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 KW |
25 | * |
26 | * All C programs have an underlying locale. Perl generally doesn't pay any | |
27 | * attention to it except within the scope of a 'use locale'. For most | |
28 | * categories, it accomplishes this by just using different operations if it is | |
29 | * in such scope than if not. However, various libc functions called by Perl | |
30 | * are affected by the LC_NUMERIC category, so there are macros in perl.h that | |
31 | * are used to toggle between the current locale and the C locale depending on | |
32 | * the desired behavior of those functions at the moment. | |
166f8a29 DM |
33 | */ |
34 | ||
98994639 HS |
35 | #include "EXTERN.h" |
36 | #define PERL_IN_LOCALE_C | |
37 | #include "perl.h" | |
38 | ||
b310b053 JH |
39 | #ifdef I_LANGINFO |
40 | # include <langinfo.h> | |
41 | #endif | |
42 | ||
a4af207c JH |
43 | #include "reentr.h" |
44 | ||
8ef6e574 KW |
45 | #ifdef USE_LOCALE |
46 | ||
98994639 | 47 | /* |
0d071d52 KW |
48 | * Standardize the locale name from a string returned by 'setlocale', possibly |
49 | * modifying that string. | |
98994639 | 50 | * |
0ef2a2b2 | 51 | * The typical return value of setlocale() is either |
98994639 HS |
52 | * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL |
53 | * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL | |
54 | * (the space-separated values represent the various sublocales, | |
0ef2a2b2 | 55 | * in some unspecified order). This is not handled by this function. |
98994639 HS |
56 | * |
57 | * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n", | |
0ef2a2b2 KW |
58 | * which is harmful for further use of the string in setlocale(). This |
59 | * function removes the trailing new line and everything up through the '=' | |
98994639 HS |
60 | * |
61 | */ | |
62 | STATIC char * | |
63 | S_stdize_locale(pTHX_ char *locs) | |
64 | { | |
7452cf6a | 65 | const char * const s = strchr(locs, '='); |
98994639 HS |
66 | bool okay = TRUE; |
67 | ||
7918f24d NC |
68 | PERL_ARGS_ASSERT_STDIZE_LOCALE; |
69 | ||
8772537c AL |
70 | if (s) { |
71 | const char * const t = strchr(s, '.'); | |
98994639 | 72 | okay = FALSE; |
8772537c AL |
73 | if (t) { |
74 | const char * const u = strchr(t, '\n'); | |
75 | if (u && (u[1] == 0)) { | |
76 | const STRLEN len = u - s; | |
77 | Move(s + 1, locs, len, char); | |
78 | locs[len] = 0; | |
79 | okay = TRUE; | |
98994639 HS |
80 | } |
81 | } | |
82 | } | |
83 | ||
84 | if (!okay) | |
85 | Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs); | |
86 | ||
87 | return locs; | |
88 | } | |
89 | ||
8ef6e574 KW |
90 | #endif |
91 | ||
98994639 HS |
92 | void |
93 | Perl_set_numeric_radix(pTHX) | |
94 | { | |
95 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 96 | dVAR; |
98994639 | 97 | # ifdef HAS_LOCALECONV |
7452cf6a | 98 | const struct lconv* const lc = localeconv(); |
98994639 | 99 | |
98994639 HS |
100 | if (lc && lc->decimal_point) { |
101 | if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) { | |
102 | SvREFCNT_dec(PL_numeric_radix_sv); | |
a0714e2c | 103 | PL_numeric_radix_sv = NULL; |
98994639 HS |
104 | } |
105 | else { | |
106 | if (PL_numeric_radix_sv) | |
107 | sv_setpv(PL_numeric_radix_sv, lc->decimal_point); | |
108 | else | |
109 | PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0); | |
28acfe03 KW |
110 | if (! is_ascii_string((U8 *) lc->decimal_point, 0) |
111 | && is_utf8_string((U8 *) lc->decimal_point, 0) | |
c1284011 | 112 | && _is_cur_LC_category_utf8(LC_NUMERIC)) |
28acfe03 KW |
113 | { |
114 | SvUTF8_on(PL_numeric_radix_sv); | |
115 | } | |
98994639 HS |
116 | } |
117 | } | |
118 | else | |
a0714e2c | 119 | PL_numeric_radix_sv = NULL; |
69014004 KW |
120 | |
121 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is %s\n", | |
122 | (PL_numeric_radix_sv) | |
123 | ? lc->decimal_point | |
124 | : "NULL")); | |
125 | ||
98994639 HS |
126 | # endif /* HAS_LOCALECONV */ |
127 | #endif /* USE_LOCALE_NUMERIC */ | |
128 | } | |
129 | ||
98994639 | 130 | void |
8772537c | 131 | Perl_new_numeric(pTHX_ const char *newnum) |
98994639 HS |
132 | { |
133 | #ifdef USE_LOCALE_NUMERIC | |
0d071d52 KW |
134 | |
135 | /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell | |
136 | * core Perl this and that 'newnum' is the name of the new locale. | |
137 | * It installs this locale as the current underlying default. | |
138 | * | |
139 | * The default locale and the C locale can be toggled between by use of the | |
140 | * set_numeric_local() and set_numeric_standard() functions, which should | |
141 | * probably not be called directly, but only via macros like | |
142 | * SET_NUMERIC_STANDARD() in perl.h. | |
143 | * | |
144 | * The toggling is necessary mainly so that a non-dot radix decimal point | |
145 | * character can be output, while allowing internal calculations to use a | |
146 | * dot. | |
147 | * | |
148 | * This sets several interpreter-level variables: | |
149 | * PL_numeric_name The default locale's name: a copy of 'newnum' | |
150 | * PL_numeric_local A boolean indicating if the toggled state is such | |
7738054c KW |
151 | * that the current locale is the program's underlying |
152 | * locale | |
153 | * PL_numeric_standard An int indicating if the toggled state is such | |
154 | * that the current locale is the C locale. If non-zero, | |
155 | * it is in C; if > 1, it means it may not be toggled away | |
156 | * from C. | |
0d071d52 KW |
157 | * Note that both of the last two variables can be true at the same time, |
158 | * if the underlying locale is C. (Toggling is a no-op under these | |
159 | * circumstances.) | |
160 | * | |
161 | * Any code changing the locale (outside this file) should use | |
162 | * POSIX::setlocale, which calls this function. Therefore this function | |
163 | * should be called directly only from this file and from | |
164 | * POSIX::setlocale() */ | |
165 | ||
b03f34cf | 166 | char *save_newnum; |
97aff369 | 167 | dVAR; |
98994639 HS |
168 | |
169 | if (! newnum) { | |
43c5f42d NC |
170 | Safefree(PL_numeric_name); |
171 | PL_numeric_name = NULL; | |
98994639 HS |
172 | PL_numeric_standard = TRUE; |
173 | PL_numeric_local = TRUE; | |
174 | return; | |
175 | } | |
176 | ||
b03f34cf KW |
177 | save_newnum = stdize_locale(savepv(newnum)); |
178 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { | |
98994639 | 179 | Safefree(PL_numeric_name); |
b03f34cf | 180 | PL_numeric_name = save_newnum; |
b03f34cf | 181 | } |
98994639 | 182 | |
e19f01cb KW |
183 | PL_numeric_standard = ((*save_newnum == 'C' && save_newnum[1] == '\0') |
184 | || strEQ(save_newnum, "POSIX")); | |
185 | PL_numeric_local = TRUE; | |
4c28b29c KW |
186 | |
187 | /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't | |
188 | * have to worry about the radix being a non-dot. (Core operations that | |
189 | * need the underlying locale change to it temporarily). */ | |
190 | set_numeric_standard(); | |
191 | ||
e19f01cb | 192 | set_numeric_radix(); |
6959d69d | 193 | |
98994639 HS |
194 | #endif /* USE_LOCALE_NUMERIC */ |
195 | } | |
196 | ||
197 | void | |
198 | Perl_set_numeric_standard(pTHX) | |
199 | { | |
200 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 201 | dVAR; |
98994639 | 202 | |
0d071d52 KW |
203 | /* Toggle the LC_NUMERIC locale to C, if not already there. Probably |
204 | * should use the macros like SET_NUMERIC_STANDARD() in perl.h instead of | |
205 | * calling this directly. */ | |
206 | ||
aaaeb297 | 207 | if (_NOT_IN_NUMERIC_STANDARD) { |
98994639 HS |
208 | setlocale(LC_NUMERIC, "C"); |
209 | PL_numeric_standard = TRUE; | |
210 | PL_numeric_local = FALSE; | |
211 | set_numeric_radix(); | |
212 | } | |
69014004 KW |
213 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
214 | "Underlying LC_NUMERIC locale now is C\n")); | |
98994639 HS |
215 | |
216 | #endif /* USE_LOCALE_NUMERIC */ | |
217 | } | |
218 | ||
219 | void | |
220 | Perl_set_numeric_local(pTHX) | |
221 | { | |
222 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 223 | dVAR; |
98994639 | 224 | |
0d071d52 KW |
225 | /* Toggle the LC_NUMERIC locale to the current underlying default, if not |
226 | * already there. Probably should use the macros like SET_NUMERIC_LOCAL() | |
227 | * in perl.h instead of calling this directly. */ | |
228 | ||
aaaeb297 | 229 | if (_NOT_IN_NUMERIC_LOCAL) { |
98994639 HS |
230 | setlocale(LC_NUMERIC, PL_numeric_name); |
231 | PL_numeric_standard = FALSE; | |
232 | PL_numeric_local = TRUE; | |
233 | set_numeric_radix(); | |
234 | } | |
69014004 KW |
235 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
236 | "Underlying LC_NUMERIC locale now is %s\n", | |
237 | PL_numeric_name)); | |
98994639 HS |
238 | |
239 | #endif /* USE_LOCALE_NUMERIC */ | |
240 | } | |
241 | ||
242 | /* | |
243 | * Set up for a new ctype locale. | |
244 | */ | |
245 | void | |
8772537c | 246 | Perl_new_ctype(pTHX_ const char *newctype) |
98994639 HS |
247 | { |
248 | #ifdef USE_LOCALE_CTYPE | |
0d071d52 KW |
249 | |
250 | /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell | |
251 | * core Perl this and that 'newctype' is the name of the new locale. | |
252 | * | |
253 | * This function sets up the folding arrays for all 256 bytes, assuming | |
254 | * that tofold() is tolc() since fold case is not a concept in POSIX, | |
255 | * | |
256 | * Any code changing the locale (outside this file) should use | |
257 | * POSIX::setlocale, which calls this function. Therefore this function | |
258 | * should be called directly only from this file and from | |
259 | * POSIX::setlocale() */ | |
260 | ||
27da23d5 | 261 | dVAR; |
68067e4e | 262 | UV i; |
98994639 | 263 | |
7918f24d NC |
264 | PERL_ARGS_ASSERT_NEW_CTYPE; |
265 | ||
c1284011 | 266 | PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE); |
31f05a37 KW |
267 | |
268 | /* A UTF-8 locale gets standard rules. But note that code still has to | |
269 | * handle this specially because of the three problematic code points */ | |
270 | if (PL_in_utf8_CTYPE_locale) { | |
271 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); | |
272 | } | |
273 | else { | |
baa60164 KW |
274 | for (i = 0; i < 256; i++) { |
275 | if (isUPPER_LC((U8) i)) | |
276 | PL_fold_locale[i] = (U8) toLOWER_LC((U8) i); | |
277 | else if (isLOWER_LC((U8) i)) | |
278 | PL_fold_locale[i] = (U8) toUPPER_LC((U8) i); | |
279 | else | |
280 | PL_fold_locale[i] = (U8) i; | |
281 | } | |
31f05a37 | 282 | } |
98994639 HS |
283 | |
284 | #endif /* USE_LOCALE_CTYPE */ | |
7918f24d | 285 | PERL_ARGS_ASSERT_NEW_CTYPE; |
8772537c | 286 | PERL_UNUSED_ARG(newctype); |
96a5add6 | 287 | PERL_UNUSED_CONTEXT; |
98994639 HS |
288 | } |
289 | ||
98994639 | 290 | void |
8772537c | 291 | Perl_new_collate(pTHX_ const char *newcoll) |
98994639 HS |
292 | { |
293 | #ifdef USE_LOCALE_COLLATE | |
0d071d52 KW |
294 | |
295 | /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell | |
296 | * core Perl this and that 'newcoll' is the name of the new locale. | |
297 | * | |
298 | * Any code changing the locale (outside this file) should use | |
299 | * POSIX::setlocale, which calls this function. Therefore this function | |
300 | * should be called directly only from this file and from | |
301 | * POSIX::setlocale() */ | |
302 | ||
97aff369 | 303 | dVAR; |
98994639 HS |
304 | |
305 | if (! newcoll) { | |
306 | if (PL_collation_name) { | |
307 | ++PL_collation_ix; | |
308 | Safefree(PL_collation_name); | |
309 | PL_collation_name = NULL; | |
310 | } | |
311 | PL_collation_standard = TRUE; | |
312 | PL_collxfrm_base = 0; | |
313 | PL_collxfrm_mult = 2; | |
314 | return; | |
315 | } | |
316 | ||
317 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { | |
318 | ++PL_collation_ix; | |
319 | Safefree(PL_collation_name); | |
320 | PL_collation_name = stdize_locale(savepv(newcoll)); | |
770526c1 NC |
321 | PL_collation_standard = ((*newcoll == 'C' && newcoll[1] == '\0') |
322 | || strEQ(newcoll, "POSIX")); | |
98994639 HS |
323 | |
324 | { | |
325 | /* 2: at most so many chars ('a', 'b'). */ | |
326 | /* 50: surely no system expands a char more. */ | |
327 | #define XFRMBUFSIZE (2 * 50) | |
328 | char xbuf[XFRMBUFSIZE]; | |
8772537c AL |
329 | const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE); |
330 | const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE); | |
331 | const SSize_t mult = fb - fa; | |
e0601d3c | 332 | if (mult < 1 && !(fa == 0 && fb == 0)) |
ad49ad39 NC |
333 | Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf, |
334 | (UV) fa, (UV) fb); | |
eb160463 | 335 | PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0; |
98994639 HS |
336 | PL_collxfrm_mult = mult; |
337 | } | |
338 | } | |
339 | ||
340 | #endif /* USE_LOCALE_COLLATE */ | |
341 | } | |
342 | ||
b385bb4d KW |
343 | #ifdef WIN32 |
344 | ||
345 | char * | |
346 | Perl_my_setlocale(pTHX_ int category, const char* locale) | |
347 | { | |
348 | /* This, for Windows, emulates POSIX setlocale() behavior. There is no | |
349 | * difference unless the input locale is "", which means on Windows to get | |
350 | * the machine default, which is set via the computer's "Regional and | |
351 | * Language Options" (or its current equivalent). In POSIX, it instead | |
352 | * means to find the locale from the user's environment. This routine | |
353 | * looks in the environment, and, if anything is found, uses that instead | |
354 | * of going to the machine default. If there is no environment override, | |
355 | * the machine default is used, as normal, by calling the real setlocale() | |
356 | * with "". The POSIX behavior is to use the LC_ALL variable if set; | |
357 | * otherwise to use the particular category's variable if set; otherwise to | |
358 | * use the LANG variable. */ | |
359 | ||
481465ea | 360 | bool override_LC_ALL = 0; |
89f7b9aa KW |
361 | char * result; |
362 | ||
b385bb4d KW |
363 | if (locale && strEQ(locale, "")) { |
364 | # ifdef LC_ALL | |
365 | locale = PerlEnv_getenv("LC_ALL"); | |
366 | if (! locale) { | |
367 | #endif | |
368 | switch (category) { | |
369 | # ifdef LC_ALL | |
370 | case LC_ALL: | |
481465ea | 371 | override_LC_ALL = TRUE; |
b385bb4d KW |
372 | break; /* We already know its variable isn't set */ |
373 | # endif | |
374 | # ifdef USE_LOCALE_TIME | |
375 | case LC_TIME: | |
376 | locale = PerlEnv_getenv("LC_TIME"); | |
377 | break; | |
378 | # endif | |
379 | # ifdef USE_LOCALE_CTYPE | |
380 | case LC_CTYPE: | |
381 | locale = PerlEnv_getenv("LC_CTYPE"); | |
382 | break; | |
383 | # endif | |
384 | # ifdef USE_LOCALE_COLLATE | |
385 | case LC_COLLATE: | |
386 | locale = PerlEnv_getenv("LC_COLLATE"); | |
387 | break; | |
388 | # endif | |
389 | # ifdef USE_LOCALE_MONETARY | |
390 | case LC_MONETARY: | |
391 | locale = PerlEnv_getenv("LC_MONETARY"); | |
392 | break; | |
393 | # endif | |
394 | # ifdef USE_LOCALE_NUMERIC | |
395 | case LC_NUMERIC: | |
396 | locale = PerlEnv_getenv("LC_NUMERIC"); | |
397 | break; | |
398 | # endif | |
399 | # ifdef USE_LOCALE_MESSAGES | |
400 | case LC_MESSAGES: | |
401 | locale = PerlEnv_getenv("LC_MESSAGES"); | |
402 | break; | |
403 | # endif | |
404 | default: | |
405 | /* This is a category, like PAPER_SIZE that we don't | |
406 | * know about; and so can't provide a wrapper. */ | |
407 | break; | |
408 | } | |
409 | if (! locale) { | |
410 | locale = PerlEnv_getenv("LANG"); | |
481465ea | 411 | if (! locale) { |
b385bb4d KW |
412 | locale = ""; |
413 | } | |
414 | } | |
415 | # ifdef LC_ALL | |
416 | } | |
417 | # endif | |
418 | } | |
419 | ||
89f7b9aa KW |
420 | result = setlocale(category, locale); |
421 | ||
481465ea | 422 | if (! override_LC_ALL) { |
89f7b9aa KW |
423 | return result; |
424 | } | |
425 | ||
426 | /* Here the input locale was LC_ALL, and we have set it to what is in the | |
481465ea KW |
427 | * LANG variable or the system default if there is no LANG. But these have |
428 | * lower priority than the other LC_foo variables, so override it for each | |
429 | * one that is set. (If they are set to "", it means to use the same thing | |
430 | * we just set LC_ALL to, so can skip) */ | |
89f7b9aa KW |
431 | # ifdef USE_LOCALE_TIME |
432 | result = PerlEnv_getenv("LC_TIME"); | |
730252b2 | 433 | if (result && strNE(result, "")) { |
89f7b9aa KW |
434 | setlocale(LC_TIME, result); |
435 | } | |
436 | # endif | |
437 | # ifdef USE_LOCALE_CTYPE | |
438 | result = PerlEnv_getenv("LC_CTYPE"); | |
730252b2 | 439 | if (result && strNE(result, "")) { |
89f7b9aa KW |
440 | setlocale(LC_CTYPE, result); |
441 | } | |
442 | # endif | |
443 | # ifdef USE_LOCALE_COLLATE | |
444 | result = PerlEnv_getenv("LC_COLLATE"); | |
730252b2 | 445 | if (result && strNE(result, "")) { |
89f7b9aa KW |
446 | setlocale(LC_COLLATE, result); |
447 | } | |
448 | # endif | |
449 | # ifdef USE_LOCALE_MONETARY | |
450 | result = PerlEnv_getenv("LC_MONETARY"); | |
730252b2 | 451 | if (result && strNE(result, "")) { |
89f7b9aa KW |
452 | setlocale(LC_MONETARY, result); |
453 | } | |
454 | # endif | |
455 | # ifdef USE_LOCALE_NUMERIC | |
456 | result = PerlEnv_getenv("LC_NUMERIC"); | |
730252b2 | 457 | if (result && strNE(result, "")) { |
89f7b9aa KW |
458 | setlocale(LC_NUMERIC, result); |
459 | } | |
460 | # endif | |
461 | # ifdef USE_LOCALE_MESSAGES | |
462 | result = PerlEnv_getenv("LC_MESSAGES"); | |
730252b2 | 463 | if (result && strNE(result, "")) { |
89f7b9aa KW |
464 | setlocale(LC_MESSAGES, result); |
465 | } | |
466 | # endif | |
467 | ||
468 | return setlocale(LC_ALL, NULL); | |
469 | ||
b385bb4d KW |
470 | } |
471 | ||
472 | #endif | |
473 | ||
474 | ||
98994639 HS |
475 | /* |
476 | * Initialize locale awareness. | |
477 | */ | |
478 | int | |
479 | Perl_init_i18nl10n(pTHX_ int printwarn) | |
480 | { | |
0e92a118 KW |
481 | /* printwarn is |
482 | * | |
483 | * 0 if not to output warning when setup locale is bad | |
484 | * 1 if to output warning based on value of PERL_BADLANG | |
485 | * >1 if to output regardless of PERL_BADLANG | |
486 | * | |
487 | * returns | |
98994639 | 488 | * 1 = set ok or not applicable, |
0e92a118 KW |
489 | * 0 = fallback to a locale of lower priority |
490 | * -1 = fallback to all locales failed, not even to the C locale | |
98994639 HS |
491 | */ |
492 | ||
0e92a118 KW |
493 | int ok = 1; |
494 | ||
98994639 | 495 | #if defined(USE_LOCALE) |
97aff369 | 496 | dVAR; |
98994639 HS |
497 | |
498 | #ifdef USE_LOCALE_CTYPE | |
499 | char *curctype = NULL; | |
500 | #endif /* USE_LOCALE_CTYPE */ | |
501 | #ifdef USE_LOCALE_COLLATE | |
502 | char *curcoll = NULL; | |
503 | #endif /* USE_LOCALE_COLLATE */ | |
504 | #ifdef USE_LOCALE_NUMERIC | |
505 | char *curnum = NULL; | |
506 | #endif /* USE_LOCALE_NUMERIC */ | |
507 | #ifdef __GLIBC__ | |
7452cf6a | 508 | char * const language = PerlEnv_getenv("LANGUAGE"); |
98994639 | 509 | #endif |
65ebb059 | 510 | |
ccd65d51 KW |
511 | /* NULL uses the existing already set up locale */ |
512 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) | |
513 | ? NULL | |
514 | : ""; | |
c3fcd832 KW |
515 | const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */ |
516 | unsigned int trial_locales_count; | |
7452cf6a AL |
517 | char * const lc_all = PerlEnv_getenv("LC_ALL"); |
518 | char * const lang = PerlEnv_getenv("LANG"); | |
98994639 | 519 | bool setlocale_failure = FALSE; |
65ebb059 KW |
520 | unsigned int i; |
521 | char *p; | |
522 | const bool locwarn = (printwarn > 1 || | |
523 | (printwarn && | |
524 | (!(p = PerlEnv_getenv("PERL_BADLANG")) || atoi(p)))); | |
0e92a118 | 525 | bool done = FALSE; |
6bce99ee JH |
526 | #ifdef WIN32 |
527 | /* In some systems you can find out the system default locale | |
528 | * and use that as the fallback locale. */ | |
529 | # define SYSTEM_DEFAULT_LOCALE | |
530 | #endif | |
531 | #ifdef SYSTEM_DEFAULT_LOCALE | |
65ebb059 | 532 | const char *system_default_locale = NULL; |
6bce99ee | 533 | #endif |
98994639 | 534 | |
0e92a118 KW |
535 | #ifndef LOCALE_ENVIRON_REQUIRED |
536 | PERL_UNUSED_VAR(done); | |
537 | #else | |
98994639 HS |
538 | |
539 | /* | |
540 | * Ultrix setlocale(..., "") fails if there are no environment | |
541 | * variables from which to get a locale name. | |
542 | */ | |
543 | ||
b3e384bf | 544 | # ifdef LC_ALL |
98994639 | 545 | if (lang) { |
b385bb4d | 546 | if (my_setlocale(LC_ALL, setlocale_init)) |
98994639 HS |
547 | done = TRUE; |
548 | else | |
549 | setlocale_failure = TRUE; | |
550 | } | |
551 | if (!setlocale_failure) { | |
b3e384bf | 552 | # ifdef USE_LOCALE_CTYPE |
56279a21 | 553 | Safefree(curctype); |
98994639 | 554 | if (! (curctype = |
b385bb4d | 555 | my_setlocale(LC_CTYPE, |
98994639 | 556 | (!done && (lang || PerlEnv_getenv("LC_CTYPE"))) |
ccd65d51 | 557 | ? setlocale_init : NULL))) |
98994639 HS |
558 | setlocale_failure = TRUE; |
559 | else | |
560 | curctype = savepv(curctype); | |
b3e384bf KW |
561 | # endif /* USE_LOCALE_CTYPE */ |
562 | # ifdef USE_LOCALE_COLLATE | |
56279a21 | 563 | Safefree(curcoll); |
98994639 | 564 | if (! (curcoll = |
b385bb4d | 565 | my_setlocale(LC_COLLATE, |
98994639 | 566 | (!done && (lang || PerlEnv_getenv("LC_COLLATE"))) |
ccd65d51 | 567 | ? setlocale_init : NULL))) |
98994639 HS |
568 | setlocale_failure = TRUE; |
569 | else | |
570 | curcoll = savepv(curcoll); | |
b3e384bf KW |
571 | # endif /* USE_LOCALE_COLLATE */ |
572 | # ifdef USE_LOCALE_NUMERIC | |
56279a21 | 573 | Safefree(curnum); |
98994639 | 574 | if (! (curnum = |
b385bb4d | 575 | my_setlocale(LC_NUMERIC, |
98994639 | 576 | (!done && (lang || PerlEnv_getenv("LC_NUMERIC"))) |
ccd65d51 | 577 | ? setlocale_init : NULL))) |
98994639 HS |
578 | setlocale_failure = TRUE; |
579 | else | |
580 | curnum = savepv(curnum); | |
b3e384bf | 581 | # endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
582 | # ifdef USE_LOCALE_MESSAGES |
583 | if (! my_setlocale(LC_MESSAGES, | |
584 | (!done && (lang || PerlEnv_getenv("LC_MESSAGES"))) | |
585 | ? setlocale_init : NULL)) | |
586 | { | |
587 | setlocale_failure = TRUE; | |
588 | } | |
589 | # endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
590 | # ifdef USE_LOCALE_MONETARY |
591 | if (! my_setlocale(LC_MONETARY, | |
592 | (!done && (lang || PerlEnv_getenv("LC_MONETARY"))) | |
593 | ? setlocale_init : NULL)) | |
594 | { | |
595 | setlocale_failure = TRUE; | |
596 | } | |
597 | # endif /* USE_LOCALE_MONETARY */ | |
98994639 HS |
598 | } |
599 | ||
b3e384bf | 600 | # endif /* LC_ALL */ |
98994639 HS |
601 | |
602 | #endif /* !LOCALE_ENVIRON_REQUIRED */ | |
603 | ||
65ebb059 KW |
604 | /* We try each locale in the list until we get one that works, or exhaust |
605 | * the list */ | |
c3fcd832 KW |
606 | trial_locales[0] = setlocale_init; |
607 | trial_locales_count = 1; | |
65ebb059 KW |
608 | for (i= 0; i < trial_locales_count; i++) { |
609 | const char * trial_locale = trial_locales[i]; | |
610 | ||
611 | if (i > 0) { | |
612 | ||
613 | /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED | |
614 | * when i==0, but I (khw) don't think that behavior makes much | |
615 | * sense */ | |
616 | setlocale_failure = FALSE; | |
617 | ||
6bce99ee JH |
618 | #ifdef SYSTEM_DEFAULT_LOCALE |
619 | # ifdef WIN32 | |
65ebb059 KW |
620 | /* On Windows machines, an entry of "" after the 0th means to use |
621 | * the system default locale, which we now proceed to get. */ | |
622 | if (strEQ(trial_locale, "")) { | |
623 | unsigned int j; | |
624 | ||
625 | /* Note that this may change the locale, but we are going to do | |
626 | * that anyway just below */ | |
627 | system_default_locale = setlocale(LC_ALL, ""); | |
628 | ||
629 | /* Skip if invalid or it's already on the list of locales to | |
630 | * try */ | |
631 | if (! system_default_locale) { | |
632 | goto next_iteration; | |
633 | } | |
634 | for (j = 0; j < trial_locales_count; j++) { | |
635 | if (strEQ(system_default_locale, trial_locales[j])) { | |
636 | goto next_iteration; | |
637 | } | |
638 | } | |
639 | ||
640 | trial_locale = system_default_locale; | |
641 | } | |
6bce99ee JH |
642 | # endif /* WIN32 */ |
643 | #endif /* SYSTEM_DEFAULT_LOCALE */ | |
65ebb059 KW |
644 | } |
645 | ||
98994639 | 646 | #ifdef LC_ALL |
7cd8b568 | 647 | if (! my_setlocale(LC_ALL, trial_locale)) { |
49c85077 | 648 | setlocale_failure = TRUE; |
7cd8b568 KW |
649 | } |
650 | else { | |
651 | /* Since LC_ALL succeeded, it should have changed all the other | |
652 | * categories it can to its value; so we massage things so that the | |
653 | * setlocales below just return their category's current values. | |
654 | * This adequately handles the case in NetBSD where LC_COLLATE may | |
655 | * not be defined for a locale, and setting it individually will | |
656 | * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to | |
657 | * the POSIX locale. */ | |
658 | trial_locale = NULL; | |
659 | } | |
98994639 HS |
660 | #endif /* LC_ALL */ |
661 | ||
49c85077 | 662 | if (!setlocale_failure) { |
98994639 | 663 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
664 | Safefree(curctype); |
665 | if (! (curctype = my_setlocale(LC_CTYPE, trial_locale))) | |
666 | setlocale_failure = TRUE; | |
667 | else | |
668 | curctype = savepv(curctype); | |
98994639 HS |
669 | #endif /* USE_LOCALE_CTYPE */ |
670 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
671 | Safefree(curcoll); |
672 | if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale))) | |
673 | setlocale_failure = TRUE; | |
674 | else | |
675 | curcoll = savepv(curcoll); | |
98994639 HS |
676 | #endif /* USE_LOCALE_COLLATE */ |
677 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
678 | Safefree(curnum); |
679 | if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale))) | |
680 | setlocale_failure = TRUE; | |
681 | else | |
682 | curnum = savepv(curnum); | |
98994639 | 683 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
684 | #ifdef USE_LOCALE_MESSAGES |
685 | if (! (my_setlocale(LC_MESSAGES, trial_locale))) | |
686 | setlocale_failure = TRUE; | |
687 | #endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
688 | #ifdef USE_LOCALE_MONETARY |
689 | if (! (my_setlocale(LC_MONETARY, trial_locale))) | |
690 | setlocale_failure = TRUE; | |
691 | #endif /* USE_LOCALE_MONETARY */ | |
692 | ||
49c85077 KW |
693 | if (! setlocale_failure) { /* Success */ |
694 | break; | |
695 | } | |
65ebb059 | 696 | } |
98994639 | 697 | |
49c85077 KW |
698 | /* Here, something failed; will need to try a fallback. */ |
699 | ok = 0; | |
65ebb059 | 700 | |
49c85077 KW |
701 | if (i == 0) { |
702 | unsigned int j; | |
98994639 | 703 | |
65ebb059 | 704 | if (locwarn) { /* Output failure info only on the first one */ |
98994639 HS |
705 | #ifdef LC_ALL |
706 | ||
49c85077 KW |
707 | PerlIO_printf(Perl_error_log, |
708 | "perl: warning: Setting locale failed.\n"); | |
98994639 HS |
709 | |
710 | #else /* !LC_ALL */ | |
711 | ||
49c85077 KW |
712 | PerlIO_printf(Perl_error_log, |
713 | "perl: warning: Setting locale failed for the categories:\n\t"); | |
98994639 | 714 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
715 | if (! curctype) |
716 | PerlIO_printf(Perl_error_log, "LC_CTYPE "); | |
98994639 HS |
717 | #endif /* USE_LOCALE_CTYPE */ |
718 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
719 | if (! curcoll) |
720 | PerlIO_printf(Perl_error_log, "LC_COLLATE "); | |
98994639 HS |
721 | #endif /* USE_LOCALE_COLLATE */ |
722 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
723 | if (! curnum) |
724 | PerlIO_printf(Perl_error_log, "LC_NUMERIC "); | |
98994639 | 725 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d | 726 | PerlIO_printf(Perl_error_log, "and possibly others\n"); |
98994639 HS |
727 | |
728 | #endif /* LC_ALL */ | |
729 | ||
49c85077 KW |
730 | PerlIO_printf(Perl_error_log, |
731 | "perl: warning: Please check that your locale settings:\n"); | |
98994639 HS |
732 | |
733 | #ifdef __GLIBC__ | |
49c85077 KW |
734 | PerlIO_printf(Perl_error_log, |
735 | "\tLANGUAGE = %c%s%c,\n", | |
736 | language ? '"' : '(', | |
737 | language ? language : "unset", | |
738 | language ? '"' : ')'); | |
98994639 HS |
739 | #endif |
740 | ||
49c85077 KW |
741 | PerlIO_printf(Perl_error_log, |
742 | "\tLC_ALL = %c%s%c,\n", | |
743 | lc_all ? '"' : '(', | |
744 | lc_all ? lc_all : "unset", | |
745 | lc_all ? '"' : ')'); | |
98994639 HS |
746 | |
747 | #if defined(USE_ENVIRON_ARRAY) | |
49c85077 KW |
748 | { |
749 | char **e; | |
750 | for (e = environ; *e; e++) { | |
751 | if (strnEQ(*e, "LC_", 3) | |
752 | && strnNE(*e, "LC_ALL=", 7) | |
753 | && (p = strchr(*e, '='))) | |
754 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", | |
755 | (int)(p - *e), *e, p + 1); | |
756 | } | |
757 | } | |
98994639 | 758 | #else |
49c85077 KW |
759 | PerlIO_printf(Perl_error_log, |
760 | "\t(possibly more locale environment variables)\n"); | |
98994639 HS |
761 | #endif |
762 | ||
49c85077 KW |
763 | PerlIO_printf(Perl_error_log, |
764 | "\tLANG = %c%s%c\n", | |
765 | lang ? '"' : '(', | |
766 | lang ? lang : "unset", | |
767 | lang ? '"' : ')'); | |
98994639 | 768 | |
49c85077 KW |
769 | PerlIO_printf(Perl_error_log, |
770 | " are supported and installed on your system.\n"); | |
771 | } | |
98994639 | 772 | |
65ebb059 KW |
773 | /* Calculate what fallback locales to try. We have avoided this |
774 | * until we have to, becuase failure is quite unlikely. This will | |
775 | * usually change the upper bound of the loop we are in. | |
776 | * | |
777 | * Since the system's default way of setting the locale has not | |
778 | * found one that works, We use Perl's defined ordering: LC_ALL, | |
779 | * LANG, and the C locale. We don't try the same locale twice, so | |
780 | * don't add to the list if already there. (On POSIX systems, the | |
781 | * LC_ALL element will likely be a repeat of the 0th element "", | |
782 | * but there's no harm done by doing it explicitly */ | |
783 | if (lc_all) { | |
784 | for (j = 0; j < trial_locales_count; j++) { | |
785 | if (strEQ(lc_all, trial_locales[j])) { | |
786 | goto done_lc_all; | |
787 | } | |
788 | } | |
789 | trial_locales[trial_locales_count++] = lc_all; | |
790 | } | |
791 | done_lc_all: | |
98994639 | 792 | |
65ebb059 KW |
793 | if (lang) { |
794 | for (j = 0; j < trial_locales_count; j++) { | |
795 | if (strEQ(lang, trial_locales[j])) { | |
796 | goto done_lang; | |
797 | } | |
798 | } | |
799 | trial_locales[trial_locales_count++] = lang; | |
800 | } | |
801 | done_lang: | |
802 | ||
803 | #if defined(WIN32) && defined(LC_ALL) | |
804 | /* For Windows, we also try the system default locale before "C". | |
805 | * (If there exists a Windows without LC_ALL we skip this because | |
806 | * it gets too complicated. For those, the "C" is the next | |
807 | * fallback possibility). The "" is the same as the 0th element of | |
808 | * the array, but the code at the loop above knows to treat it | |
809 | * differently when not the 0th */ | |
810 | trial_locales[trial_locales_count++] = ""; | |
811 | #endif | |
812 | ||
813 | for (j = 0; j < trial_locales_count; j++) { | |
814 | if (strEQ("C", trial_locales[j])) { | |
815 | goto done_C; | |
816 | } | |
817 | } | |
818 | trial_locales[trial_locales_count++] = "C"; | |
98994639 | 819 | |
65ebb059 KW |
820 | done_C: ; |
821 | } /* end of first time through the loop */ | |
98994639 | 822 | |
65ebb059 KW |
823 | #ifdef WIN32 |
824 | next_iteration: ; | |
825 | #endif | |
826 | ||
827 | } /* end of looping through the trial locales */ | |
828 | ||
829 | if (ok < 1) { /* If we tried to fallback */ | |
830 | const char* msg; | |
831 | if (! setlocale_failure) { /* fallback succeeded */ | |
832 | msg = "Falling back to"; | |
833 | } | |
834 | else { /* fallback failed */ | |
98994639 | 835 | |
65ebb059 KW |
836 | /* We dropped off the end of the loop, so have to decrement i to |
837 | * get back to the value the last time through */ | |
838 | i--; | |
98994639 | 839 | |
65ebb059 KW |
840 | ok = -1; |
841 | msg = "Failed to fall back to"; | |
842 | ||
843 | /* To continue, we should use whatever values we've got */ | |
98994639 | 844 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
845 | Safefree(curctype); |
846 | curctype = savepv(setlocale(LC_CTYPE, NULL)); | |
98994639 HS |
847 | #endif /* USE_LOCALE_CTYPE */ |
848 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
849 | Safefree(curcoll); |
850 | curcoll = savepv(setlocale(LC_COLLATE, NULL)); | |
98994639 HS |
851 | #endif /* USE_LOCALE_COLLATE */ |
852 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
853 | Safefree(curnum); |
854 | curnum = savepv(setlocale(LC_NUMERIC, NULL)); | |
98994639 | 855 | #endif /* USE_LOCALE_NUMERIC */ |
65ebb059 KW |
856 | } |
857 | ||
858 | if (locwarn) { | |
859 | const char * description; | |
860 | const char * name = ""; | |
861 | if (strEQ(trial_locales[i], "C")) { | |
862 | description = "the standard locale"; | |
863 | name = "C"; | |
864 | } | |
6bce99ee | 865 | #ifdef SYSTEM_DEFAULT_LOCALE |
65ebb059 KW |
866 | else if (strEQ(trial_locales[i], "")) { |
867 | description = "the system default locale"; | |
868 | if (system_default_locale) { | |
869 | name = system_default_locale; | |
870 | } | |
871 | } | |
6bce99ee | 872 | #endif /* SYSTEM_DEFAULT_LOCALE */ |
65ebb059 KW |
873 | else { |
874 | description = "a fallback locale"; | |
875 | name = trial_locales[i]; | |
876 | } | |
877 | if (name && strNE(name, "")) { | |
878 | PerlIO_printf(Perl_error_log, | |
879 | "perl: warning: %s %s (\"%s\").\n", msg, description, name); | |
880 | } | |
881 | else { | |
882 | PerlIO_printf(Perl_error_log, | |
883 | "perl: warning: %s %s.\n", msg, description); | |
884 | } | |
885 | } | |
886 | } /* End of tried to fallback */ | |
98994639 HS |
887 | |
888 | #ifdef USE_LOCALE_CTYPE | |
889 | new_ctype(curctype); | |
890 | #endif /* USE_LOCALE_CTYPE */ | |
891 | ||
892 | #ifdef USE_LOCALE_COLLATE | |
893 | new_collate(curcoll); | |
894 | #endif /* USE_LOCALE_COLLATE */ | |
895 | ||
896 | #ifdef USE_LOCALE_NUMERIC | |
897 | new_numeric(curnum); | |
898 | #endif /* USE_LOCALE_NUMERIC */ | |
b310b053 | 899 | |
8ef6e574 | 900 | #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE) |
49c85077 KW |
901 | /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE |
902 | * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by | |
903 | * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the | |
904 | * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open | |
905 | * discipline. */ | |
c1284011 | 906 | PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE); |
49c85077 | 907 | |
a05d7ebb | 908 | /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO. |
fde18df1 JH |
909 | This is an alternative to using the -C command line switch |
910 | (the -C if present will override this). */ | |
911 | { | |
dd374669 | 912 | const char *p = PerlEnv_getenv("PERL_UNICODE"); |
a05d7ebb | 913 | PL_unicode = p ? parse_unicode_opts(&p) : 0; |
5a22a2bb NC |
914 | if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG) |
915 | PL_utf8cache = -1; | |
b310b053 | 916 | } |
ec71e770 | 917 | #endif |
b310b053 | 918 | |
98994639 | 919 | #ifdef USE_LOCALE_CTYPE |
43c5f42d | 920 | Safefree(curctype); |
98994639 HS |
921 | #endif /* USE_LOCALE_CTYPE */ |
922 | #ifdef USE_LOCALE_COLLATE | |
43c5f42d | 923 | Safefree(curcoll); |
98994639 HS |
924 | #endif /* USE_LOCALE_COLLATE */ |
925 | #ifdef USE_LOCALE_NUMERIC | |
43c5f42d | 926 | Safefree(curnum); |
98994639 | 927 | #endif /* USE_LOCALE_NUMERIC */ |
8ef6e574 KW |
928 | |
929 | #endif /* USE_LOCALE */ | |
930 | ||
98994639 HS |
931 | return ok; |
932 | } | |
933 | ||
49c85077 | 934 | |
98994639 HS |
935 | #ifdef USE_LOCALE_COLLATE |
936 | ||
937 | /* | |
938 | * mem_collxfrm() is a bit like strxfrm() but with two important | |
939 | * differences. First, it handles embedded NULs. Second, it allocates | |
940 | * a bit more memory than needed for the transformed data itself. | |
941 | * The real transformed data begins at offset sizeof(collationix). | |
942 | * Please see sv_collxfrm() to see how this is used. | |
943 | */ | |
166f8a29 | 944 | |
98994639 HS |
945 | char * |
946 | Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen) | |
947 | { | |
97aff369 | 948 | dVAR; |
98994639 HS |
949 | char *xbuf; |
950 | STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */ | |
951 | ||
7918f24d NC |
952 | PERL_ARGS_ASSERT_MEM_COLLXFRM; |
953 | ||
98994639 HS |
954 | /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */ |
955 | /* the +1 is for the terminating NUL. */ | |
956 | ||
957 | xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1; | |
a02a5408 | 958 | Newx(xbuf, xAlloc, char); |
98994639 HS |
959 | if (! xbuf) |
960 | goto bad; | |
961 | ||
962 | *(U32*)xbuf = PL_collation_ix; | |
963 | xout = sizeof(PL_collation_ix); | |
964 | for (xin = 0; xin < len; ) { | |
224e8ef5 | 965 | Size_t xused; |
98994639 HS |
966 | |
967 | for (;;) { | |
968 | xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout); | |
224e8ef5 | 969 | if (xused >= PERL_INT_MAX) |
98994639 | 970 | goto bad; |
eb160463 | 971 | if ((STRLEN)xused < xAlloc - xout) |
98994639 HS |
972 | break; |
973 | xAlloc = (2 * xAlloc) + 1; | |
974 | Renew(xbuf, xAlloc, char); | |
975 | if (! xbuf) | |
976 | goto bad; | |
977 | } | |
978 | ||
979 | xin += strlen(s + xin) + 1; | |
980 | xout += xused; | |
981 | ||
982 | /* Embedded NULs are understood but silently skipped | |
983 | * because they make no sense in locale collation. */ | |
984 | } | |
985 | ||
986 | xbuf[xout] = '\0'; | |
987 | *xlen = xout - sizeof(PL_collation_ix); | |
988 | return xbuf; | |
989 | ||
990 | bad: | |
991 | Safefree(xbuf); | |
992 | *xlen = 0; | |
993 | return NULL; | |
994 | } | |
995 | ||
996 | #endif /* USE_LOCALE_COLLATE */ | |
997 | ||
8ef6e574 KW |
998 | #ifdef USE_LOCALE |
999 | ||
c1284011 KW |
1000 | bool |
1001 | Perl__is_cur_LC_category_utf8(pTHX_ int category) | |
7d74bb61 KW |
1002 | { |
1003 | /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE | |
1004 | * otherwise. 'category' may not be LC_ALL. If the platform doesn't have | |
119ee68b KW |
1005 | * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence |
1006 | * could give the wrong result. It errs on the side of not being a UTF-8 | |
1007 | * locale. */ | |
7d74bb61 KW |
1008 | |
1009 | char *save_input_locale = NULL; | |
7d74bb61 KW |
1010 | STRLEN final_pos; |
1011 | ||
8ef6e574 | 1012 | #ifdef LC_ALL |
7d74bb61 | 1013 | assert(category != LC_ALL); |
8ef6e574 | 1014 | #endif |
7d74bb61 KW |
1015 | |
1016 | /* First dispose of the trivial cases */ | |
b07fffd1 | 1017 | save_input_locale = setlocale(category, NULL); |
7d74bb61 | 1018 | if (! save_input_locale) { |
69014004 KW |
1019 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1020 | "Could not find current locale for category %d\n", | |
1021 | category)); | |
7d74bb61 KW |
1022 | return FALSE; /* XXX maybe should croak */ |
1023 | } | |
b07fffd1 | 1024 | save_input_locale = stdize_locale(savepv(save_input_locale)); |
7d74bb61 KW |
1025 | if ((*save_input_locale == 'C' && save_input_locale[1] == '\0') |
1026 | || strEQ(save_input_locale, "POSIX")) | |
1027 | { | |
69014004 KW |
1028 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1029 | "Current locale for category %d is %s\n", | |
1030 | category, save_input_locale)); | |
b07fffd1 | 1031 | Safefree(save_input_locale); |
7d74bb61 KW |
1032 | return FALSE; |
1033 | } | |
1034 | ||
1d958db2 KW |
1035 | #if defined(USE_LOCALE_CTYPE) \ |
1036 | && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET))) | |
7d74bb61 | 1037 | |
1d958db2 | 1038 | { /* Next try nl_langinfo or MB_CUR_MAX if available */ |
7d74bb61 KW |
1039 | |
1040 | char *save_ctype_locale = NULL; | |
119ee68b | 1041 | bool is_utf8; |
7d74bb61 | 1042 | |
119ee68b | 1043 | if (category != LC_CTYPE) { /* These work only on LC_CTYPE */ |
7d74bb61 KW |
1044 | |
1045 | /* Get the current LC_CTYPE locale */ | |
1046 | save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL))); | |
1047 | if (! save_ctype_locale) { | |
69014004 KW |
1048 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1049 | "Could not find current locale for LC_CTYPE\n")); | |
7d74bb61 KW |
1050 | goto cant_use_nllanginfo; |
1051 | } | |
1052 | ||
1053 | /* If LC_CTYPE and the desired category use the same locale, this | |
1054 | * means that finding the value for LC_CTYPE is the same as finding | |
1055 | * the value for the desired category. Otherwise, switch LC_CTYPE | |
1056 | * to the desired category's locale */ | |
1057 | if (strEQ(save_ctype_locale, save_input_locale)) { | |
1058 | Safefree(save_ctype_locale); | |
1059 | save_ctype_locale = NULL; | |
1060 | } | |
1061 | else if (! setlocale(LC_CTYPE, save_input_locale)) { | |
69014004 KW |
1062 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1063 | "Could not change LC_CTYPE locale to %s\n", | |
1064 | save_input_locale)); | |
7d74bb61 KW |
1065 | Safefree(save_ctype_locale); |
1066 | goto cant_use_nllanginfo; | |
1067 | } | |
1068 | } | |
1069 | ||
69014004 KW |
1070 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n", |
1071 | save_input_locale)); | |
1072 | ||
7d74bb61 | 1073 | /* Here the current LC_CTYPE is set to the locale of the category whose |
1d958db2 KW |
1074 | * information is desired. This means that nl_langinfo() and MB_CUR_MAX |
1075 | * should give the correct results */ | |
119ee68b | 1076 | |
1d958db2 KW |
1077 | # if defined(HAS_NL_LANGINFO) && defined(CODESET) |
1078 | { | |
1079 | char *codeset = savepv(nl_langinfo(CODESET)); | |
1080 | if (codeset && strNE(codeset, "")) { | |
119ee68b | 1081 | |
1d958db2 KW |
1082 | /* If we switched LC_CTYPE, switch back */ |
1083 | if (save_ctype_locale) { | |
1084 | setlocale(LC_CTYPE, save_ctype_locale); | |
1085 | Safefree(save_ctype_locale); | |
1086 | } | |
1087 | ||
1088 | is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8")) | |
1089 | || foldEQ(codeset, STR_WITH_LEN("UTF8")); | |
1090 | ||
69014004 KW |
1091 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1092 | "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n", | |
1093 | codeset, is_utf8)); | |
1d958db2 KW |
1094 | Safefree(codeset); |
1095 | Safefree(save_input_locale); | |
1096 | return is_utf8; | |
1097 | } | |
0ac78434 | 1098 | Safefree(codeset); |
119ee68b KW |
1099 | } |
1100 | ||
1d958db2 KW |
1101 | # endif |
1102 | # ifdef MB_CUR_MAX | |
1103 | ||
1104 | /* Here, either we don't have nl_langinfo, or it didn't return a | |
1105 | * codeset. Try MB_CUR_MAX */ | |
1106 | ||
119ee68b KW |
1107 | /* Standard UTF-8 needs at least 4 bytes to represent the maximum |
1108 | * Unicode code point. Since UTF-8 is the only non-single byte | |
1109 | * encoding we handle, we just say any such encoding is UTF-8, and if | |
1110 | * turns out to be wrong, other things will fail */ | |
1111 | is_utf8 = MB_CUR_MAX >= 4; | |
1112 | ||
69014004 KW |
1113 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1114 | "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n", | |
1115 | (int) MB_CUR_MAX, is_utf8)); | |
1116 | ||
119ee68b KW |
1117 | Safefree(save_input_locale); |
1118 | ||
1119 | # ifdef HAS_MBTOWC | |
1120 | ||
1121 | /* ... But, most system that have MB_CUR_MAX will also have mbtowc(), | |
1122 | * since they are both in the C99 standard. We can feed a known byte | |
1123 | * string to the latter function, and check that it gives the expected | |
1124 | * result */ | |
1125 | if (is_utf8) { | |
1126 | wchar_t wc; | |
a6715020 | 1127 | GCC_DIAG_IGNORE(-Wunused-result); |
1d958db2 | 1128 | (void) mbtowc(&wc, NULL, 0); /* Reset any shift state */ |
a6715020 | 1129 | GCC_DIAG_RESTORE; |
69014004 | 1130 | errno = 0; |
f019f68f | 1131 | if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)) |
119ee68b KW |
1132 | != strlen(HYPHEN_UTF8) |
1133 | || wc != (wchar_t) 0x2010) | |
1134 | { | |
1135 | is_utf8 = FALSE; | |
69014004 KW |
1136 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", wc)); |
1137 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1138 | "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n", | |
1139 | mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno)); | |
119ee68b KW |
1140 | } |
1141 | } | |
119ee68b KW |
1142 | # endif |
1143 | ||
1d958db2 KW |
1144 | /* If we switched LC_CTYPE, switch back */ |
1145 | if (save_ctype_locale) { | |
1146 | setlocale(LC_CTYPE, save_ctype_locale); | |
1147 | Safefree(save_ctype_locale); | |
119ee68b | 1148 | } |
7d74bb61 | 1149 | |
1d958db2 | 1150 | return is_utf8; |
119ee68b | 1151 | # endif |
7d74bb61 | 1152 | } |
119ee68b | 1153 | |
7d74bb61 KW |
1154 | cant_use_nllanginfo: |
1155 | ||
1156 | #endif /* HAS_NL_LANGINFO etc */ | |
1157 | ||
1158 | /* nl_langinfo not available or failed somehow. Look at the locale name to | |
751e9426 | 1159 | * see if it matches qr/UTF -? 8 /ix */ |
7d74bb61 KW |
1160 | |
1161 | final_pos = strlen(save_input_locale) - 1; | |
751e9426 KW |
1162 | if (final_pos >= 3) { |
1163 | char *name = save_input_locale; | |
1164 | ||
1165 | /* Find next 'U' or 'u' and look from there */ | |
1166 | while ((name += strcspn(name, "Uu") + 1) | |
1167 | <= save_input_locale + final_pos - 2) | |
7d74bb61 | 1168 | { |
751e9426 KW |
1169 | if (toFOLD(*(name)) != 't' |
1170 | || toFOLD(*(name + 1)) != 'f') | |
1171 | { | |
1172 | continue; | |
1173 | } | |
1174 | name += 2; | |
1175 | if (*(name) == '-') { | |
1176 | if ((name > save_input_locale + final_pos - 1)) { | |
1177 | break; | |
1178 | } | |
1179 | name++; | |
1180 | } | |
1181 | if (*(name) == '8') { | |
1182 | Safefree(save_input_locale); | |
69014004 KW |
1183 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1184 | "Locale %s ends with UTF-8 in name\n", | |
1185 | save_input_locale)); | |
751e9426 KW |
1186 | return TRUE; |
1187 | } | |
7d74bb61 | 1188 | } |
69014004 KW |
1189 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1190 | "Locale %s doesn't end with UTF-8 in name\n", | |
1191 | save_input_locale)); | |
7d74bb61 KW |
1192 | } |
1193 | ||
1194 | #ifdef WIN32 | |
1195 | /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */ | |
1196 | if (final_pos >= 4 | |
1197 | && *(save_input_locale + final_pos - 0) == '1' | |
1198 | && *(save_input_locale + final_pos - 1) == '0' | |
1199 | && *(save_input_locale + final_pos - 2) == '0' | |
1200 | && *(save_input_locale + final_pos - 3) == '5' | |
1201 | && *(save_input_locale + final_pos - 4) == '6') | |
1202 | { | |
69014004 KW |
1203 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1204 | "Locale %s ends with 10056 in name, is UTF-8 locale\n", | |
1205 | save_input_locale)); | |
d37662c0 | 1206 | Safefree(save_input_locale); |
7d74bb61 KW |
1207 | return TRUE; |
1208 | } | |
1209 | #endif | |
1210 | ||
fa9b773e KW |
1211 | /* Other common encodings are the ISO 8859 series, which aren't UTF-8 */ |
1212 | if (instr(save_input_locale, "8859")) { | |
69014004 KW |
1213 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1214 | "Locale %s has 8859 in name, not UTF-8 locale\n", | |
1215 | save_input_locale)); | |
d37662c0 | 1216 | Safefree(save_input_locale); |
fa9b773e KW |
1217 | return FALSE; |
1218 | } | |
1219 | ||
1220 | #ifdef HAS_LOCALECONV | |
1221 | ||
1222 | # ifdef USE_LOCALE_MONETARY | |
1223 | ||
1224 | /* Here, there is nothing in the locale name to indicate whether the locale | |
1225 | * is UTF-8 or not. This "name", the return of setlocale(), is actually | |
1226 | * defined to be opaque, so we can't really rely on the absence of various | |
1227 | * substrings in the name to indicate its UTF-8ness. Look at the locale's | |
1228 | * currency symbol. Often that will be in the native script, and if the | |
1229 | * symbol isn't in UTF-8, we know that the locale isn't. If it is | |
1230 | * non-ASCII UTF-8, we infer that the locale is too. | |
1231 | * To do this, like above for LC_CTYPE, we first set LC_MONETARY to the | |
1232 | * locale of the desired category, if it isn't that locale already */ | |
1233 | ||
1234 | { | |
1235 | char *save_monetary_locale = NULL; | |
1236 | bool illegal_utf8 = FALSE; | |
1237 | bool only_ascii = FALSE; | |
1238 | const struct lconv* const lc = localeconv(); | |
1239 | ||
1240 | if (category != LC_MONETARY) { | |
1241 | ||
1242 | save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY, | |
1243 | NULL))); | |
1244 | if (! save_monetary_locale) { | |
69014004 KW |
1245 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1246 | "Could not find current locale for LC_MONETARY\n")); | |
fa9b773e KW |
1247 | goto cant_use_monetary; |
1248 | } | |
1249 | ||
1250 | if (strNE(save_monetary_locale, save_input_locale)) { | |
1251 | if (! setlocale(LC_MONETARY, save_input_locale)) { | |
69014004 KW |
1252 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1253 | "Could not change LC_MONETARY locale to %s\n", | |
1254 | save_input_locale)); | |
fa9b773e KW |
1255 | Safefree(save_monetary_locale); |
1256 | goto cant_use_monetary; | |
1257 | } | |
1258 | } | |
1259 | } | |
1260 | ||
1261 | /* Here the current LC_MONETARY is set to the locale of the category | |
1262 | * whose information is desired. */ | |
1263 | ||
1264 | if (lc && lc->currency_symbol) { | |
1265 | if (! is_utf8_string((U8 *) lc->currency_symbol, 0)) { | |
69014004 KW |
1266 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1267 | "Currency symbol for %s is not legal UTF-8\n", | |
1268 | save_input_locale)); | |
fa9b773e KW |
1269 | illegal_utf8 = TRUE; |
1270 | } | |
1271 | else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) { | |
69014004 | 1272 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Currency symbol for %s contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); |
fa9b773e KW |
1273 | only_ascii = TRUE; |
1274 | } | |
1275 | } | |
1276 | ||
1277 | /* If we changed it, restore LC_MONETARY to its original locale */ | |
1278 | if (save_monetary_locale) { | |
1279 | setlocale(LC_MONETARY, save_monetary_locale); | |
1280 | Safefree(save_monetary_locale); | |
1281 | } | |
1282 | ||
1283 | Safefree(save_input_locale); | |
1284 | ||
1285 | /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; otherwise | |
1286 | * assume the locale is UTF-8 if and only if the symbol is non-ascii | |
1287 | * UTF-8. (We can't really tell if the locale is UTF-8 or not if the | |
1288 | * symbol is just a '$', so we err on the side of it not being UTF-8) | |
1289 | * */ | |
69014004 KW |
1290 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\tis_utf8=%d\n", (illegal_utf8) |
1291 | ? FALSE | |
1292 | : ! only_ascii)); | |
fa9b773e KW |
1293 | return (illegal_utf8) |
1294 | ? FALSE | |
1295 | : ! only_ascii; | |
1296 | ||
1297 | } | |
1298 | cant_use_monetary: | |
1299 | ||
1300 | # endif /* USE_LOCALE_MONETARY */ | |
1301 | #endif /* HAS_LOCALECONV */ | |
1302 | ||
1303 | #if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES) | |
1304 | ||
1305 | /* This code is ifdefd out because it was found to not be necessary in testing | |
1306 | * on our dromedary test machine, which has over 700 locales. There, looking | |
1307 | * at just the currency symbol gave essentially the same results as doing this | |
1308 | * extra work. Executing this also caused segfaults in miniperl. I left it in | |
1309 | * so as to avoid rewriting it if real-world experience indicates that | |
1310 | * dromedary is an outlier. Essentially, instead of returning abpve if we | |
1311 | * haven't found illegal utf8, we continue on and examine all the strerror() | |
1312 | * messages on the platform for utf8ness. If all are ASCII, we still don't | |
1313 | * know the answer; but otherwise we have a pretty good indication of the | |
1314 | * utf8ness. The reason this doesn't necessarily help much is that the | |
1315 | * messages may not have been translated into the locale. The currency symbol | |
1316 | * is much more likely to have been translated. The code below would need to | |
1317 | * be altered somewhat to just be a continuation of testing the currency | |
1318 | * symbol. */ | |
1319 | int e; | |
1320 | unsigned int failures = 0, non_ascii = 0; | |
1321 | char *save_messages_locale = NULL; | |
1322 | ||
1323 | /* Like above for LC_CTYPE, we set LC_MESSAGES to the locale of the | |
1324 | * desired category, if it isn't that locale already */ | |
1325 | ||
1326 | if (category != LC_MESSAGES) { | |
1327 | ||
1328 | save_messages_locale = stdize_locale(savepv(setlocale(LC_MESSAGES, | |
1329 | NULL))); | |
1330 | if (! save_messages_locale) { | |
1331 | goto cant_use_messages; | |
1332 | } | |
1333 | ||
1334 | if (strEQ(save_messages_locale, save_input_locale)) { | |
1335 | Safefree(save_input_locale); | |
1336 | } | |
1337 | else if (! setlocale(LC_MESSAGES, save_input_locale)) { | |
1338 | Safefree(save_messages_locale); | |
1339 | goto cant_use_messages; | |
1340 | } | |
1341 | } | |
1342 | ||
1343 | /* Here the current LC_MESSAGES is set to the locale of the category | |
1344 | * whose information is desired. Look through all the messages */ | |
1345 | ||
1346 | for (e = 0; | |
1347 | #ifdef HAS_SYS_ERRLIST | |
1348 | e <= sys_nerr | |
1349 | #endif | |
1350 | ; e++) | |
1351 | { | |
1352 | const U8* const errmsg = (U8 *) Strerror(e) ; | |
1353 | if (!errmsg) | |
1354 | break; | |
1355 | if (! is_utf8_string(errmsg, 0)) { | |
1356 | failures++; | |
1357 | break; | |
1358 | } | |
1359 | else if (! is_ascii_string(errmsg, 0)) { | |
1360 | non_ascii++; | |
1361 | } | |
1362 | } | |
1363 | ||
1364 | /* And, if we changed it, restore LC_MESSAGES to its original locale */ | |
1365 | if (save_messages_locale) { | |
1366 | setlocale(LC_MESSAGES, save_messages_locale); | |
1367 | Safefree(save_messages_locale); | |
1368 | } | |
1369 | ||
1370 | /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid, | |
1371 | * any non-ascii means it is one; otherwise we assume it isn't */ | |
1372 | return (failures) ? FALSE : non_ascii; | |
1373 | ||
1374 | } | |
1375 | cant_use_messages: | |
1376 | ||
1377 | #endif | |
1378 | ||
69014004 KW |
1379 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1380 | "Assuming locale %s is not a UTF-8 locale\n", | |
1381 | save_input_locale)); | |
fa9b773e | 1382 | Safefree(save_input_locale); |
7d74bb61 KW |
1383 | return FALSE; |
1384 | } | |
1385 | ||
8ef6e574 | 1386 | #endif |
7d74bb61 | 1387 | |
66610fdd RGS |
1388 | /* |
1389 | * Local variables: | |
1390 | * c-indentation-style: bsd | |
1391 | * c-basic-offset: 4 | |
14d04a33 | 1392 | * indent-tabs-mode: nil |
66610fdd RGS |
1393 | * End: |
1394 | * | |
14d04a33 | 1395 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 1396 | */ |