Commit | Line | Data |
---|---|---|
98994639 HS |
1 | /* locale.c |
2 | * | |
1129b882 NC |
3 | * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, |
4 | * 2002, 2003, 2005, 2006, 2007, 2008 by Larry Wall and others | |
98994639 HS |
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 | /* | |
4ac71550 | 12 | * A Elbereth Gilthoniel, |
cdad3b53 | 13 | * silivren penna míriel |
4ac71550 | 14 | * o menel aglar elenath! |
cdad3b53 | 15 | * Na-chaered palan-díriel |
4ac71550 TC |
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"] | |
98994639 HS |
21 | */ |
22 | ||
166f8a29 DM |
23 | /* utility functions for handling locale-specific stuff like what |
24 | * character represents the decimal point. | |
0d071d52 | 25 | * |
7d4bcc4a KW |
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 | |
0d071d52 KW |
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 | |
a9ad02a8 KW |
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'. | |
e9bc6d6b KW |
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. | |
166f8a29 DM |
44 | */ |
45 | ||
98994639 HS |
46 | #include "EXTERN.h" |
47 | #define PERL_IN_LOCALE_C | |
f7416781 | 48 | #include "perl_langinfo.h" |
98994639 HS |
49 | #include "perl.h" |
50 | ||
a4af207c JH |
51 | #include "reentr.h" |
52 | ||
0dec74cd KW |
53 | #ifdef I_WCHAR |
54 | # include <wchar.h> | |
55 | #endif | |
5b64f24c KW |
56 | #ifdef I_WCTYPE |
57 | # include <wctype.h> | |
58 | #endif | |
0dec74cd | 59 | |
2fcc0ca9 KW |
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 */ | |
8c3a0f6c | 63 | #if ! defined(DEBUGGING) |
5a4b0634 KW |
64 | # define debug_initialization 0 |
65 | # define DEBUG_INITIALIZATION_set(v) | |
66 | #else | |
2fcc0ca9 | 67 | static bool debug_initialization = FALSE; |
5a4b0634 | 68 | # define DEBUG_INITIALIZATION_set(v) (debug_initialization = v) |
2fcc0ca9 KW |
69 | #endif |
70 | ||
48015184 KW |
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 | ||
291a84fb KW |
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 */ | |
ff1b739b KW |
80 | #define STRLENs(s) (sizeof("" s "") - 1) |
81 | ||
63e5b0d7 KW |
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 | ||
8ef6e574 KW |
97 | #ifdef USE_LOCALE |
98 | ||
47280b20 KW |
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: | |
98994639 | 106 | * |
47280b20 | 107 | * \vC\a0\vPOSIX\a0\vam_ET\a0\vaf_ZA.utf8\a1\ven_US.UTF-8\a1\0 |
98994639 | 108 | * |
47280b20 KW |
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 | ||
98994639 HS |
122 | STATIC char * |
123 | S_stdize_locale(pTHX_ char *locs) | |
124 | { | |
47280b20 KW |
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 | ||
7452cf6a | 139 | const char * const s = strchr(locs, '='); |
98994639 HS |
140 | bool okay = TRUE; |
141 | ||
7918f24d NC |
142 | PERL_ARGS_ASSERT_STDIZE_LOCALE; |
143 | ||
8772537c AL |
144 | if (s) { |
145 | const char * const t = strchr(s, '.'); | |
98994639 | 146 | okay = FALSE; |
8772537c AL |
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; | |
98994639 HS |
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 | ||
e5f10d49 KW |
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 | |
9821811f KW |
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 | |
36dbc955 KW |
203 | # ifdef USE_LOCALE_SYNTAX |
204 | LC_SYNTAX, | |
205 | # endif | |
206 | # ifdef USE_LOCALE_TOD | |
207 | LC_TOD, | |
208 | # endif | |
e5f10d49 KW |
209 | # ifdef LC_ALL |
210 | LC_ALL, | |
211 | # endif | |
212 | -1 /* Placeholder because C doesn't allow a | |
213 | trailing comma, and it would get complicated | |
214 | with all the #ifdef's */ | |
215 | }; | |
216 | ||
217 | /* The top-most real element is LC_ALL */ | |
218 | ||
4ef8bdf9 | 219 | const char * const category_names[] = { |
e5f10d49 KW |
220 | |
221 | # ifdef USE_LOCALE_NUMERIC | |
222 | "LC_NUMERIC", | |
223 | # endif | |
224 | # ifdef USE_LOCALE_CTYPE | |
225 | "LC_CTYPE", | |
226 | # endif | |
227 | # ifdef USE_LOCALE_COLLATE | |
228 | "LC_COLLATE", | |
229 | # endif | |
230 | # ifdef USE_LOCALE_TIME | |
231 | "LC_TIME", | |
232 | # endif | |
233 | # ifdef USE_LOCALE_MESSAGES | |
234 | "LC_MESSAGES", | |
235 | # endif | |
236 | # ifdef USE_LOCALE_MONETARY | |
237 | "LC_MONETARY", | |
238 | # endif | |
9821811f KW |
239 | # ifdef USE_LOCALE_ADDRESS |
240 | "LC_ADDRESS", | |
241 | # endif | |
242 | # ifdef USE_LOCALE_IDENTIFICATION | |
243 | "LC_IDENTIFICATION", | |
244 | # endif | |
245 | # ifdef USE_LOCALE_MEASUREMENT | |
246 | "LC_MEASUREMENT", | |
247 | # endif | |
248 | # ifdef USE_LOCALE_PAPER | |
249 | "LC_PAPER", | |
250 | # endif | |
251 | # ifdef USE_LOCALE_TELEPHONE | |
252 | "LC_TELEPHONE", | |
253 | # endif | |
36dbc955 KW |
254 | # ifdef USE_LOCALE_SYNTAX |
255 | "LC_SYNTAX", | |
256 | # endif | |
257 | # ifdef USE_LOCALE_TOD | |
258 | "LC_TOD", | |
259 | # endif | |
e5f10d49 KW |
260 | # ifdef LC_ALL |
261 | "LC_ALL", | |
262 | # endif | |
263 | NULL /* Placeholder */ | |
264 | }; | |
265 | ||
266 | # ifdef LC_ALL | |
267 | ||
268 | /* On systems with LC_ALL, it is kept in the highest index position. (-2 | |
269 | * to account for the final unused placeholder element.) */ | |
270 | # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 2) | |
271 | ||
272 | # else | |
273 | ||
274 | /* On systems without LC_ALL, we pretend it is there, one beyond the real | |
275 | * top element, hence in the unused placeholder element. */ | |
276 | # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 1) | |
277 | ||
278 | # endif | |
279 | ||
280 | /* Pretending there is an LC_ALL element just above allows us to avoid most | |
281 | * special cases. Most loops through these arrays in the code below are | |
282 | * written like 'for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++)'. They will work | |
283 | * on either type of system. But the code must be written to not access the | |
948523db KW |
284 | * element at 'LC_ALL_INDEX' except on platforms that have it. This can be |
285 | * checked for at compile time by using the #define LC_ALL_INDEX which is only | |
286 | * defined if we do have LC_ALL. */ | |
e5f10d49 | 287 | |
b09aaf40 KW |
288 | STATIC const char * |
289 | S_category_name(const int category) | |
290 | { | |
291 | unsigned int i; | |
292 | ||
293 | #ifdef LC_ALL | |
294 | ||
295 | if (category == LC_ALL) { | |
296 | return "LC_ALL"; | |
297 | } | |
298 | ||
299 | #endif | |
300 | ||
301 | for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) { | |
302 | if (category == categories[i]) { | |
303 | return category_names[i]; | |
304 | } | |
305 | } | |
306 | ||
307 | { | |
308 | const char suffix[] = " (unknown)"; | |
309 | int temp = category; | |
310 | Size_t length = sizeof(suffix) + 1; | |
311 | char * unknown; | |
312 | dTHX; | |
313 | ||
314 | if (temp < 0) { | |
315 | length++; | |
316 | temp = - temp; | |
317 | } | |
318 | ||
319 | /* Calculate the number of digits */ | |
320 | while (temp >= 10) { | |
321 | temp /= 10; | |
322 | length++; | |
323 | } | |
324 | ||
325 | Newx(unknown, length, char); | |
326 | my_snprintf(unknown, length, "%d%s", category, suffix); | |
327 | SAVEFREEPV(unknown); | |
328 | return unknown; | |
329 | } | |
330 | } | |
331 | ||
948523db KW |
332 | /* Now create LC_foo_INDEX #defines for just those categories on this system */ |
333 | # ifdef USE_LOCALE_NUMERIC | |
334 | # define LC_NUMERIC_INDEX 0 | |
335 | # define _DUMMY_NUMERIC LC_NUMERIC_INDEX | |
336 | # else | |
337 | # define _DUMMY_NUMERIC -1 | |
338 | # endif | |
339 | # ifdef USE_LOCALE_CTYPE | |
340 | # define LC_CTYPE_INDEX _DUMMY_NUMERIC + 1 | |
341 | # define _DUMMY_CTYPE LC_CTYPE_INDEX | |
342 | # else | |
343 | # define _DUMMY_CTYPE _DUMMY_NUMERIC | |
344 | # endif | |
345 | # ifdef USE_LOCALE_COLLATE | |
346 | # define LC_COLLATE_INDEX _DUMMY_CTYPE + 1 | |
347 | # define _DUMMY_COLLATE LC_COLLATE_INDEX | |
348 | # else | |
8203b3c3 | 349 | # define _DUMMY_COLLATE _DUMMY_CTYPE |
948523db KW |
350 | # endif |
351 | # ifdef USE_LOCALE_TIME | |
352 | # define LC_TIME_INDEX _DUMMY_COLLATE + 1 | |
353 | # define _DUMMY_TIME LC_TIME_INDEX | |
354 | # else | |
355 | # define _DUMMY_TIME _DUMMY_COLLATE | |
356 | # endif | |
357 | # ifdef USE_LOCALE_MESSAGES | |
358 | # define LC_MESSAGES_INDEX _DUMMY_TIME + 1 | |
359 | # define _DUMMY_MESSAGES LC_MESSAGES_INDEX | |
360 | # else | |
361 | # define _DUMMY_MESSAGES _DUMMY_TIME | |
362 | # endif | |
363 | # ifdef USE_LOCALE_MONETARY | |
364 | # define LC_MONETARY_INDEX _DUMMY_MESSAGES + 1 | |
365 | # define _DUMMY_MONETARY LC_MONETARY_INDEX | |
366 | # else | |
367 | # define _DUMMY_MONETARY _DUMMY_MESSAGES | |
368 | # endif | |
9821811f KW |
369 | # ifdef USE_LOCALE_ADDRESS |
370 | # define LC_ADDRESS_INDEX _DUMMY_MONETARY + 1 | |
371 | # define _DUMMY_ADDRESS LC_ADDRESS_INDEX | |
372 | # else | |
373 | # define _DUMMY_ADDRESS _DUMMY_MONETARY | |
374 | # endif | |
375 | # ifdef USE_LOCALE_IDENTIFICATION | |
376 | # define LC_IDENTIFICATION_INDEX _DUMMY_ADDRESS + 1 | |
377 | # define _DUMMY_IDENTIFICATION LC_IDENTIFICATION_INDEX | |
378 | # else | |
379 | # define _DUMMY_IDENTIFICATION _DUMMY_ADDRESS | |
380 | # endif | |
381 | # ifdef USE_LOCALE_MEASUREMENT | |
382 | # define LC_MEASUREMENT_INDEX _DUMMY_IDENTIFICATION + 1 | |
383 | # define _DUMMY_MEASUREMENT LC_MEASUREMENT_INDEX | |
384 | # else | |
385 | # define _DUMMY_MEASUREMENT _DUMMY_IDENTIFICATION | |
386 | # endif | |
387 | # ifdef USE_LOCALE_PAPER | |
388 | # define LC_PAPER_INDEX _DUMMY_MEASUREMENT + 1 | |
389 | # define _DUMMY_PAPER LC_PAPER_INDEX | |
390 | # else | |
391 | # define _DUMMY_PAPER _DUMMY_MEASUREMENT | |
392 | # endif | |
393 | # ifdef USE_LOCALE_TELEPHONE | |
394 | # define LC_TELEPHONE_INDEX _DUMMY_PAPER + 1 | |
395 | # define _DUMMY_TELEPHONE LC_TELEPHONE_INDEX | |
396 | # else | |
397 | # define _DUMMY_TELEPHONE _DUMMY_PAPER | |
398 | # endif | |
36dbc955 KW |
399 | # ifdef USE_LOCALE_SYNTAX |
400 | # define LC_SYNTAX_INDEX _DUMMY_TELEPHONE + 1 | |
401 | # define _DUMMY_SYNTAX LC_SYNTAX_INDEX | |
402 | # else | |
403 | # define _DUMMY_SYNTAX _DUMMY_TELEPHONE | |
404 | # endif | |
405 | # ifdef USE_LOCALE_TOD | |
406 | # define LC_TOD_INDEX _DUMMY_SYNTAX + 1 | |
407 | # define _DUMMY_TOD LC_TOD_INDEX | |
408 | # else | |
409 | # define _DUMMY_TOD _DUMMY_SYNTAX | |
410 | # endif | |
948523db | 411 | # ifdef LC_ALL |
36dbc955 | 412 | # define LC_ALL_INDEX _DUMMY_TOD + 1 |
948523db KW |
413 | # endif |
414 | #endif /* ifdef USE_LOCALE */ | |
8ef6e574 | 415 | |
d2b24094 | 416 | /* Windows requres a customized base-level setlocale() */ |
e9bc6d6b KW |
417 | #ifdef WIN32 |
418 | # define my_setlocale(cat, locale) win32_setlocale(cat, locale) | |
419 | #else | |
420 | # define my_setlocale(cat, locale) setlocale(cat, locale) | |
421 | #endif | |
422 | ||
423 | #ifndef USE_POSIX_2008_LOCALE | |
424 | ||
425 | /* "do_setlocale_c" is intended to be called when the category is a constant | |
426 | * known at compile time; "do_setlocale_r", not known until run time */ | |
427 | # define do_setlocale_c(cat, locale) my_setlocale(cat, locale) | |
428 | # define do_setlocale_r(cat, locale) my_setlocale(cat, locale) | |
45cef8fb | 429 | # define FIX_GLIBC_LC_MESSAGES_BUG(i) |
e9bc6d6b KW |
430 | |
431 | #else /* Below uses POSIX 2008 */ | |
432 | ||
433 | /* We emulate setlocale with our own function. LC_foo is not valid for the | |
434 | * POSIX 2008 functions. Instead LC_foo_MASK is used, which we use an array | |
435 | * lookup to convert to. At compile time we have defined LC_foo_INDEX as the | |
436 | * proper offset into the array 'category_masks[]'. At runtime, we have to | |
437 | * search through the array (as the actual numbers may not be small contiguous | |
438 | * positive integers which would lend themselves to array lookup). */ | |
439 | # define do_setlocale_c(cat, locale) \ | |
440 | emulate_setlocale(cat, locale, cat ## _INDEX, TRUE) | |
441 | # define do_setlocale_r(cat, locale) emulate_setlocale(cat, locale, 0, FALSE) | |
442 | ||
45cef8fb KW |
443 | # if ! defined(__GLIBC__) || ! defined(USE_LOCALE_MESSAGES) |
444 | ||
445 | # define FIX_GLIBC_LC_MESSAGES_BUG(i) | |
446 | ||
447 | # else /* Invalidate glibc cache of loaded translations, see [perl #134264] */ | |
448 | ||
449 | # include <libintl.h> | |
450 | # define FIX_GLIBC_LC_MESSAGES_BUG(i) \ | |
451 | STMT_START { \ | |
452 | if ((i) == LC_MESSAGES_INDEX) { \ | |
453 | textdomain(textdomain(NULL)); \ | |
454 | } \ | |
455 | } STMT_END | |
456 | ||
457 | # endif | |
458 | ||
e9bc6d6b KW |
459 | /* A third array, parallel to the ones above to map from category to its |
460 | * equivalent mask */ | |
461 | const int category_masks[] = { | |
462 | # ifdef USE_LOCALE_NUMERIC | |
463 | LC_NUMERIC_MASK, | |
464 | # endif | |
465 | # ifdef USE_LOCALE_CTYPE | |
466 | LC_CTYPE_MASK, | |
467 | # endif | |
468 | # ifdef USE_LOCALE_COLLATE | |
469 | LC_COLLATE_MASK, | |
470 | # endif | |
471 | # ifdef USE_LOCALE_TIME | |
472 | LC_TIME_MASK, | |
473 | # endif | |
474 | # ifdef USE_LOCALE_MESSAGES | |
475 | LC_MESSAGES_MASK, | |
476 | # endif | |
477 | # ifdef USE_LOCALE_MONETARY | |
478 | LC_MONETARY_MASK, | |
479 | # endif | |
480 | # ifdef USE_LOCALE_ADDRESS | |
481 | LC_ADDRESS_MASK, | |
482 | # endif | |
483 | # ifdef USE_LOCALE_IDENTIFICATION | |
484 | LC_IDENTIFICATION_MASK, | |
485 | # endif | |
486 | # ifdef USE_LOCALE_MEASUREMENT | |
487 | LC_MEASUREMENT_MASK, | |
488 | # endif | |
489 | # ifdef USE_LOCALE_PAPER | |
490 | LC_PAPER_MASK, | |
491 | # endif | |
492 | # ifdef USE_LOCALE_TELEPHONE | |
493 | LC_TELEPHONE_MASK, | |
494 | # endif | |
36dbc955 KW |
495 | # ifdef USE_LOCALE_SYNTAX |
496 | LC_SYNTAX_MASK, | |
497 | # endif | |
498 | # ifdef USE_LOCALE_TOD | |
499 | LC_TOD_MASK, | |
500 | # endif | |
e9bc6d6b KW |
501 | /* LC_ALL can't be turned off by a Configure |
502 | * option, and in Posix 2008, should always be | |
503 | * here, so compile it in unconditionally. | |
504 | * This could catch some glitches at compile | |
505 | * time */ | |
506 | LC_ALL_MASK | |
507 | }; | |
508 | ||
509 | STATIC const char * | |
510 | S_emulate_setlocale(const int category, | |
511 | const char * locale, | |
512 | unsigned int index, | |
513 | const bool is_index_valid | |
514 | ) | |
515 | { | |
516 | /* This function effectively performs a setlocale() on just the current | |
517 | * thread; thus it is thread-safe. It does this by using the POSIX 2008 | |
518 | * locale functions to emulate the behavior of setlocale(). Similar to | |
519 | * regular setlocale(), the return from this function points to memory that | |
520 | * can be overwritten by other system calls, so needs to be copied | |
521 | * immediately if you need to retain it. The difference here is that | |
522 | * system calls besides another setlocale() can overwrite it. | |
523 | * | |
524 | * By doing this, most locale-sensitive functions become thread-safe. The | |
525 | * exceptions are mostly those that return a pointer to static memory. | |
526 | * | |
527 | * This function takes the same parameters, 'category' and 'locale', that | |
528 | * the regular setlocale() function does, but it also takes two additional | |
529 | * ones. This is because the 2008 functions don't use a category; instead | |
530 | * they use a corresponding mask. Because this function operates in both | |
531 | * worlds, it may need one or the other or both. This function can | |
532 | * calculate the mask from the input category, but to avoid this | |
533 | * calculation, if the caller knows at compile time what the mask is, it | |
534 | * can pass it, setting 'is_index_valid' to TRUE; otherwise the mask | |
535 | * parameter is ignored. | |
536 | * | |
537 | * POSIX 2008, for some sick reason, chose not to provide a method to find | |
538 | * the category name of a locale. Some vendors have created a | |
539 | * querylocale() function to do just that. This function is a lot simpler | |
540 | * to implement on systems that have this. Otherwise, we have to keep | |
541 | * track of what the locale has been set to, so that we can return its | |
542 | * name to emulate setlocale(). It's also possible for C code in some | |
543 | * library to change the locale without us knowing it, though as of | |
544 | * September 2017, there are no occurrences in CPAN of uselocale(). Some | |
545 | * libraries do use setlocale(), but that changes the global locale, and | |
546 | * threads using per-thread locales will just ignore those changes. | |
547 | * Another problem is that without querylocale(), we have to guess at what | |
548 | * was meant by setting a locale of "". We handle this by not actually | |
549 | * ever setting to "" (unless querylocale exists), but to emulate what we | |
550 | * think should happen for "". | |
551 | */ | |
552 | ||
553 | int mask; | |
554 | locale_t old_obj; | |
555 | locale_t new_obj; | |
556 | dTHX; | |
557 | ||
558 | # ifdef DEBUGGING | |
559 | ||
560 | if (DEBUG_Lv_TEST || debug_initialization) { | |
561 | 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); | |
562 | } | |
563 | ||
564 | # endif | |
565 | ||
566 | /* If the input mask might be incorrect, calculate the correct one */ | |
567 | if (! is_index_valid) { | |
568 | unsigned int i; | |
569 | ||
570 | # ifdef DEBUGGING | |
571 | ||
572 | if (DEBUG_Lv_TEST || debug_initialization) { | |
573 | PerlIO_printf(Perl_debug_log, "%s:%d: finding index of category %d (%s)\n", __FILE__, __LINE__, category, category_name(category)); | |
574 | } | |
575 | ||
576 | # endif | |
577 | ||
578 | for (i = 0; i <= LC_ALL_INDEX; i++) { | |
579 | if (category == categories[i]) { | |
580 | index = i; | |
581 | goto found_index; | |
582 | } | |
583 | } | |
584 | ||
585 | /* Here, we don't know about this category, so can't handle it. | |
586 | * Fallback to the early POSIX usages */ | |
587 | Perl_warner(aTHX_ packWARN(WARN_LOCALE), | |
588 | "Unknown locale category %d; can't set it to %s\n", | |
589 | category, locale); | |
590 | return NULL; | |
591 | ||
592 | found_index: ; | |
593 | ||
594 | # ifdef DEBUGGING | |
595 | ||
596 | if (DEBUG_Lv_TEST || debug_initialization) { | |
597 | PerlIO_printf(Perl_debug_log, "%s:%d: index is %d for %s\n", __FILE__, __LINE__, index, category_name(category)); | |
598 | } | |
599 | ||
600 | # endif | |
601 | ||
602 | } | |
603 | ||
604 | mask = category_masks[index]; | |
605 | ||
606 | # ifdef DEBUGGING | |
607 | ||
608 | if (DEBUG_Lv_TEST || debug_initialization) { | |
609 | PerlIO_printf(Perl_debug_log, "%s:%d: category name is %s; mask is 0x%x\n", __FILE__, __LINE__, category_names[index], mask); | |
610 | } | |
611 | ||
612 | # endif | |
613 | ||
614 | /* If just querying what the existing locale is ... */ | |
615 | if (locale == NULL) { | |
616 | locale_t cur_obj = uselocale((locale_t) 0); | |
617 | ||
618 | # ifdef DEBUGGING | |
619 | ||
620 | if (DEBUG_Lv_TEST || debug_initialization) { | |
621 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale querying %p\n", __FILE__, __LINE__, cur_obj); | |
622 | } | |
623 | ||
624 | # endif | |
625 | ||
626 | if (cur_obj == LC_GLOBAL_LOCALE) { | |
627 | return my_setlocale(category, NULL); | |
628 | } | |
629 | ||
630 | # ifdef HAS_QUERYLOCALE | |
631 | ||
632 | return (char *) querylocale(mask, cur_obj); | |
633 | ||
d2b24094 | 634 | # else |
ecdda939 KW |
635 | |
636 | /* If this assert fails, adjust the size of curlocales in intrpvar.h */ | |
637 | STATIC_ASSERT_STMT(C_ARRAY_LENGTH(PL_curlocales) > LC_ALL_INDEX); | |
638 | ||
f33fd0eb KW |
639 | # if defined(_NL_LOCALE_NAME) \ |
640 | && defined(DEBUGGING) \ | |
641 | /* On systems that accept any locale name, the real underlying \ | |
642 | * locale is often returned by this internal function, so we \ | |
643 | * can't use it */ \ | |
8e13243a | 644 | && ! defined(SETLOCALE_ACCEPTS_ANY_LOCALE_NAME) |
e9bc6d6b KW |
645 | { |
646 | /* Internal glibc for querylocale(), but doesn't handle | |
647 | * empty-string ("") locale properly; who knows what other | |
3a732259 | 648 | * glitches. Check for it now, under debug. */ |
e9bc6d6b KW |
649 | |
650 | char * temp_name = nl_langinfo_l(_NL_LOCALE_NAME(category), | |
651 | uselocale((locale_t) 0)); | |
652 | /* | |
653 | PerlIO_printf(Perl_debug_log, "%s:%d: temp_name=%s\n", __FILE__, __LINE__, temp_name ? temp_name : "NULL"); | |
654 | PerlIO_printf(Perl_debug_log, "%s:%d: index=%d\n", __FILE__, __LINE__, index); | |
655 | PerlIO_printf(Perl_debug_log, "%s:%d: PL_curlocales[index]=%s\n", __FILE__, __LINE__, PL_curlocales[index]); | |
656 | */ | |
657 | if (temp_name && PL_curlocales[index] && strNE(temp_name, "")) { | |
658 | if ( strNE(PL_curlocales[index], temp_name) | |
659 | && ! ( isNAME_C_OR_POSIX(temp_name) | |
660 | && isNAME_C_OR_POSIX(PL_curlocales[index]))) { | |
661 | ||
662 | # ifdef USE_C_BACKTRACE | |
663 | ||
664 | dump_c_backtrace(Perl_debug_log, 20, 1); | |
665 | ||
666 | # endif | |
667 | ||
668 | Perl_croak(aTHX_ "panic: Mismatch between what Perl thinks %s is" | |
669 | " (%s) and what internal glibc thinks" | |
670 | " (%s)\n", category_names[index], | |
671 | PL_curlocales[index], temp_name); | |
672 | } | |
673 | ||
674 | return temp_name; | |
675 | } | |
676 | } | |
677 | ||
678 | # endif | |
679 | ||
680 | /* Without querylocale(), we have to use our record-keeping we've | |
681 | * done. */ | |
682 | ||
683 | if (category != LC_ALL) { | |
684 | ||
685 | # ifdef DEBUGGING | |
686 | ||
687 | if (DEBUG_Lv_TEST || debug_initialization) { | |
688 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, PL_curlocales[index]); | |
689 | } | |
690 | ||
691 | # endif | |
692 | ||
693 | return PL_curlocales[index]; | |
694 | } | |
695 | else { /* For LC_ALL */ | |
696 | unsigned int i; | |
697 | Size_t names_len = 0; | |
698 | char * all_string; | |
7c93a471 | 699 | bool are_all_categories_the_same_locale = TRUE; |
e9bc6d6b KW |
700 | |
701 | /* If we have a valid LC_ALL value, just return it */ | |
702 | if (PL_curlocales[LC_ALL_INDEX]) { | |
703 | ||
704 | # ifdef DEBUGGING | |
705 | ||
706 | if (DEBUG_Lv_TEST || debug_initialization) { | |
707 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, PL_curlocales[LC_ALL_INDEX]); | |
708 | } | |
709 | ||
710 | # endif | |
711 | ||
712 | return PL_curlocales[LC_ALL_INDEX]; | |
713 | } | |
714 | ||
715 | /* Otherwise, we need to construct a string of name=value pairs. | |
716 | * We use the glibc syntax, like | |
717 | * LC_NUMERIC=C;LC_TIME=en_US.UTF-8;... | |
7c93a471 KW |
718 | * First calculate the needed size. Along the way, check if all |
719 | * the locale names are the same */ | |
e9bc6d6b KW |
720 | for (i = 0; i < LC_ALL_INDEX; i++) { |
721 | ||
722 | # ifdef DEBUGGING | |
723 | ||
724 | if (DEBUG_Lv_TEST || debug_initialization) { | |
725 | 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]); | |
726 | } | |
727 | ||
728 | # endif | |
729 | ||
730 | names_len += strlen(category_names[i]) | |
731 | + 1 /* '=' */ | |
732 | + strlen(PL_curlocales[i]) | |
733 | + 1; /* ';' */ | |
7c93a471 KW |
734 | |
735 | if (i > 0 && strNE(PL_curlocales[i], PL_curlocales[i-1])) { | |
736 | are_all_categories_the_same_locale = FALSE; | |
737 | } | |
738 | } | |
739 | ||
740 | /* If they are the same, we don't actually have to construct the | |
741 | * string; we just make the entry in LC_ALL_INDEX valid, and be | |
742 | * that single name */ | |
743 | if (are_all_categories_the_same_locale) { | |
744 | PL_curlocales[LC_ALL_INDEX] = savepv(PL_curlocales[0]); | |
745 | return PL_curlocales[LC_ALL_INDEX]; | |
e9bc6d6b | 746 | } |
7c93a471 | 747 | |
e9bc6d6b KW |
748 | names_len++; /* Trailing '\0' */ |
749 | SAVEFREEPV(Newx(all_string, names_len, char)); | |
750 | *all_string = '\0'; | |
751 | ||
752 | /* Then fill in the string */ | |
753 | for (i = 0; i < LC_ALL_INDEX; i++) { | |
754 | ||
755 | # ifdef DEBUGGING | |
756 | ||
757 | if (DEBUG_Lv_TEST || debug_initialization) { | |
758 | 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]); | |
759 | } | |
760 | ||
761 | # endif | |
762 | ||
763 | my_strlcat(all_string, category_names[i], names_len); | |
764 | my_strlcat(all_string, "=", names_len); | |
765 | my_strlcat(all_string, PL_curlocales[i], names_len); | |
766 | my_strlcat(all_string, ";", names_len); | |
767 | } | |
768 | ||
769 | # ifdef DEBUGGING | |
770 | ||
771 | if (DEBUG_L_TEST || debug_initialization) { | |
772 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale returning %s\n", __FILE__, __LINE__, all_string); | |
773 | } | |
774 | ||
775 | #endif | |
776 | ||
777 | return all_string; | |
778 | } | |
779 | ||
780 | # ifdef EINVAL | |
781 | ||
782 | SETERRNO(EINVAL, LIB_INVARG); | |
783 | ||
784 | # endif | |
785 | ||
786 | return NULL; | |
787 | ||
d2b24094 KW |
788 | # endif |
789 | ||
6aba5c5e | 790 | } /* End of this being setlocale(LC_foo, NULL) */ |
e9bc6d6b | 791 | |
f8115d60 | 792 | /* Here, we are switching locales. */ |
e9bc6d6b KW |
793 | |
794 | # ifndef HAS_QUERYLOCALE | |
795 | ||
796 | if (strEQ(locale, "")) { | |
797 | ||
798 | /* For non-querylocale() systems, we do the setting of "" ourselves to | |
799 | * be sure that we really know what's going on. We follow the Linux | |
800 | * documented behavior (but if that differs from the actual behavior, | |
801 | * this won't work exactly as the OS implements). We go out and | |
802 | * examine the environment based on our understanding of how the system | |
803 | * works, and use that to figure things out */ | |
804 | ||
805 | const char * const lc_all = PerlEnv_getenv("LC_ALL"); | |
806 | ||
807 | /* Use any "LC_ALL" environment variable, as it overrides everything | |
808 | * else. */ | |
809 | if (lc_all && strNE(lc_all, "")) { | |
810 | locale = lc_all; | |
811 | } | |
812 | else { | |
813 | ||
814 | /* Otherwise, we need to dig deeper. Unless overridden, the | |
815 | * default is the LANG environment variable; if it doesn't exist, | |
816 | * then "C" */ | |
817 | ||
818 | const char * default_name; | |
819 | ||
207cc8df | 820 | default_name = PerlEnv_getenv("LANG"); |
e9bc6d6b KW |
821 | |
822 | if (! default_name || strEQ(default_name, "")) { | |
929a7f8c | 823 | default_name = "C"; |
e9bc6d6b | 824 | } |
e9bc6d6b KW |
825 | |
826 | if (category != LC_ALL) { | |
827 | const char * const name = PerlEnv_getenv(category_names[index]); | |
828 | ||
829 | /* Here we are setting a single category. Assume will have the | |
830 | * default name */ | |
831 | locale = default_name; | |
832 | ||
833 | /* But then look for an overriding environment variable */ | |
834 | if (name && strNE(name, "")) { | |
835 | locale = name; | |
836 | } | |
837 | } | |
838 | else { | |
839 | bool did_override = FALSE; | |
840 | unsigned int i; | |
841 | ||
842 | /* Here, we are getting LC_ALL. Any categories that don't have | |
843 | * a corresponding environment variable set should be set to | |
844 | * LANG, or to "C" if there is no LANG. If no individual | |
845 | * categories differ from this, we can just set LC_ALL. This | |
846 | * is buggy on systems that have extra categories that we don't | |
847 | * know about. If there is an environment variable that sets | |
848 | * that category, we won't know to look for it, and so our use | |
849 | * of LANG or "C" improperly overrides it. On the other hand, | |
850 | * if we don't do what is done here, and there is no | |
851 | * environment variable, the category's locale should be set to | |
852 | * LANG or "C". So there is no good solution. khw thinks the | |
853 | * best is to look at systems to see what categories they have, | |
854 | * and include them, and then to assume that we know the | |
855 | * complete set */ | |
856 | ||
857 | for (i = 0; i < LC_ALL_INDEX; i++) { | |
858 | const char * const env_override | |
24f3e849 | 859 | = PerlEnv_getenv(category_names[i]); |
e9bc6d6b KW |
860 | const char * this_locale = ( env_override |
861 | && strNE(env_override, "")) | |
862 | ? env_override | |
863 | : default_name; | |
6435f98d KW |
864 | if (! emulate_setlocale(categories[i], this_locale, i, TRUE)) |
865 | { | |
6435f98d KW |
866 | return NULL; |
867 | } | |
e9bc6d6b KW |
868 | |
869 | if (strNE(this_locale, default_name)) { | |
870 | did_override = TRUE; | |
871 | } | |
e9bc6d6b KW |
872 | } |
873 | ||
874 | /* If all the categories are the same, we can set LC_ALL to | |
875 | * that */ | |
876 | if (! did_override) { | |
877 | locale = default_name; | |
878 | } | |
879 | else { | |
880 | ||
881 | /* Here, LC_ALL is no longer valid, as some individual | |
882 | * categories don't match it. We call ourselves | |
883 | * recursively, as that will execute the code that | |
884 | * generates the proper locale string for this situation. | |
885 | * We don't do the remainder of this function, as that is | |
886 | * to update our records, and we've just done that for the | |
887 | * individual categories in the loop above, and doing so | |
888 | * would cause LC_ALL to be done as well */ | |
889 | return emulate_setlocale(LC_ALL, NULL, LC_ALL_INDEX, TRUE); | |
890 | } | |
891 | } | |
892 | } | |
6aba5c5e | 893 | } /* End of this being setlocale(LC_foo, "") */ |
e9bc6d6b KW |
894 | else if (strchr(locale, ';')) { |
895 | ||
896 | /* LC_ALL may actually incude a conglomeration of various categories. | |
897 | * Without querylocale, this code uses the glibc (as of this writing) | |
898 | * syntax for representing that, but that is not a stable API, and | |
899 | * other platforms do it differently, so we have to handle all cases | |
900 | * ourselves */ | |
901 | ||
f0023d47 | 902 | unsigned int i; |
e9bc6d6b KW |
903 | const char * s = locale; |
904 | const char * e = locale + strlen(locale); | |
905 | const char * p = s; | |
906 | const char * category_end; | |
907 | const char * name_start; | |
908 | const char * name_end; | |
909 | ||
f0023d47 KW |
910 | /* If the string that gives what to set doesn't include all categories, |
911 | * the omitted ones get set to "C". To get this behavior, first set | |
912 | * all the individual categories to "C", and override the furnished | |
913 | * ones below */ | |
914 | for (i = 0; i < LC_ALL_INDEX; i++) { | |
915 | if (! emulate_setlocale(categories[i], "C", i, TRUE)) { | |
916 | return NULL; | |
917 | } | |
918 | } | |
919 | ||
e9bc6d6b | 920 | while (s < e) { |
e9bc6d6b KW |
921 | |
922 | /* Parse through the category */ | |
923 | while (isWORDCHAR(*p)) { | |
924 | p++; | |
925 | } | |
926 | category_end = p; | |
927 | ||
928 | if (*p++ != '=') { | |
929 | Perl_croak(aTHX_ | |
930 | "panic: %s: %d: Unexpected character in locale name '%02X", | |
931 | __FILE__, __LINE__, *(p-1)); | |
932 | } | |
933 | ||
934 | /* Parse through the locale name */ | |
935 | name_start = p; | |
bee74f4b KW |
936 | while (p < e && *p != ';') { |
937 | if (! isGRAPH(*p)) { | |
938 | Perl_croak(aTHX_ | |
939 | "panic: %s: %d: Unexpected character in locale name '%02X", | |
940 | __FILE__, __LINE__, *(p-1)); | |
941 | } | |
e9bc6d6b KW |
942 | p++; |
943 | } | |
944 | name_end = p; | |
945 | ||
bee74f4b KW |
946 | /* Space past the semi-colon */ |
947 | if (p < e) { | |
948 | p++; | |
e9bc6d6b KW |
949 | } |
950 | ||
951 | /* Find the index of the category name in our lists */ | |
952 | for (i = 0; i < LC_ALL_INDEX; i++) { | |
3c76a412 | 953 | char * individ_locale; |
e9bc6d6b KW |
954 | |
955 | /* Keep going if this isn't the index. The strnNE() avoids a | |
956 | * Perl_form(), but would fail if ever a category name could be | |
957 | * a substring of another one, like if there were a | |
958 | * "LC_TIME_DATE" */ | |
959 | if strnNE(s, category_names[i], category_end - s) { | |
960 | continue; | |
961 | } | |
962 | ||
963 | /* If this index is for the single category we're changing, we | |
964 | * have found the locale to set it to. */ | |
965 | if (category == categories[i]) { | |
966 | locale = Perl_form(aTHX_ "%.*s", | |
967 | (int) (name_end - name_start), | |
968 | name_start); | |
969 | goto ready_to_set; | |
970 | } | |
971 | ||
3c76a412 | 972 | assert(category == LC_ALL); |
bee74f4b KW |
973 | individ_locale = Perl_form(aTHX_ "%.*s", |
974 | (int) (name_end - name_start), name_start); | |
6435f98d KW |
975 | if (! emulate_setlocale(categories[i], individ_locale, i, TRUE)) |
976 | { | |
977 | return NULL; | |
978 | } | |
e9bc6d6b KW |
979 | } |
980 | ||
981 | s = p; | |
982 | } | |
983 | ||
984 | /* Here we have set all the individual categories by recursive calls. | |
985 | * These collectively should have fixed up LC_ALL, so can just query | |
986 | * what that now is */ | |
987 | assert(category == LC_ALL); | |
988 | ||
989 | return do_setlocale_c(LC_ALL, NULL); | |
6aba5c5e KW |
990 | } /* End of this being setlocale(LC_ALL, |
991 | "LC_CTYPE=foo;LC_NUMERIC=bar;...") */ | |
e9bc6d6b KW |
992 | |
993 | ready_to_set: ; | |
994 | ||
f8115d60 KW |
995 | /* Here at the end of having to deal with the absence of querylocale(). |
996 | * Some cases have already been fully handled by recursive calls to this | |
997 | * function. But at this point, we haven't dealt with those, but are now | |
998 | * prepared to, knowing what the locale name to set this category to is. | |
999 | * This would have come for free if this system had had querylocale() */ | |
1000 | ||
e9bc6d6b KW |
1001 | # endif /* end of ! querylocale */ |
1002 | ||
f8115d60 KW |
1003 | assert(PL_C_locale_obj); |
1004 | ||
1005 | /* Switching locales generally entails freeing the current one's space (at | |
1006 | * the C library's discretion). We need to stop using that locale before | |
1007 | * the switch. So switch to a known locale object that we don't otherwise | |
1008 | * mess with. This returns the locale object in effect at the time of the | |
1009 | * switch. */ | |
1010 | old_obj = uselocale(PL_C_locale_obj); | |
1011 | ||
1012 | # ifdef DEBUGGING | |
1013 | ||
1014 | if (DEBUG_Lv_TEST || debug_initialization) { | |
1015 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale was using %p\n", __FILE__, __LINE__, old_obj); | |
1016 | } | |
1017 | ||
1018 | # endif | |
1019 | ||
1020 | if (! old_obj) { | |
1021 | ||
1022 | # ifdef DEBUGGING | |
1023 | ||
1024 | if (DEBUG_L_TEST || debug_initialization) { | |
1025 | dSAVE_ERRNO; | |
1026 | PerlIO_printf(Perl_debug_log, "%s:%d: emulate_setlocale switching to C failed: %d\n", __FILE__, __LINE__, GET_ERRNO); | |
1027 | RESTORE_ERRNO; | |
1028 | } | |
1029 | ||
1030 | # endif | |
1031 | ||
1032 | return NULL; | |
1033 | } | |
1034 | ||
1035 | # ifdef DEBUGGING | |
1036 | ||
1037 | if (DEBUG_Lv_TEST || debug_initialization) { | |
b22e9937 KW |
1038 | PerlIO_printf(Perl_debug_log, |
1039 | "%s:%d: emulate_setlocale now using %p\n", | |
1040 | __FILE__, __LINE__, PL_C_locale_obj); | |
f8115d60 KW |
1041 | } |
1042 | ||
1043 | # endif | |
1044 | ||
6aba5c5e KW |
1045 | /* If this call is to switch to the LC_ALL C locale, it already exists, and |
1046 | * in fact, we already have switched to it (in preparation for what | |
1047 | * normally is to come). But since we're already there, continue to use | |
70bd6bc8 KW |
1048 | * it instead of trying to create a new locale */ |
1049 | if (mask == LC_ALL_MASK && isNAME_C_OR_POSIX(locale)) { | |
1050 | ||
1051 | # ifdef DEBUGGING | |
1052 | ||
1053 | if (DEBUG_Lv_TEST || debug_initialization) { | |
1054 | PerlIO_printf(Perl_debug_log, | |
1055 | "%s:%d: will stay in C object\n", __FILE__, __LINE__); | |
1056 | } | |
1057 | ||
1058 | # endif | |
1059 | ||
1060 | new_obj = PL_C_locale_obj; | |
1061 | ||
1062 | /* We already had switched to the C locale in preparation for freeing | |
b22e9937 | 1063 | * 'old_obj' */ |
70bd6bc8 KW |
1064 | if (old_obj != LC_GLOBAL_LOCALE && old_obj != PL_C_locale_obj) { |
1065 | freelocale(old_obj); | |
1066 | } | |
1067 | } | |
1068 | else { | |
b22e9937 KW |
1069 | /* If we weren't in a thread safe locale, set so that newlocale() below |
1070 | * which uses 'old_obj', uses an empty one. Same for our reserved C | |
1071 | * object. The latter is defensive coding, so that, even if there is | |
1072 | * some bug, we will never end up trying to modify either of these, as | |
1073 | * if passed to newlocale(), they can be. */ | |
1074 | if (old_obj == LC_GLOBAL_LOCALE || old_obj == PL_C_locale_obj) { | |
1075 | old_obj = (locale_t) 0; | |
1076 | } | |
f8115d60 | 1077 | |
b22e9937 KW |
1078 | /* Ready to create a new locale by modification of the exising one */ |
1079 | new_obj = newlocale(mask, locale, old_obj); | |
e9bc6d6b | 1080 | |
b22e9937 KW |
1081 | if (! new_obj) { |
1082 | dSAVE_ERRNO; | |
e9bc6d6b KW |
1083 | |
1084 | # ifdef DEBUGGING | |
1085 | ||
b22e9937 KW |
1086 | if (DEBUG_L_TEST || debug_initialization) { |
1087 | PerlIO_printf(Perl_debug_log, | |
1088 | "%s:%d: emulate_setlocale creating new object" | |
1089 | " failed: %d\n", __FILE__, __LINE__, GET_ERRNO); | |
1090 | } | |
e9bc6d6b KW |
1091 | |
1092 | # endif | |
1093 | ||
b22e9937 | 1094 | if (! uselocale(old_obj)) { |
e9bc6d6b KW |
1095 | |
1096 | # ifdef DEBUGGING | |
1097 | ||
b22e9937 KW |
1098 | if (DEBUG_L_TEST || debug_initialization) { |
1099 | PerlIO_printf(Perl_debug_log, | |
1100 | "%s:%d: switching back failed: %d\n", | |
1101 | __FILE__, __LINE__, GET_ERRNO); | |
1102 | } | |
e9bc6d6b KW |
1103 | |
1104 | # endif | |
1105 | ||
b22e9937 KW |
1106 | } |
1107 | RESTORE_ERRNO; | |
1108 | return NULL; | |
e9bc6d6b | 1109 | } |
e9bc6d6b KW |
1110 | |
1111 | # ifdef DEBUGGING | |
1112 | ||
b22e9937 KW |
1113 | if (DEBUG_Lv_TEST || debug_initialization) { |
1114 | PerlIO_printf(Perl_debug_log, | |
1115 | "%s:%d: emulate_setlocale created %p", | |
1116 | __FILE__, __LINE__, new_obj); | |
1117 | if (old_obj) { | |
1118 | PerlIO_printf(Perl_debug_log, | |
1119 | "; should have freed %p", old_obj); | |
1120 | } | |
1121 | PerlIO_printf(Perl_debug_log, "\n"); | |
19ee3daf | 1122 | } |
e9bc6d6b KW |
1123 | |
1124 | # endif | |
1125 | ||
b22e9937 KW |
1126 | /* And switch into it */ |
1127 | if (! uselocale(new_obj)) { | |
1128 | dSAVE_ERRNO; | |
e9bc6d6b KW |
1129 | |
1130 | # ifdef DEBUGGING | |
1131 | ||
b22e9937 KW |
1132 | if (DEBUG_L_TEST || debug_initialization) { |
1133 | PerlIO_printf(Perl_debug_log, | |
1134 | "%s:%d: emulate_setlocale switching to new object" | |
1135 | " failed\n", __FILE__, __LINE__); | |
1136 | } | |
e9bc6d6b KW |
1137 | |
1138 | # endif | |
1139 | ||
b22e9937 | 1140 | if (! uselocale(old_obj)) { |
e9bc6d6b KW |
1141 | |
1142 | # ifdef DEBUGGING | |
1143 | ||
b22e9937 KW |
1144 | if (DEBUG_L_TEST || debug_initialization) { |
1145 | PerlIO_printf(Perl_debug_log, | |
1146 | "%s:%d: switching back failed: %d\n", | |
1147 | __FILE__, __LINE__, GET_ERRNO); | |
1148 | } | |
e9bc6d6b KW |
1149 | |
1150 | # endif | |
1151 | ||
b22e9937 KW |
1152 | } |
1153 | freelocale(new_obj); | |
1154 | RESTORE_ERRNO; | |
1155 | return NULL; | |
e9bc6d6b | 1156 | } |
70bd6bc8 | 1157 | } |
e9bc6d6b KW |
1158 | |
1159 | # ifdef DEBUGGING | |
1160 | ||
1161 | if (DEBUG_Lv_TEST || debug_initialization) { | |
b22e9937 KW |
1162 | PerlIO_printf(Perl_debug_log, |
1163 | "%s:%d: emulate_setlocale now using %p\n", | |
1164 | __FILE__, __LINE__, new_obj); | |
e9bc6d6b KW |
1165 | } |
1166 | ||
1167 | # endif | |
1168 | ||
1169 | /* We are done, except for updating our records (if the system doesn't keep | |
1170 | * them) and in the case of locale "", we don't actually know what the | |
1171 | * locale that got switched to is, as it came from the environment. So | |
1172 | * have to find it */ | |
1173 | ||
1174 | # ifdef HAS_QUERYLOCALE | |
1175 | ||
1176 | if (strEQ(locale, "")) { | |
1177 | locale = querylocale(mask, new_obj); | |
1178 | } | |
1179 | ||
1180 | # else | |
1181 | ||
1182 | /* Here, 'locale' is the return value */ | |
1183 | ||
1184 | /* Without querylocale(), we have to update our records */ | |
1185 | ||
1186 | if (category == LC_ALL) { | |
1187 | unsigned int i; | |
1188 | ||
1189 | /* For LC_ALL, we change all individual categories to correspond */ | |
1190 | /* PL_curlocales is a parallel array, so has same | |
1191 | * length as 'categories' */ | |
1192 | for (i = 0; i <= LC_ALL_INDEX; i++) { | |
1193 | Safefree(PL_curlocales[i]); | |
1194 | PL_curlocales[i] = savepv(locale); | |
1195 | } | |
45cef8fb KW |
1196 | |
1197 | FIX_GLIBC_LC_MESSAGES_BUG(LC_MESSAGES_INDEX); | |
e9bc6d6b KW |
1198 | } |
1199 | else { | |
1200 | ||
1201 | /* For a single category, if it's not the same as the one in LC_ALL, we | |
1202 | * nullify LC_ALL */ | |
1203 | ||
1204 | if (PL_curlocales[LC_ALL_INDEX] && strNE(PL_curlocales[LC_ALL_INDEX], locale)) { | |
1205 | Safefree(PL_curlocales[LC_ALL_INDEX]); | |
1206 | PL_curlocales[LC_ALL_INDEX] = NULL; | |
1207 | } | |
1208 | ||
1209 | /* Then update the category's record */ | |
1210 | Safefree(PL_curlocales[index]); | |
1211 | PL_curlocales[index] = savepv(locale); | |
45cef8fb KW |
1212 | |
1213 | FIX_GLIBC_LC_MESSAGES_BUG(index); | |
e9bc6d6b KW |
1214 | } |
1215 | ||
1216 | # endif | |
1217 | ||
1218 | return locale; | |
1219 | } | |
1220 | ||
1221 | #endif /* USE_POSIX_2008_LOCALE */ | |
1222 | ||
1223 | #if 0 /* Code that was to emulate thread-safe locales on platforms that | |
1224 | didn't natively support them */ | |
1225 | ||
1226 | /* The way this would work is that we would keep a per-thread list of the | |
1227 | * correct locale for that thread. Any operation that was locale-sensitive | |
1228 | * would have to be changed so that it would look like this: | |
1229 | * | |
1230 | * LOCALE_LOCK; | |
1231 | * setlocale to the correct locale for this operation | |
1232 | * do operation | |
1233 | * LOCALE_UNLOCK | |
1234 | * | |
1235 | * This leaves the global locale in the most recently used operation's, but it | |
1236 | * was locked long enough to get the result. If that result is static, it | |
1237 | * needs to be copied before the unlock. | |
1238 | * | |
1239 | * Macros could be written like SETUP_LOCALE_DEPENDENT_OP(category) that did | |
1240 | * the setup, but are no-ops when not needed, and similarly, | |
1241 | * END_LOCALE_DEPENDENT_OP for the tear-down | |
1242 | * | |
1243 | * But every call to a locale-sensitive function would have to be changed, and | |
1244 | * if a module didn't cooperate by using the mutex, things would break. | |
1245 | * | |
1246 | * This code was abandoned before being completed or tested, and is left as-is | |
1247 | */ | |
1248 | ||
1249 | # define do_setlocale_c(cat, locale) locking_setlocale(cat, locale, cat ## _INDEX, TRUE) | |
1250 | # define do_setlocale_r(cat, locale) locking_setlocale(cat, locale, 0, FALSE) | |
1251 | ||
1252 | STATIC char * | |
1253 | S_locking_setlocale(pTHX_ | |
1254 | const int category, | |
1255 | const char * locale, | |
1256 | int index, | |
1257 | const bool is_index_valid | |
1258 | ) | |
1259 | { | |
1260 | /* This function kind of performs a setlocale() on just the current thread; | |
1261 | * thus it is kind of thread-safe. It does this by keeping a thread-level | |
1262 | * array of the current locales for each category. Every time a locale is | |
1263 | * switched to, it does the switch globally, but updates the thread's | |
1264 | * array. A query as to what the current locale is just returns the | |
1265 | * appropriate element from the array, and doesn't actually call the system | |
1266 | * setlocale(). The saving into the array is done in an uninterruptible | |
1267 | * section of code, so is unaffected by whatever any other threads might be | |
1268 | * doing. | |
1269 | * | |
1270 | * All locale-sensitive operations must work by first starting a critical | |
1271 | * section, then switching to the thread's locale as kept by this function, | |
1272 | * and then doing the operation, then ending the critical section. Thus, | |
1273 | * each gets done in the appropriate locale. simulating thread-safety. | |
1274 | * | |
1275 | * This function takes the same parameters, 'category' and 'locale', that | |
1276 | * the regular setlocale() function does, but it also takes two additional | |
1277 | * ones. This is because as described earlier. If we know on input the | |
1278 | * index corresponding to the category into the array where we store the | |
1279 | * current locales, we don't have to calculate it. If the caller knows at | |
06cc5386 | 1280 | * compile time what the index is, it can pass it, setting |
e9bc6d6b KW |
1281 | * 'is_index_valid' to TRUE; otherwise the index parameter is ignored. |
1282 | * | |
1283 | */ | |
1284 | ||
1285 | /* If the input index might be incorrect, calculate the correct one */ | |
1286 | if (! is_index_valid) { | |
1287 | unsigned int i; | |
1288 | ||
1289 | if (DEBUG_Lv_TEST || debug_initialization) { | |
1290 | PerlIO_printf(Perl_debug_log, "%s:%d: converting category %d to index\n", __FILE__, __LINE__, category); | |
1291 | } | |
1292 | ||
1293 | for (i = 0; i <= LC_ALL_INDEX; i++) { | |
1294 | if (category == categories[i]) { | |
1295 | index = i; | |
1296 | goto found_index; | |
1297 | } | |
1298 | } | |
1299 | ||
1300 | /* Here, we don't know about this category, so can't handle it. | |
1301 | * XXX best we can do is to unsafely set this | |
1302 | * XXX warning */ | |
1303 | ||
1304 | return my_setlocale(category, locale); | |
1305 | ||
1306 | found_index: ; | |
1307 | ||
1308 | if (DEBUG_Lv_TEST || debug_initialization) { | |
1309 | PerlIO_printf(Perl_debug_log, "%s:%d: index is 0x%x\n", __FILE__, __LINE__, index); | |
1310 | } | |
1311 | } | |
1312 | ||
1313 | /* For a query, just return what's in our records */ | |
1314 | if (new_locale == NULL) { | |
1315 | return curlocales[index]; | |
1316 | } | |
1317 | ||
1318 | ||
1319 | /* Otherwise, we need to do the switch, and save the result, all in a | |
1320 | * critical section */ | |
1321 | ||
1322 | Safefree(curlocales[[index]]); | |
1323 | ||
1324 | /* It might be that this is called from an already-locked section of code. | |
1325 | * We would have to detect and skip the LOCK/UNLOCK if so */ | |
1326 | LOCALE_LOCK; | |
1327 | ||
1328 | curlocales[index] = savepv(my_setlocale(category, new_locale)); | |
1329 | ||
1330 | if (strEQ(new_locale, "")) { | |
1331 | ||
1332 | #ifdef LC_ALL | |
1333 | ||
1334 | /* The locale values come from the environment, and may not all be the | |
1335 | * same, so for LC_ALL, we have to update all the others, while the | |
1336 | * mutex is still locked */ | |
1337 | ||
1338 | if (category == LC_ALL) { | |
1339 | unsigned int i; | |
1340 | for (i = 0; i < LC_ALL_INDEX) { | |
1341 | curlocales[i] = my_setlocale(categories[i], NULL); | |
1342 | } | |
1343 | } | |
1344 | } | |
1345 | ||
1346 | #endif | |
1347 | ||
1348 | LOCALE_UNLOCK; | |
1349 | ||
1350 | return curlocales[index]; | |
1351 | } | |
1352 | ||
1353 | #endif | |
d36adde0 | 1354 | #ifdef USE_LOCALE |
837ce802 | 1355 | |
a4f00dcc | 1356 | STATIC void |
86799d2d | 1357 | S_set_numeric_radix(pTHX_ const bool use_locale) |
98994639 | 1358 | { |
86799d2d KW |
1359 | /* If 'use_locale' is FALSE, set to use a dot for the radix character. If |
1360 | * TRUE, use the radix character derived from the current locale */ | |
7d4bcc4a | 1361 | |
86799d2d KW |
1362 | #if defined(USE_LOCALE_NUMERIC) && ( defined(HAS_LOCALECONV) \ |
1363 | || defined(HAS_NL_LANGINFO)) | |
98994639 | 1364 | |
87f8e8e7 | 1365 | const char * radix = (use_locale) |
4e6826bf | 1366 | ? my_nl_langinfo(RADIXCHAR, FALSE) |
87f8e8e7 KW |
1367 | /* FALSE => already in dest locale */ |
1368 | : "."; | |
2213a3be | 1369 | |
87f8e8e7 | 1370 | sv_setpv(PL_numeric_radix_sv, radix); |
86799d2d | 1371 | |
87f8e8e7 KW |
1372 | /* If this is valid UTF-8 that isn't totally ASCII, and we are in |
1373 | * a UTF-8 locale, then mark the radix as being in UTF-8 */ | |
1374 | if (is_utf8_non_invariant_string((U8 *) SvPVX(PL_numeric_radix_sv), | |
7a393424 | 1375 | SvCUR(PL_numeric_radix_sv)) |
87f8e8e7 KW |
1376 | && _is_cur_LC_category_utf8(LC_NUMERIC)) |
1377 | { | |
1378 | SvUTF8_on(PL_numeric_radix_sv); | |
1379 | } | |
86799d2d KW |
1380 | |
1381 | # ifdef DEBUGGING | |
7d4bcc4a | 1382 | |
2fcc0ca9 KW |
1383 | if (DEBUG_L_TEST || debug_initialization) { |
1384 | PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n", | |
3ca88433 KW |
1385 | SvPVX(PL_numeric_radix_sv), |
1386 | cBOOL(SvUTF8(PL_numeric_radix_sv))); | |
2fcc0ca9 | 1387 | } |
69014004 | 1388 | |
86799d2d | 1389 | # endif |
d36adde0 KW |
1390 | #else |
1391 | ||
1392 | PERL_UNUSED_ARG(use_locale); | |
1393 | ||
86799d2d | 1394 | #endif /* USE_LOCALE_NUMERIC and can find the radix char */ |
7d4bcc4a | 1395 | |
98994639 HS |
1396 | } |
1397 | ||
398eeea9 KW |
1398 | STATIC void |
1399 | S_new_numeric(pTHX_ const char *newnum) | |
98994639 | 1400 | { |
7d4bcc4a KW |
1401 | |
1402 | #ifndef USE_LOCALE_NUMERIC | |
1403 | ||
1404 | PERL_UNUSED_ARG(newnum); | |
1405 | ||
1406 | #else | |
0d071d52 | 1407 | |
291a84fb | 1408 | /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell |
0d071d52 KW |
1409 | * core Perl this and that 'newnum' is the name of the new locale. |
1410 | * It installs this locale as the current underlying default. | |
1411 | * | |
1412 | * The default locale and the C locale can be toggled between by use of the | |
5792c642 KW |
1413 | * set_numeric_underlying() and set_numeric_standard() functions, which |
1414 | * should probably not be called directly, but only via macros like | |
0d071d52 KW |
1415 | * SET_NUMERIC_STANDARD() in perl.h. |
1416 | * | |
1417 | * The toggling is necessary mainly so that a non-dot radix decimal point | |
1418 | * character can be output, while allowing internal calculations to use a | |
1419 | * dot. | |
1420 | * | |
1421 | * This sets several interpreter-level variables: | |
bb304765 | 1422 | * PL_numeric_name The underlying locale's name: a copy of 'newnum' |
892e6465 | 1423 | * PL_numeric_underlying A boolean indicating if the toggled state is such |
7738054c KW |
1424 | * that the current locale is the program's underlying |
1425 | * locale | |
1426 | * PL_numeric_standard An int indicating if the toggled state is such | |
4c68b815 KW |
1427 | * that the current locale is the C locale or |
1428 | * indistinguishable from the C locale. If non-zero, it | |
1429 | * is in C; if > 1, it means it may not be toggled away | |
7738054c | 1430 | * from C. |
4c68b815 KW |
1431 | * PL_numeric_underlying_is_standard A bool kept by this function |
1432 | * indicating that the underlying locale and the standard | |
1433 | * C locale are indistinguishable for the purposes of | |
1434 | * LC_NUMERIC. This happens when both of the above two | |
1435 | * variables are true at the same time. (Toggling is a | |
1436 | * no-op under these circumstances.) This variable is | |
1437 | * used to avoid having to recalculate. | |
398eeea9 | 1438 | */ |
0d071d52 | 1439 | |
b03f34cf | 1440 | char *save_newnum; |
98994639 HS |
1441 | |
1442 | if (! newnum) { | |
43c5f42d NC |
1443 | Safefree(PL_numeric_name); |
1444 | PL_numeric_name = NULL; | |
98994639 | 1445 | PL_numeric_standard = TRUE; |
892e6465 | 1446 | PL_numeric_underlying = TRUE; |
4c68b815 | 1447 | PL_numeric_underlying_is_standard = TRUE; |
98994639 HS |
1448 | return; |
1449 | } | |
1450 | ||
b03f34cf | 1451 | save_newnum = stdize_locale(savepv(newnum)); |
892e6465 | 1452 | PL_numeric_underlying = TRUE; |
4c68b815 KW |
1453 | PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum); |
1454 | ||
69c5e0db KW |
1455 | #ifndef TS_W32_BROKEN_LOCALECONV |
1456 | ||
4c68b815 | 1457 | /* If its name isn't C nor POSIX, it could still be indistinguishable from |
69c5e0db KW |
1458 | * them. But on broken Windows systems calling my_nl_langinfo() for |
1459 | * THOUSEP can currently (but rarely) cause a race, so avoid doing that, | |
1460 | * and just always change the locale if not C nor POSIX on those systems */ | |
4c68b815 | 1461 | if (! PL_numeric_standard) { |
4e6826bf | 1462 | PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(RADIXCHAR, |
4c68b815 | 1463 | FALSE /* Don't toggle locale */ )) |
4e6826bf | 1464 | && strEQ("", my_nl_langinfo(THOUSEP, FALSE))); |
4c68b815 | 1465 | } |
abe1abcf | 1466 | |
69c5e0db KW |
1467 | #endif |
1468 | ||
4c68b815 | 1469 | /* Save the new name if it isn't the same as the previous one, if any */ |
b03f34cf | 1470 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { |
98994639 | 1471 | Safefree(PL_numeric_name); |
b03f34cf | 1472 | PL_numeric_name = save_newnum; |
b03f34cf | 1473 | } |
abe1abcf KW |
1474 | else { |
1475 | Safefree(save_newnum); | |
1476 | } | |
4c28b29c | 1477 | |
4c68b815 KW |
1478 | PL_numeric_underlying_is_standard = PL_numeric_standard; |
1479 | ||
7e5377f7 | 1480 | # ifdef HAS_POSIX_2008_LOCALE |
e1aa2579 KW |
1481 | |
1482 | PL_underlying_numeric_obj = newlocale(LC_NUMERIC_MASK, | |
1483 | PL_numeric_name, | |
1484 | PL_underlying_numeric_obj); | |
1485 | ||
1486 | #endif | |
1487 | ||
4c68b815 KW |
1488 | if (DEBUG_L_TEST || debug_initialization) { |
1489 | PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name); | |
1490 | } | |
1491 | ||
4c28b29c KW |
1492 | /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't |
1493 | * have to worry about the radix being a non-dot. (Core operations that | |
1494 | * need the underlying locale change to it temporarily). */ | |
84f1d29f KW |
1495 | if (PL_numeric_standard) { |
1496 | set_numeric_radix(0); | |
1497 | } | |
1498 | else { | |
1499 | set_numeric_standard(); | |
1500 | } | |
4c28b29c | 1501 | |
98994639 | 1502 | #endif /* USE_LOCALE_NUMERIC */ |
7d4bcc4a | 1503 | |
98994639 HS |
1504 | } |
1505 | ||
1506 | void | |
1507 | Perl_set_numeric_standard(pTHX) | |
1508 | { | |
7d4bcc4a | 1509 | |
98994639 | 1510 | #ifdef USE_LOCALE_NUMERIC |
7d4bcc4a | 1511 | |
28c1bf33 KW |
1512 | /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like |
1513 | * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The | |
1514 | * macro avoids calling this routine if toggling isn't necessary according | |
1515 | * to our records (which could be wrong if some XS code has changed the | |
1516 | * locale behind our back) */ | |
0d071d52 | 1517 | |
7d4bcc4a KW |
1518 | # ifdef DEBUGGING |
1519 | ||
2fcc0ca9 KW |
1520 | if (DEBUG_L_TEST || debug_initialization) { |
1521 | PerlIO_printf(Perl_debug_log, | |
7bb8f702 | 1522 | "Setting LC_NUMERIC locale to standard C\n"); |
2fcc0ca9 | 1523 | } |
98994639 | 1524 | |
7d4bcc4a | 1525 | # endif |
7bb8f702 KW |
1526 | |
1527 | do_setlocale_c(LC_NUMERIC, "C"); | |
1528 | PL_numeric_standard = TRUE; | |
1529 | PL_numeric_underlying = PL_numeric_underlying_is_standard; | |
1530 | set_numeric_radix(0); | |
1531 | ||
98994639 | 1532 | #endif /* USE_LOCALE_NUMERIC */ |
7d4bcc4a | 1533 | |
98994639 HS |
1534 | } |
1535 | ||
1536 | void | |
5792c642 | 1537 | Perl_set_numeric_underlying(pTHX) |
98994639 | 1538 | { |
7d4bcc4a | 1539 | |
98994639 | 1540 | #ifdef USE_LOCALE_NUMERIC |
7d4bcc4a | 1541 | |
28c1bf33 | 1542 | /* Toggle the LC_NUMERIC locale to the current underlying default. Most |
7d4bcc4a KW |
1543 | * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h |
1544 | * instead of calling this directly. The macro avoids calling this routine | |
1545 | * if toggling isn't necessary according to our records (which could be | |
1546 | * wrong if some XS code has changed the locale behind our back) */ | |
a9b8c0d8 | 1547 | |
7d4bcc4a KW |
1548 | # ifdef DEBUGGING |
1549 | ||
2fcc0ca9 KW |
1550 | if (DEBUG_L_TEST || debug_initialization) { |
1551 | PerlIO_printf(Perl_debug_log, | |
7bb8f702 | 1552 | "Setting LC_NUMERIC locale to %s\n", |
2fcc0ca9 KW |
1553 | PL_numeric_name); |
1554 | } | |
98994639 | 1555 | |
7d4bcc4a | 1556 | # endif |
7bb8f702 KW |
1557 | |
1558 | do_setlocale_c(LC_NUMERIC, PL_numeric_name); | |
1559 | PL_numeric_standard = PL_numeric_underlying_is_standard; | |
1560 | PL_numeric_underlying = TRUE; | |
1561 | set_numeric_radix(! PL_numeric_standard); | |
1562 | ||
98994639 | 1563 | #endif /* USE_LOCALE_NUMERIC */ |
7d4bcc4a | 1564 | |
98994639 HS |
1565 | } |
1566 | ||
1567 | /* | |
1568 | * Set up for a new ctype locale. | |
1569 | */ | |
a4f00dcc KW |
1570 | STATIC void |
1571 | S_new_ctype(pTHX_ const char *newctype) | |
98994639 | 1572 | { |
7d4bcc4a KW |
1573 | |
1574 | #ifndef USE_LOCALE_CTYPE | |
1575 | ||
7d4bcc4a KW |
1576 | PERL_UNUSED_ARG(newctype); |
1577 | PERL_UNUSED_CONTEXT; | |
1578 | ||
1579 | #else | |
0d071d52 | 1580 | |
291a84fb | 1581 | /* Called after each libc setlocale() call affecting LC_CTYPE, to tell |
0d071d52 KW |
1582 | * core Perl this and that 'newctype' is the name of the new locale. |
1583 | * | |
1584 | * This function sets up the folding arrays for all 256 bytes, assuming | |
1585 | * that tofold() is tolc() since fold case is not a concept in POSIX, | |
1586 | * | |
1587 | * Any code changing the locale (outside this file) should use | |
9aac5db8 KW |
1588 | * Perl_setlocale or POSIX::setlocale, which call this function. Therefore |
1589 | * this function should be called directly only from this file and from | |
0d071d52 KW |
1590 | * POSIX::setlocale() */ |
1591 | ||
013f4e03 | 1592 | unsigned int i; |
98994639 | 1593 | |
8b7358b9 KW |
1594 | /* Don't check for problems if we are suppressing the warnings */ |
1595 | bool check_for_problems = ckWARN_d(WARN_LOCALE) || UNLIKELY(DEBUG_L_TEST); | |
30d8090d | 1596 | bool maybe_utf8_turkic = FALSE; |
8b7358b9 | 1597 | |
7918f24d NC |
1598 | PERL_ARGS_ASSERT_NEW_CTYPE; |
1599 | ||
215c5139 KW |
1600 | /* We will replace any bad locale warning with 1) nothing if the new one is |
1601 | * ok; or 2) a new warning for the bad new locale */ | |
1602 | if (PL_warn_locale) { | |
1603 | SvREFCNT_dec_NN(PL_warn_locale); | |
1604 | PL_warn_locale = NULL; | |
1605 | } | |
1606 | ||
c1284011 | 1607 | PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE); |
31f05a37 KW |
1608 | |
1609 | /* A UTF-8 locale gets standard rules. But note that code still has to | |
1610 | * handle this specially because of the three problematic code points */ | |
1611 | if (PL_in_utf8_CTYPE_locale) { | |
1612 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); | |
30d8090d KW |
1613 | |
1614 | /* UTF-8 locales can have special handling for 'I' and 'i' if they are | |
1615 | * Turkic. Make sure these two are the only anomalies. (We don't use | |
1616 | * towupper and towlower because they aren't in C89.) */ | |
5b64f24c KW |
1617 | |
1618 | #if defined(HAS_TOWUPPER) && defined (HAS_TOWLOWER) | |
1619 | ||
1620 | if (towupper('i') == 0x130 && towlower('I') == 0x131) { | |
1621 | ||
1622 | #else | |
1623 | ||
30d8090d | 1624 | if (toupper('i') == 'i' && tolower('I') == 'I') { |
5b64f24c KW |
1625 | |
1626 | #endif | |
30d8090d KW |
1627 | check_for_problems = TRUE; |
1628 | maybe_utf8_turkic = TRUE; | |
1629 | } | |
31f05a37 | 1630 | } |
8b7358b9 KW |
1631 | |
1632 | /* We don't populate the other lists if a UTF-8 locale, but do check that | |
1633 | * everything works as expected, unless checking turned off */ | |
1634 | if (check_for_problems || ! PL_in_utf8_CTYPE_locale) { | |
8c6180a9 KW |
1635 | /* Assume enough space for every character being bad. 4 spaces each |
1636 | * for the 94 printable characters that are output like "'x' "; and 5 | |
1637 | * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating | |
1638 | * NUL */ | |
8b7358b9 | 1639 | char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ] = { '\0' }; |
8c6180a9 KW |
1640 | bool multi_byte_locale = FALSE; /* Assume is a single-byte locale |
1641 | to start */ | |
1642 | unsigned int bad_count = 0; /* Count of bad characters */ | |
1643 | ||
baa60164 | 1644 | for (i = 0; i < 256; i++) { |
8b7358b9 | 1645 | if (! PL_in_utf8_CTYPE_locale) { |
bd6d0898 KW |
1646 | if (isupper(i)) |
1647 | PL_fold_locale[i] = (U8) tolower(i); | |
1648 | else if (islower(i)) | |
1649 | PL_fold_locale[i] = (U8) toupper(i); | |
1650 | else | |
1651 | PL_fold_locale[i] = (U8) i; | |
8b7358b9 | 1652 | } |
8c6180a9 KW |
1653 | |
1654 | /* If checking for locale problems, see if the native ASCII-range | |
1655 | * printables plus \n and \t are in their expected categories in | |
1656 | * the new locale. If not, this could mean big trouble, upending | |
1657 | * Perl's and most programs' assumptions, like having a | |
1658 | * metacharacter with special meaning become a \w. Fortunately, | |
1659 | * it's very rare to find locales that aren't supersets of ASCII | |
1660 | * nowadays. It isn't a problem for most controls to be changed | |
1661 | * into something else; we check only \n and \t, though perhaps \r | |
1662 | * could be an issue as well. */ | |
7d4bcc4a | 1663 | if ( check_for_problems |
8c6180a9 KW |
1664 | && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n')) |
1665 | { | |
8b7358b9 | 1666 | bool is_bad = FALSE; |
6726b4f4 | 1667 | char name[4] = { '\0' }; |
8b7358b9 KW |
1668 | |
1669 | /* Convert the name into a string */ | |
6726b4f4 | 1670 | if (isGRAPH_A(i)) { |
8b7358b9 KW |
1671 | name[0] = i; |
1672 | name[1] = '\0'; | |
1673 | } | |
1674 | else if (i == '\n') { | |
6726b4f4 KW |
1675 | my_strlcpy(name, "\\n", sizeof(name)); |
1676 | } | |
1677 | else if (i == '\t') { | |
1678 | my_strlcpy(name, "\\t", sizeof(name)); | |
8b7358b9 KW |
1679 | } |
1680 | else { | |
6726b4f4 KW |
1681 | assert(i == ' '); |
1682 | my_strlcpy(name, "' '", sizeof(name)); | |
8b7358b9 KW |
1683 | } |
1684 | ||
1685 | /* Check each possibe class */ | |
1686 | if (UNLIKELY(cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC_A(i)))) { | |
1687 | is_bad = TRUE; | |
1688 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1689 | "isalnum('%s') unexpectedly is %d\n", | |
1690 | name, cBOOL(isalnum(i)))); | |
1691 | } | |
1692 | if (UNLIKELY(cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i)))) { | |
1693 | is_bad = TRUE; | |
1694 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1695 | "isalpha('%s') unexpectedly is %d\n", | |
1696 | name, cBOOL(isalpha(i)))); | |
1697 | } | |
1698 | if (UNLIKELY(cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i)))) { | |
1699 | is_bad = TRUE; | |
1700 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1701 | "isdigit('%s') unexpectedly is %d\n", | |
1702 | name, cBOOL(isdigit(i)))); | |
1703 | } | |
1704 | if (UNLIKELY(cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i)))) { | |
1705 | is_bad = TRUE; | |
1706 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1707 | "isgraph('%s') unexpectedly is %d\n", | |
1708 | name, cBOOL(isgraph(i)))); | |
1709 | } | |
1710 | if (UNLIKELY(cBOOL(islower(i)) != cBOOL(isLOWER_A(i)))) { | |
1711 | is_bad = TRUE; | |
1712 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1713 | "islower('%s') unexpectedly is %d\n", | |
1714 | name, cBOOL(islower(i)))); | |
1715 | } | |
1716 | if (UNLIKELY(cBOOL(isprint(i)) != cBOOL(isPRINT_A(i)))) { | |
1717 | is_bad = TRUE; | |
1718 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1719 | "isprint('%s') unexpectedly is %d\n", | |
1720 | name, cBOOL(isprint(i)))); | |
1721 | } | |
1722 | if (UNLIKELY(cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i)))) { | |
1723 | is_bad = TRUE; | |
1724 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1725 | "ispunct('%s') unexpectedly is %d\n", | |
1726 | name, cBOOL(ispunct(i)))); | |
1727 | } | |
1728 | if (UNLIKELY(cBOOL(isspace(i)) != cBOOL(isSPACE_A(i)))) { | |
1729 | is_bad = TRUE; | |
1730 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1731 | "isspace('%s') unexpectedly is %d\n", | |
1732 | name, cBOOL(isspace(i)))); | |
1733 | } | |
1734 | if (UNLIKELY(cBOOL(isupper(i)) != cBOOL(isUPPER_A(i)))) { | |
1735 | is_bad = TRUE; | |
1736 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1737 | "isupper('%s') unexpectedly is %d\n", | |
1738 | name, cBOOL(isupper(i)))); | |
1739 | } | |
1740 | if (UNLIKELY(cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i)))) { | |
1741 | is_bad = TRUE; | |
1742 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1743 | "isxdigit('%s') unexpectedly is %d\n", | |
1744 | name, cBOOL(isxdigit(i)))); | |
1745 | } | |
65c15174 | 1746 | if (UNLIKELY(tolower(i) != (int) toLOWER_A(i))) { |
8b7358b9 KW |
1747 | is_bad = TRUE; |
1748 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1749 | "tolower('%s')=0x%x instead of the expected 0x%x\n", | |
1750 | name, tolower(i), (int) toLOWER_A(i))); | |
1751 | } | |
65c15174 | 1752 | if (UNLIKELY(toupper(i) != (int) toUPPER_A(i))) { |
8b7358b9 KW |
1753 | is_bad = TRUE; |
1754 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1755 | "toupper('%s')=0x%x instead of the expected 0x%x\n", | |
1756 | name, toupper(i), (int) toUPPER_A(i))); | |
1757 | } | |
1758 | if (UNLIKELY((i == '\n' && ! isCNTRL_LC(i)))) { | |
1759 | is_bad = TRUE; | |
1760 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1761 | "'\\n' (=%02X) is not a control\n", (int) i)); | |
1762 | } | |
1763 | ||
1764 | /* Add to the list; Separate multiple entries with a blank */ | |
1765 | if (is_bad) { | |
1766 | if (bad_count) { | |
1767 | my_strlcat(bad_chars_list, " ", sizeof(bad_chars_list)); | |
8c6180a9 | 1768 | } |
8b7358b9 KW |
1769 | my_strlcat(bad_chars_list, name, sizeof(bad_chars_list)); |
1770 | bad_count++; | |
8c6180a9 KW |
1771 | } |
1772 | } | |
1773 | } | |
1774 | ||
30d8090d KW |
1775 | if (bad_count == 2 && maybe_utf8_turkic) { |
1776 | bad_count = 0; | |
1777 | *bad_chars_list = '\0'; | |
1778 | PL_fold_locale['I'] = 'I'; | |
1779 | PL_fold_locale['i'] = 'i'; | |
1780 | PL_in_utf8_turkic_locale = TRUE; | |
1781 | DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s is turkic\n", | |
1782 | __FILE__, __LINE__, newctype)); | |
1783 | } | |
1784 | else { | |
b8df1494 | 1785 | PL_in_utf8_turkic_locale = FALSE; |
30d8090d | 1786 | } |
b8df1494 | 1787 | |
7d4bcc4a KW |
1788 | # ifdef MB_CUR_MAX |
1789 | ||
8c6180a9 | 1790 | /* We only handle single-byte locales (outside of UTF-8 ones; so if |
d35fca5f | 1791 | * this locale requires more than one byte, there are going to be |
8c6180a9 | 1792 | * problems. */ |
9c8a6dc2 KW |
1793 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, |
1794 | "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n", | |
1795 | __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX)); | |
1796 | ||
8b7358b9 KW |
1797 | if ( check_for_problems && MB_CUR_MAX > 1 |
1798 | && ! PL_in_utf8_CTYPE_locale | |
ba1a4362 KW |
1799 | |
1800 | /* Some platforms return MB_CUR_MAX > 1 for even the "C" | |
1801 | * locale. Just assume that the implementation for them (plus | |
1802 | * for POSIX) is correct and the > 1 value is spurious. (Since | |
1803 | * these are specially handled to never be considered UTF-8 | |
1804 | * locales, as long as this is the only problem, everything | |
1805 | * should work fine */ | |
1806 | && strNE(newctype, "C") && strNE(newctype, "POSIX")) | |
1807 | { | |
8c6180a9 KW |
1808 | multi_byte_locale = TRUE; |
1809 | } | |
7d4bcc4a KW |
1810 | |
1811 | # endif | |
8c6180a9 | 1812 | |
30d8090d KW |
1813 | /* If we found problems and we want them output, do so */ |
1814 | if ( (UNLIKELY(bad_count) || UNLIKELY(multi_byte_locale)) | |
1815 | && (LIKELY(ckWARN_d(WARN_LOCALE)) || UNLIKELY(DEBUG_L_TEST))) | |
1816 | { | |
8b7358b9 KW |
1817 | if (UNLIKELY(bad_count) && PL_in_utf8_CTYPE_locale) { |
1818 | PL_warn_locale = Perl_newSVpvf(aTHX_ | |
1819 | "Locale '%s' contains (at least) the following characters" | |
578a6a87 KW |
1820 | " which have\nunexpected meanings: %s\nThe Perl program" |
1821 | " will use the expected meanings", | |
8b7358b9 | 1822 | newctype, bad_chars_list); |
8b7358b9 KW |
1823 | } |
1824 | else { | |
1825 | PL_warn_locale = Perl_newSVpvf(aTHX_ | |
8c6180a9 | 1826 | "Locale '%s' may not work well.%s%s%s\n", |
780fcc9f | 1827 | newctype, |
8c6180a9 KW |
1828 | (multi_byte_locale) |
1829 | ? " Some characters in it are not recognized by" | |
1830 | " Perl." | |
1831 | : "", | |
1832 | (bad_count) | |
1833 | ? "\nThe following characters (and maybe others)" | |
1834 | " may not have the same meaning as the Perl" | |
1835 | " program expects:\n" | |
1836 | : "", | |
1837 | (bad_count) | |
1838 | ? bad_chars_list | |
1839 | : "" | |
1840 | ); | |
8b7358b9 KW |
1841 | } |
1842 | ||
b119c2be KW |
1843 | # ifdef HAS_NL_LANGINFO |
1844 | ||
1845 | Perl_sv_catpvf(aTHX_ PL_warn_locale, "; codeset=%s", | |
1846 | /* parameter FALSE is a don't care here */ | |
4e6826bf | 1847 | my_nl_langinfo(CODESET, FALSE)); |
b119c2be KW |
1848 | |
1849 | # endif | |
1850 | ||
8b7358b9 KW |
1851 | Perl_sv_catpvf(aTHX_ PL_warn_locale, "\n"); |
1852 | ||
cc9eaeb0 | 1853 | /* If we are actually in the scope of the locale or are debugging, |
bddebb56 KW |
1854 | * output the message now. If not in that scope, we save the |
1855 | * message to be output at the first operation using this locale, | |
1856 | * if that actually happens. Most programs don't use locales, so | |
1857 | * they are immune to bad ones. */ | |
cc9eaeb0 | 1858 | if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) { |
780fcc9f | 1859 | |
780fcc9f KW |
1860 | /* The '0' below suppresses a bogus gcc compiler warning */ |
1861 | Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0); | |
bddebb56 | 1862 | |
bddebb56 KW |
1863 | if (IN_LC(LC_CTYPE)) { |
1864 | SvREFCNT_dec_NN(PL_warn_locale); | |
1865 | PL_warn_locale = NULL; | |
1866 | } | |
780fcc9f | 1867 | } |
baa60164 | 1868 | } |
31f05a37 | 1869 | } |
98994639 HS |
1870 | |
1871 | #endif /* USE_LOCALE_CTYPE */ | |
7d4bcc4a | 1872 | |
98994639 HS |
1873 | } |
1874 | ||
98994639 | 1875 | void |
2726666d KW |
1876 | Perl__warn_problematic_locale() |
1877 | { | |
2726666d KW |
1878 | |
1879 | #ifdef USE_LOCALE_CTYPE | |
1880 | ||
5f04a188 KW |
1881 | dTHX; |
1882 | ||
1883 | /* Internal-to-core function that outputs the message in PL_warn_locale, | |
1884 | * and then NULLS it. Should be called only through the macro | |
1885 | * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */ | |
1886 | ||
2726666d | 1887 | if (PL_warn_locale) { |
2726666d KW |
1888 | Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE), |
1889 | SvPVX(PL_warn_locale), | |
1890 | 0 /* dummy to avoid compiler warning */ ); | |
2726666d KW |
1891 | SvREFCNT_dec_NN(PL_warn_locale); |
1892 | PL_warn_locale = NULL; | |
1893 | } | |
1894 | ||
1895 | #endif | |
1896 | ||
1897 | } | |
1898 | ||
a4f00dcc KW |
1899 | STATIC void |
1900 | S_new_collate(pTHX_ const char *newcoll) | |
98994639 | 1901 | { |
7d4bcc4a KW |
1902 | |
1903 | #ifndef USE_LOCALE_COLLATE | |
1904 | ||
1905 | PERL_UNUSED_ARG(newcoll); | |
1906 | PERL_UNUSED_CONTEXT; | |
1907 | ||
1908 | #else | |
0d071d52 | 1909 | |
291a84fb | 1910 | /* Called after each libc setlocale() call affecting LC_COLLATE, to tell |
0d071d52 KW |
1911 | * core Perl this and that 'newcoll' is the name of the new locale. |
1912 | * | |
d35fca5f KW |
1913 | * The design of locale collation is that every locale change is given an |
1914 | * index 'PL_collation_ix'. The first time a string particpates in an | |
1915 | * operation that requires collation while locale collation is active, it | |
1916 | * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That | |
1917 | * magic includes the collation index, and the transformation of the string | |
1918 | * by strxfrm(), q.v. That transformation is used when doing comparisons, | |
1919 | * instead of the string itself. If a string changes, the magic is | |
1920 | * cleared. The next time the locale changes, the index is incremented, | |
1921 | * and so we know during a comparison that the transformation is not | |
1922 | * necessarily still valid, and so is recomputed. Note that if the locale | |
1923 | * changes enough times, the index could wrap (a U32), and it is possible | |
1924 | * that a transformation would improperly be considered valid, leading to | |
1925 | * an unlikely bug */ | |
0d071d52 | 1926 | |
98994639 HS |
1927 | if (! newcoll) { |
1928 | if (PL_collation_name) { | |
1929 | ++PL_collation_ix; | |
1930 | Safefree(PL_collation_name); | |
1931 | PL_collation_name = NULL; | |
1932 | } | |
1933 | PL_collation_standard = TRUE; | |
00bf60ca | 1934 | is_standard_collation: |
98994639 HS |
1935 | PL_collxfrm_base = 0; |
1936 | PL_collxfrm_mult = 2; | |
165a1c52 | 1937 | PL_in_utf8_COLLATE_locale = FALSE; |
f28f4d2a | 1938 | PL_strxfrm_NUL_replacement = '\0'; |
a4a439fb | 1939 | PL_strxfrm_max_cp = 0; |
98994639 HS |
1940 | return; |
1941 | } | |
1942 | ||
d35fca5f | 1943 | /* If this is not the same locale as currently, set the new one up */ |
98994639 HS |
1944 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { |
1945 | ++PL_collation_ix; | |
1946 | Safefree(PL_collation_name); | |
1947 | PL_collation_name = stdize_locale(savepv(newcoll)); | |
a39edc4c | 1948 | PL_collation_standard = isNAME_C_OR_POSIX(newcoll); |
00bf60ca KW |
1949 | if (PL_collation_standard) { |
1950 | goto is_standard_collation; | |
1951 | } | |
98994639 | 1952 | |
165a1c52 | 1953 | PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE); |
f28f4d2a | 1954 | PL_strxfrm_NUL_replacement = '\0'; |
a4a439fb | 1955 | PL_strxfrm_max_cp = 0; |
165a1c52 | 1956 | |
59c018b9 KW |
1957 | /* A locale collation definition includes primary, secondary, tertiary, |
1958 | * etc. weights for each character. To sort, the primary weights are | |
1959 | * used, and only if they compare equal, then the secondary weights are | |
1960 | * used, and only if they compare equal, then the tertiary, etc. | |
1961 | * | |
1962 | * strxfrm() works by taking the input string, say ABC, and creating an | |
1963 | * output transformed string consisting of first the primary weights, | |
1964 | * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the | |
1965 | * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters | |
1966 | * may not have weights at every level. In our example, let's say B | |
1967 | * doesn't have a tertiary weight, and A doesn't have a secondary | |
1968 | * weight. The constructed string is then going to be | |
1969 | * A¹B¹C¹ B²C² A³C³ .... | |
1970 | * This has the desired effect that strcmp() will look at the secondary | |
1971 | * or tertiary weights only if the strings compare equal at all higher | |
1972 | * priority weights. The spaces shown here, like in | |
c342d20e | 1973 | * "A¹B¹C¹ A²B²C² " |
59c018b9 KW |
1974 | * are not just for readability. In the general case, these must |
1975 | * actually be bytes, which we will call here 'separator weights'; and | |
1976 | * they must be smaller than any other weight value, but since these | |
1977 | * are C strings, only the terminating one can be a NUL (some | |
1978 | * implementations may include a non-NUL separator weight just before | |
1979 | * the NUL). Implementations tend to reserve 01 for the separator | |
1980 | * weights. They are needed so that a shorter string's secondary | |
1981 | * weights won't be misconstrued as primary weights of a longer string, | |
1982 | * etc. By making them smaller than any other weight, the shorter | |
1983 | * string will sort first. (Actually, if all secondary weights are | |
1984 | * smaller than all primary ones, there is no need for a separator | |
1985 | * weight between those two levels, etc.) | |
1986 | * | |
1987 | * The length of the transformed string is roughly a linear function of | |
1988 | * the input string. It's not exactly linear because some characters | |
1989 | * don't have weights at all levels. When we call strxfrm() we have to | |
1990 | * allocate some memory to hold the transformed string. The | |
1991 | * calculations below try to find coefficients 'm' and 'b' for this | |
1992 | * locale so that m*x + b equals how much space we need, given the size | |
1993 | * of the input string in 'x'. If we calculate too small, we increase | |
1994 | * the size as needed, and call strxfrm() again, but it is better to | |
1995 | * get it right the first time to avoid wasted expensive string | |
1996 | * transformations. */ | |
1997 | ||
98994639 | 1998 | { |
79f120c8 KW |
1999 | /* We use the string below to find how long the tranformation of it |
2000 | * is. Almost all locales are supersets of ASCII, or at least the | |
2001 | * ASCII letters. We use all of them, half upper half lower, | |
2002 | * because if we used fewer, we might hit just the ones that are | |
2003 | * outliers in a particular locale. Most of the strings being | |
2004 | * collated will contain a preponderance of letters, and even if | |
2005 | * they are above-ASCII, they are likely to have the same number of | |
2006 | * weight levels as the ASCII ones. It turns out that digits tend | |
2007 | * to have fewer levels, and some punctuation has more, but those | |
2008 | * are relatively sparse in text, and khw believes this gives a | |
2009 | * reasonable result, but it could be changed if experience so | |
2010 | * dictates. */ | |
2011 | const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz"; | |
2012 | char * x_longer; /* Transformed 'longer' */ | |
2013 | Size_t x_len_longer; /* Length of 'x_longer' */ | |
2014 | ||
2015 | char * x_shorter; /* We also transform a substring of 'longer' */ | |
2016 | Size_t x_len_shorter; | |
2017 | ||
a4a439fb | 2018 | /* _mem_collxfrm() is used get the transformation (though here we |
79f120c8 KW |
2019 | * are interested only in its length). It is used because it has |
2020 | * the intelligence to handle all cases, but to work, it needs some | |
2021 | * values of 'm' and 'b' to get it started. For the purposes of | |
2022 | * this calculation we use a very conservative estimate of 'm' and | |
2023 | * 'b'. This assumes a weight can be multiple bytes, enough to | |
2024 | * hold any UV on the platform, and there are 5 levels, 4 weight | |
2025 | * bytes, and a trailing NUL. */ | |
2026 | PL_collxfrm_base = 5; | |
2027 | PL_collxfrm_mult = 5 * sizeof(UV); | |
2028 | ||
2029 | /* Find out how long the transformation really is */ | |
a4a439fb KW |
2030 | x_longer = _mem_collxfrm(longer, |
2031 | sizeof(longer) - 1, | |
2032 | &x_len_longer, | |
2033 | ||
2034 | /* We avoid converting to UTF-8 in the | |
2035 | * called function by telling it the | |
2036 | * string is in UTF-8 if the locale is a | |
2037 | * UTF-8 one. Since the string passed | |
2038 | * here is invariant under UTF-8, we can | |
2039 | * claim it's UTF-8 even though it isn't. | |
2040 | * */ | |
2041 | PL_in_utf8_COLLATE_locale); | |
79f120c8 KW |
2042 | Safefree(x_longer); |
2043 | ||
2044 | /* Find out how long the transformation of a substring of 'longer' | |
2045 | * is. Together the lengths of these transformations are | |
2046 | * sufficient to calculate 'm' and 'b'. The substring is all of | |
2047 | * 'longer' except the first character. This minimizes the chances | |
2048 | * of being swayed by outliers */ | |
a4a439fb | 2049 | x_shorter = _mem_collxfrm(longer + 1, |
79f120c8 | 2050 | sizeof(longer) - 2, |
a4a439fb KW |
2051 | &x_len_shorter, |
2052 | PL_in_utf8_COLLATE_locale); | |
79f120c8 KW |
2053 | Safefree(x_shorter); |
2054 | ||
2055 | /* If the results are nonsensical for this simple test, the whole | |
2056 | * locale definition is suspect. Mark it so that locale collation | |
2057 | * is not active at all for it. XXX Should we warn? */ | |
2058 | if ( x_len_shorter == 0 | |
2059 | || x_len_longer == 0 | |
2060 | || x_len_shorter >= x_len_longer) | |
2061 | { | |
2062 | PL_collxfrm_mult = 0; | |
2063 | PL_collxfrm_base = 0; | |
2064 | } | |
2065 | else { | |
2066 | SSize_t base; /* Temporary */ | |
2067 | ||
2068 | /* We have both: m * strlen(longer) + b = x_len_longer | |
2069 | * m * strlen(shorter) + b = x_len_shorter; | |
2070 | * subtracting yields: | |
2071 | * m * (strlen(longer) - strlen(shorter)) | |
2072 | * = x_len_longer - x_len_shorter | |
2073 | * But we have set things up so that 'shorter' is 1 byte smaller | |
2074 | * than 'longer'. Hence: | |
2075 | * m = x_len_longer - x_len_shorter | |
2076 | * | |
2077 | * But if something went wrong, make sure the multiplier is at | |
2078 | * least 1. | |
2079 | */ | |
2080 | if (x_len_longer > x_len_shorter) { | |
2081 | PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter; | |
2082 | } | |
2083 | else { | |
2084 | PL_collxfrm_mult = 1; | |
2085 | } | |
2086 | ||
2087 | /* mx + b = len | |
2088 | * so: b = len - mx | |
2089 | * but in case something has gone wrong, make sure it is | |
2090 | * non-negative */ | |
2091 | base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1); | |
2092 | if (base < 0) { | |
2093 | base = 0; | |
2094 | } | |
2095 | ||
2096 | /* Add 1 for the trailing NUL */ | |
2097 | PL_collxfrm_base = base + 1; | |
2098 | } | |
58eebef2 | 2099 | |
7d4bcc4a KW |
2100 | # ifdef DEBUGGING |
2101 | ||
58eebef2 KW |
2102 | if (DEBUG_L_TEST || debug_initialization) { |
2103 | PerlIO_printf(Perl_debug_log, | |
b07929e4 KW |
2104 | "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, " |
2105 | "x_len_longer=%zu," | |
2106 | " collate multipler=%zu, collate base=%zu\n", | |
58eebef2 KW |
2107 | __FILE__, __LINE__, |
2108 | PL_in_utf8_COLLATE_locale, | |
2109 | x_len_shorter, x_len_longer, | |
2110 | PL_collxfrm_mult, PL_collxfrm_base); | |
2111 | } | |
7d4bcc4a KW |
2112 | # endif |
2113 | ||
98994639 HS |
2114 | } |
2115 | } | |
2116 | ||
2117 | #endif /* USE_LOCALE_COLLATE */ | |
7d4bcc4a | 2118 | |
98994639 HS |
2119 | } |
2120 | ||
d36adde0 KW |
2121 | #endif |
2122 | ||
d2b24094 | 2123 | #ifdef WIN32 |
b8cc575c | 2124 | |
6c332036 TC |
2125 | #define USE_WSETLOCALE |
2126 | ||
2127 | #ifdef USE_WSETLOCALE | |
2128 | ||
2129 | STATIC char * | |
2130 | S_wrap_wsetlocale(pTHX_ int category, const char *locale) { | |
2131 | wchar_t *wlocale; | |
2132 | wchar_t *wresult; | |
2133 | char *result; | |
2134 | ||
2135 | if (locale) { | |
2136 | int req_size = | |
2137 | MultiByteToWideChar(CP_UTF8, 0, locale, -1, NULL, 0); | |
2138 | ||
2139 | if (!req_size) { | |
2140 | errno = EINVAL; | |
2141 | return NULL; | |
2142 | } | |
2143 | ||
2144 | Newx(wlocale, req_size, wchar_t); | |
2145 | if (!MultiByteToWideChar(CP_UTF8, 0, locale, -1, wlocale, req_size)) { | |
2146 | Safefree(wlocale); | |
2147 | errno = EINVAL; | |
2148 | return NULL; | |
2149 | } | |
2150 | } | |
2151 | else { | |
2152 | wlocale = NULL; | |
2153 | } | |
2154 | wresult = _wsetlocale(category, wlocale); | |
2155 | Safefree(wlocale); | |
2156 | if (wresult) { | |
2157 | int req_size = | |
2158 | WideCharToMultiByte(CP_UTF8, 0, wresult, -1, NULL, 0, NULL, NULL); | |
2159 | Newx(result, req_size, char); | |
2160 | SAVEFREEPV(result); /* is there something better we can do here? */ | |
2161 | if (!WideCharToMultiByte(CP_UTF8, 0, wresult, -1, | |
2162 | result, req_size, NULL, NULL)) { | |
2163 | errno = EINVAL; | |
2164 | return NULL; | |
2165 | } | |
2166 | } | |
2167 | else { | |
2168 | result = NULL; | |
2169 | } | |
2170 | ||
2171 | return result; | |
2172 | } | |
2173 | ||
2174 | #endif | |
2175 | ||
a4f00dcc | 2176 | STATIC char * |
b8cc575c | 2177 | S_win32_setlocale(pTHX_ int category, const char* locale) |
b385bb4d KW |
2178 | { |
2179 | /* This, for Windows, emulates POSIX setlocale() behavior. There is no | |
7d4bcc4a KW |
2180 | * difference between the two unless the input locale is "", which normally |
2181 | * means on Windows to get the machine default, which is set via the | |
2182 | * computer's "Regional and Language Options" (or its current equivalent). | |
2183 | * In POSIX, it instead means to find the locale from the user's | |
2184 | * environment. This routine changes the Windows behavior to first look in | |
2185 | * the environment, and, if anything is found, use that instead of going to | |
2186 | * the machine default. If there is no environment override, the machine | |
2187 | * default is used, by calling the real setlocale() with "". | |
2188 | * | |
2189 | * The POSIX behavior is to use the LC_ALL variable if set; otherwise to | |
2190 | * use the particular category's variable if set; otherwise to use the LANG | |
2191 | * variable. */ | |
b385bb4d | 2192 | |
175c4cf9 | 2193 | bool override_LC_ALL = FALSE; |
89f7b9aa | 2194 | char * result; |
e5f10d49 | 2195 | unsigned int i; |
89f7b9aa | 2196 | |
b385bb4d | 2197 | if (locale && strEQ(locale, "")) { |
7d4bcc4a KW |
2198 | |
2199 | # ifdef LC_ALL | |
2200 | ||
b385bb4d KW |
2201 | locale = PerlEnv_getenv("LC_ALL"); |
2202 | if (! locale) { | |
e5f10d49 KW |
2203 | if (category == LC_ALL) { |
2204 | override_LC_ALL = TRUE; | |
2205 | } | |
2206 | else { | |
7d4bcc4a KW |
2207 | |
2208 | # endif | |
7d4bcc4a | 2209 | |
e5f10d49 KW |
2210 | for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) { |
2211 | if (category == categories[i]) { | |
2212 | locale = PerlEnv_getenv(category_names[i]); | |
2213 | goto found_locale; | |
2214 | } | |
2215 | } | |
7d4bcc4a | 2216 | |
b385bb4d | 2217 | locale = PerlEnv_getenv("LANG"); |
481465ea | 2218 | if (! locale) { |
b385bb4d KW |
2219 | locale = ""; |
2220 | } | |
e5f10d49 KW |
2221 | |
2222 | found_locale: ; | |
7d4bcc4a KW |
2223 | |
2224 | # ifdef LC_ALL | |
2225 | ||
e5f10d49 | 2226 | } |
b385bb4d | 2227 | } |
7d4bcc4a KW |
2228 | |
2229 | # endif | |
2230 | ||
b385bb4d KW |
2231 | } |
2232 | ||
6c332036 TC |
2233 | #ifdef USE_WSETLOCALE |
2234 | result = S_wrap_wsetlocale(aTHX_ category, locale); | |
2235 | #else | |
89f7b9aa | 2236 | result = setlocale(category, locale); |
6c332036 | 2237 | #endif |
48015184 KW |
2238 | DEBUG_L(STMT_START { |
2239 | dSAVE_ERRNO; | |
2240 | PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__, | |
2241 | setlocale_debug_string(category, locale, result)); | |
2242 | RESTORE_ERRNO; | |
2243 | } STMT_END); | |
89f7b9aa | 2244 | |
481465ea | 2245 | if (! override_LC_ALL) { |
89f7b9aa KW |
2246 | return result; |
2247 | } | |
2248 | ||
dfd77d7a | 2249 | /* Here the input category was LC_ALL, and we have set it to what is in the |
481465ea KW |
2250 | * LANG variable or the system default if there is no LANG. But these have |
2251 | * lower priority than the other LC_foo variables, so override it for each | |
2252 | * one that is set. (If they are set to "", it means to use the same thing | |
2253 | * we just set LC_ALL to, so can skip) */ | |
7d4bcc4a | 2254 | |
948523db | 2255 | for (i = 0; i < LC_ALL_INDEX; i++) { |
e5f10d49 KW |
2256 | result = PerlEnv_getenv(category_names[i]); |
2257 | if (result && strNE(result, "")) { | |
6c332036 | 2258 | #ifdef USE_WSETLOCALE |
17a1e9ac | 2259 | S_wrap_wsetlocale(aTHX_ categories[i], result); |
6c332036 | 2260 | #else |
17a1e9ac | 2261 | setlocale(categories[i], result); |
6c332036 | 2262 | #endif |
e5f10d49 KW |
2263 | DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", |
2264 | __FILE__, __LINE__, | |
2265 | setlocale_debug_string(categories[i], result, "not captured"))); | |
2266 | } | |
89f7b9aa | 2267 | } |
7d4bcc4a | 2268 | |
bbc98134 | 2269 | result = setlocale(LC_ALL, NULL); |
48015184 KW |
2270 | DEBUG_L(STMT_START { |
2271 | dSAVE_ERRNO; | |
2272 | PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", | |
bbc98134 | 2273 | __FILE__, __LINE__, |
48015184 KW |
2274 | setlocale_debug_string(LC_ALL, NULL, result)); |
2275 | RESTORE_ERRNO; | |
2276 | } STMT_END); | |
89f7b9aa | 2277 | |
bbc98134 | 2278 | return result; |
b385bb4d KW |
2279 | } |
2280 | ||
2281 | #endif | |
2282 | ||
9aac5db8 | 2283 | /* |
9aac5db8 KW |
2284 | =for apidoc Perl_setlocale |
2285 | ||
2286 | This is an (almost) drop-in replacement for the system L<C<setlocale(3)>>, | |
2287 | taking the same parameters, and returning the same information, except that it | |
9487427b KW |
2288 | returns the correct underlying C<LC_NUMERIC> locale. Regular C<setlocale> will |
2289 | instead return C<C> if the underlying locale has a non-dot decimal point | |
2290 | character, or a non-empty thousands separator for displaying floating point | |
2291 | numbers. This is because perl keeps that locale category such that it has a | |
2292 | dot and empty separator, changing the locale briefly during the operations | |
2293 | where the underlying one is required. C<Perl_setlocale> knows about this, and | |
2294 | compensates; regular C<setlocale> doesn't. | |
9aac5db8 | 2295 | |
e9bc6d6b | 2296 | Another reason it isn't completely a drop-in replacement is that it is |
9aac5db8 | 2297 | declared to return S<C<const char *>>, whereas the system setlocale omits the |
163deff3 KW |
2298 | C<const> (presumably because its API was specified long ago, and can't be |
2299 | updated; it is illegal to change the information C<setlocale> returns; doing | |
2300 | so leads to segfaults.) | |
9aac5db8 | 2301 | |
e9bc6d6b KW |
2302 | Finally, C<Perl_setlocale> works under all circumstances, whereas plain |
2303 | C<setlocale> can be completely ineffective on some platforms under some | |
2304 | configurations. | |
2305 | ||
9aac5db8 | 2306 | C<Perl_setlocale> should not be used to change the locale except on systems |
e9bc6d6b KW |
2307 | where the predefined variable C<${^SAFE_LOCALES}> is 1. On some such systems, |
2308 | the system C<setlocale()> is ineffective, returning the wrong information, and | |
2309 | failing to actually change the locale. C<Perl_setlocale>, however works | |
2310 | properly in all circumstances. | |
9aac5db8 KW |
2311 | |
2312 | The return points to a per-thread static buffer, which is overwritten the next | |
2313 | time C<Perl_setlocale> is called from the same thread. | |
2314 | ||
2315 | =cut | |
2316 | ||
2317 | */ | |
2318 | ||
2319 | const char * | |
2320 | Perl_setlocale(const int category, const char * locale) | |
a4f00dcc KW |
2321 | { |
2322 | /* This wraps POSIX::setlocale() */ | |
2323 | ||
52129632 | 2324 | #ifndef USE_LOCALE |
d36adde0 KW |
2325 | |
2326 | PERL_UNUSED_ARG(category); | |
2327 | PERL_UNUSED_ARG(locale); | |
2328 | ||
2329 | return "C"; | |
2330 | ||
2331 | #else | |
2332 | ||
9aac5db8 KW |
2333 | const char * retval; |
2334 | const char * newlocale; | |
2335 | dSAVEDERRNO; | |
a4f00dcc | 2336 | dTHX; |
d36adde0 | 2337 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; |
a4f00dcc | 2338 | |
a4f00dcc KW |
2339 | #ifdef USE_LOCALE_NUMERIC |
2340 | ||
291a84fb KW |
2341 | /* A NULL locale means only query what the current one is. We have the |
2342 | * LC_NUMERIC name saved, because we are normally switched into the C | |
9487427b KW |
2343 | * (or equivalent) locale for it. For an LC_ALL query, switch back to get |
2344 | * the correct results. All other categories don't require special | |
2345 | * handling */ | |
a4f00dcc KW |
2346 | if (locale == NULL) { |
2347 | if (category == LC_NUMERIC) { | |
9aac5db8 KW |
2348 | |
2349 | /* We don't have to copy this return value, as it is a per-thread | |
2350 | * variable, and won't change until a future setlocale */ | |
2351 | return PL_numeric_name; | |
a4f00dcc KW |
2352 | } |
2353 | ||
7d4bcc4a | 2354 | # ifdef LC_ALL |
a4f00dcc | 2355 | |
9aac5db8 KW |
2356 | else if (category == LC_ALL) { |
2357 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); | |
a4f00dcc KW |
2358 | } |
2359 | ||
7d4bcc4a | 2360 | # endif |
a4f00dcc KW |
2361 | |
2362 | } | |
2363 | ||
2364 | #endif | |
2365 | ||
33e5a354 KW |
2366 | retval = save_to_buffer(do_setlocale_r(category, locale), |
2367 | &PL_setlocale_buf, &PL_setlocale_bufsize, 0); | |
9aac5db8 KW |
2368 | SAVE_ERRNO; |
2369 | ||
2370 | #if defined(USE_LOCALE_NUMERIC) && defined(LC_ALL) | |
2371 | ||
2372 | if (locale == NULL && category == LC_ALL) { | |
2373 | RESTORE_LC_NUMERIC(); | |
2374 | } | |
2375 | ||
2376 | #endif | |
a4f00dcc KW |
2377 | |
2378 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
2379 | "%s:%d: %s\n", __FILE__, __LINE__, | |
2380 | setlocale_debug_string(category, locale, retval))); | |
7d4bcc4a | 2381 | |
9aac5db8 KW |
2382 | RESTORE_ERRNO; |
2383 | ||
2384 | if (! retval) { | |
a4f00dcc KW |
2385 | return NULL; |
2386 | } | |
2387 | ||
9aac5db8 | 2388 | /* If locale == NULL, we are just querying the state */ |
a4f00dcc | 2389 | if (locale == NULL) { |
a4f00dcc KW |
2390 | return retval; |
2391 | } | |
a4f00dcc | 2392 | |
1159483a KW |
2393 | /* Now that have switched locales, we have to update our records to |
2394 | * correspond. */ | |
a4f00dcc | 2395 | |
1159483a | 2396 | switch (category) { |
a4f00dcc | 2397 | |
1159483a | 2398 | #ifdef USE_LOCALE_CTYPE |
a4f00dcc | 2399 | |
1159483a KW |
2400 | case LC_CTYPE: |
2401 | new_ctype(retval); | |
2402 | break; | |
a4f00dcc | 2403 | |
1159483a | 2404 | #endif |
a4f00dcc KW |
2405 | #ifdef USE_LOCALE_COLLATE |
2406 | ||
1159483a KW |
2407 | case LC_COLLATE: |
2408 | new_collate(retval); | |
2409 | break; | |
a4f00dcc | 2410 | |
1159483a | 2411 | #endif |
a4f00dcc KW |
2412 | #ifdef USE_LOCALE_NUMERIC |
2413 | ||
1159483a KW |
2414 | case LC_NUMERIC: |
2415 | new_numeric(retval); | |
2416 | break; | |
a4f00dcc | 2417 | |
1159483a KW |
2418 | #endif |
2419 | #ifdef LC_ALL | |
a4f00dcc | 2420 | |
1159483a | 2421 | case LC_ALL: |
a4f00dcc | 2422 | |
1159483a KW |
2423 | /* LC_ALL updates all the things we care about. The values may not |
2424 | * be the same as 'retval', as the locale "" may have set things | |
2425 | * individually */ | |
a4f00dcc | 2426 | |
1159483a | 2427 | # ifdef USE_LOCALE_CTYPE |
a4f00dcc | 2428 | |
b0bde642 | 2429 | newlocale = savepv(do_setlocale_c(LC_CTYPE, NULL)); |
1159483a | 2430 | new_ctype(newlocale); |
b0bde642 | 2431 | Safefree(newlocale); |
a4f00dcc | 2432 | |
1159483a KW |
2433 | # endif /* USE_LOCALE_CTYPE */ |
2434 | # ifdef USE_LOCALE_COLLATE | |
2435 | ||
b0bde642 | 2436 | newlocale = savepv(do_setlocale_c(LC_COLLATE, NULL)); |
1159483a | 2437 | new_collate(newlocale); |
b0bde642 | 2438 | Safefree(newlocale); |
a4f00dcc | 2439 | |
7d4bcc4a | 2440 | # endif |
1159483a | 2441 | # ifdef USE_LOCALE_NUMERIC |
a4f00dcc | 2442 | |
b0bde642 | 2443 | newlocale = savepv(do_setlocale_c(LC_NUMERIC, NULL)); |
1159483a | 2444 | new_numeric(newlocale); |
b0bde642 | 2445 | Safefree(newlocale); |
a4f00dcc | 2446 | |
1159483a KW |
2447 | # endif /* USE_LOCALE_NUMERIC */ |
2448 | #endif /* LC_ALL */ | |
a4f00dcc | 2449 | |
1159483a KW |
2450 | default: |
2451 | break; | |
a4f00dcc KW |
2452 | } |
2453 | ||
2454 | return retval; | |
2455 | ||
d36adde0 KW |
2456 | #endif |
2457 | ||
f7416781 KW |
2458 | } |
2459 | ||
2460 | PERL_STATIC_INLINE const char * | |
2461 | S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset) | |
2462 | { | |
2463 | /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size', | |
2464 | * growing it if necessary */ | |
2465 | ||
0b6697c5 | 2466 | Size_t string_size; |
f7416781 KW |
2467 | |
2468 | PERL_ARGS_ASSERT_SAVE_TO_BUFFER; | |
2469 | ||
0b6697c5 KW |
2470 | if (! string) { |
2471 | return NULL; | |
2472 | } | |
2473 | ||
2474 | string_size = strlen(string) + offset + 1; | |
2475 | ||
f7416781 KW |
2476 | if (*buf_size == 0) { |
2477 | Newx(*buf, string_size, char); | |
2478 | *buf_size = string_size; | |
2479 | } | |
2480 | else if (string_size > *buf_size) { | |
2481 | Renew(*buf, string_size, char); | |
2482 | *buf_size = string_size; | |
2483 | } | |
2484 | ||
2485 | Copy(string, *buf + offset, string_size - offset, char); | |
2486 | return *buf; | |
2487 | } | |
2488 | ||
2489 | /* | |
2490 | ||
f7416781 KW |
2491 | =for apidoc Perl_langinfo |
2492 | ||
ce181bf6 | 2493 | This is an (almost) drop-in replacement for the system C<L<nl_langinfo(3)>>, |
f7416781 KW |
2494 | taking the same C<item> parameter values, and returning the same information. |
2495 | But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks | |
2496 | of Perl's locale handling from your code, and can be used on systems that lack | |
2497 | a native C<nl_langinfo>. | |
2498 | ||
2499 | Expanding on these: | |
2500 | ||
2501 | =over | |
2502 | ||
2503 | =item * | |
2504 | ||
ce181bf6 KW |
2505 | The reason it isn't quite a drop-in replacement is actually an advantage. The |
2506 | only difference is that it returns S<C<const char *>>, whereas plain | |
2507 | C<nl_langinfo()> returns S<C<char *>>, but you are (only by documentation) | |
2508 | forbidden to write into the buffer. By declaring this C<const>, the compiler | |
2509 | enforces this restriction, so if it is violated, you know at compilation time, | |
2510 | rather than getting segfaults at runtime. | |
2511 | ||
2512 | =item * | |
2513 | ||
69e2275e | 2514 | It delivers the correct results for the C<RADIXCHAR> and C<THOUSEP> items, |
f7416781 KW |
2515 | without you having to write extra code. The reason for the extra code would be |
2516 | because these are from the C<LC_NUMERIC> locale category, which is normally | |
9487427b KW |
2517 | kept set by Perl so that the radix is a dot, and the separator is the empty |
2518 | string, no matter what the underlying locale is supposed to be, and so to get | |
2519 | the expected results, you have to temporarily toggle into the underlying | |
2520 | locale, and later toggle back. (You could use plain C<nl_langinfo> and | |
2521 | C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this but then you wouldn't get | |
2522 | the other advantages of C<Perl_langinfo()>; not keeping C<LC_NUMERIC> in the C | |
2523 | (or equivalent) locale would break a lot of CPAN, which is expecting the radix | |
2524 | (decimal point) character to be a dot.) | |
ce181bf6 KW |
2525 | |
2526 | =item * | |
2527 | ||
2528 | The system function it replaces can have its static return buffer trashed, | |
f1460a66 | 2529 | not only by a subsequent call to that function, but by a C<freelocale>, |
ce181bf6 KW |
2530 | C<setlocale>, or other locale change. The returned buffer of this function is |
2531 | not changed until the next call to it, so the buffer is never in a trashed | |
2532 | state. | |
f7416781 KW |
2533 | |
2534 | =item * | |
2535 | ||
ce181bf6 KW |
2536 | Its return buffer is per-thread, so it also is never overwritten by a call to |
2537 | this function from another thread; unlike the function it replaces. | |
2538 | ||
2539 | =item * | |
2540 | ||
2541 | But most importantly, it works on systems that don't have C<nl_langinfo>, such | |
2542 | as Windows, hence makes your code more portable. Of the fifty-some possible | |
2543 | items specified by the POSIX 2008 standard, | |
f7416781 | 2544 | L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>, |
8d72e74e KW |
2545 | only one is completely unimplemented, though on non-Windows platforms, another |
2546 | significant one is also not implemented). It uses various techniques to | |
2547 | recover the other items, including calling C<L<localeconv(3)>>, and | |
2548 | C<L<strftime(3)>>, both of which are specified in C89, so should be always be | |
2549 | available. Later C<strftime()> versions have additional capabilities; C<""> is | |
2550 | returned for those not available on your system. | |
ce181bf6 KW |
2551 | |
2552 | It is important to note that when called with an item that is recovered by | |
2553 | using C<localeconv>, the buffer from any previous explicit call to | |
2554 | C<localeconv> will be overwritten. This means you must save that buffer's | |
c3997e39 KW |
2555 | contents if you need to access them after a call to this function. (But note |
2556 | that you might not want to be using C<localeconv()> directly anyway, because of | |
2557 | issues like the ones listed in the second item of this list (above) for | |
2558 | C<RADIXCHAR> and C<THOUSEP>. You can use the methods given in L<perlcall> to | |
2559 | call L<POSIX/localeconv> and avoid all the issues, but then you have a hash to | |
2560 | unpack). | |
9f0bc911 | 2561 | |
6201c220 KW |
2562 | The details for those items which may deviate from what this emulation returns |
2563 | and what a native C<nl_langinfo()> would return are specified in | |
2564 | L<I18N::Langinfo>. | |
f7416781 | 2565 | |
ce181bf6 KW |
2566 | =back |
2567 | ||
f7416781 KW |
2568 | When using C<Perl_langinfo> on systems that don't have a native |
2569 | C<nl_langinfo()>, you must | |
2570 | ||
2571 | #include "perl_langinfo.h" | |
2572 | ||
2573 | before the C<perl.h> C<#include>. You can replace your C<langinfo.h> | |
2574 | C<#include> with this one. (Doing it this way keeps out the symbols that plain | |
c3997e39 KW |
2575 | C<langinfo.h> would try to import into the namespace for code that doesn't need |
2576 | it.) | |
f7416781 | 2577 | |
f7416781 KW |
2578 | The original impetus for C<Perl_langinfo()> was so that code that needs to |
2579 | find out the current currency symbol, floating point radix character, or digit | |
2580 | grouping separator can use, on all systems, the simpler and more | |
2581 | thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a | |
2582 | pain to make thread-friendly. For other fields returned by C<localeconv>, it | |
2583 | is better to use the methods given in L<perlcall> to call | |
2584 | L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly. | |
2585 | ||
2586 | =cut | |
2587 | ||
2588 | */ | |
2589 | ||
2590 | const char * | |
2591 | #ifdef HAS_NL_LANGINFO | |
2592 | Perl_langinfo(const nl_item item) | |
2593 | #else | |
2594 | Perl_langinfo(const int item) | |
2595 | #endif | |
2596 | { | |
f61748ac KW |
2597 | return my_nl_langinfo(item, TRUE); |
2598 | } | |
2599 | ||
59a15af0 | 2600 | STATIC const char * |
f61748ac KW |
2601 | #ifdef HAS_NL_LANGINFO |
2602 | S_my_nl_langinfo(const nl_item item, bool toggle) | |
2603 | #else | |
2604 | S_my_nl_langinfo(const int item, bool toggle) | |
2605 | #endif | |
2606 | { | |
ae74815b | 2607 | dTHX; |
0b6697c5 | 2608 | const char * retval; |
f7416781 | 2609 | |
d36adde0 KW |
2610 | #ifdef USE_LOCALE_NUMERIC |
2611 | ||
5a854ab3 KW |
2612 | /* We only need to toggle into the underlying LC_NUMERIC locale for these |
2613 | * two items, and only if not already there */ | |
4e6826bf | 2614 | if (toggle && (( item != RADIXCHAR && item != THOUSEP) |
5a854ab3 | 2615 | || PL_numeric_underlying)) |
d36adde0 KW |
2616 | |
2617 | #endif /* No toggling needed if not using LC_NUMERIC */ | |
2618 | ||
5a854ab3 | 2619 | toggle = FALSE; |
5a854ab3 | 2620 | |
ab340fff | 2621 | #if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */ |
ee90eb2d | 2622 | # if ! defined(HAS_THREAD_SAFE_NL_LANGINFO_L) \ |
80306bb6 | 2623 | || ! defined(HAS_POSIX_2008_LOCALE) \ |
9289d4dc | 2624 | || ! defined(HAS_DUPLOCALE) |
f7416781 | 2625 | |
ab340fff | 2626 | /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC |
ae74815b KW |
2627 | * for those items dependent on it. This must be copied to a buffer before |
2628 | * switching back, as some systems destroy the buffer when setlocale() is | |
2629 | * called */ | |
f7416781 | 2630 | |
038d3702 KW |
2631 | { |
2632 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; | |
2633 | ||
b2fee59c | 2634 | if (toggle) { |
038d3702 | 2635 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
b2fee59c | 2636 | } |
f7416781 | 2637 | |
5acc4454 KW |
2638 | LOCALE_LOCK; /* Prevent interference from another thread executing |
2639 | this code section (the only call to nl_langinfo in | |
2640 | the core) */ | |
2641 | ||
ee90eb2d | 2642 | |
628ff75a KW |
2643 | /* Copy to a per-thread buffer, which is also one that won't be |
2644 | * destroyed by a subsequent setlocale(), such as the | |
2645 | * RESTORE_LC_NUMERIC may do just below. */ | |
0b6697c5 KW |
2646 | retval = save_to_buffer(nl_langinfo(item), |
2647 | &PL_langinfo_buf, &PL_langinfo_bufsize, 0); | |
f7416781 | 2648 | |
5acc4454 KW |
2649 | LOCALE_UNLOCK; |
2650 | ||
b2fee59c | 2651 | if (toggle) { |
038d3702 | 2652 | RESTORE_LC_NUMERIC(); |
b2fee59c | 2653 | } |
f7416781 KW |
2654 | } |
2655 | ||
ab340fff KW |
2656 | # else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */ |
2657 | ||
5a854ab3 | 2658 | { |
b2fee59c KW |
2659 | bool do_free = FALSE; |
2660 | locale_t cur = uselocale((locale_t) 0); | |
ab340fff | 2661 | |
b2fee59c KW |
2662 | if (cur == LC_GLOBAL_LOCALE) { |
2663 | cur = duplocale(LC_GLOBAL_LOCALE); | |
2664 | do_free = TRUE; | |
2665 | } | |
ab340fff | 2666 | |
d36adde0 KW |
2667 | # ifdef USE_LOCALE_NUMERIC |
2668 | ||
b2fee59c | 2669 | if (toggle) { |
e1aa2579 KW |
2670 | if (PL_underlying_numeric_obj) { |
2671 | cur = PL_underlying_numeric_obj; | |
2672 | } | |
2673 | else { | |
2674 | cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur); | |
2675 | do_free = TRUE; | |
2676 | } | |
b2fee59c | 2677 | } |
ab340fff | 2678 | |
d36adde0 KW |
2679 | # endif |
2680 | ||
628ff75a KW |
2681 | /* We have to save it to a buffer, because the freelocale() just below |
2682 | * can invalidate the internal one */ | |
0b6697c5 KW |
2683 | retval = save_to_buffer(nl_langinfo_l(item, cur), |
2684 | &PL_langinfo_buf, &PL_langinfo_bufsize, 0); | |
ee90eb2d | 2685 | |
b2fee59c KW |
2686 | if (do_free) { |
2687 | freelocale(cur); | |
2688 | } | |
5a854ab3 | 2689 | } |
ab340fff | 2690 | |
c1566110 KW |
2691 | # endif |
2692 | ||
0b6697c5 | 2693 | if (strEQ(retval, "")) { |
4e6826bf | 2694 | if (item == YESSTR) { |
c1566110 KW |
2695 | return "yes"; |
2696 | } | |
4e6826bf | 2697 | if (item == NOSTR) { |
c1566110 KW |
2698 | return "no"; |
2699 | } | |
2700 | } | |
2701 | ||
0b6697c5 | 2702 | return retval; |
ab340fff | 2703 | |
f7416781 | 2704 | #else /* Below, emulate nl_langinfo as best we can */ |
43dd6b15 KW |
2705 | |
2706 | { | |
2707 | ||
f7416781 KW |
2708 | # ifdef HAS_LOCALECONV |
2709 | ||
43dd6b15 | 2710 | const struct lconv* lc; |
628ff75a | 2711 | const char * temp; |
038d3702 | 2712 | DECLARATION_FOR_LC_NUMERIC_MANIPULATION; |
f7416781 | 2713 | |
69c5e0db KW |
2714 | # ifdef TS_W32_BROKEN_LOCALECONV |
2715 | ||
2716 | const char * save_global; | |
2717 | const char * save_thread; | |
2718 | int needed_size; | |
2719 | char * ptr; | |
2720 | char * e; | |
2721 | char * item_start; | |
2722 | ||
2723 | # endif | |
f7416781 KW |
2724 | # endif |
2725 | # ifdef HAS_STRFTIME | |
2726 | ||
43dd6b15 KW |
2727 | struct tm tm; |
2728 | bool return_format = FALSE; /* Return the %format, not the value */ | |
2729 | const char * format; | |
f7416781 KW |
2730 | |
2731 | # endif | |
2732 | ||
43dd6b15 KW |
2733 | /* We copy the results to a per-thread buffer, even if not |
2734 | * multi-threaded. This is in part to simplify this code, and partly | |
2735 | * because we need a buffer anyway for strftime(), and partly because a | |
2736 | * call of localeconv() could otherwise wipe out the buffer, and the | |
2737 | * programmer would not be expecting this, as this is a nl_langinfo() | |
2738 | * substitute after all, so s/he might be thinking their localeconv() | |
2739 | * is safe until another localeconv() call. */ | |
f7416781 | 2740 | |
43dd6b15 KW |
2741 | switch (item) { |
2742 | Size_t len; | |
f7416781 | 2743 | |
8d72e74e | 2744 | /* This is unimplemented */ |
4e6826bf | 2745 | case ERA: /* For use with strftime() %E modifier */ |
f7416781 | 2746 | |
43dd6b15 KW |
2747 | default: |
2748 | return ""; | |
f7416781 | 2749 | |
43dd6b15 | 2750 | /* We use only an English set, since we don't know any more */ |
4e6826bf KW |
2751 | case YESEXPR: return "^[+1yY]"; |
2752 | case YESSTR: return "yes"; | |
2753 | case NOEXPR: return "^[-0nN]"; | |
2754 | case NOSTR: return "no"; | |
2755 | ||
8d72e74e KW |
2756 | case CODESET: |
2757 | ||
2758 | # ifndef WIN32 | |
2759 | ||
2760 | /* On non-windows, this is unimplemented, in part because of | |
2761 | * inconsistencies between vendors. The Darwin native | |
2762 | * nl_langinfo() implementation simply looks at everything past | |
2763 | * any dot in the name, but that doesn't work for other | |
2764 | * vendors. Many Linux locales that don't have UTF-8 in their | |
2765 | * names really are UTF-8, for example; z/OS locales that do | |
2766 | * have UTF-8 in their names, aren't really UTF-8 */ | |
2767 | return ""; | |
2768 | ||
2769 | # else | |
2770 | ||
2771 | { /* But on Windows, the name does seem to be consistent, so | |
2772 | use that. */ | |
2773 | const char * p; | |
2774 | const char * first; | |
2775 | Size_t offset = 0; | |
2776 | const char * name = my_setlocale(LC_CTYPE, NULL); | |
2777 | ||
2778 | if (isNAME_C_OR_POSIX(name)) { | |
2779 | return "ANSI_X3.4-1968"; | |
2780 | } | |
2781 | ||
2782 | /* Find the dot in the locale name */ | |
2783 | first = (const char *) strchr(name, '.'); | |
2784 | if (! first) { | |
2785 | first = name; | |
2786 | goto has_nondigit; | |
2787 | } | |
2788 | ||
2789 | /* Look at everything past the dot */ | |
2790 | first++; | |
2791 | p = first; | |
2792 | ||
2793 | while (*p) { | |
2794 | if (! isDIGIT(*p)) { | |
2795 | goto has_nondigit; | |
2796 | } | |
2797 | ||
2798 | p++; | |
2799 | } | |
2800 | ||
2801 | /* Here everything past the dot is a digit. Treat it as a | |
2802 | * code page */ | |
18503cbd | 2803 | retval = save_to_buffer("CP", &PL_langinfo_buf, |
32a62865 | 2804 | &PL_langinfo_bufsize, 0); |
8d72e74e | 2805 | offset = STRLENs("CP"); |
f7416781 | 2806 | |
8d72e74e KW |
2807 | has_nondigit: |
2808 | ||
2809 | retval = save_to_buffer(first, &PL_langinfo_buf, | |
2810 | &PL_langinfo_bufsize, offset); | |
2811 | } | |
2812 | ||
2813 | break; | |
2814 | ||
2815 | # endif | |
f7416781 KW |
2816 | # ifdef HAS_LOCALECONV |
2817 | ||
4e6826bf | 2818 | case CRNCYSTR: |
f7416781 | 2819 | |
43dd6b15 KW |
2820 | /* We don't bother with localeconv_l() because any system that |
2821 | * has it is likely to also have nl_langinfo() */ | |
291a84fb | 2822 | |
69c5e0db KW |
2823 | LOCALE_LOCK_V; /* Prevent interference with other threads |
2824 | using localeconv() */ | |
2825 | ||
2826 | # ifdef TS_W32_BROKEN_LOCALECONV | |
2827 | ||
2828 | /* This is a workaround for a Windows bug prior to VS 15. | |
2829 | * What we do here is, while locked, switch to the global | |
2830 | * locale so localeconv() works; then switch back just before | |
2831 | * the unlock. This can screw things up if some thread is | |
2832 | * already using the global locale while assuming no other is. | |
2833 | * A different workaround would be to call GetCurrencyFormat on | |
2834 | * a known value, and parse it; patches welcome | |
2835 | * | |
2836 | * We have to use LC_ALL instead of LC_MONETARY because of | |
2837 | * another bug in Windows */ | |
2838 | ||
2839 | save_thread = savepv(my_setlocale(LC_ALL, NULL)); | |
2840 | _configthreadlocale(_DISABLE_PER_THREAD_LOCALE); | |
2841 | save_global= savepv(my_setlocale(LC_ALL, NULL)); | |
2842 | my_setlocale(LC_ALL, save_thread); | |
2843 | ||
2844 | # endif | |
5acc4454 | 2845 | |
43dd6b15 KW |
2846 | lc = localeconv(); |
2847 | if ( ! lc | |
2848 | || ! lc->currency_symbol | |
2849 | || strEQ("", lc->currency_symbol)) | |
2850 | { | |
69c5e0db | 2851 | LOCALE_UNLOCK_V; |
43dd6b15 KW |
2852 | return ""; |
2853 | } | |
f7416781 | 2854 | |
43dd6b15 | 2855 | /* Leave the first spot empty to be filled in below */ |
0b6697c5 KW |
2856 | retval = save_to_buffer(lc->currency_symbol, &PL_langinfo_buf, |
2857 | &PL_langinfo_bufsize, 1); | |
43dd6b15 KW |
2858 | if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, "")) |
2859 | { /* khw couldn't figure out how the localedef specifications | |
2860 | would show that the $ should replace the radix; this is | |
2861 | just a guess as to how it might work.*/ | |
a34edee3 | 2862 | PL_langinfo_buf[0] = '.'; |
43dd6b15 KW |
2863 | } |
2864 | else if (lc->p_cs_precedes) { | |
a34edee3 | 2865 | PL_langinfo_buf[0] = '-'; |
43dd6b15 KW |
2866 | } |
2867 | else { | |
a34edee3 | 2868 | PL_langinfo_buf[0] = '+'; |
43dd6b15 | 2869 | } |
f7416781 | 2870 | |
69c5e0db KW |
2871 | # ifdef TS_W32_BROKEN_LOCALECONV |
2872 | ||
2873 | my_setlocale(LC_ALL, save_global); | |
2874 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); | |
2875 | my_setlocale(LC_ALL, save_thread); | |
2876 | Safefree(save_global); | |
2877 | Safefree(save_thread); | |
2878 | ||
2879 | # endif | |
2880 | ||
2881 | LOCALE_UNLOCK_V; | |
43dd6b15 | 2882 | break; |
f7416781 | 2883 | |
69c5e0db KW |
2884 | # ifdef TS_W32_BROKEN_LOCALECONV |
2885 | ||
4e6826bf | 2886 | case RADIXCHAR: |
69c5e0db KW |
2887 | |
2888 | /* For this, we output a known simple floating point number to | |
2889 | * a buffer, and parse it, looking for the radix */ | |
2890 | ||
2891 | if (toggle) { | |
2892 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); | |
2893 | } | |
2894 | ||
2895 | if (PL_langinfo_bufsize < 10) { | |
2896 | PL_langinfo_bufsize = 10; | |
2897 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); | |
2898 | } | |
2899 | ||
2900 | needed_size = my_snprintf(PL_langinfo_buf, PL_langinfo_bufsize, | |
2901 | "%.1f", 1.5); | |
2902 | if (needed_size >= (int) PL_langinfo_bufsize) { | |
2903 | PL_langinfo_bufsize = needed_size + 1; | |
2904 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); | |
2905 | needed_size = my_snprintf(PL_langinfo_buf, PL_langinfo_bufsize, | |
2906 | "%.1f", 1.5); | |
2907 | assert(needed_size < (int) PL_langinfo_bufsize); | |
2908 | } | |
2909 | ||
2910 | ptr = PL_langinfo_buf; | |
2911 | e = PL_langinfo_buf + PL_langinfo_bufsize; | |
2912 | ||
2913 | /* Find the '1' */ | |
2914 | while (ptr < e && *ptr != '1') { | |
2915 | ptr++; | |
2916 | } | |
2917 | ptr++; | |
2918 | ||
2919 | /* Find the '5' */ | |
2920 | item_start = ptr; | |
2921 | while (ptr < e && *ptr != '5') { | |
2922 | ptr++; | |
2923 | } | |
2924 | ||
2925 | /* Everything in between is the radix string */ | |
2926 | if (ptr >= e) { | |
2927 | PL_langinfo_buf[0] = '?'; | |
2928 | PL_langinfo_buf[1] = '\0'; | |
2929 | } | |
2930 | else { | |
2931 | *ptr = '\0'; | |
2932 | Move(item_start, PL_langinfo_buf, ptr - PL_langinfo_buf, char); | |
2933 | } | |
2934 | ||
2935 | if (toggle) { | |
2936 | RESTORE_LC_NUMERIC(); | |
2937 | } | |
2938 | ||
2939 | retval = PL_langinfo_buf; | |
2940 | break; | |
2941 | ||
2942 | # else | |
2943 | ||
2944 | case RADIXCHAR: /* No special handling needed */ | |
2945 | ||
2946 | # endif | |
2947 | ||
4e6826bf | 2948 | case THOUSEP: |
f7416781 | 2949 | |
43dd6b15 | 2950 | if (toggle) { |
038d3702 | 2951 | STORE_LC_NUMERIC_FORCE_TO_UNDERLYING(); |
c0d737a8 | 2952 | } |
f7416781 | 2953 | |
69c5e0db KW |
2954 | LOCALE_LOCK_V; /* Prevent interference with other threads |
2955 | using localeconv() */ | |
2956 | ||
2957 | # ifdef TS_W32_BROKEN_LOCALECONV | |
2958 | ||
2959 | /* This should only be for the thousands separator. A | |
2960 | * different work around would be to use GetNumberFormat on a | |
2961 | * known value and parse the result to find the separator */ | |
2962 | save_thread = savepv(my_setlocale(LC_ALL, NULL)); | |
2963 | _configthreadlocale(_DISABLE_PER_THREAD_LOCALE); | |
2964 | save_global = savepv(my_setlocale(LC_ALL, NULL)); | |
2965 | my_setlocale(LC_ALL, save_thread); | |
2966 | # if 0 | |
2967 | /* This is the start of code that for broken Windows replaces | |
2968 | * the above and below code, and instead calls | |
2969 | * GetNumberFormat() and then would parse that to find the | |
2970 | * thousands separator. It needs to handle UTF-16 vs -8 | |
2971 | * issues. */ | |
2972 | ||
2973 | needed_size = GetNumberFormatEx(PL_numeric_name, 0, "1234.5", NULL, PL_langinfo_buf, PL_langinfo_bufsize); | |
2974 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
2975 | "%s: %d: return from GetNumber, count=%d, val=%s\n", | |
2976 | __FILE__, __LINE__, needed_size, PL_langinfo_buf)); | |
2977 | ||
2978 | # endif | |
2979 | # endif | |
5acc4454 | 2980 | |
43dd6b15 KW |
2981 | lc = localeconv(); |
2982 | if (! lc) { | |
628ff75a | 2983 | temp = ""; |
33394adc | 2984 | } |
43dd6b15 | 2985 | else { |
4e6826bf | 2986 | temp = (item == RADIXCHAR) |
43dd6b15 KW |
2987 | ? lc->decimal_point |
2988 | : lc->thousands_sep; | |
628ff75a KW |
2989 | if (! temp) { |
2990 | temp = ""; | |
43dd6b15 KW |
2991 | } |
2992 | } | |
f7416781 | 2993 | |
0b6697c5 KW |
2994 | retval = save_to_buffer(temp, &PL_langinfo_buf, |
2995 | &PL_langinfo_bufsize, 0); | |
f7416781 | 2996 | |
69c5e0db KW |
2997 | # ifdef TS_W32_BROKEN_LOCALECONV |
2998 | ||
2999 | my_setlocale(LC_ALL, save_global); | |
3000 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); | |
3001 | my_setlocale(LC_ALL, save_thread); | |
3002 | Safefree(save_global); | |
3003 | Safefree(save_thread); | |
3004 | ||
3005 | # endif | |
3006 | ||
3007 | LOCALE_UNLOCK_V; | |
5acc4454 | 3008 | |
43dd6b15 | 3009 | if (toggle) { |
038d3702 | 3010 | RESTORE_LC_NUMERIC(); |
43dd6b15 | 3011 | } |
f7416781 | 3012 | |
43dd6b15 | 3013 | break; |
f7416781 KW |
3014 | |
3015 | # endif | |
3016 | # ifdef HAS_STRFTIME | |
3017 | ||
43dd6b15 KW |
3018 | /* These are defined by C89, so we assume that strftime supports |
3019 | * them, and so are returned unconditionally; they may not be what | |
3020 | * the locale actually says, but should give good enough results | |
3021 | * for someone using them as formats (as opposed to trying to parse | |
3022 | * them to figure out what the locale says). The other format | |
3023 | * items are actually tested to verify they work on the platform */ | |
4e6826bf KW |
3024 | case D_FMT: return "%x"; |
3025 | case T_FMT: return "%X"; | |
3026 | case D_T_FMT: return "%c"; | |
43dd6b15 KW |
3027 | |
3028 | /* These formats are only available in later strfmtime's */ | |
4e6826bf | 3029 | case ERA_D_FMT: case ERA_T_FMT: case ERA_D_T_FMT: case T_FMT_AMPM: |
43dd6b15 KW |
3030 | |
3031 | /* The rest can be gotten from most versions of strftime(). */ | |
4e6826bf KW |
3032 | case ABDAY_1: case ABDAY_2: case ABDAY_3: |
3033 | case ABDAY_4: case ABDAY_5: case ABDAY_6: case ABDAY_7: | |
3034 | case ALT_DIGITS: | |
3035 | case AM_STR: case PM_STR: | |
3036 | case ABMON_1: case ABMON_2: case ABMON_3: case ABMON_4: | |
3037 | case ABMON_5: case ABMON_6: case ABMON_7: case ABMON_8: | |
3038 | case ABMON_9: case ABMON_10: case ABMON_11: case ABMON_12: | |
3039 | case DAY_1: case DAY_2: case DAY_3: case DAY_4: | |
3040 | case DAY_5: case DAY_6: case DAY_7: | |
3041 | case MON_1: case MON_2: case MON_3: case MON_4: | |
3042 | case MON_5: case MON_6: case MON_7: case MON_8: | |
3043 | case MON_9: case MON_10: case MON_11: case MON_12: | |
43dd6b15 KW |
3044 | |
3045 | LOCALE_LOCK; | |
3046 | ||
3047 | init_tm(&tm); /* Precaution against core dumps */ | |
3048 | tm.tm_sec = 30; | |
3049 | tm.tm_min = 30; | |
3050 | tm.tm_hour = 6; | |
3051 | tm.tm_year = 2017 - 1900; | |
3052 | tm.tm_wday = 0; | |
3053 | tm.tm_mon = 0; | |
3054 | switch (item) { | |
3055 | default: | |
3056 | LOCALE_UNLOCK; | |
3057 | Perl_croak(aTHX_ | |
3058 | "panic: %s: %d: switch case: %d problem", | |
3059 | __FILE__, __LINE__, item); | |
3060 | NOT_REACHED; /* NOTREACHED */ | |
3061 | ||
4e6826bf KW |
3062 | case PM_STR: tm.tm_hour = 18; |
3063 | case AM_STR: | |
43dd6b15 KW |
3064 | format = "%p"; |
3065 | break; | |
3066 | ||
4e6826bf KW |
3067 | case ABDAY_7: tm.tm_wday++; |
3068 | case ABDAY_6: tm.tm_wday++; | |
3069 | case ABDAY_5: tm.tm_wday++; | |
3070 | case ABDAY_4: tm.tm_wday++; | |
3071 | case ABDAY_3: tm.tm_wday++; | |
3072 | case ABDAY_2: tm.tm_wday++; | |
3073 | case ABDAY_1: | |
43dd6b15 KW |
3074 | format = "%a"; |
3075 | break; | |
3076 | ||
4e6826bf KW |
3077 | case DAY_7: tm.tm_wday++; |
3078 | case DAY_6: tm.tm_wday++; | |
3079 | case DAY_5: tm.tm_wday++; | |
3080 | case DAY_4: tm.tm_wday++; | |
3081 | case DAY_3: tm.tm_wday++; | |
3082 | case DAY_2: tm.tm_wday++; | |
3083 | case DAY_1: | |
43dd6b15 KW |
3084 | format = "%A"; |
3085 | break; | |
3086 | ||
4e6826bf KW |
3087 | case ABMON_12: tm.tm_mon++; |
3088 | case ABMON_11: tm.tm_mon++; | |
3089 | case ABMON_10: tm.tm_mon++; | |
3090 | case ABMON_9: tm.tm_mon++; | |
3091 | case ABMON_8: tm.tm_mon++; | |
3092 | case ABMON_7: tm.tm_mon++; | |
3093 | case ABMON_6: tm.tm_mon++; | |
3094 | case ABMON_5: tm.tm_mon++; | |
3095 | case ABMON_4: tm.tm_mon++; | |
3096 | case ABMON_3: tm.tm_mon++; | |
3097 | case ABMON_2: tm.tm_mon++; | |
3098 | case ABMON_1: | |
43dd6b15 KW |
3099 | format = "%b"; |
3100 | break; | |
3101 | ||
4e6826bf KW |
3102 | case MON_12: tm.tm_mon++; |
3103 | case MON_11: tm.tm_mon++; | |
3104 | case MON_10: tm.tm_mon++; | |
3105 | case MON_9: tm.tm_mon++; | |
3106 | case MON_8: tm.tm_mon++; | |
3107 | case MON_7: tm.tm_mon++; | |
3108 | case MON_6: tm.tm_mon++; | |
3109 | case MON_5: tm.tm_mon++; | |
3110 | case MON_4: tm.tm_mon++; | |
3111 | case MON_3: tm.tm_mon++; | |
3112 | case MON_2: tm.tm_mon++; | |
3113 | case MON_1: | |
43dd6b15 KW |
3114 | format = "%B"; |
3115 | break; | |
3116 | ||
4e6826bf | 3117 | case T_FMT_AMPM: |
43dd6b15 KW |
3118 | format = "%r"; |
3119 | return_format = TRUE; | |
3120 | break; | |
3121 | ||
4e6826bf | 3122 | case ERA_D_FMT: |
43dd6b15 KW |
3123 | format = "%Ex"; |
3124 | return_format = TRUE; | |
3125 | break; | |
3126 | ||
4e6826bf | 3127 | case ERA_T_FMT: |
43dd6b15 KW |
3128 | format = "%EX"; |
3129 | return_format = TRUE; | |
3130 | break; | |
3131 | ||
4e6826bf | 3132 | case ERA_D_T_FMT: |
43dd6b15 KW |
3133 | format = "%Ec"; |
3134 | return_format = TRUE; | |
3135 | break; | |
3136 | ||
4e6826bf | 3137 | case ALT_DIGITS: |
43dd6b15 KW |
3138 | tm.tm_wday = 0; |
3139 | format = "%Ow"; /* Find the alternate digit for 0 */ | |
3140 | break; | |
3141 | } | |
f7416781 | 3142 | |
43dd6b15 KW |
3143 | /* We can't use my_strftime() because it doesn't look at |
3144 | * tm_wday */ | |
3145 | while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize, | |
3146 | format, &tm)) | |
3147 | { | |
3148 | /* A zero return means one of: | |
3149 | * a) there wasn't enough space in PL_langinfo_buf | |
3150 | * b) the format, like a plain %p, returns empty | |
3151 | * c) it was an illegal format, though some | |
3152 | * implementations of strftime will just return the | |
3153 | * illegal format as a plain character sequence. | |
3154 | * | |
3155 | * To quickly test for case 'b)', try again but precede | |
3156 | * the format with a plain character. If that result is | |
3157 | * still empty, the problem is either 'a)' or 'c)' */ | |
3158 | ||
3159 | Size_t format_size = strlen(format) + 1; | |
3160 | Size_t mod_size = format_size + 1; | |
3161 | char * mod_format; | |
3162 | char * temp_result; | |
3163 | ||
3164 | Newx(mod_format, mod_size, char); | |
3165 | Newx(temp_result, PL_langinfo_bufsize, char); | |
6873aa47 | 3166 | *mod_format = ' '; |
43dd6b15 KW |
3167 | my_strlcpy(mod_format + 1, format, mod_size); |
3168 | len = strftime(temp_result, | |
3169 | PL_langinfo_bufsize, | |
3170 | mod_format, &tm); | |
3171 | Safefree(mod_format); | |
3172 | Safefree(temp_result); | |
3173 | ||
3174 | /* If 'len' is non-zero, it means that we had a case like | |
3175 | * %p which means the current locale doesn't use a.m. or | |
3176 | * p.m., and that is valid */ | |
3177 | if (len == 0) { | |
3178 | ||
3179 | /* Here, still didn't work. If we get well beyond a | |
3180 | * reasonable size, bail out to prevent an infinite | |
3181 | * loop. */ | |
3182 | ||
3183 | if (PL_langinfo_bufsize > 100 * format_size) { | |
3184 | *PL_langinfo_buf = '\0'; | |
3185 | } | |
3186 | else { | |
3187 | /* Double the buffer size to retry; Add 1 in case | |
3188 | * original was 0, so we aren't stuck at 0. */ | |
3189 | PL_langinfo_bufsize *= 2; | |
3190 | PL_langinfo_bufsize++; | |
3191 | Renew(PL_langinfo_buf, PL_langinfo_bufsize, char); | |
3192 | continue; | |
3193 | } | |
3194 | } | |
f7416781 | 3195 | |
f7416781 | 3196 | break; |
43dd6b15 | 3197 | } |
f7416781 | 3198 | |
43dd6b15 KW |
3199 | /* Here, we got a result. |
3200 | * | |
3201 | * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the | |
3202 | * alternate format for wday 0. If the value is the same as | |
3203 | * the normal 0, there isn't an alternate, so clear the buffer. | |
3204 | * */ | |
4e6826bf | 3205 | if ( item == ALT_DIGITS |
43dd6b15 KW |
3206 | && strEQ(PL_langinfo_buf, "0")) |
3207 | { | |
3208 | *PL_langinfo_buf = '\0'; | |
3209 | } | |
f7416781 | 3210 | |
43dd6b15 KW |
3211 | /* ALT_DIGITS is problematic. Experiments on it showed that |
3212 | * strftime() did not always work properly when going from | |
3213 | * alt-9 to alt-10. Only a few locales have this item defined, | |
3214 | * and in all of them on Linux that khw was able to find, | |
3215 | * nl_langinfo() merely returned the alt-0 character, possibly | |
3216 | * doubled. Most Unicode digits are in blocks of 10 | |
3217 | * consecutive code points, so that is sufficient information | |
3218 | * for those scripts, as we can infer alt-1, alt-2, .... But | |
3219 | * for a Japanese locale, a CJK ideographic 0 is returned, and | |
3220 | * the CJK digits are not in code point order, so you can't | |
3221 | * really infer anything. The localedef for this locale did | |
3222 | * specify the succeeding digits, so that strftime() works | |
3223 | * properly on them, without needing to infer anything. But | |
3224 | * the nl_langinfo() return did not give sufficient information | |
3225 | * for the caller to understand what's going on. So until | |
3226 | * there is evidence that it should work differently, this | |
3227 | * returns the alt-0 string for ALT_DIGITS. | |
3228 | * | |
3229 | * wday was chosen because its range is all a single digit. | |
3230 | * Things like tm_sec have two digits as the minimum: '00' */ | |
f7416781 | 3231 | |
43dd6b15 | 3232 | LOCALE_UNLOCK; |
f7416781 | 3233 | |
d1b150d9 KW |
3234 | retval = PL_langinfo_buf; |
3235 | ||
43dd6b15 KW |
3236 | /* If to return the format, not the value, overwrite the buffer |
3237 | * with it. But some strftime()s will keep the original format | |
3238 | * if illegal, so change those to "" */ | |
3239 | if (return_format) { | |
3240 | if (strEQ(PL_langinfo_buf, format)) { | |
f7416781 KW |
3241 | *PL_langinfo_buf = '\0'; |
3242 | } | |
43dd6b15 | 3243 | else { |
0b6697c5 KW |
3244 | retval = save_to_buffer(format, &PL_langinfo_buf, |
3245 | &PL_langinfo_bufsize, 0); | |
f7416781 KW |
3246 | } |
3247 | } | |
3248 | ||
3249 | break; | |
f7416781 KW |
3250 | |
3251 | # endif | |
3252 | ||
43dd6b15 | 3253 | } |
f7416781 KW |
3254 | } |
3255 | ||
0b6697c5 | 3256 | return retval; |
f7416781 KW |
3257 | |
3258 | #endif | |
3259 | ||
a4f00dcc | 3260 | } |
b385bb4d | 3261 | |
98994639 HS |
3262 | /* |
3263 | * Initialize locale awareness. | |
3264 | */ | |
3265 | int | |
3266 | Perl_init_i18nl10n(pTHX_ int printwarn) | |
3267 | { | |
0e92a118 KW |
3268 | /* printwarn is |
3269 | * | |
3270 | * 0 if not to output warning when setup locale is bad | |
3271 | * 1 if to output warning based on value of PERL_BADLANG | |
3272 | * >1 if to output regardless of PERL_BADLANG | |
3273 | * | |
3274 | * returns | |
98994639 | 3275 | * 1 = set ok or not applicable, |
0e92a118 KW |
3276 | * 0 = fallback to a locale of lower priority |
3277 | * -1 = fallback to all locales failed, not even to the C locale | |
6b058d42 KW |
3278 | * |
3279 | * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is | |
3280 | * set, debugging information is output. | |
3281 | * | |
3282 | * This looks more complicated than it is, mainly due to the #ifdefs. | |
3283 | * | |
3284 | * We try to set LC_ALL to the value determined by the environment. If | |
3285 | * there is no LC_ALL on this platform, we try the individual categories we | |
3286 | * know about. If this works, we are done. | |
3287 | * | |
3288 | * But if it doesn't work, we have to do something else. We search the | |
3289 | * environment variables ourselves instead of relying on the system to do | |
3290 | * it. We look at, in order, LC_ALL, LANG, a system default locale (if we | |
3291 | * think there is one), and the ultimate fallback "C". This is all done in | |
3292 | * the same loop as above to avoid duplicating code, but it makes things | |
7d4bcc4a KW |
3293 | * more complex. The 'trial_locales' array is initialized with just one |
3294 | * element; it causes the behavior described in the paragraph above this to | |
3295 | * happen. If that fails, we add elements to 'trial_locales', and do extra | |
3296 | * loop iterations to cause the behavior described in this paragraph. | |
6b058d42 KW |
3297 | * |
3298 | * On Ultrix, the locale MUST come from the environment, so there is | |
3299 | * preliminary code to set it. I (khw) am not sure that it is necessary, | |
3300 | * and that this couldn't be folded into the loop, but barring any real | |
3301 | * platforms to test on, it's staying as-is | |
3302 | * | |
3303 | * A slight complication is that in embedded Perls, the locale may already | |
3304 | * be set-up, and we don't want to get it from the normal environment | |
3305 | * variables. This is handled by having a special environment variable | |
3306 | * indicate we're in this situation. We simply set setlocale's 2nd | |
3307 | * parameter to be a NULL instead of "". That indicates to setlocale that | |
3308 | * it is not to change anything, but to return the current value, | |
3309 | * effectively initializing perl's db to what the locale already is. | |
3310 | * | |
3311 | * We play the same trick with NULL if a LC_ALL succeeds. We call | |
3312 | * setlocale() on the individual categores with NULL to get their existing | |
3313 | * values for our db, instead of trying to change them. | |
3314 | * */ | |
98994639 | 3315 | |
1565c085 | 3316 | |
0e92a118 KW |
3317 | int ok = 1; |
3318 | ||
7d4bcc4a KW |
3319 | #ifndef USE_LOCALE |
3320 | ||
3321 | PERL_UNUSED_ARG(printwarn); | |
3322 | ||
3323 | #else /* USE_LOCALE */ | |
7d4bcc4a KW |
3324 | # ifdef __GLIBC__ |
3325 | ||
24f3e849 | 3326 | const char * const language = PerlEnv_getenv("LANGUAGE"); |
7d4bcc4a KW |
3327 | |
3328 | # endif | |
65ebb059 | 3329 | |
ccd65d51 KW |
3330 | /* NULL uses the existing already set up locale */ |
3331 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) | |
3332 | ? NULL | |
3333 | : ""; | |
c3fcd832 KW |
3334 | const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */ |
3335 | unsigned int trial_locales_count; | |
24f3e849 KW |
3336 | const char * const lc_all = PerlEnv_getenv("LC_ALL"); |
3337 | const char * const lang = PerlEnv_getenv("LANG"); | |
98994639 | 3338 | bool setlocale_failure = FALSE; |
65ebb059 | 3339 | unsigned int i; |
175c4cf9 KW |
3340 | |
3341 | /* A later getenv() could zap this, so only use here */ | |
3342 | const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG"); | |
3343 | ||
3344 | const bool locwarn = (printwarn > 1 | |
e5f10d49 KW |
3345 | || ( printwarn |
3346 | && ( ! bad_lang_use_once | |
22ff3130 | 3347 | || ( |
e5f10d49 KW |
3348 | /* disallow with "" or "0" */ |
3349 | *bad_lang_use_once | |
3350 | && strNE("0", bad_lang_use_once))))); | |
ea92aad8 | 3351 | |
291a84fb | 3352 | /* setlocale() return vals; not copied so must be looked at immediately */ |
8de4332b | 3353 | const char * sl_result[NOMINAL_LC_ALL_INDEX + 1]; |
291a84fb KW |
3354 | |
3355 | /* current locale for given category; should have been copied so aren't | |
3356 | * volatile */ | |
8de4332b | 3357 | const char * curlocales[NOMINAL_LC_ALL_INDEX + 1]; |
291a84fb | 3358 | |
7d4bcc4a KW |
3359 | # ifdef WIN32 |
3360 | ||
6bce99ee JH |
3361 | /* In some systems you can find out the system default locale |
3362 | * and use that as the fallback locale. */ | |
7d4bcc4a KW |
3363 | # define SYSTEM_DEFAULT_LOCALE |
3364 | # endif | |
3365 | # ifdef SYSTEM_DEFAULT_LOCALE | |
3366 | ||
65ebb059 | 3367 | const char *system_default_locale = NULL; |
98994639 | 3368 | |
7d4bcc4a | 3369 | # endif |
948523db KW |
3370 | |
3371 | # ifndef DEBUGGING | |
3372 | # define DEBUG_LOCALE_INIT(a,b,c) | |
3373 | # else | |
7d4bcc4a | 3374 | |
8298454c | 3375 | DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))); |
7d4bcc4a KW |
3376 | |
3377 | # define DEBUG_LOCALE_INIT(category, locale, result) \ | |
2fcc0ca9 KW |
3378 | STMT_START { \ |
3379 | if (debug_initialization) { \ | |
3380 | PerlIO_printf(Perl_debug_log, \ | |
3381 | "%s:%d: %s\n", \ | |
3382 | __FILE__, __LINE__, \ | |
a4f00dcc | 3383 | setlocale_debug_string(category, \ |
2fcc0ca9 KW |
3384 | locale, \ |
3385 | result)); \ | |
3386 | } \ | |
3387 | } STMT_END | |
2fcc0ca9 | 3388 | |
948523db KW |
3389 | /* Make sure the parallel arrays are properly set up */ |
3390 | # ifdef USE_LOCALE_NUMERIC | |
3391 | assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC); | |
3392 | assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC")); | |
e9bc6d6b KW |
3393 | # ifdef USE_POSIX_2008_LOCALE |
3394 | assert(category_masks[LC_NUMERIC_INDEX] == LC_NUMERIC_MASK); | |
3395 | # endif | |
948523db KW |
3396 | # endif |
3397 | # ifdef USE_LOCALE_CTYPE | |
3398 | assert(categories[LC_CTYPE_INDEX] == LC_CTYPE); | |
3399 | assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE")); | |
e9bc6d6b KW |
3400 | # ifdef USE_POSIX_2008_LOCALE |
3401 | assert(category_masks[LC_CTYPE_INDEX] == LC_CTYPE_MASK); | |
3402 | # endif | |
948523db KW |
3403 | # endif |
3404 | # ifdef USE_LOCALE_COLLATE | |
3405 | assert(categories[LC_COLLATE_INDEX] == LC_COLLATE); | |
3406 | assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE")); | |
e9bc6d6b KW |
3407 | # ifdef USE_POSIX_2008_LOCALE |
3408 | assert(category_masks[LC_COLLATE_INDEX] == LC_COLLATE_MASK); | |
3409 | # endif | |
948523db KW |
3410 | # endif |
3411 | # ifdef USE_LOCALE_TIME | |
3412 | assert(categories[LC_TIME_INDEX] == LC_TIME); | |
3413 | assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME")); | |
e9bc6d6b KW |
3414 | # ifdef USE_POSIX_2008_LOCALE |
3415 | assert(category_masks[LC_TIME_INDEX] == LC_TIME_MASK); | |
3416 | # endif | |
948523db KW |
3417 | # endif |
3418 | # ifdef USE_LOCALE_MESSAGES | |
3419 | assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES); | |
3420 | assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES")); | |
e9bc6d6b KW |
3421 | # ifdef USE_POSIX_2008_LOCALE |
3422 | assert(category_masks[LC_MESSAGES_INDEX] == LC_MESSAGES_MASK); | |
3423 | # endif | |
948523db KW |
3424 | # endif |
3425 | # ifdef USE_LOCALE_MONETARY | |
3426 | assert(categories[LC_MONETARY_INDEX] == LC_MONETARY); | |
3427 | assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY")); | |
e9bc6d6b KW |
3428 | # ifdef USE_POSIX_2008_LOCALE |
3429 | assert(category_masks[LC_MONETARY_INDEX] == LC_MONETARY_MASK); | |
3430 | # endif | |
948523db | 3431 | # endif |
9821811f KW |
3432 | # ifdef USE_LOCALE_ADDRESS |
3433 | assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS); | |
3434 | assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS")); | |
e9bc6d6b KW |
3435 | # ifdef USE_POSIX_2008_LOCALE |
3436 | assert(category_masks[LC_ADDRESS_INDEX] == LC_ADDRESS_MASK); | |
3437 | # endif | |
9821811f KW |
3438 | # endif |
3439 | # ifdef USE_LOCALE_IDENTIFICATION | |
3440 | assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION); | |
3441 | assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION")); | |
e9bc6d6b KW |
3442 | # ifdef USE_POSIX_2008_LOCALE |
3443 | assert(category_masks[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION_MASK); | |
3444 | # endif | |
9821811f KW |
3445 | # endif |
3446 | # ifdef USE_LOCALE_MEASUREMENT | |
3447 | assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT); | |
3448 | assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT")); | |
e9bc6d6b KW |
3449 | # ifdef USE_POSIX_2008_LOCALE |
3450 | assert(category_masks[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT_MASK); | |
3451 | # endif | |
9821811f KW |
3452 | # endif |
3453 | # ifdef USE_LOCALE_PAPER | |
3454 | assert(categories[LC_PAPER_INDEX] == LC_PAPER); | |
3455 | assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER")); | |
e9bc6d6b KW |
3456 | # ifdef USE_POSIX_2008_LOCALE |
3457 | assert(category_masks[LC_PAPER_INDEX] == LC_PAPER_MASK); | |
3458 | # endif | |
9821811f KW |
3459 | # endif |
3460 | # ifdef USE_LOCALE_TELEPHONE | |
3461 | assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE); | |
3462 | assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE")); | |
e9bc6d6b KW |
3463 | # ifdef USE_POSIX_2008_LOCALE |
3464 | assert(category_masks[LC_TELEPHONE_INDEX] == LC_TELEPHONE_MASK); | |
3465 | # endif | |
9821811f | 3466 | # endif |
36dbc955 KW |
3467 | # ifdef USE_LOCALE_SYNTAX |
3468 | assert(categories[LC_SYNTAX_INDEX] == LC_SYNTAX); | |
3469 | assert(strEQ(category_names[LC_SYNTAX_INDEX], "LC_SYNTAX")); | |
3470 | # ifdef USE_POSIX_2008_LOCALE | |
3471 | assert(category_masks[LC_SYNTAX_INDEX] == LC_SYNTAX_MASK); | |
3472 | # endif | |
3473 | # endif | |
3474 | # ifdef USE_LOCALE_TOD | |
3475 | assert(categories[LC_TOD_INDEX] == LC_TOD); | |
3476 | assert(strEQ(category_names[LC_TOD_INDEX], "LC_TOD")); | |
3477 | # ifdef USE_POSIX_2008_LOCALE | |
3478 | assert(category_masks[LC_TOD_INDEX] == LC_TOD_MASK); | |
3479 | # endif | |
3480 | # endif | |
948523db KW |
3481 | # ifdef LC_ALL |
3482 | assert(categories[LC_ALL_INDEX] == LC_ALL); | |
3483 | assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL")); | |
3484 | assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX); | |
e9bc6d6b KW |
3485 | # ifdef USE_POSIX_2008_LOCALE |
3486 | assert(category_masks[LC_ALL_INDEX] == LC_ALL_MASK); | |
3487 | # endif | |
948523db KW |
3488 | # endif |
3489 | # endif /* DEBUGGING */ | |
47280b20 | 3490 | |
5a6637f0 KW |
3491 | /* Initialize the per-thread mbrFOO() state variables. See POSIX.xs for |
3492 | * why these particular incantations are used. */ | |
d2c9cb53 KW |
3493 | #ifdef HAS_MBRLEN |
3494 | memzero(&PL_mbrlen_ps, sizeof(PL_mbrlen_ps)); | |
3495 | #endif | |
5a6637f0 KW |
3496 | #ifdef HAS_MBRTOWC |
3497 | memzero(&PL_mbrtowc_ps, sizeof(PL_mbrtowc_ps)); | |
3498 | #endif | |
3499 | #ifdef HAS_WCTOMBR | |
3500 | wcrtomb(NULL, L'\0', &PL_wcrtomb_ps); | |
3501 | #endif | |
d2c9cb53 | 3502 | |
47280b20 KW |
3503 | /* Initialize the cache of the program's UTF-8ness for the always known |
3504 | * locales C and POSIX */ | |
3505 | my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness, | |
3506 | sizeof(PL_locale_utf8ness)); | |
3507 | ||
e9bc6d6b KW |
3508 | # ifdef USE_THREAD_SAFE_LOCALE |
3509 | # ifdef WIN32 | |
3510 | ||
3511 | _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); | |
3512 | ||
3513 | # endif | |
3514 | # endif | |
39e69e77 | 3515 | # ifdef USE_POSIX_2008_LOCALE |
e9bc6d6b KW |
3516 | |
3517 | PL_C_locale_obj = newlocale(LC_ALL_MASK, "C", (locale_t) 0); | |
3518 | if (! PL_C_locale_obj) { | |
3519 | Perl_croak_nocontext( | |
3520 | "panic: Cannot create POSIX 2008 C locale object; errno=%d", errno); | |
3521 | } | |
3522 | if (DEBUG_Lv_TEST || debug_initialization) { | |
3523 | PerlIO_printf(Perl_debug_log, "%s:%d: created C object %p\n", __FILE__, __LINE__, PL_C_locale_obj); | |
3524 | } | |
3525 | ||
3526 | # endif | |
3527 | ||
78f7c09f KW |
3528 | # ifdef USE_LOCALE_NUMERIC |
3529 | ||
3ca88433 KW |
3530 | PL_numeric_radix_sv = newSVpvs("."); |
3531 | ||
78f7c09f KW |
3532 | # endif |
3533 | ||
e9bc6d6b KW |
3534 | # if defined(USE_POSIX_2008_LOCALE) && ! defined(HAS_QUERYLOCALE) |
3535 | ||
3536 | /* Initialize our records. If we have POSIX 2008, we have LC_ALL */ | |
3537 | do_setlocale_c(LC_ALL, my_setlocale(LC_ALL, NULL)); | |
3538 | ||
3539 | # endif | |
b56c4436 | 3540 | # ifdef LOCALE_ENVIRON_REQUIRED |
98994639 HS |
3541 | |
3542 | /* | |
3543 | * Ultrix setlocale(..., "") fails if there are no environment | |
3544 | * variables from which to get a locale name. | |
3545 | */ | |
3546 | ||
b56c4436 KW |
3547 | # ifndef LC_ALL |
3548 | # error Ultrix without LC_ALL not implemented | |
3549 | # else | |
7d4bcc4a | 3550 | |
b56c4436 KW |
3551 | { |
3552 | bool done = FALSE; | |
ea92aad8 KW |
3553 | if (lang) { |
3554 | sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init); | |
3555 | DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]); | |
3556 | if (sl_result[LC_ALL_INDEX]) | |
3557 | done = TRUE; | |
3558 | else | |
e5f10d49 | 3559 | setlocale_failure = TRUE; |
ea92aad8 KW |
3560 | } |
3561 | if (! setlocale_failure) { | |
3562 | const char * locale_param; | |
3563 | for (i = 0; i < LC_ALL_INDEX; i++) { | |
3564 | locale_param = (! done && (lang || PerlEnv_getenv(category_names[i]))) | |
3565 | ? setlocale_init | |
3566 | : NULL; | |
3567 | sl_result[i] = do_setlocale_r(categories[i], locale_param); | |
3568 | if (! sl_result[i]) { | |
3569 | setlocale_failure = TRUE; | |
3570 | } | |
3571 | DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]); | |
e5f10d49 | 3572 | } |
c835d6be | 3573 | } |
7d4bcc4a KW |
3574 | } |
3575 | ||
3576 | # endif /* LC_ALL */ | |
e5f10d49 | 3577 | # endif /* LOCALE_ENVIRON_REQUIRED */ |
98994639 | 3578 | |
65ebb059 | 3579 | /* We try each locale in the list until we get one that works, or exhaust |
20a240df KW |
3580 | * the list. Normally the loop is executed just once. But if setting the |
3581 | * locale fails, inside the loop we add fallback trials to the array and so | |
3582 | * will execute the loop multiple times */ | |
c3fcd832 KW |
3583 | trial_locales[0] = setlocale_init; |
3584 | trial_locales_count = 1; | |
7d4bcc4a | 3585 | |
65ebb059 KW |
3586 | for (i= 0; i < trial_locales_count; i++) { |
3587 | const char * trial_locale = trial_locales[i]; | |
3588 | ||
3589 | if (i > 0) { | |
3590 | ||
3591 | /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED | |
3592 | * when i==0, but I (khw) don't think that behavior makes much | |
3593 | * sense */ | |
3594 | setlocale_failure = FALSE; | |
3595 | ||
7d4bcc4a | 3596 | # ifdef SYSTEM_DEFAULT_LOCALE |
291a84fb | 3597 | # ifdef WIN32 /* Note that assumes Win32 has LC_ALL */ |
7d4bcc4a | 3598 | |
65ebb059 KW |
3599 | /* On Windows machines, an entry of "" after the 0th means to use |
3600 | * the system default locale, which we now proceed to get. */ | |
3601 | if (strEQ(trial_locale, "")) { | |
3602 | unsigned int j; | |
3603 | ||
3604 | /* Note that this may change the locale, but we are going to do | |
3605 | * that anyway just below */ | |
837ce802 | 3606 | system_default_locale = do_setlocale_c(LC_ALL, ""); |
5d1187d1 | 3607 | DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale); |
65ebb059 | 3608 | |
7d4bcc4a | 3609 | /* Skip if invalid or if it's already on the list of locales to |
65ebb059 KW |
3610 | * try */ |
3611 | if (! system_default_locale) { | |
3612 | goto next_iteration; | |
3613 | } | |
3614 | for (j = 0; j < trial_locales_count; j++) { | |
3615 | if (strEQ(system_default_locale, trial_locales[j])) { | |
3616 | goto next_iteration; | |
3617 | } | |
3618 | } | |
3619 | ||
3620 | trial_locale = system_default_locale; | |
3621 | } | |
ec0202b5 KW |
3622 | # else |
3623 | # error SYSTEM_DEFAULT_LOCALE only implemented for Win32 | |
3624 | # endif | |
7d4bcc4a | 3625 | # endif /* SYSTEM_DEFAULT_LOCALE */ |
291a84fb KW |
3626 | |
3627 | } /* For i > 0 */ | |
65ebb059 | 3628 | |
7d4bcc4a KW |
3629 | # ifdef LC_ALL |
3630 | ||
948523db KW |
3631 | sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale); |
3632 | DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]); | |
3633 | if (! sl_result[LC_ALL_INDEX]) { | |
49c85077 | 3634 | setlocale_failure = TRUE; |
7cd8b568 KW |
3635 | } |
3636 | else { | |
3637 | /* Since LC_ALL succeeded, it should have changed all the other | |
3638 | * categories it can to its value; so we massage things so that the | |
3639 | * setlocales below just return their category's current values. | |
3640 | * This adequately handles the case in NetBSD where LC_COLLATE may | |
3641 | * not be defined for a locale, and setting it individually will | |
7d4bcc4a | 3642 | * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to |
7cd8b568 KW |
3643 | * the POSIX locale. */ |
3644 | trial_locale = NULL; | |
3645 | } | |
7d4bcc4a KW |
3646 | |
3647 | # endif /* LC_ALL */ | |
98994639 | 3648 | |
e5f10d49 KW |
3649 | if (! setlocale_failure) { |
3650 | unsigned int j; | |
3651 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { | |
3652 | curlocales[j] | |
3653 | = savepv(do_setlocale_r(categories[j], trial_locale)); | |
3654 | if (! curlocales[j]) { | |
3655 | setlocale_failure = TRUE; | |
3656 | } | |
3657 | DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]); | |
3658 | } | |
c835d6be | 3659 | |
e5f10d49 KW |
3660 | if (! setlocale_failure) { /* All succeeded */ |
3661 | break; /* Exit trial_locales loop */ | |
49c85077 | 3662 | } |
65ebb059 | 3663 | } |
98994639 | 3664 | |
49c85077 KW |
3665 | /* Here, something failed; will need to try a fallback. */ |
3666 | ok = 0; | |
65ebb059 | 3667 | |
49c85077 KW |
3668 | if (i == 0) { |
3669 | unsigned int j; | |
98994639 | 3670 | |
65ebb059 | 3671 | if (locwarn) { /* Output failure info only on the first one */ |
7d4bcc4a KW |
3672 | |
3673 | # ifdef LC_ALL | |
98994639 | 3674 | |
49c85077 KW |
3675 | PerlIO_printf(Perl_error_log, |
3676 | "perl: warning: Setting locale failed.\n"); | |
98994639 | 3677 | |
7d4bcc4a | 3678 | # else /* !LC_ALL */ |
98994639 | 3679 | |
49c85077 KW |
3680 | PerlIO_printf(Perl_error_log, |
3681 | "perl: warning: Setting locale failed for the categories:\n\t"); | |
7d4bcc4a | 3682 | |
e5f10d49 KW |
3683 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { |
3684 | if (! curlocales[j]) { | |
3685 | PerlIO_printf(Perl_error_log, category_names[j]); | |
3686 | } | |
3687 | else { | |
3688 | Safefree(curlocales[j]); | |
3689 | } | |
3690 | } | |
7d4bcc4a | 3691 | |
7d4bcc4a | 3692 | # endif /* LC_ALL */ |
98994639 | 3693 | |
49c85077 KW |
3694 | PerlIO_printf(Perl_error_log, |
3695 | "perl: warning: Please check that your locale settings:\n"); | |
98994639 | 3696 | |
7d4bcc4a KW |
3697 | # ifdef __GLIBC__ |
3698 | ||
49c85077 KW |
3699 | PerlIO_printf(Perl_error_log, |
3700 | "\tLANGUAGE = %c%s%c,\n", | |
3701 | language ? '"' : '(', | |
3702 | language ? language : "unset", | |
3703 | language ? '"' : ')'); | |
7d4bcc4a | 3704 | # endif |
98994639 | 3705 | |
49c85077 KW |
3706 | PerlIO_printf(Perl_error_log, |
3707 | "\tLC_ALL = %c%s%c,\n", | |
3708 | lc_all ? '"' : '(', | |
3709 | lc_all ? lc_all : "unset", | |
3710 | lc_all ? '"' : ')'); | |
98994639 | 3711 | |
7d4bcc4a KW |
3712 | # if defined(USE_ENVIRON_ARRAY) |
3713 | ||
49c85077 | 3714 | { |
cd999af9 | 3715 | char **e; |
d5e32b93 KW |
3716 | |
3717 | /* Look through the environment for any variables of the | |
3718 | * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was | |
3719 | * already handled above. These are assumed to be locale | |
3720 | * settings. Output them and their values. */ | |
cd999af9 | 3721 | for (e = environ; *e; e++) { |
d5e32b93 KW |
3722 | const STRLEN prefix_len = sizeof("LC_") - 1; |
3723 | STRLEN uppers_len; | |
3724 | ||
cd999af9 | 3725 | if ( strBEGINs(*e, "LC_") |
c8b388b0 | 3726 | && ! strBEGINs(*e, "LC_ALL=") |
d5e32b93 KW |
3727 | && (uppers_len = strspn(*e + prefix_len, |
3728 | "ABCDEFGHIJKLMNOPQRSTUVWXYZ")) | |
3729 | && ((*e)[prefix_len + uppers_len] == '=')) | |
cd999af9 KW |
3730 | { |
3731 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", | |
d5e32b93 KW |
3732 | (int) (prefix_len + uppers_len), *e, |
3733 | *e + prefix_len + uppers_len + 1); | |
cd999af9 KW |
3734 | } |
3735 | } | |
49c85077 | 3736 | } |
7d4bcc4a KW |
3737 | |
3738 | # else | |
3739 | ||
49c85077 KW |
3740 | PerlIO_printf(Perl_error_log, |
3741 | "\t(possibly more locale environment variables)\n"); | |
7d4bcc4a KW |
3742 | |
3743 | # endif | |
98994639 | 3744 | |
49c85077 KW |
3745 | PerlIO_printf(Perl_error_log, |
3746 | "\tLANG = %c%s%c\n", | |
3747 | lang ? '"' : '(', | |
3748 | lang ? lang : "unset", | |
3749 | lang ? '"' : ')'); | |
98994639 | 3750 | |
49c85077 KW |
3751 | PerlIO_printf(Perl_error_log, |
3752 | " are supported and installed on your system.\n"); | |
3753 | } | |
98994639 | 3754 | |
65ebb059 | 3755 | /* Calculate what fallback locales to try. We have avoided this |
f6bab5f6 | 3756 | * until we have to, because failure is quite unlikely. This will |
65ebb059 KW |
3757 | * usually change the upper bound of the loop we are in. |
3758 | * | |
3759 | * Since the system's default way of setting the locale has not | |
3760 | * found one that works, We use Perl's defined ordering: LC_ALL, | |
3761 | * LANG, and the C locale. We don't try the same locale twice, so | |
3762 | * don't add to the list if already there. (On POSIX systems, the | |
3763 | * LC_ALL element will likely be a repeat of the 0th element "", | |
6b058d42 KW |
3764 | * but there's no harm done by doing it explicitly. |
3765 | * | |
3766 | * Note that this tries the LC_ALL environment variable even on | |
3767 | * systems which have no LC_ALL locale setting. This may or may | |
3768 | * not have been originally intentional, but there's no real need | |
3769 | * to change the behavior. */ | |
65ebb059 KW |
3770 | if (lc_all) { |
3771 | for (j = 0; j < trial_locales_count; j++) { | |
3772 | if (strEQ(lc_all, trial_locales[j])) { | |
3773 | goto done_lc_all; | |
3774 | } | |
3775 | } | |
3776 | trial_locales[trial_locales_count++] = lc_all; | |
3777 | } | |
3778 | done_lc_all: | |
98994639 | 3779 | |
65ebb059 KW |
3780 | if (lang) { |
3781 | for (j = 0; j < trial_locales_count; j++) { | |
3782 | if (strEQ(lang, trial_locales[j])) { | |
3783 | goto done_lang; | |
3784 | } | |
3785 | } | |
3786 | trial_locales[trial_locales_count++] = lang; | |
3787 | } | |
3788 | done_lang: | |
3789 | ||
7d4bcc4a KW |
3790 | # if defined(WIN32) && defined(LC_ALL) |
3791 | ||
65ebb059 KW |
3792 | /* For Windows, we also try the system default locale before "C". |
3793 | * (If there exists a Windows without LC_ALL we skip this because | |
3794 | * it gets too complicated. For those, the "C" is the next | |
3795 | * fallback possibility). The "" is the same as the 0th element of | |
3796 | * the array, but the code at the loop above knows to treat it | |
3797 | * differently when not the 0th */ | |
3798 | trial_locales[trial_locales_count++] = ""; | |
7d4bcc4a KW |
3799 | |
3800 | # endif | |
65ebb059 KW |
3801 | |
3802 | for (j = 0; j < trial_locales_count; j++) { | |
3803 | if (strEQ("C", trial_locales[j])) { | |
3804 | goto done_C; | |
3805 | } | |
3806 | } | |
3807 | trial_locales[trial_locales_count++] = "C"; | |
98994639 | 3808 | |
65ebb059 KW |
3809 | done_C: ; |
3810 | } /* end of first time through the loop */ | |
98994639 | 3811 | |
7d4bcc4a KW |
3812 | # ifdef WIN32 |
3813 | ||
65ebb059 | 3814 | next_iteration: ; |
7d4bcc4a KW |
3815 | |
3816 | # endif | |
65ebb059 KW |
3817 | |
3818 | } /* end of looping through the trial locales */ | |
3819 | ||
3820 | if (ok < 1) { /* If we tried to fallback */ | |
3821 | const char* msg; | |
3822 | if (! setlocale_failure) { /* fallback succeeded */ | |
3823 | msg = "Falling back to"; | |
3824 | } | |
3825 | else { /* fallback failed */ | |
e5f10d49 | 3826 | unsigned int j; |
98994639 | 3827 | |
65ebb059 KW |
3828 | /* We dropped off the end of the loop, so have to decrement i to |
3829 | * get back to the value the last time through */ | |
3830 | i--; | |
98994639 | 3831 | |
65ebb059 KW |
3832 | ok = -1; |
3833 | msg = "Failed to fall back to"; | |
3834 | ||
3835 | /* To continue, we should use whatever values we've got */ | |
7d4bcc4a | 3836 | |
e5f10d49 KW |
3837 | for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) { |
3838 | Safefree(curlocales[j]); | |
3839 | curlocales[j] = savepv(do_setlocale_r(categories[j], NULL)); | |
3840 | DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]); | |
3841 | } | |
65ebb059 KW |
3842 | } |
3843 | ||
3844 | if (locwarn) { | |
3845 | const char * description; | |
3846 | const char * name = ""; | |
3847 | if (strEQ(trial_locales[i], "C")) { | |
3848 | description = "the standard locale"; | |
3849 | name = "C"; | |
3850 | } | |
7d4bcc4a KW |
3851 | |
3852 | # ifdef SYSTEM_DEFAULT_LOCALE | |
3853 | ||
65ebb059 KW |
3854 | else if (strEQ(trial_locales[i], "")) { |
3855 | description = "the system default locale"; | |
3856 | if (system_default_locale) { | |
3857 | name = system_default_locale; | |
3858 | } | |
3859 | } | |