 4146 | |
| 4147 | for (i = 0; i < len; i+= UTF8SKIP(t + i)) { |
| 4148 | U8 cur_char = t[i]; |
| 4149 | if (UTF8_IS_INVARIANT(cur_char)) { |
| 4150 | s[d++] = cur_char; |
| 4151 | } |
| 4152 | else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) { |
| 4153 | s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]); |
| 4154 | } |
| 4155 | else { /* Replace illegal cp with highest collating |
| 4156 | one */ |
| 4157 | s[d++] = PL_strxfrm_max_cp; |
| 4158 | } |
| 4159 | } |
| 4160 | s[d++] = '\0'; |
| 4161 | Renew(s, d, char); /* Free up unused space */ |
| 4162 | } |
| 4163 | } |
| 4164 | } |
| 4165 | |
| 4166 | /* Here, we have constructed a modified version of the input. It could |
| 4167 | * be that we already had a modified copy before we did this version. |
| 4168 | * If so, that copy is no longer needed */ |
| 4169 | if (t != input_string) { |
| 4170 | Safefree(t); |
| 4171 | } |
| 4172 | } |
| 4173 | |
| 4174 | length_in_chars = (utf8) |
| 4175 | ? utf8_length((U8 *) s, (U8 *) s + len) |
| 4176 | : len; |
| 4177 | |
| 4178 | /* The first element in the output is the collation id, used by |
| 4179 | * sv_collxfrm(); then comes the space for the transformed string. The |
| 4180 | * equation should give us a good estimate as to how much is needed */ |
| 4181 | xAlloc = COLLXFRM_HDR_LEN |
| 4182 | + PL_collxfrm_base |
| 4183 | + (PL_collxfrm_mult * length_in_chars); |
| 4184 | Newx(xbuf, xAlloc, char); |
| 4185 | if (UNLIKELY(! xbuf)) { |
| 4186 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4187 | "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc)); |
| 4188 | goto bad; |
| 4189 | } |
| 4190 | |
| 4191 | /* Store the collation id */ |
| 4192 | *(U32*)xbuf = PL_collation_ix; |
| 4193 | |
| 4194 | /* Then the transformation of the input. We loop until successful, or we |
| 4195 | * give up */ |
| 4196 | for (;;) { |
| 4197 | |
| 4198 | *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN); |
| 4199 | |
| 4200 | /* If the transformed string occupies less space than we told strxfrm() |
| 4201 | * was available, it means it successfully transformed the whole |
| 4202 | * string. */ |
| 4203 | if (*xlen < xAlloc - COLLXFRM_HDR_LEN) { |
| 4204 | |
| 4205 | /* Some systems include a trailing NUL in the returned length. |
| 4206 | * Ignore it, using a loop in case multiple trailing NULs are |
| 4207 | * returned. */ |
| 4208 | while ( (*xlen) > 0 |
| 4209 | && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0') |
| 4210 | { |
| 4211 | (*xlen)--; |
| 4212 | } |
| 4213 | |
| 4214 | /* If the first try didn't get it, it means our prediction was low. |
| 4215 | * Modify the coefficients so that we predict a larger value in any |
| 4216 | * future transformations */ |
| 4217 | if (! first_time) { |
| 4218 | STRLEN needed = *xlen + 1; /* +1 For trailing NUL */ |
| 4219 | STRLEN computed_guess = PL_collxfrm_base |
| 4220 | + (PL_collxfrm_mult * length_in_chars); |
| 4221 | |
| 4222 | /* On zero-length input, just keep current slope instead of |
| 4223 | * dividing by 0 */ |
| 4224 | const STRLEN new_m = (length_in_chars != 0) |
| 4225 | ? needed / length_in_chars |
| 4226 | : PL_collxfrm_mult; |
| 4227 | |
| 4228 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 4229 | "%s: %d: initial size of %zu bytes for a length " |
| 4230 | "%zu string was insufficient, %zu needed\n", |
| 4231 | __FILE__, __LINE__, |
| 4232 | computed_guess, length_in_chars, needed)); |
| 4233 | |
| 4234 | /* If slope increased, use it, but discard this result for |
| 4235 | * length 1 strings, as we can't be sure that it's a real slope |
| 4236 | * change */ |
| 4237 | if (length_in_chars > 1 && new_m > PL_collxfrm_mult) { |
| 4238 | |
| 4239 | # ifdef DEBUGGING |
| 4240 | |
| 4241 | STRLEN old_m = PL_collxfrm_mult; |
| 4242 | STRLEN old_b = PL_collxfrm_base; |
| 4243 | |
| 4244 | # endif |
| 4245 | |
| 4246 | PL_collxfrm_mult = new_m; |
| 4247 | PL_collxfrm_base = 1; /* +1 For trailing NUL */ |
| 4248 | computed_guess = PL_collxfrm_base |
| 4249 | + (PL_collxfrm_mult * length_in_chars); |
| 4250 | if (computed_guess < needed) { |
| 4251 | PL_collxfrm_base += needed - computed_guess; |
| 4252 | } |
| 4253 | |
| 4254 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 4255 | "%s: %d: slope is now %zu; was %zu, base " |
| 4256 | "is now %zu; was %zu\n", |
| 4257 | __FILE__, __LINE__, |
| 4258 | PL_collxfrm_mult, old_m, |
| 4259 | PL_collxfrm_base, old_b)); |
| 4260 | } |
| 4261 | else { /* Slope didn't change, but 'b' did */ |
| 4262 | const STRLEN new_b = needed |
| 4263 | - computed_guess |
| 4264 | + PL_collxfrm_base; |
| 4265 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 4266 | "%s: %d: base is now %zu; was %zu\n", |
| 4267 | __FILE__, __LINE__, |
| 4268 | new_b, PL_collxfrm_base)); |
| 4269 | PL_collxfrm_base = new_b; |
| 4270 | } |
| 4271 | } |
| 4272 | |
| 4273 | break; |
| 4274 | } |
| 4275 | |
| 4276 | if (UNLIKELY(*xlen >= PERL_INT_MAX)) { |
| 4277 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4278 | "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n", |
| 4279 | *xlen, PERL_INT_MAX)); |
| 4280 | goto bad; |
| 4281 | } |
| 4282 | |
| 4283 | /* A well-behaved strxfrm() returns exactly how much space it needs |
| 4284 | * (usually not including the trailing NUL) when it fails due to not |
| 4285 | * enough space being provided. Assume that this is the case unless |
| 4286 | * it's been proven otherwise */ |
| 4287 | if (LIKELY(PL_strxfrm_is_behaved) && first_time) { |
| 4288 | xAlloc = *xlen + COLLXFRM_HDR_LEN + 1; |
| 4289 | } |
| 4290 | else { /* Here, either: |
| 4291 | * 1) The strxfrm() has previously shown bad behavior; or |
| 4292 | * 2) It isn't the first time through the loop, which means |
| 4293 | * that the strxfrm() is now showing bad behavior, because |
| 4294 | * we gave it what it said was needed in the previous |
| 4295 | * iteration, and it came back saying it needed still more. |
| 4296 | * (Many versions of cygwin fit this. When the buffer size |
| 4297 | * isn't sufficient, they return the input size instead of |
| 4298 | * how much is needed.) |
| 4299 | * Increase the buffer size by a fixed percentage and try again. |
| 4300 | * */ |
| 4301 | xAlloc += (xAlloc / 4) + 1; |
| 4302 | PL_strxfrm_is_behaved = FALSE; |
| 4303 | |
| 4304 | # ifdef DEBUGGING |
| 4305 | |
| 4306 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 4307 | PerlIO_printf(Perl_debug_log, |
| 4308 | "_mem_collxfrm required more space than previously calculated" |
| 4309 | " for locale %s, trying again with new guess=%d+%zu\n", |
| 4310 | PL_collation_name, (int) COLLXFRM_HDR_LEN, |
| 4311 | xAlloc - COLLXFRM_HDR_LEN); |
| 4312 | } |
| 4313 | |
| 4314 | # endif |
| 4315 | |
| 4316 | } |
| 4317 | |
| 4318 | Renew(xbuf, xAlloc, char); |
| 4319 | if (UNLIKELY(! xbuf)) { |
| 4320 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4321 | "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc)); |
| 4322 | goto bad; |
| 4323 | } |
| 4324 | |
| 4325 | first_time = FALSE; |
| 4326 | } |
| 4327 | |
| 4328 | |
| 4329 | # ifdef DEBUGGING |
| 4330 | |
| 4331 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 4332 | |
| 4333 | print_collxfrm_input_and_return(s, s + len, xlen, utf8); |
| 4334 | PerlIO_printf(Perl_debug_log, "Its xfrm is:"); |
| 4335 | PerlIO_printf(Perl_debug_log, "%s\n", |
| 4336 | _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN, |
| 4337 | *xlen, 1)); |
| 4338 | } |
| 4339 | |
| 4340 | # endif |
| 4341 | |
| 4342 | /* Free up unneeded space; retain ehough for trailing NUL */ |
| 4343 | Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char); |
| 4344 | |
| 4345 | if (s != input_string) { |
| 4346 | Safefree(s); |
| 4347 | } |
| 4348 | |
| 4349 | return xbuf; |
| 4350 | |
| 4351 | bad: |
| 4352 | Safefree(xbuf); |
| 4353 | if (s != input_string) { |
| 4354 | Safefree(s); |
| 4355 | } |
| 4356 | *xlen = 0; |
| 4357 | |
| 4358 | # ifdef DEBUGGING |
| 4359 | |
| 4360 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 4361 | print_collxfrm_input_and_return(s, s + len, NULL, utf8); |
| 4362 | } |
| 4363 | |
| 4364 | # endif |
| 4365 | |
| 4366 | return NULL; |
| 4367 | } |
| 4368 | |
| 4369 | # ifdef DEBUGGING |
| 4370 | |
| 4371 | STATIC void |
| 4372 | S_print_collxfrm_input_and_return(pTHX_ |
| 4373 | const char * const s, |
| 4374 | const char * const e, |
| 4375 | const STRLEN * const xlen, |
| 4376 | const bool is_utf8) |
| 4377 | { |
| 4378 | |
| 4379 | PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN; |
| 4380 | |
| 4381 | PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ", |
| 4382 | (UV)PL_collation_ix); |
| 4383 | if (xlen) { |
| 4384 | PerlIO_printf(Perl_debug_log, "%zu", *xlen); |
| 4385 | } |
| 4386 | else { |
| 4387 | PerlIO_printf(Perl_debug_log, "NULL"); |
| 4388 | } |
| 4389 | PerlIO_printf(Perl_debug_log, " for locale '%s', string='", |
| 4390 | PL_collation_name); |
| 4391 | print_bytes_for_locale(s, e, is_utf8); |
| 4392 | |
| 4393 | PerlIO_printf(Perl_debug_log, "'\n"); |
| 4394 | } |
| 4395 | |
| 4396 | # endif /* DEBUGGING */ |
| 4397 | #endif /* USE_LOCALE_COLLATE */ |
| 4398 | #ifdef USE_LOCALE |
| 4399 | # ifdef DEBUGGING |
| 4400 | |
| 4401 | STATIC void |
| 4402 | S_print_bytes_for_locale(pTHX_ |
| 4403 | const char * const s, |
| 4404 | const char * const e, |
| 4405 | const bool is_utf8) |
| 4406 | { |
| 4407 | const char * t = s; |
| 4408 | bool prev_was_printable = TRUE; |
| 4409 | bool first_time = TRUE; |
| 4410 | |
| 4411 | PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE; |
| 4412 | |
| 4413 | while (t < e) { |
| 4414 | UV cp = (is_utf8) |
| 4415 | ? utf8_to_uvchr_buf((U8 *) t, e, NULL) |
| 4416 | : * (U8 *) t; |
| 4417 | if (isPRINT(cp)) { |
| 4418 | if (! prev_was_printable) { |
| 4419 | PerlIO_printf(Perl_debug_log, " "); |
| 4420 | } |
| 4421 | PerlIO_printf(Perl_debug_log, "%c", (U8) cp); |
| 4422 | prev_was_printable = TRUE; |
| 4423 | } |
| 4424 | else { |
| 4425 | if (! first_time) { |
| 4426 | PerlIO_printf(Perl_debug_log, " "); |
| 4427 | } |
| 4428 | PerlIO_printf(Perl_debug_log, "%02" UVXf, cp); |
| 4429 | prev_was_printable = FALSE; |
| 4430 | } |
| 4431 | t += (is_utf8) ? UTF8SKIP(t) : 1; |
| 4432 | first_time = FALSE; |
| 4433 | } |
| 4434 | } |
| 4435 | |
| 4436 | # endif /* #ifdef DEBUGGING */ |
| 4437 | |
| 4438 | STATIC const char * |
| 4439 | S_switch_category_locale_to_template(pTHX_ const int switch_category, const int template_category, const char * template_locale) |
| 4440 | { |
| 4441 | /* Changes the locale for LC_'switch_category" to that of |
| 4442 | * LC_'template_category', if they aren't already the same. If not NULL, |
| 4443 | * 'template_locale' is the locale that 'template_category' is in. |
| 4444 | * |
| 4445 | * Returns a copy of the name of the original locale for 'switch_category' |
| 4446 | * so can be switched back to with the companion function |
| 4447 | * restore_switched_locale(), (NULL if no restoral is necessary.) */ |
| 4448 | |
| 4449 | char * restore_to_locale = NULL; |
| 4450 | |
| 4451 | if (switch_category == template_category) { /* No changes needed */ |
| 4452 | return NULL; |
| 4453 | } |
| 4454 | |
| 4455 | /* Find the original locale of the category we may need to change, so that |
| 4456 | * it can be restored to later */ |
| 4457 | restore_to_locale = stdize_locale(savepv(do_setlocale_r(switch_category, |
| 4458 | NULL))); |
| 4459 | if (! restore_to_locale) { |
| 4460 | Perl_croak(aTHX_ |
| 4461 | "panic: %s: %d: Could not find current %s locale, errno=%d\n", |
| 4462 | __FILE__, __LINE__, category_name(switch_category), errno); |
| 4463 | } |
| 4464 | |
| 4465 | /* If the locale of the template category wasn't passed in, find it now */ |
| 4466 | if (template_locale == NULL) { |
| 4467 | template_locale = do_setlocale_r(template_category, NULL); |
| 4468 | if (! template_locale) { |
| 4469 | Perl_croak(aTHX_ |
| 4470 | "panic: %s: %d: Could not find current %s locale, errno=%d\n", |
| 4471 | __FILE__, __LINE__, category_name(template_category), errno); |
| 4472 | } |
| 4473 | } |
| 4474 | |
| 4475 | /* It the locales are the same, there's nothing to do */ |
| 4476 | if (strEQ(restore_to_locale, template_locale)) { |
| 4477 | Safefree(restore_to_locale); |
| 4478 | |
| 4479 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s locale unchanged as %s\n", |
| 4480 | category_name(switch_category), restore_to_locale)); |
| 4481 | |
| 4482 | return NULL; |
| 4483 | } |
| 4484 | |
| 4485 | /* Finally, change the locale to the template one */ |
| 4486 | if (! do_setlocale_r(switch_category, template_locale)) { |
| 4487 | Perl_croak(aTHX_ |
| 4488 | "panic: %s: %d: Could not change %s locale to %s, errno=%d\n", |
| 4489 | __FILE__, __LINE__, category_name(switch_category), |
| 4490 | template_locale, errno); |
| 4491 | } |
| 4492 | |
| 4493 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s locale switched to %s\n", |
| 4494 | category_name(switch_category), template_locale)); |
| 4495 | |
| 4496 | return restore_to_locale; |
| 4497 | } |
| 4498 | |
| 4499 | STATIC void |
| 4500 | S_restore_switched_locale(pTHX_ const int category, const char * const original_locale) |
| 4501 | { |
| 4502 | /* Restores the locale for LC_'category' to 'original_locale' (which is a |
| 4503 | * copy that will be freed by this function), or do nothing if the latter |
| 4504 | * parameter is NULL */ |
| 4505 | |
| 4506 | if (original_locale == NULL) { |
| 4507 | return; |
| 4508 | } |
| 4509 | |
| 4510 | if (! do_setlocale_r(category, original_locale)) { |
| 4511 | Perl_croak(aTHX_ |
| 4512 | "panic: %s: %d: setlocale %s restore to %s failed, errno=%d\n", |
| 4513 | __FILE__, __LINE__, |
| 4514 | category_name(category), original_locale, errno); |
| 4515 | } |
| 4516 | |
| 4517 | Safefree(original_locale); |
| 4518 | } |
| 4519 | |
| 4520 | /* is_cur_LC_category_utf8 uses a small char buffer to avoid malloc/free */ |
| 4521 | #define CUR_LC_BUFFER_SIZE 64 |
| 4522 | |
| 4523 | bool |
| 4524 | Perl__is_cur_LC_category_utf8(pTHX_ int category) |
| 4525 | { |
| 4526 | /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE |
| 4527 | * otherwise. 'category' may not be LC_ALL. If the platform doesn't have |
| 4528 | * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence |
| 4529 | * could give the wrong result. The result will very likely be correct for |
| 4530 | * languages that have commonly used non-ASCII characters, but for notably |
| 4531 | * English, it comes down to if the locale's name ends in something like |
| 4532 | * "UTF-8". It errs on the side of not being a UTF-8 locale. |
| 4533 | * |
| 4534 | * If the platform is early C89, not containing mbtowc(), or we are |
| 4535 | * compiled to not pay attention to LC_CTYPE, this employs heuristics. |
| 4536 | * These work very well for non-Latin locales or those whose currency |
| 4537 | * symbol isn't a '$' nor plain ASCII text. But without LC_CTYPE and at |
| 4538 | * least MB_CUR_MAX, English locales with an ASCII currency symbol depend |
| 4539 | * on the name containing UTF-8 or not. */ |
| 4540 | |
| 4541 | /* Name of current locale corresponding to the input category */ |
| 4542 | const char *save_input_locale = NULL; |
| 4543 | |
| 4544 | bool is_utf8 = FALSE; /* The return value */ |
| 4545 | |
| 4546 | /* The variables below are for the cache of previous lookups using this |
| 4547 | * function. The cache is a C string, described at the definition for |
| 4548 | * 'C_and_POSIX_utf8ness'. |
| 4549 | * |
| 4550 | * The first part of the cache is fixed, for the C and POSIX locales. The |
| 4551 | * varying part starts just after them. */ |
| 4552 | char * utf8ness_cache = PL_locale_utf8ness + STRLENs(C_and_POSIX_utf8ness); |
| 4553 | |
| 4554 | Size_t utf8ness_cache_size; /* Size of the varying portion */ |
| 4555 | Size_t input_name_len; /* Length in bytes of save_input_locale */ |
| 4556 | Size_t input_name_len_with_overhead; /* plus extra chars used to store |
| 4557 | the name in the cache */ |
| 4558 | char * delimited; /* The name plus the delimiters used to store |
| 4559 | it in the cache */ |
| 4560 | char buffer[CUR_LC_BUFFER_SIZE]; /* small buffer */ |
| 4561 | char * name_pos; /* position of 'delimited' in the cache, or 0 |
| 4562 | if not there */ |
| 4563 | |
| 4564 | |
| 4565 | # ifdef LC_ALL |
| 4566 | |
| 4567 | assert(category != LC_ALL); |
| 4568 | |
| 4569 | # endif |
| 4570 | |
| 4571 | /* Get the desired category's locale */ |
| 4572 | save_input_locale = stdize_locale(savepv(do_setlocale_r(category, NULL))); |
| 4573 | if (! save_input_locale) { |
| 4574 | Perl_croak(aTHX_ |
| 4575 | "panic: %s: %d: Could not find current %s locale, errno=%d\n", |
| 4576 | __FILE__, __LINE__, category_name(category), errno); |
| 4577 | } |
| 4578 | |
| 4579 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4580 | "Current locale for %s is %s\n", |
| 4581 | category_name(category), save_input_locale)); |
| 4582 | |
| 4583 | input_name_len = strlen(save_input_locale); |
| 4584 | |
| 4585 | /* In our cache, each name is accompanied by two delimiters and a single |
| 4586 | * utf8ness digit */ |
| 4587 | input_name_len_with_overhead = input_name_len + 3; |
| 4588 | |
| 4589 | if ( input_name_len_with_overhead <= CUR_LC_BUFFER_SIZE ) { |
| 4590 | /* we can use the buffer, avoid a malloc */ |
| 4591 | delimited = buffer; |
| 4592 | } else { /* need a malloc */ |
| 4593 | /* Allocate and populate space for a copy of the name surrounded by the |
| 4594 | * delimiters */ |
| 4595 | Newx(delimited, input_name_len_with_overhead, char); |
| 4596 | } |
| 4597 | |
| 4598 | delimited[0] = UTF8NESS_SEP[0]; |
| 4599 | Copy(save_input_locale, delimited + 1, input_name_len, char); |
| 4600 | delimited[input_name_len+1] = UTF8NESS_PREFIX[0]; |
| 4601 | delimited[input_name_len+2] = '\0'; |
| 4602 | |
| 4603 | /* And see if that is in the cache */ |
| 4604 | name_pos = instr(PL_locale_utf8ness, delimited); |
| 4605 | if (name_pos) { |
| 4606 | is_utf8 = *(name_pos + input_name_len_with_overhead - 1) - '0'; |
| 4607 | |
| 4608 | # ifdef DEBUGGING |
| 4609 | |
| 4610 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 4611 | PerlIO_printf(Perl_debug_log, "UTF8ness for locale %s=%d, \n", |
| 4612 | save_input_locale, is_utf8); |
| 4613 | } |
| 4614 | |
| 4615 | # endif |
| 4616 | |
| 4617 | /* And, if not already in that position, move it to the beginning of |
| 4618 | * the non-constant portion of the list, since it is the most recently |
| 4619 | * used. (We don't have to worry about overflow, since just moving |
| 4620 | * existing names around) */ |
| 4621 | if (name_pos > utf8ness_cache) { |
| 4622 | Move(utf8ness_cache, |
| 4623 | utf8ness_cache + input_name_len_with_overhead, |
| 4624 | name_pos - utf8ness_cache, char); |
| 4625 | Copy(delimited, |
| 4626 | utf8ness_cache, |
| 4627 | input_name_len_with_overhead - 1, char); |
| 4628 | utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0'; |
| 4629 | } |
| 4630 | |
| 4631 | /* free only when not using the buffer */ |
| 4632 | if ( delimited != buffer ) Safefree(delimited); |
| 4633 | Safefree(save_input_locale); |
| 4634 | return is_utf8; |
| 4635 | } |
| 4636 | |
| 4637 | /* Here we don't have stored the utf8ness for the input locale. We have to |
| 4638 | * calculate it */ |
| 4639 | |
| 4640 | # if defined(USE_LOCALE_CTYPE) \ |
| 4641 | && ( defined(HAS_NL_LANGINFO) \ |
| 4642 | || (defined(HAS_MBTOWC) || defined(HAS_MBRTOWC))) |
| 4643 | |
| 4644 | { |
| 4645 | const char *original_ctype_locale |
| 4646 | = switch_category_locale_to_template(LC_CTYPE, |
| 4647 | category, |
| 4648 | save_input_locale); |
| 4649 | |
| 4650 | /* Here the current LC_CTYPE is set to the locale of the category whose |
| 4651 | * information is desired. This means that nl_langinfo() and mbtowc() |
| 4652 | * should give the correct results */ |
| 4653 | |
| 4654 | # ifdef MB_CUR_MAX /* But we can potentially rule out UTF-8ness, avoiding |
| 4655 | calling the functions if we have this */ |
| 4656 | |
| 4657 | /* Standard UTF-8 needs at least 4 bytes to represent the maximum |
| 4658 | * Unicode code point. */ |
| 4659 | |
| 4660 | DEBUG_L(PerlIO_printf(Perl_debug_log, "%s: %d: MB_CUR_MAX=%d\n", |
| 4661 | __FILE__, __LINE__, (int) MB_CUR_MAX)); |
| 4662 | if ((unsigned) MB_CUR_MAX < STRLENs(MAX_UNICODE_UTF8)) { |
| 4663 | is_utf8 = FALSE; |
| 4664 | restore_switched_locale(LC_CTYPE, original_ctype_locale); |
| 4665 | goto finish_and_return; |
| 4666 | } |
| 4667 | |
| 4668 | # endif |
| 4669 | # if defined(HAS_NL_LANGINFO) |
| 4670 | |
| 4671 | { /* The task is easiest if the platform has this POSIX 2001 function. |
| 4672 | Except on some platforms it can wrongly return "", so have to have |
| 4673 | a fallback. And it can return that it's UTF-8, even if there are |
| 4674 | variances from that. For example, Turkish locales may use the |
| 4675 | alternate dotted I rules, and sometimes it appears to be a |
| 4676 | defective locale definition. XXX We should probably check for |
| 4677 | these in the Latin1 range and warn (but on glibc, requires |
| 4678 | iswalnum() etc. due to their not handling 80-FF correctly */ |
| 4679 | const char *codeset = my_nl_langinfo(CODESET, FALSE); |
| 4680 | /* FALSE => already in dest locale */ |
| 4681 | |
| 4682 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 4683 | "\tnllanginfo returned CODESET '%s'\n", codeset)); |
| 4684 | |
| 4685 | if (codeset && strNE(codeset, "")) { |
| 4686 | |
| 4687 | /* If the implementation of foldEQ() somehow were |
| 4688 | * to change to not go byte-by-byte, this could |
| 4689 | * read past end of string, as only one length is |
| 4690 | * checked. But currently, a premature NUL will |
| 4691 | * compare false, and it will stop there */ |
| 4692 | is_utf8 = cBOOL( foldEQ(codeset, STR_WITH_LEN("UTF-8")) |
| 4693 | || foldEQ(codeset, STR_WITH_LEN("UTF8"))); |
| 4694 | |
| 4695 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4696 | "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n", |
| 4697 | codeset, is_utf8)); |
| 4698 | restore_switched_locale(LC_CTYPE, original_ctype_locale); |
| 4699 | goto finish_and_return; |
| 4700 | } |
| 4701 | } |
| 4702 | |
| 4703 | # endif |
| 4704 | # if defined(HAS_MBTOWC) || defined(HAS_MBRTOWC) |
| 4705 | /* We can see if this is a UTF-8-like locale if have mbtowc(). It was a |
| 4706 | * late adder to C89, so very likely to have it. However, testing has |
| 4707 | * shown that, like nl_langinfo() above, there are locales that are not |
| 4708 | * strictly UTF-8 that this will return that they are */ |
| 4709 | |
| 4710 | { |
| 4711 | wchar_t wc; |
| 4712 | int len; |
| 4713 | dSAVEDERRNO; |
| 4714 | |
| 4715 | # if defined(HAS_MBRTOWC) && defined(USE_ITHREADS) |
| 4716 | |
| 4717 | mbstate_t ps; |
| 4718 | |
| 4719 | # endif |
| 4720 | |
| 4721 | /* mbrtowc() and mbtowc() convert a byte string to a wide |
| 4722 | * character. Feed a byte string to one of them and check that the |
| 4723 | * result is the expected Unicode code point */ |
| 4724 | |
| 4725 | # if defined(HAS_MBRTOWC) && defined(USE_ITHREADS) |
| 4726 | /* Prefer this function if available, as it's reentrant */ |
| 4727 | |
| 4728 | memset(&ps, 0, sizeof(ps));; |
| 4729 | PERL_UNUSED_RESULT(mbrtowc(&wc, NULL, 0, &ps)); /* Reset any shift |
| 4730 | state */ |
| 4731 | SETERRNO(0, 0); |
| 4732 | len = mbrtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8), &ps); |
| 4733 | SAVE_ERRNO; |
| 4734 | |
| 4735 | # else |
| 4736 | |
| 4737 | LOCALE_LOCK; |
| 4738 | PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */ |
| 4739 | SETERRNO(0, 0); |
| 4740 | len = mbtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8)); |
| 4741 | SAVE_ERRNO; |
| 4742 | LOCALE_UNLOCK; |
| 4743 | |
| 4744 | # endif |
| 4745 | |
| 4746 | RESTORE_ERRNO; |
| 4747 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 4748 | "\treturn from mbtowc; len=%d; code_point=%x; errno=%d\n", |
| 4749 | len, (unsigned int) wc, GET_ERRNO)); |
| 4750 | |
| 4751 | is_utf8 = cBOOL( len == STRLENs(REPLACEMENT_CHARACTER_UTF8) |
| 4752 | && wc == (wchar_t) UNICODE_REPLACEMENT); |
| 4753 | } |
| 4754 | |
| 4755 | # endif |
| 4756 | |
| 4757 | restore_switched_locale(LC_CTYPE, original_ctype_locale); |
| 4758 | goto finish_and_return; |
| 4759 | } |
| 4760 | |
| 4761 | # else |
| 4762 | |
| 4763 | /* Here, we must have a C89 compiler that doesn't have mbtowc(). Next |
| 4764 | * try looking at the currency symbol to see if it disambiguates |
| 4765 | * things. Often that will be in the native script, and if the symbol |
| 4766 | * isn't in UTF-8, we know that the locale isn't. If it is non-ASCII |
| 4767 | * UTF-8, we infer that the locale is too, as the odds of a non-UTF8 |
| 4768 | * string being valid UTF-8 are quite small */ |
| 4769 | |
| 4770 | # ifdef USE_LOCALE_MONETARY |
| 4771 | |
| 4772 | /* If have LC_MONETARY, we can look at the currency symbol. Often that |
| 4773 | * will be in the native script. We do this one first because there is |
| 4774 | * just one string to examine, so potentially avoids work */ |
| 4775 | |
| 4776 | { |
| 4777 | const char *original_monetary_locale |
| 4778 | = switch_category_locale_to_template(LC_MONETARY, |
| 4779 | category, |
| 4780 | save_input_locale); |
| 4781 | bool only_ascii = FALSE; |
| 4782 | const U8 * currency_string |
| 4783 | = (const U8 *) my_nl_langinfo(CRNCYSTR, FALSE); |
| 4784 | /* 2nd param not relevant for this item */ |
| 4785 | const U8 * first_variant; |
| 4786 | |
| 4787 | assert( *currency_string == '-' |
| 4788 | || *currency_string == '+' |
| 4789 | || *currency_string == '.'); |
| 4790 | |
| 4791 | currency_string++; |
| 4792 | |
| 4793 | if (is_utf8_invariant_string_loc(currency_string, 0, &first_variant)) |
| 4794 | { |
| 4795 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Couldn't get currency symbol for %s, or contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); |
| 4796 | only_ascii = TRUE; |
| 4797 | } |
| 4798 | else { |
| 4799 | is_utf8 = is_strict_utf8_string(first_variant, 0); |
| 4800 | } |
| 4801 | |
| 4802 | restore_switched_locale(LC_MONETARY, original_monetary_locale); |
| 4803 | |
| 4804 | if (! only_ascii) { |
| 4805 | |
| 4806 | /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; |
| 4807 | * otherwise assume the locale is UTF-8 if and only if the symbol |
| 4808 | * is non-ascii UTF-8. */ |
| 4809 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n", |
| 4810 | save_input_locale, is_utf8)); |
| 4811 | goto finish_and_return; |
| 4812 | } |
| 4813 | } |
| 4814 | |
| 4815 | # endif /* USE_LOCALE_MONETARY */ |
| 4816 | # if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME) |
| 4817 | |
| 4818 | /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try |
| 4819 | * the names of the months and weekdays, timezone, and am/pm indicator */ |
| 4820 | { |
| 4821 | const char *original_time_locale |
| 4822 | = switch_category_locale_to_template(LC_TIME, |
| 4823 | category, |
| 4824 | save_input_locale); |
| 4825 | int hour = 10; |
| 4826 | bool is_dst = FALSE; |
| 4827 | int dom = 1; |
| 4828 | int month = 0; |
| 4829 | int i; |
| 4830 | char * formatted_time; |
| 4831 | |
| 4832 | /* Here the current LC_TIME is set to the locale of the category |
| 4833 | * whose information is desired. Look at all the days of the week and |
| 4834 | * month names, and the timezone and am/pm indicator for UTF-8 variant |
| 4835 | * characters. The first such a one found will tell us if the locale |
| 4836 | * is UTF-8 or not */ |
| 4837 | |
| 4838 | for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */ |
| 4839 | formatted_time = my_strftime("%A %B %Z %p", |
| 4840 | 0, 0, hour, dom, month, 2012 - 1900, 0, 0, is_dst); |
| 4841 | if ( ! formatted_time |
| 4842 | || is_utf8_invariant_string((U8 *) formatted_time, 0)) |
| 4843 | { |
| 4844 | |
| 4845 | /* Here, we didn't find a non-ASCII. Try the next time through |
| 4846 | * with the complemented dst and am/pm, and try with the next |
| 4847 | * weekday. After we have gotten all weekdays, try the next |
| 4848 | * month */ |
| 4849 | is_dst = ! is_dst; |
| 4850 | hour = (hour + 12) % 24; |
| 4851 | dom++; |
| 4852 | if (i > 6) { |
| 4853 | month++; |
| 4854 | } |
| 4855 | continue; |
| 4856 | } |
| 4857 | |
| 4858 | /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8; |
| 4859 | * false otherwise. But first, restore LC_TIME to its original |
| 4860 | * locale if we changed it */ |
| 4861 | restore_switched_locale(LC_TIME, original_time_locale); |
| 4862 | |
| 4863 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n", |
| 4864 | save_input_locale, |
| 4865 | is_utf8_string((U8 *) formatted_time, 0))); |
| 4866 | is_utf8 = is_utf8_string((U8 *) formatted_time, 0); |
| 4867 | goto finish_and_return; |
| 4868 | } |
| 4869 | |
| 4870 | /* Falling off the end of the loop indicates all the names were just |
| 4871 | * ASCII. Go on to the next test. If we changed it, restore LC_TIME |
| 4872 | * to its original locale */ |
| 4873 | restore_switched_locale(LC_TIME, original_time_locale); |
| 4874 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "All time-related words for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); |
| 4875 | } |
| 4876 | |
| 4877 | # endif |
| 4878 | |
| 4879 | # if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST) |
| 4880 | |
| 4881 | /* This code is ifdefd out because it was found to not be necessary in testing |
| 4882 | * on our dromedary test machine, which has over 700 locales. There, this |
| 4883 | * added no value to looking at the currency symbol and the time strings. I |
| 4884 | * left it in so as to avoid rewriting it if real-world experience indicates |
| 4885 | * that dromedary is an outlier. Essentially, instead of returning abpve if we |
| 4886 | * haven't found illegal utf8, we continue on and examine all the strerror() |
| 4887 | * messages on the platform for utf8ness. If all are ASCII, we still don't |
| 4888 | * know the answer; but otherwise we have a pretty good indication of the |
| 4889 | * utf8ness. The reason this doesn't help much is that the messages may not |
| 4890 | * have been translated into the locale. The currency symbol and time strings |
| 4891 | * are much more likely to have been translated. */ |
| 4892 | { |
| 4893 | int e; |
| 4894 | bool non_ascii = FALSE; |
| 4895 | const char *original_messages_locale |
| 4896 | = switch_category_locale_to_template(LC_MESSAGES, |
| 4897 | category, |
| 4898 | save_input_locale); |
| 4899 | const char * errmsg = NULL; |
| 4900 | |
| 4901 | /* Here the current LC_MESSAGES is set to the locale of the category |
| 4902 | * whose information is desired. Look through all the messages. We |
| 4903 | * can't use Strerror() here because it may expand to code that |
| 4904 | * segfaults in miniperl */ |
| 4905 | |
| 4906 | for (e = 0; e <= sys_nerr; e++) { |
| 4907 | errno = 0; |
| 4908 | errmsg = sys_errlist[e]; |
| 4909 | if (errno || !errmsg) { |
| 4910 | break; |
| 4911 | } |
| 4912 | errmsg = savepv(errmsg); |
| 4913 | if (! is_utf8_invariant_string((U8 *) errmsg, 0)) { |
| 4914 | non_ascii = TRUE; |
| 4915 | is_utf8 = is_utf8_string((U8 *) errmsg, 0); |
| 4916 | break; |
| 4917 | } |
| 4918 | } |
| 4919 | Safefree(errmsg); |
| 4920 | |
| 4921 | restore_switched_locale(LC_MESSAGES, original_messages_locale); |
| 4922 | |
| 4923 | if (non_ascii) { |
| 4924 | |
| 4925 | /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid, |
| 4926 | * any non-ascii means it is one; otherwise we assume it isn't */ |
| 4927 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n", |
| 4928 | save_input_locale, |
| 4929 | is_utf8)); |
| 4930 | goto finish_and_return; |
| 4931 | } |
| 4932 | |
| 4933 | DEBUG_L(PerlIO_printf(Perl_debug_log, "All error messages for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); |
| 4934 | } |
| 4935 | |
| 4936 | # endif |
| 4937 | # ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a |
| 4938 | UTF-8 locale */ |
| 4939 | |
| 4940 | /* As a last resort, look at the locale name to see if it matches |
| 4941 | * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the |
| 4942 | * return of setlocale(), is actually defined to be opaque, so we can't |
| 4943 | * really rely on the absence of various substrings in the name to indicate |
| 4944 | * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to |
| 4945 | * be a UTF-8 locale. Similarly for the other common names */ |
| 4946 | |
| 4947 | { |
| 4948 | const Size_t final_pos = strlen(save_input_locale) - 1; |
| 4949 | |
| 4950 | if (final_pos >= 3) { |
| 4951 | const char *name = save_input_locale; |
| 4952 | |
| 4953 | /* Find next 'U' or 'u' and look from there */ |
| 4954 | while ((name += strcspn(name, "Uu") + 1) |
| 4955 | <= save_input_locale + final_pos - 2) |
| 4956 | { |
| 4957 | if ( isALPHA_FOLD_NE(*name, 't') |
| 4958 | || isALPHA_FOLD_NE(*(name + 1), 'f')) |
| 4959 | { |
| 4960 | continue; |
| 4961 | } |
| 4962 | name += 2; |
| 4963 | if (*(name) == '-') { |
| 4964 | if ((name > save_input_locale + final_pos - 1)) { |
| 4965 | break; |
| 4966 | } |
| 4967 | name++; |
| 4968 | } |
| 4969 | if (*(name) == '8') { |
| 4970 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4971 | "Locale %s ends with UTF-8 in name\n", |
| 4972 | save_input_locale)); |
| 4973 | is_utf8 = TRUE; |
| 4974 | goto finish_and_return; |
| 4975 | } |
| 4976 | } |
| 4977 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4978 | "Locale %s doesn't end with UTF-8 in name\n", |
| 4979 | save_input_locale)); |
| 4980 | } |
| 4981 | |
| 4982 | # ifdef WIN32 |
| 4983 | |
| 4984 | /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */ |
| 4985 | if (memENDs(save_input_locale, final_pos, "65001")) { |
| 4986 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 4987 | "Locale %s ends with 65001 in name, is UTF-8 locale\n", |
| 4988 | save_input_locale)); |
| 4989 | is_utf8 = TRUE; |
| 4990 | goto finish_and_return; |
| 4991 | } |
| 4992 | |
| 4993 | # endif |
| 4994 | } |
| 4995 | # endif |
| 4996 | |
| 4997 | /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But |
| 4998 | * since we are about to return FALSE anyway, there is no point in doing |
| 4999 | * this extra work */ |
| 5000 | |
| 5001 | # if 0 |
| 5002 | if (instr(save_input_locale, "8859")) { |
| 5003 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 5004 | "Locale %s has 8859 in name, not UTF-8 locale\n", |
| 5005 | save_input_locale)); |
| 5006 | is_utf8 = FALSE; |
| 5007 | goto finish_and_return; |
| 5008 | } |
| 5009 | # endif |
| 5010 | |
| 5011 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 5012 | "Assuming locale %s is not a UTF-8 locale\n", |
| 5013 | save_input_locale)); |
| 5014 | is_utf8 = FALSE; |
| 5015 | |
| 5016 | # endif /* the code that is compiled when no modern LC_CTYPE */ |
| 5017 | |
| 5018 | finish_and_return: |
| 5019 | |
| 5020 | /* Cache this result so we don't have to go through all this next time. */ |
| 5021 | utf8ness_cache_size = sizeof(PL_locale_utf8ness) |
| 5022 | - (utf8ness_cache - PL_locale_utf8ness); |
| 5023 | |
| 5024 | /* But we can't save it if it is too large for the total space available */ |
| 5025 | if (LIKELY(input_name_len_with_overhead < utf8ness_cache_size)) { |
| 5026 | Size_t utf8ness_cache_len = strlen(utf8ness_cache); |
| 5027 | |
| 5028 | /* Here it can fit, but we may need to clear out the oldest cached |
| 5029 | * result(s) to do so. Check */ |
| 5030 | if (utf8ness_cache_len + input_name_len_with_overhead |
| 5031 | >= utf8ness_cache_size) |
| 5032 | { |
| 5033 | /* Here we have to clear something out to make room for this. |
| 5034 | * Start looking at the rightmost place where it could fit and find |
| 5035 | * the beginning of the entry that extends past that. */ |
| 5036 | char * cutoff = (char *) my_memrchr(utf8ness_cache, |
| 5037 | UTF8NESS_SEP[0], |
| 5038 | utf8ness_cache_size |
| 5039 | - input_name_len_with_overhead); |
| 5040 | |
| 5041 | assert(cutoff); |
| 5042 | assert(cutoff >= utf8ness_cache); |
| 5043 | |
| 5044 | /* This and all subsequent entries must be removed */ |
| 5045 | *cutoff = '\0'; |
| 5046 | utf8ness_cache_len = strlen(utf8ness_cache); |
| 5047 | } |
| 5048 | |
| 5049 | /* Make space for the new entry */ |
| 5050 | Move(utf8ness_cache, |
| 5051 | utf8ness_cache + input_name_len_with_overhead, |
| 5052 | utf8ness_cache_len + 1 /* Incl. trailing NUL */, char); |
| 5053 | |
| 5054 | /* And insert it */ |
| 5055 | Copy(delimited, utf8ness_cache, input_name_len_with_overhead - 1, char); |
| 5056 | utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0'; |
| 5057 | |
| 5058 | if ((PL_locale_utf8ness[strlen(PL_locale_utf8ness)-1] & ~1) != '0') { |
| 5059 | Perl_croak(aTHX_ |
| 5060 | "panic: %s: %d: Corrupt utf8ness_cache=%s\nlen=%zu," |
| 5061 | " inserted_name=%s, its_len=%zu\n", |
| 5062 | __FILE__, __LINE__, |
| 5063 | PL_locale_utf8ness, strlen(PL_locale_utf8ness), |
| 5064 | delimited, input_name_len_with_overhead); |
| 5065 | } |
| 5066 | } |
| 5067 | |
| 5068 | # ifdef DEBUGGING |
| 5069 | |
| 5070 | if (DEBUG_Lv_TEST) { |
| 5071 | const char * s = PL_locale_utf8ness; |
| 5072 | |
| 5073 | /* Audit the structure */ |
| 5074 | while (s < PL_locale_utf8ness + strlen(PL_locale_utf8ness)) { |
| 5075 | const char *e; |
| 5076 | |
| 5077 | if (*s != UTF8NESS_SEP[0]) { |
| 5078 | Perl_croak(aTHX_ |
| 5079 | "panic: %s: %d: Corrupt utf8ness_cache: missing" |
| 5080 | " separator %.*s<-- HERE %s\n", |
| 5081 | __FILE__, __LINE__, |
| 5082 | (int) (s - PL_locale_utf8ness), PL_locale_utf8ness, |
| 5083 | s); |
| 5084 | } |
| 5085 | s++; |
| 5086 | e = strchr(s, UTF8NESS_PREFIX[0]); |
| 5087 | if (! e) { |
| 5088 | e = PL_locale_utf8ness + strlen(PL_locale_utf8ness); |
| 5089 | Perl_croak(aTHX_ |
| 5090 | "panic: %s: %d: Corrupt utf8ness_cache: missing" |
| 5091 | " separator %.*s<-- HERE %s\n", |
| 5092 | __FILE__, __LINE__, |
| 5093 | (int) (e - PL_locale_utf8ness), PL_locale_utf8ness, |
| 5094 | e); |
| 5095 | } |
| 5096 | e++; |
| 5097 | if (*e != '0' && *e != '1') { |
| 5098 | Perl_croak(aTHX_ |
| 5099 | "panic: %s: %d: Corrupt utf8ness_cache: utf8ness" |
| 5100 | " must be [01] %.*s<-- HERE %s\n", |
| 5101 | __FILE__, __LINE__, |
| 5102 | (int) (e + 1 - PL_locale_utf8ness), |
| 5103 | PL_locale_utf8ness, e + 1); |
| 5104 | } |
| 5105 | if (ninstr(PL_locale_utf8ness, s, s-1, e)) { |
| 5106 | Perl_croak(aTHX_ |
| 5107 | "panic: %s: %d: Corrupt utf8ness_cache: entry" |
| 5108 | " has duplicate %.*s<-- HERE %s\n", |
| 5109 | __FILE__, __LINE__, |
| 5110 | (int) (e - PL_locale_utf8ness), PL_locale_utf8ness, |
| 5111 | e); |
| 5112 | } |
| 5113 | s = e + 1; |
| 5114 | } |
| 5115 | } |
| 5116 | |
| 5117 | if (DEBUG_Lv_TEST || debug_initialization) { |
| 5118 | |
| 5119 | PerlIO_printf(Perl_debug_log, |
| 5120 | "PL_locale_utf8ness is now %s; returning %d\n", |
| 5121 | PL_locale_utf8ness, is_utf8); |
| 5122 | } |
| 5123 | |
| 5124 | # endif |
| 5125 | |
| 5126 | /* free only when not using the buffer */ |
| 5127 | if ( delimited != buffer ) Safefree(delimited); |
| 5128 | Safefree(save_input_locale); |
| 5129 | return is_utf8; |
| 5130 | } |
| 5131 | |
| 5132 | #endif |
| 5133 | |
| 5134 | bool |
| 5135 | Perl__is_in_locale_category(pTHX_ const bool compiling, const int category) |
| 5136 | { |
| 5137 | dVAR; |
| 5138 | /* Internal function which returns if we are in the scope of a pragma that |
| 5139 | * enables the locale category 'category'. 'compiling' should indicate if |
| 5140 | * this is during the compilation phase (TRUE) or not (FALSE). */ |
| 5141 | |
| 5142 | const COP * const cop = (compiling) ? &PL_compiling : PL_curcop; |
| 5143 | |
| 5144 | SV *these_categories = cop_hints_fetch_pvs(cop, "locale", 0); |
| 5145 | if (! these_categories || these_categories == &PL_sv_placeholder) { |
| 5146 | return FALSE; |
| 5147 | } |
| 5148 | |
| 5149 | /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get |
| 5150 | * a valid unsigned */ |
| 5151 | assert(category >= -1); |
| 5152 | return cBOOL(SvUV(these_categories) & (1U << (category + 1))); |
| 5153 | } |
| 5154 | |
| 5155 | char * |
| 5156 | Perl_my_strerror(pTHX_ const int errnum) |
| 5157 | { |
| 5158 | /* Returns a mortalized copy of the text of the error message associated |
| 5159 | * with 'errnum'. It uses the current locale's text unless the platform |
| 5160 | * doesn't have the LC_MESSAGES category or we are not being called from |
| 5161 | * within the scope of 'use locale'. In the former case, it uses whatever |
| 5162 | * strerror returns; in the latter case it uses the text from the C locale. |
| 5163 | * |
| 5164 | * The function just calls strerror(), but temporarily switches, if needed, |
| 5165 | * to the C locale */ |
| 5166 | |
| 5167 | char *errstr; |
| 5168 | dVAR; |
| 5169 | |
| 5170 | #ifndef USE_LOCALE_MESSAGES |
| 5171 | |
| 5172 | /* If platform doesn't have messages category, we don't do any switching to |
| 5173 | * the C locale; we just use whatever strerror() returns */ |
| 5174 | |
| 5175 | errstr = savepv(Strerror(errnum)); |
| 5176 | |
| 5177 | #else /* Has locale messages */ |
| 5178 | |
| 5179 | const bool within_locale_scope = IN_LC(LC_MESSAGES); |
| 5180 | |
| 5181 | # ifndef USE_ITHREADS |
| 5182 | |
| 5183 | /* This function is trivial without threads. */ |
| 5184 | if (within_locale_scope) { |
| 5185 | errstr = savepv(strerror(errnum)); |
| 5186 | } |
| 5187 | else { |
| 5188 | const char * save_locale = savepv(do_setlocale_c(LC_MESSAGES, NULL)); |
| 5189 | |
| 5190 | do_setlocale_c(LC_MESSAGES, "C"); |
| 5191 | errstr = savepv(strerror(errnum)); |
| 5192 | do_setlocale_c(LC_MESSAGES, save_locale); |
| 5193 | Safefree(save_locale); |
| 5194 | } |
| 5195 | |
| 5196 | # elif defined(HAS_POSIX_2008_LOCALE) \ |
| 5197 | && defined(HAS_STRERROR_L) \ |
| 5198 | && defined(HAS_DUPLOCALE) |
| 5199 | |
| 5200 | /* This function is also trivial if we don't have to worry about thread |
| 5201 | * safety and have strerror_l(), as it handles the switch of locales so we |
| 5202 | * don't have to deal with that. We don't have to worry about thread |
| 5203 | * safety if strerror_r() is also available. Both it and strerror_l() are |
| 5204 | * thread-safe. Plain strerror() isn't thread safe. But on threaded |
| 5205 | * builds when strerror_r() is available, the apparent call to strerror() |
| 5206 | * below is actually a macro that behind-the-scenes calls strerror_r(). */ |
| 5207 | |
| 5208 | # ifdef HAS_STRERROR_R |
| 5209 | |
| 5210 | if (within_locale_scope) { |
| 5211 | errstr = savepv(strerror(errnum)); |
| 5212 | } |
| 5213 | else { |
| 5214 | errstr = savepv(strerror_l(errnum, PL_C_locale_obj)); |
| 5215 | } |
| 5216 | |
| 5217 | # else |
| 5218 | |
| 5219 | /* Here we have strerror_l(), but not strerror_r() and we are on a |
| 5220 | * threaded-build. We use strerror_l() for everything, constructing a |
| 5221 | * locale to pass to it if necessary */ |
| 5222 | |
| 5223 | bool do_free = FALSE; |
| 5224 | locale_t locale_to_use; |
| 5225 | |
| 5226 | if (within_locale_scope) { |
| 5227 | locale_to_use = uselocale((locale_t) 0); |
| 5228 | if (locale_to_use == LC_GLOBAL_LOCALE) { |
| 5229 | locale_to_use = duplocale(LC_GLOBAL_LOCALE); |
| 5230 | do_free = TRUE; |
| 5231 | } |
| 5232 | } |
| 5233 | else { /* Use C locale if not within 'use locale' scope */ |
| 5234 | locale_to_use = PL_C_locale_obj; |
| 5235 | } |
| 5236 | |
| 5237 | errstr = savepv(strerror_l(errnum, locale_to_use)); |
| 5238 | |
| 5239 | if (do_free) { |
| 5240 | freelocale(locale_to_use); |
| 5241 | } |
| 5242 | |
| 5243 | # endif |
| 5244 | # else /* Doesn't have strerror_l() */ |
| 5245 | |
| 5246 | const char * save_locale = NULL; |
| 5247 | bool locale_is_C = FALSE; |
| 5248 | |
| 5249 | /* We have a critical section to prevent another thread from executing this |
| 5250 | * same code at the same time. (On thread-safe perls, the LOCK is a |
| 5251 | * no-op.) Since this is the only place in core that changes LC_MESSAGES |
| 5252 | * (unless the user has called setlocale(), this works to prevent races. */ |
| 5253 | LOCALE_LOCK; |
| 5254 | |
| 5255 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 5256 | "my_strerror called with errnum %d\n", errnum)); |
| 5257 | if (! within_locale_scope) { |
| 5258 | save_locale = do_setlocale_c(LC_MESSAGES, NULL); |
| 5259 | if (! save_locale) { |
| 5260 | Perl_croak(aTHX_ |
| 5261 | "panic: %s: %d: Could not find current LC_MESSAGES locale," |
| 5262 | " errno=%d\n", __FILE__, __LINE__, errno); |
| 5263 | } |
| 5264 | else { |
| 5265 | locale_is_C = isNAME_C_OR_POSIX(save_locale); |
| 5266 | |
| 5267 | /* Switch to the C locale if not already in it */ |
| 5268 | if (! locale_is_C) { |
| 5269 | |
| 5270 | /* The setlocale() just below likely will zap 'save_locale', so |
| 5271 | * create a copy. */ |
| 5272 | save_locale = savepv(save_locale); |
| 5273 | do_setlocale_c(LC_MESSAGES, "C"); |
| 5274 | } |
| 5275 | } |
| 5276 | } /* end of ! within_locale_scope */ |
| 5277 | else { |
| 5278 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s: %d: WITHIN locale scope\n", |
| 5279 | __FILE__, __LINE__)); |
| 5280 | } |
| 5281 | |
| 5282 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 5283 | "Any locale change has been done; about to call Strerror\n")); |
| 5284 | errstr = savepv(Strerror(errnum)); |
| 5285 | |
| 5286 | if (! within_locale_scope) { |
| 5287 | if (save_locale && ! locale_is_C) { |
| 5288 | if (! do_setlocale_c(LC_MESSAGES, save_locale)) { |
| 5289 | Perl_croak(aTHX_ |
| 5290 | "panic: %s: %d: setlocale restore failed, errno=%d\n", |
| 5291 | __FILE__, __LINE__, errno); |
| 5292 | } |
| 5293 | Safefree(save_locale); |
| 5294 | } |
| 5295 | } |
| 5296 | |
| 5297 | LOCALE_UNLOCK; |
| 5298 | |
| 5299 | # endif /* End of doesn't have strerror_l */ |
| 5300 | # ifdef DEBUGGING |
| 5301 | |
| 5302 | if (DEBUG_Lv_TEST) { |
| 5303 | PerlIO_printf(Perl_debug_log, "Strerror returned; saving a copy: '"); |
| 5304 | print_bytes_for_locale(errstr, errstr + strlen(errstr), 0); |
| 5305 | PerlIO_printf(Perl_debug_log, "'\n"); |
| 5306 | } |
| 5307 | |
| 5308 | # endif |
| 5309 | #endif /* End of does have locale messages */ |
| 5310 | |
| 5311 | SAVEFREEPV(errstr); |
| 5312 | return errstr; |
| 5313 | } |
| 5314 | |
| 5315 | /* |
| 5316 | |
| 5317 | =for apidoc switch_to_global_locale |
| 5318 | |
| 5319 | On systems without locale support, or on typical single-threaded builds, or on |
| 5320 | platforms that do not support per-thread locale operations, this function does |
| 5321 | nothing. On such systems that do have locale support, only a locale global to |
| 5322 | the whole program is available. |
| 5323 | |
| 5324 | On multi-threaded builds on systems that do have per-thread locale operations, |
| 5325 | this function converts the thread it is running in to use the global locale. |
| 5326 | This is for code that has not yet or cannot be updated to handle multi-threaded |
| 5327 | locale operation. As long as only a single thread is so-converted, everything |
| 5328 | works fine, as all the other threads continue to ignore the global one, so only |
| 5329 | this thread looks at it. |
| 5330 | |
| 5331 | However, on Windows systems this isn't quite true prior to Visual Studio 15, |
| 5332 | at which point Microsoft fixed a bug. A race can occur if you use the |
| 5333 | following operations on earlier Windows platforms: |
| 5334 | |
| 5335 | =over |
| 5336 | |
| 5337 | =item L<POSIX::localeconv|POSIX/localeconv> |
| 5338 | |
| 5339 | =item L<I18N::Langinfo>, items C<CRNCYSTR> and C<THOUSEP> |
| 5340 | |
| 5341 | =item L<perlapi/Perl_langinfo>, items C<CRNCYSTR> and C<THOUSEP> |
| 5342 | |
| 5343 | =back |
| 5344 | |
| 5345 | The first item is not fixable (except by upgrading to a later Visual Studio |
| 5346 | release), but it would be possible to work around the latter two items by using |
| 5347 | the Windows API functions C<GetNumberFormat> and C<GetCurrencyFormat>; patches |
| 5348 | welcome. |
| 5349 | |
| 5350 | Without this function call, threads that use the L<C<setlocale(3)>> system |
| 5351 | function will not work properly, as all the locale-sensitive functions will |
| 5352 | look at the per-thread locale, and C<setlocale> will have no effect on this |
| 5353 | thread. |
| 5354 | |
| 5355 | Perl code should convert to either call |
| 5356 | L<C<Perl_setlocale>|perlapi/Perl_setlocale> (which is a drop-in for the system |
| 5357 | C<setlocale>) or use the methods given in L<perlcall> to call |
| 5358 | L<C<POSIX::setlocale>|POSIX/setlocale>. Either one will transparently properly |
| 5359 | handle all cases of single- vs multi-thread, POSIX 2008-supported or not. |
| 5360 | |
| 5361 | Non-Perl libraries, such as C<gtk>, that call the system C<setlocale> can |
| 5362 | continue to work if this function is called before transferring control to the |
| 5363 | library. |
| 5364 | |
| 5365 | Upon return from the code that needs to use the global locale, |
| 5366 | L<C<sync_locale()>|perlapi/sync_locale> should be called to restore the safe |
| 5367 | multi-thread operation. |
| 5368 | |
| 5369 | =cut |
| 5370 | */ |
| 5371 | |
| 5372 | void |
| 5373 | Perl_switch_to_global_locale() |
| 5374 | { |
| 5375 | |
| 5376 | #ifdef USE_THREAD_SAFE_LOCALE |
| 5377 | # ifdef WIN32 |
| 5378 | |
| 5379 | _configthreadlocale(_DISABLE_PER_THREAD_LOCALE); |
| 5380 | |
| 5381 | # else |
| 5382 | # ifdef HAS_QUERYLOCALE |
| 5383 | |
| 5384 | setlocale(LC_ALL, querylocale(LC_ALL_MASK, uselocale((locale_t) 0))); |
| 5385 | |
| 5386 | # else |
| 5387 | |
| 5388 | { |
| 5389 | unsigned int i; |
| 5390 | |
| 5391 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 5392 | setlocale(categories[i], do_setlocale_r(categories[i], NULL)); |
| 5393 | } |
| 5394 | } |
| 5395 | |
| 5396 | # endif |
| 5397 | |
| 5398 | uselocale(LC_GLOBAL_LOCALE); |
| 5399 | |
| 5400 | # endif |
| 5401 | #endif |
| 5402 | |
| 5403 | } |
| 5404 | |
| 5405 | /* |
| 5406 | |
| 5407 | =for apidoc sync_locale |
| 5408 | |
| 5409 | L<C<Perl_setlocale>|perlapi/Perl_setlocale> can be used at any time to query or |
| 5410 | change the locale (though changing the locale is antisocial and dangerous on |
| 5411 | multi-threaded systems that don't have multi-thread safe locale operations. |
| 5412 | (See L<perllocale/Multi-threaded operation>). Using the system |
| 5413 | L<C<setlocale(3)>> should be avoided. Nevertheless, certain non-Perl libraries |
| 5414 | called from XS, such as C<Gtk> do so, and this can't be changed. When the |
| 5415 | locale is changed by XS code that didn't use |
| 5416 | L<C<Perl_setlocale>|perlapi/Perl_setlocale>, Perl needs to be told that the |
| 5417 | locale has changed. Use this function to do so, before returning to Perl. |
| 5418 | |
| 5419 | The return value is a boolean: TRUE if the global locale at the time of call |
| 5420 | was in effect; and FALSE if a per-thread locale was in effect. This can be |
| 5421 | used by the caller that needs to restore things as-they-were to decide whether |
| 5422 | or not to call |
| 5423 | L<C<Perl_switch_to_global_locale>|perlapi/switch_to_global_locale>. |
| 5424 | |
| 5425 | =cut |
| 5426 | */ |
| 5427 | |
| 5428 | bool |
| 5429 | Perl_sync_locale() |
| 5430 | { |
| 5431 | |
| 5432 | #ifndef USE_LOCALE |
| 5433 | |
| 5434 | return TRUE; |
| 5435 | |
| 5436 | #else |
| 5437 | |
| 5438 | const char * newlocale; |
| 5439 | dTHX; |
| 5440 | |
| 5441 | # ifdef USE_POSIX_2008_LOCALE |
| 5442 | |
| 5443 | bool was_in_global_locale = FALSE; |
| 5444 | locale_t cur_obj = uselocale((locale_t) 0); |
| 5445 | |
| 5446 | /* On Windows, unless the foreign code has turned off the thread-safe |
| 5447 | * locale setting, any plain setlocale() will have affected what we see, so |
| 5448 | * no need to worry. Otherwise, If the foreign code has done a plain |
| 5449 | * setlocale(), it will only affect the global locale on POSIX systems, but |
| 5450 | * will affect the */ |
| 5451 | if (cur_obj == LC_GLOBAL_LOCALE) { |
| 5452 | |
| 5453 | # ifdef HAS_QUERY_LOCALE |
| 5454 | |
| 5455 | do_setlocale_c(LC_ALL, setlocale(LC_ALL, NULL)); |
| 5456 | |
| 5457 | # else |
| 5458 | |
| 5459 | unsigned int i; |
| 5460 | |
| 5461 | /* We can't trust that we can read the LC_ALL format on the |
| 5462 | * platform, so do them individually */ |
| 5463 | for (i = 0; i < LC_ALL_INDEX; i++) { |
| 5464 | do_setlocale_r(categories[i], setlocale(categories[i], NULL)); |
| 5465 | } |
| 5466 | |
| 5467 | # endif |
| 5468 | |
| 5469 | was_in_global_locale = TRUE; |
| 5470 | } |
| 5471 | |
| 5472 | # else |
| 5473 | |
| 5474 | bool was_in_global_locale = TRUE; |
| 5475 | |
| 5476 | # endif |
| 5477 | # ifdef USE_LOCALE_CTYPE |
| 5478 | |
| 5479 | newlocale = savepv(do_setlocale_c(LC_CTYPE, NULL)); |
| 5480 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 5481 | "%s:%d: %s\n", __FILE__, __LINE__, |
| 5482 | setlocale_debug_string(LC_CTYPE, NULL, newlocale))); |
| 5483 | new_ctype(newlocale); |
| 5484 | Safefree(newlocale); |
| 5485 | |
| 5486 | # endif /* USE_LOCALE_CTYPE */ |
| 5487 | # ifdef USE_LOCALE_COLLATE |
| 5488 | |
| 5489 | newlocale = savepv(do_setlocale_c(LC_COLLATE, NULL)); |
| 5490 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 5491 | "%s:%d: %s\n", __FILE__, __LINE__, |
| 5492 | setlocale_debug_string(LC_COLLATE, NULL, newlocale))); |
| 5493 | new_collate(newlocale); |
| 5494 | Safefree(newlocale); |
| 5495 | |
| 5496 | # endif |
| 5497 | # ifdef USE_LOCALE_NUMERIC |
| 5498 | |
| 5499 | newlocale = savepv(do_setlocale_c(LC_NUMERIC, NULL)); |
| 5500 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
| 5501 | "%s:%d: %s\n", __FILE__, __LINE__, |
| 5502 | setlocale_debug_string(LC_NUMERIC, NULL, newlocale))); |
| 5503 | new_numeric(newlocale); |
| 5504 | Safefree(newlocale); |
| 5505 | |
| 5506 | # endif /* USE_LOCALE_NUMERIC */ |
| 5507 | |
| 5508 | return was_in_global_locale; |
| 5509 | |
| 5510 | #endif |
| 5511 | |
| 5512 | } |
| 5513 | |
| 5514 | #if defined(DEBUGGING) && defined(USE_LOCALE) |
| 5515 | |
| 5516 | STATIC char * |
| 5517 | S_setlocale_debug_string(const int category, /* category number, |
| 5518 | like LC_ALL */ |
| 5519 | const char* const locale, /* locale name */ |
| 5520 | |
| 5521 | /* return value from setlocale() when attempting to |
| 5522 | * set 'category' to 'locale' */ |
| 5523 | const char* const retval) |
| 5524 | { |
| 5525 | /* Returns a pointer to a NUL-terminated string in static storage with |
| 5526 | * added text about the info passed in. This is not thread safe and will |
| 5527 | * be overwritten by the next call, so this should be used just to |
| 5528 | * formulate a string to immediately print or savepv() on. */ |
| 5529 | |
| 5530 | /* initialise to a non-null value to keep it out of BSS and so keep |
| 5531 | * -DPERL_GLOBAL_STRUCT_PRIVATE happy */ |
| 5532 | static char ret[256] = "If you can read this, thank your buggy C" |
| 5533 | " library strlcpy(), and change your hints file" |
| 5534 | " to undef it"; |
| 5535 | |
| 5536 | my_strlcpy(ret, "setlocale(", sizeof(ret)); |
| 5537 | my_strlcat(ret, category_name(category), sizeof(ret)); |
| 5538 | my_strlcat(ret, ", ", sizeof(ret)); |
| 5539 | |
| 5540 | if (locale) { |
| 5541 | my_strlcat(ret, "\"", sizeof(ret)); |
| 5542 | my_strlcat(ret, locale, sizeof(ret)); |
| 5543 | my_strlcat(ret, "\"", sizeof(ret)); |
| 5544 | } |
| 5545 | else { |
| 5546 | my_strlcat(ret, "NULL", sizeof(ret)); |
| 5547 | } |
| 5548 | |
| 5549 | my_strlcat(ret, ") returned ", sizeof(ret)); |
| 5550 | |
| 5551 | if (retval) { |
| 5552 | my_strlcat(ret, "\"", sizeof(ret)); |
| 5553 | my_strlcat(ret, retval, sizeof(ret)); |
| 5554 | my_strlcat(ret, "\"", sizeof(ret)); |
| 5555 | } |
| 5556 | else { |
| 5557 | my_strlcat(ret, "NULL", sizeof(ret)); |
| 5558 | } |
| 5559 | |
| 5560 | assert(strlen(ret) < sizeof(ret)); |
| 5561 | |
| 5562 | return ret; |
| 5563 | } |
| 5564 | |
| 5565 | #endif |
| 5566 | |
| 5567 | void |
| 5568 | Perl_thread_locale_init() |
| 5569 | { |
| 5570 | /* Called from a thread on startup*/ |
| 5571 | |
| 5572 | #ifdef USE_THREAD_SAFE_LOCALE |
| 5573 | |
| 5574 | dTHX_DEBUGGING; |
| 5575 | |
| 5576 | /* C starts the new thread in the global C locale. If we are thread-safe, |
| 5577 | * we want to not be in the global locale */ |
| 5578 | |
| 5579 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
| 5580 | "%s:%d: new thread, initial locale is %s; calling setlocale\n", |
| 5581 | __FILE__, __LINE__, setlocale(LC_ALL, NULL))); |
| 5582 | |
| 5583 | # ifdef WIN32 |
| 5584 | |
| 5585 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); |
| 5586 | |
| 5587 | # else |
| 5588 | |
| 5589 | Perl_setlocale(LC_ALL, "C"); |
| 5590 | |
| 5591 | # endif |
| 5592 | #endif |
| 5593 | |
| 5594 | } |
| 5595 | |
| 5596 | void |
| 5597 | Perl_thread_locale_term() |
| 5598 | { |
| 5599 | /* Called from a thread as it gets ready to terminate */ |
| 5600 | |
| 5601 | #ifdef USE_THREAD_SAFE_LOCALE |
| 5602 | |
| 5603 | /* C starts the new thread in the global C locale. If we are thread-safe, |
| 5604 | * we want to not be in the global locale */ |
| 5605 | |
| 5606 | # ifndef WIN32 |
| 5607 | |
| 5608 | { /* Free up */ |
| 5609 | dVAR; |
| 5610 | locale_t cur_obj = uselocale(LC_GLOBAL_LOCALE); |
| 5611 | if (cur_obj != LC_GLOBAL_LOCALE && cur_obj != PL_C_locale_obj) { |
| 5612 | freelocale(cur_obj); |
| 5613 | } |
| 5614 | } |
| 5615 | |
| 5616 | # endif |
| 5617 | #endif |
| 5618 | |
| 5619 | } |
| 5620 | |
| 5621 | /* |
| 5622 | * ex: set ts=8 sts=4 sw=4 et: |
| 5623 | */ |