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) | |
112 | && is_cur_LC_category_utf8(LC_NUMERIC)) | |
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 | |
151 | * that the current locale is the default locale | |
152 | * PL_numeric_standard A boolean indicating if the toggled state is such | |
153 | * that the current locale is the C locale | |
154 | * Note that both of the last two variables can be true at the same time, | |
155 | * if the underlying locale is C. (Toggling is a no-op under these | |
156 | * circumstances.) | |
157 | * | |
158 | * Any code changing the locale (outside this file) should use | |
159 | * POSIX::setlocale, which calls this function. Therefore this function | |
160 | * should be called directly only from this file and from | |
161 | * POSIX::setlocale() */ | |
162 | ||
b03f34cf | 163 | char *save_newnum; |
97aff369 | 164 | dVAR; |
98994639 HS |
165 | |
166 | if (! newnum) { | |
43c5f42d NC |
167 | Safefree(PL_numeric_name); |
168 | PL_numeric_name = NULL; | |
98994639 HS |
169 | PL_numeric_standard = TRUE; |
170 | PL_numeric_local = TRUE; | |
171 | return; | |
172 | } | |
173 | ||
b03f34cf KW |
174 | save_newnum = stdize_locale(savepv(newnum)); |
175 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { | |
98994639 | 176 | Safefree(PL_numeric_name); |
b03f34cf | 177 | PL_numeric_name = save_newnum; |
b03f34cf | 178 | } |
98994639 | 179 | |
e19f01cb KW |
180 | PL_numeric_standard = ((*save_newnum == 'C' && save_newnum[1] == '\0') |
181 | || strEQ(save_newnum, "POSIX")); | |
182 | PL_numeric_local = TRUE; | |
183 | set_numeric_radix(); | |
6959d69d | 184 | |
98994639 HS |
185 | #endif /* USE_LOCALE_NUMERIC */ |
186 | } | |
187 | ||
188 | void | |
189 | Perl_set_numeric_standard(pTHX) | |
190 | { | |
191 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 192 | dVAR; |
98994639 | 193 | |
0d071d52 KW |
194 | /* Toggle the LC_NUMERIC locale to C, if not already there. Probably |
195 | * should use the macros like SET_NUMERIC_STANDARD() in perl.h instead of | |
196 | * calling this directly. */ | |
197 | ||
98994639 HS |
198 | if (! PL_numeric_standard) { |
199 | setlocale(LC_NUMERIC, "C"); | |
200 | PL_numeric_standard = TRUE; | |
201 | PL_numeric_local = FALSE; | |
202 | set_numeric_radix(); | |
203 | } | |
69014004 KW |
204 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
205 | "Underlying LC_NUMERIC locale now is C\n")); | |
98994639 HS |
206 | |
207 | #endif /* USE_LOCALE_NUMERIC */ | |
208 | } | |
209 | ||
210 | void | |
211 | Perl_set_numeric_local(pTHX) | |
212 | { | |
213 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 214 | dVAR; |
98994639 | 215 | |
0d071d52 KW |
216 | /* Toggle the LC_NUMERIC locale to the current underlying default, if not |
217 | * already there. Probably should use the macros like SET_NUMERIC_LOCAL() | |
218 | * in perl.h instead of calling this directly. */ | |
219 | ||
98994639 HS |
220 | if (! PL_numeric_local) { |
221 | setlocale(LC_NUMERIC, PL_numeric_name); | |
222 | PL_numeric_standard = FALSE; | |
223 | PL_numeric_local = TRUE; | |
224 | set_numeric_radix(); | |
225 | } | |
69014004 KW |
226 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
227 | "Underlying LC_NUMERIC locale now is %s\n", | |
228 | PL_numeric_name)); | |
98994639 HS |
229 | |
230 | #endif /* USE_LOCALE_NUMERIC */ | |
231 | } | |
232 | ||
233 | /* | |
234 | * Set up for a new ctype locale. | |
235 | */ | |
236 | void | |
8772537c | 237 | Perl_new_ctype(pTHX_ const char *newctype) |
98994639 HS |
238 | { |
239 | #ifdef USE_LOCALE_CTYPE | |
0d071d52 KW |
240 | |
241 | /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell | |
242 | * core Perl this and that 'newctype' is the name of the new locale. | |
243 | * | |
244 | * This function sets up the folding arrays for all 256 bytes, assuming | |
245 | * that tofold() is tolc() since fold case is not a concept in POSIX, | |
246 | * | |
247 | * Any code changing the locale (outside this file) should use | |
248 | * POSIX::setlocale, which calls this function. Therefore this function | |
249 | * should be called directly only from this file and from | |
250 | * POSIX::setlocale() */ | |
251 | ||
27da23d5 | 252 | dVAR; |
68067e4e | 253 | UV i; |
98994639 | 254 | |
7918f24d NC |
255 | PERL_ARGS_ASSERT_NEW_CTYPE; |
256 | ||
31f05a37 KW |
257 | PL_in_utf8_CTYPE_locale = is_cur_LC_category_utf8(LC_CTYPE); |
258 | ||
259 | /* A UTF-8 locale gets standard rules. But note that code still has to | |
260 | * handle this specially because of the three problematic code points */ | |
261 | if (PL_in_utf8_CTYPE_locale) { | |
262 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); | |
263 | } | |
264 | else { | |
baa60164 KW |
265 | for (i = 0; i < 256; i++) { |
266 | if (isUPPER_LC((U8) i)) | |
267 | PL_fold_locale[i] = (U8) toLOWER_LC((U8) i); | |
268 | else if (isLOWER_LC((U8) i)) | |
269 | PL_fold_locale[i] = (U8) toUPPER_LC((U8) i); | |
270 | else | |
271 | PL_fold_locale[i] = (U8) i; | |
272 | } | |
31f05a37 | 273 | } |
98994639 HS |
274 | |
275 | #endif /* USE_LOCALE_CTYPE */ | |
7918f24d | 276 | PERL_ARGS_ASSERT_NEW_CTYPE; |
8772537c | 277 | PERL_UNUSED_ARG(newctype); |
96a5add6 | 278 | PERL_UNUSED_CONTEXT; |
98994639 HS |
279 | } |
280 | ||
98994639 | 281 | void |
8772537c | 282 | Perl_new_collate(pTHX_ const char *newcoll) |
98994639 HS |
283 | { |
284 | #ifdef USE_LOCALE_COLLATE | |
0d071d52 KW |
285 | |
286 | /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell | |
287 | * core Perl this and that 'newcoll' is the name of the new locale. | |
288 | * | |
289 | * Any code changing the locale (outside this file) should use | |
290 | * POSIX::setlocale, which calls this function. Therefore this function | |
291 | * should be called directly only from this file and from | |
292 | * POSIX::setlocale() */ | |
293 | ||
97aff369 | 294 | dVAR; |
98994639 HS |
295 | |
296 | if (! newcoll) { | |
297 | if (PL_collation_name) { | |
298 | ++PL_collation_ix; | |
299 | Safefree(PL_collation_name); | |
300 | PL_collation_name = NULL; | |
301 | } | |
302 | PL_collation_standard = TRUE; | |
303 | PL_collxfrm_base = 0; | |
304 | PL_collxfrm_mult = 2; | |
305 | return; | |
306 | } | |
307 | ||
308 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { | |
309 | ++PL_collation_ix; | |
310 | Safefree(PL_collation_name); | |
311 | PL_collation_name = stdize_locale(savepv(newcoll)); | |
770526c1 NC |
312 | PL_collation_standard = ((*newcoll == 'C' && newcoll[1] == '\0') |
313 | || strEQ(newcoll, "POSIX")); | |
98994639 HS |
314 | |
315 | { | |
316 | /* 2: at most so many chars ('a', 'b'). */ | |
317 | /* 50: surely no system expands a char more. */ | |
318 | #define XFRMBUFSIZE (2 * 50) | |
319 | char xbuf[XFRMBUFSIZE]; | |
8772537c AL |
320 | const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE); |
321 | const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE); | |
322 | const SSize_t mult = fb - fa; | |
e0601d3c | 323 | if (mult < 1 && !(fa == 0 && fb == 0)) |
ad49ad39 NC |
324 | Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf, |
325 | (UV) fa, (UV) fb); | |
eb160463 | 326 | PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0; |
98994639 HS |
327 | PL_collxfrm_mult = mult; |
328 | } | |
329 | } | |
330 | ||
331 | #endif /* USE_LOCALE_COLLATE */ | |
332 | } | |
333 | ||
b385bb4d KW |
334 | #ifdef WIN32 |
335 | ||
336 | char * | |
337 | Perl_my_setlocale(pTHX_ int category, const char* locale) | |
338 | { | |
339 | /* This, for Windows, emulates POSIX setlocale() behavior. There is no | |
340 | * difference unless the input locale is "", which means on Windows to get | |
341 | * the machine default, which is set via the computer's "Regional and | |
342 | * Language Options" (or its current equivalent). In POSIX, it instead | |
343 | * means to find the locale from the user's environment. This routine | |
344 | * looks in the environment, and, if anything is found, uses that instead | |
345 | * of going to the machine default. If there is no environment override, | |
346 | * the machine default is used, as normal, by calling the real setlocale() | |
347 | * with "". The POSIX behavior is to use the LC_ALL variable if set; | |
348 | * otherwise to use the particular category's variable if set; otherwise to | |
349 | * use the LANG variable. */ | |
350 | ||
481465ea | 351 | bool override_LC_ALL = 0; |
89f7b9aa KW |
352 | char * result; |
353 | ||
b385bb4d KW |
354 | if (locale && strEQ(locale, "")) { |
355 | # ifdef LC_ALL | |
356 | locale = PerlEnv_getenv("LC_ALL"); | |
357 | if (! locale) { | |
358 | #endif | |
359 | switch (category) { | |
360 | # ifdef LC_ALL | |
361 | case LC_ALL: | |
481465ea | 362 | override_LC_ALL = TRUE; |
b385bb4d KW |
363 | break; /* We already know its variable isn't set */ |
364 | # endif | |
365 | # ifdef USE_LOCALE_TIME | |
366 | case LC_TIME: | |
367 | locale = PerlEnv_getenv("LC_TIME"); | |
368 | break; | |
369 | # endif | |
370 | # ifdef USE_LOCALE_CTYPE | |
371 | case LC_CTYPE: | |
372 | locale = PerlEnv_getenv("LC_CTYPE"); | |
373 | break; | |
374 | # endif | |
375 | # ifdef USE_LOCALE_COLLATE | |
376 | case LC_COLLATE: | |
377 | locale = PerlEnv_getenv("LC_COLLATE"); | |
378 | break; | |
379 | # endif | |
380 | # ifdef USE_LOCALE_MONETARY | |
381 | case LC_MONETARY: | |
382 | locale = PerlEnv_getenv("LC_MONETARY"); | |
383 | break; | |
384 | # endif | |
385 | # ifdef USE_LOCALE_NUMERIC | |
386 | case LC_NUMERIC: | |
387 | locale = PerlEnv_getenv("LC_NUMERIC"); | |
388 | break; | |
389 | # endif | |
390 | # ifdef USE_LOCALE_MESSAGES | |
391 | case LC_MESSAGES: | |
392 | locale = PerlEnv_getenv("LC_MESSAGES"); | |
393 | break; | |
394 | # endif | |
395 | default: | |
396 | /* This is a category, like PAPER_SIZE that we don't | |
397 | * know about; and so can't provide a wrapper. */ | |
398 | break; | |
399 | } | |
400 | if (! locale) { | |
401 | locale = PerlEnv_getenv("LANG"); | |
481465ea | 402 | if (! locale) { |
b385bb4d KW |
403 | locale = ""; |
404 | } | |
405 | } | |
406 | # ifdef LC_ALL | |
407 | } | |
408 | # endif | |
409 | } | |
410 | ||
89f7b9aa KW |
411 | result = setlocale(category, locale); |
412 | ||
481465ea | 413 | if (! override_LC_ALL) { |
89f7b9aa KW |
414 | return result; |
415 | } | |
416 | ||
417 | /* Here the input locale was LC_ALL, and we have set it to what is in the | |
481465ea KW |
418 | * LANG variable or the system default if there is no LANG. But these have |
419 | * lower priority than the other LC_foo variables, so override it for each | |
420 | * one that is set. (If they are set to "", it means to use the same thing | |
421 | * we just set LC_ALL to, so can skip) */ | |
89f7b9aa KW |
422 | # ifdef USE_LOCALE_TIME |
423 | result = PerlEnv_getenv("LC_TIME"); | |
730252b2 | 424 | if (result && strNE(result, "")) { |
89f7b9aa KW |
425 | setlocale(LC_TIME, result); |
426 | } | |
427 | # endif | |
428 | # ifdef USE_LOCALE_CTYPE | |
429 | result = PerlEnv_getenv("LC_CTYPE"); | |
730252b2 | 430 | if (result && strNE(result, "")) { |
89f7b9aa KW |
431 | setlocale(LC_CTYPE, result); |
432 | } | |
433 | # endif | |
434 | # ifdef USE_LOCALE_COLLATE | |
435 | result = PerlEnv_getenv("LC_COLLATE"); | |
730252b2 | 436 | if (result && strNE(result, "")) { |
89f7b9aa KW |
437 | setlocale(LC_COLLATE, result); |
438 | } | |
439 | # endif | |
440 | # ifdef USE_LOCALE_MONETARY | |
441 | result = PerlEnv_getenv("LC_MONETARY"); | |
730252b2 | 442 | if (result && strNE(result, "")) { |
89f7b9aa KW |
443 | setlocale(LC_MONETARY, result); |
444 | } | |
445 | # endif | |
446 | # ifdef USE_LOCALE_NUMERIC | |
447 | result = PerlEnv_getenv("LC_NUMERIC"); | |
730252b2 | 448 | if (result && strNE(result, "")) { |
89f7b9aa KW |
449 | setlocale(LC_NUMERIC, result); |
450 | } | |
451 | # endif | |
452 | # ifdef USE_LOCALE_MESSAGES | |
453 | result = PerlEnv_getenv("LC_MESSAGES"); | |
730252b2 | 454 | if (result && strNE(result, "")) { |
89f7b9aa KW |
455 | setlocale(LC_MESSAGES, result); |
456 | } | |
457 | # endif | |
458 | ||
459 | return setlocale(LC_ALL, NULL); | |
460 | ||
b385bb4d KW |
461 | } |
462 | ||
463 | #endif | |
464 | ||
465 | ||
98994639 HS |
466 | /* |
467 | * Initialize locale awareness. | |
468 | */ | |
469 | int | |
470 | Perl_init_i18nl10n(pTHX_ int printwarn) | |
471 | { | |
0e92a118 KW |
472 | /* printwarn is |
473 | * | |
474 | * 0 if not to output warning when setup locale is bad | |
475 | * 1 if to output warning based on value of PERL_BADLANG | |
476 | * >1 if to output regardless of PERL_BADLANG | |
477 | * | |
478 | * returns | |
98994639 | 479 | * 1 = set ok or not applicable, |
0e92a118 KW |
480 | * 0 = fallback to a locale of lower priority |
481 | * -1 = fallback to all locales failed, not even to the C locale | |
98994639 HS |
482 | */ |
483 | ||
0e92a118 KW |
484 | int ok = 1; |
485 | ||
98994639 | 486 | #if defined(USE_LOCALE) |
97aff369 | 487 | dVAR; |
98994639 HS |
488 | |
489 | #ifdef USE_LOCALE_CTYPE | |
490 | char *curctype = NULL; | |
491 | #endif /* USE_LOCALE_CTYPE */ | |
492 | #ifdef USE_LOCALE_COLLATE | |
493 | char *curcoll = NULL; | |
494 | #endif /* USE_LOCALE_COLLATE */ | |
495 | #ifdef USE_LOCALE_NUMERIC | |
496 | char *curnum = NULL; | |
497 | #endif /* USE_LOCALE_NUMERIC */ | |
498 | #ifdef __GLIBC__ | |
7452cf6a | 499 | char * const language = PerlEnv_getenv("LANGUAGE"); |
98994639 | 500 | #endif |
65ebb059 | 501 | |
ccd65d51 KW |
502 | /* NULL uses the existing already set up locale */ |
503 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) | |
504 | ? NULL | |
505 | : ""; | |
c3fcd832 KW |
506 | const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */ |
507 | unsigned int trial_locales_count; | |
7452cf6a AL |
508 | char * const lc_all = PerlEnv_getenv("LC_ALL"); |
509 | char * const lang = PerlEnv_getenv("LANG"); | |
98994639 | 510 | bool setlocale_failure = FALSE; |
65ebb059 KW |
511 | unsigned int i; |
512 | char *p; | |
513 | const bool locwarn = (printwarn > 1 || | |
514 | (printwarn && | |
515 | (!(p = PerlEnv_getenv("PERL_BADLANG")) || atoi(p)))); | |
0e92a118 | 516 | bool done = FALSE; |
65ebb059 | 517 | const char *system_default_locale = NULL; |
0e92a118 | 518 | |
98994639 | 519 | |
0e92a118 KW |
520 | #ifndef LOCALE_ENVIRON_REQUIRED |
521 | PERL_UNUSED_VAR(done); | |
522 | #else | |
98994639 HS |
523 | |
524 | /* | |
525 | * Ultrix setlocale(..., "") fails if there are no environment | |
526 | * variables from which to get a locale name. | |
527 | */ | |
528 | ||
b3e384bf | 529 | # ifdef LC_ALL |
98994639 | 530 | if (lang) { |
b385bb4d | 531 | if (my_setlocale(LC_ALL, setlocale_init)) |
98994639 HS |
532 | done = TRUE; |
533 | else | |
534 | setlocale_failure = TRUE; | |
535 | } | |
536 | if (!setlocale_failure) { | |
b3e384bf | 537 | # ifdef USE_LOCALE_CTYPE |
56279a21 | 538 | Safefree(curctype); |
98994639 | 539 | if (! (curctype = |
b385bb4d | 540 | my_setlocale(LC_CTYPE, |
98994639 | 541 | (!done && (lang || PerlEnv_getenv("LC_CTYPE"))) |
ccd65d51 | 542 | ? setlocale_init : NULL))) |
98994639 HS |
543 | setlocale_failure = TRUE; |
544 | else | |
545 | curctype = savepv(curctype); | |
b3e384bf KW |
546 | # endif /* USE_LOCALE_CTYPE */ |
547 | # ifdef USE_LOCALE_COLLATE | |
56279a21 | 548 | Safefree(curcoll); |
98994639 | 549 | if (! (curcoll = |
b385bb4d | 550 | my_setlocale(LC_COLLATE, |
98994639 | 551 | (!done && (lang || PerlEnv_getenv("LC_COLLATE"))) |
ccd65d51 | 552 | ? setlocale_init : NULL))) |
98994639 HS |
553 | setlocale_failure = TRUE; |
554 | else | |
555 | curcoll = savepv(curcoll); | |
b3e384bf KW |
556 | # endif /* USE_LOCALE_COLLATE */ |
557 | # ifdef USE_LOCALE_NUMERIC | |
56279a21 | 558 | Safefree(curnum); |
98994639 | 559 | if (! (curnum = |
b385bb4d | 560 | my_setlocale(LC_NUMERIC, |
98994639 | 561 | (!done && (lang || PerlEnv_getenv("LC_NUMERIC"))) |
ccd65d51 | 562 | ? setlocale_init : NULL))) |
98994639 HS |
563 | setlocale_failure = TRUE; |
564 | else | |
565 | curnum = savepv(curnum); | |
b3e384bf | 566 | # endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
567 | # ifdef USE_LOCALE_MESSAGES |
568 | if (! my_setlocale(LC_MESSAGES, | |
569 | (!done && (lang || PerlEnv_getenv("LC_MESSAGES"))) | |
570 | ? setlocale_init : NULL)) | |
571 | { | |
572 | setlocale_failure = TRUE; | |
573 | } | |
574 | # endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
575 | # ifdef USE_LOCALE_MONETARY |
576 | if (! my_setlocale(LC_MONETARY, | |
577 | (!done && (lang || PerlEnv_getenv("LC_MONETARY"))) | |
578 | ? setlocale_init : NULL)) | |
579 | { | |
580 | setlocale_failure = TRUE; | |
581 | } | |
582 | # endif /* USE_LOCALE_MONETARY */ | |
98994639 HS |
583 | } |
584 | ||
b3e384bf | 585 | # endif /* LC_ALL */ |
98994639 HS |
586 | |
587 | #endif /* !LOCALE_ENVIRON_REQUIRED */ | |
588 | ||
65ebb059 KW |
589 | /* We try each locale in the list until we get one that works, or exhaust |
590 | * the list */ | |
c3fcd832 KW |
591 | trial_locales[0] = setlocale_init; |
592 | trial_locales_count = 1; | |
65ebb059 KW |
593 | for (i= 0; i < trial_locales_count; i++) { |
594 | const char * trial_locale = trial_locales[i]; | |
595 | ||
596 | if (i > 0) { | |
597 | ||
598 | /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED | |
599 | * when i==0, but I (khw) don't think that behavior makes much | |
600 | * sense */ | |
601 | setlocale_failure = FALSE; | |
602 | ||
603 | #ifdef WIN32 | |
604 | ||
605 | /* On Windows machines, an entry of "" after the 0th means to use | |
606 | * the system default locale, which we now proceed to get. */ | |
607 | if (strEQ(trial_locale, "")) { | |
608 | unsigned int j; | |
609 | ||
610 | /* Note that this may change the locale, but we are going to do | |
611 | * that anyway just below */ | |
612 | system_default_locale = setlocale(LC_ALL, ""); | |
613 | ||
614 | /* Skip if invalid or it's already on the list of locales to | |
615 | * try */ | |
616 | if (! system_default_locale) { | |
617 | goto next_iteration; | |
618 | } | |
619 | for (j = 0; j < trial_locales_count; j++) { | |
620 | if (strEQ(system_default_locale, trial_locales[j])) { | |
621 | goto next_iteration; | |
622 | } | |
623 | } | |
624 | ||
625 | trial_locale = system_default_locale; | |
626 | } | |
627 | #endif | |
628 | } | |
629 | ||
98994639 | 630 | #ifdef LC_ALL |
7cd8b568 | 631 | if (! my_setlocale(LC_ALL, trial_locale)) { |
49c85077 | 632 | setlocale_failure = TRUE; |
7cd8b568 KW |
633 | } |
634 | else { | |
635 | /* Since LC_ALL succeeded, it should have changed all the other | |
636 | * categories it can to its value; so we massage things so that the | |
637 | * setlocales below just return their category's current values. | |
638 | * This adequately handles the case in NetBSD where LC_COLLATE may | |
639 | * not be defined for a locale, and setting it individually will | |
640 | * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to | |
641 | * the POSIX locale. */ | |
642 | trial_locale = NULL; | |
643 | } | |
98994639 HS |
644 | #endif /* LC_ALL */ |
645 | ||
49c85077 | 646 | if (!setlocale_failure) { |
98994639 | 647 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
648 | Safefree(curctype); |
649 | if (! (curctype = my_setlocale(LC_CTYPE, trial_locale))) | |
650 | setlocale_failure = TRUE; | |
651 | else | |
652 | curctype = savepv(curctype); | |
98994639 HS |
653 | #endif /* USE_LOCALE_CTYPE */ |
654 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
655 | Safefree(curcoll); |
656 | if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale))) | |
657 | setlocale_failure = TRUE; | |
658 | else | |
659 | curcoll = savepv(curcoll); | |
98994639 HS |
660 | #endif /* USE_LOCALE_COLLATE */ |
661 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
662 | Safefree(curnum); |
663 | if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale))) | |
664 | setlocale_failure = TRUE; | |
665 | else | |
666 | curnum = savepv(curnum); | |
98994639 | 667 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
668 | #ifdef USE_LOCALE_MESSAGES |
669 | if (! (my_setlocale(LC_MESSAGES, trial_locale))) | |
670 | setlocale_failure = TRUE; | |
671 | #endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
672 | #ifdef USE_LOCALE_MONETARY |
673 | if (! (my_setlocale(LC_MONETARY, trial_locale))) | |
674 | setlocale_failure = TRUE; | |
675 | #endif /* USE_LOCALE_MONETARY */ | |
676 | ||
49c85077 KW |
677 | if (! setlocale_failure) { /* Success */ |
678 | break; | |
679 | } | |
65ebb059 | 680 | } |
98994639 | 681 | |
49c85077 KW |
682 | /* Here, something failed; will need to try a fallback. */ |
683 | ok = 0; | |
65ebb059 | 684 | |
49c85077 KW |
685 | if (i == 0) { |
686 | unsigned int j; | |
98994639 | 687 | |
65ebb059 | 688 | if (locwarn) { /* Output failure info only on the first one */ |
98994639 HS |
689 | #ifdef LC_ALL |
690 | ||
49c85077 KW |
691 | PerlIO_printf(Perl_error_log, |
692 | "perl: warning: Setting locale failed.\n"); | |
98994639 HS |
693 | |
694 | #else /* !LC_ALL */ | |
695 | ||
49c85077 KW |
696 | PerlIO_printf(Perl_error_log, |
697 | "perl: warning: Setting locale failed for the categories:\n\t"); | |
98994639 | 698 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
699 | if (! curctype) |
700 | PerlIO_printf(Perl_error_log, "LC_CTYPE "); | |
98994639 HS |
701 | #endif /* USE_LOCALE_CTYPE */ |
702 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
703 | if (! curcoll) |
704 | PerlIO_printf(Perl_error_log, "LC_COLLATE "); | |
98994639 HS |
705 | #endif /* USE_LOCALE_COLLATE */ |
706 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
707 | if (! curnum) |
708 | PerlIO_printf(Perl_error_log, "LC_NUMERIC "); | |
98994639 | 709 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d | 710 | PerlIO_printf(Perl_error_log, "and possibly others\n"); |
98994639 HS |
711 | |
712 | #endif /* LC_ALL */ | |
713 | ||
49c85077 KW |
714 | PerlIO_printf(Perl_error_log, |
715 | "perl: warning: Please check that your locale settings:\n"); | |
98994639 HS |
716 | |
717 | #ifdef __GLIBC__ | |
49c85077 KW |
718 | PerlIO_printf(Perl_error_log, |
719 | "\tLANGUAGE = %c%s%c,\n", | |
720 | language ? '"' : '(', | |
721 | language ? language : "unset", | |
722 | language ? '"' : ')'); | |
98994639 HS |
723 | #endif |
724 | ||
49c85077 KW |
725 | PerlIO_printf(Perl_error_log, |
726 | "\tLC_ALL = %c%s%c,\n", | |
727 | lc_all ? '"' : '(', | |
728 | lc_all ? lc_all : "unset", | |
729 | lc_all ? '"' : ')'); | |
98994639 HS |
730 | |
731 | #if defined(USE_ENVIRON_ARRAY) | |
49c85077 KW |
732 | { |
733 | char **e; | |
734 | for (e = environ; *e; e++) { | |
735 | if (strnEQ(*e, "LC_", 3) | |
736 | && strnNE(*e, "LC_ALL=", 7) | |
737 | && (p = strchr(*e, '='))) | |
738 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", | |
739 | (int)(p - *e), *e, p + 1); | |
740 | } | |
741 | } | |
98994639 | 742 | #else |
49c85077 KW |
743 | PerlIO_printf(Perl_error_log, |
744 | "\t(possibly more locale environment variables)\n"); | |
98994639 HS |
745 | #endif |
746 | ||
49c85077 KW |
747 | PerlIO_printf(Perl_error_log, |
748 | "\tLANG = %c%s%c\n", | |
749 | lang ? '"' : '(', | |
750 | lang ? lang : "unset", | |
751 | lang ? '"' : ')'); | |
98994639 | 752 | |
49c85077 KW |
753 | PerlIO_printf(Perl_error_log, |
754 | " are supported and installed on your system.\n"); | |
755 | } | |
98994639 | 756 | |
65ebb059 KW |
757 | /* Calculate what fallback locales to try. We have avoided this |
758 | * until we have to, becuase failure is quite unlikely. This will | |
759 | * usually change the upper bound of the loop we are in. | |
760 | * | |
761 | * Since the system's default way of setting the locale has not | |
762 | * found one that works, We use Perl's defined ordering: LC_ALL, | |
763 | * LANG, and the C locale. We don't try the same locale twice, so | |
764 | * don't add to the list if already there. (On POSIX systems, the | |
765 | * LC_ALL element will likely be a repeat of the 0th element "", | |
766 | * but there's no harm done by doing it explicitly */ | |
767 | if (lc_all) { | |
768 | for (j = 0; j < trial_locales_count; j++) { | |
769 | if (strEQ(lc_all, trial_locales[j])) { | |
770 | goto done_lc_all; | |
771 | } | |
772 | } | |
773 | trial_locales[trial_locales_count++] = lc_all; | |
774 | } | |
775 | done_lc_all: | |
98994639 | 776 | |
65ebb059 KW |
777 | if (lang) { |
778 | for (j = 0; j < trial_locales_count; j++) { | |
779 | if (strEQ(lang, trial_locales[j])) { | |
780 | goto done_lang; | |
781 | } | |
782 | } | |
783 | trial_locales[trial_locales_count++] = lang; | |
784 | } | |
785 | done_lang: | |
786 | ||
787 | #if defined(WIN32) && defined(LC_ALL) | |
788 | /* For Windows, we also try the system default locale before "C". | |
789 | * (If there exists a Windows without LC_ALL we skip this because | |
790 | * it gets too complicated. For those, the "C" is the next | |
791 | * fallback possibility). The "" is the same as the 0th element of | |
792 | * the array, but the code at the loop above knows to treat it | |
793 | * differently when not the 0th */ | |
794 | trial_locales[trial_locales_count++] = ""; | |
795 | #endif | |
796 | ||
797 | for (j = 0; j < trial_locales_count; j++) { | |
798 | if (strEQ("C", trial_locales[j])) { | |
799 | goto done_C; | |
800 | } | |
801 | } | |
802 | trial_locales[trial_locales_count++] = "C"; | |
98994639 | 803 | |
65ebb059 KW |
804 | done_C: ; |
805 | } /* end of first time through the loop */ | |
98994639 | 806 | |
65ebb059 KW |
807 | #ifdef WIN32 |
808 | next_iteration: ; | |
809 | #endif | |
810 | ||
811 | } /* end of looping through the trial locales */ | |
812 | ||
813 | if (ok < 1) { /* If we tried to fallback */ | |
814 | const char* msg; | |
815 | if (! setlocale_failure) { /* fallback succeeded */ | |
816 | msg = "Falling back to"; | |
817 | } | |
818 | else { /* fallback failed */ | |
98994639 | 819 | |
65ebb059 KW |
820 | /* We dropped off the end of the loop, so have to decrement i to |
821 | * get back to the value the last time through */ | |
822 | i--; | |
98994639 | 823 | |
65ebb059 KW |
824 | ok = -1; |
825 | msg = "Failed to fall back to"; | |
826 | ||
827 | /* To continue, we should use whatever values we've got */ | |
98994639 | 828 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
829 | Safefree(curctype); |
830 | curctype = savepv(setlocale(LC_CTYPE, NULL)); | |
98994639 HS |
831 | #endif /* USE_LOCALE_CTYPE */ |
832 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
833 | Safefree(curcoll); |
834 | curcoll = savepv(setlocale(LC_COLLATE, NULL)); | |
98994639 HS |
835 | #endif /* USE_LOCALE_COLLATE */ |
836 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
837 | Safefree(curnum); |
838 | curnum = savepv(setlocale(LC_NUMERIC, NULL)); | |
98994639 | 839 | #endif /* USE_LOCALE_NUMERIC */ |
65ebb059 KW |
840 | } |
841 | ||
842 | if (locwarn) { | |
843 | const char * description; | |
844 | const char * name = ""; | |
845 | if (strEQ(trial_locales[i], "C")) { | |
846 | description = "the standard locale"; | |
847 | name = "C"; | |
848 | } | |
849 | else if (strEQ(trial_locales[i], "")) { | |
850 | description = "the system default locale"; | |
851 | if (system_default_locale) { | |
852 | name = system_default_locale; | |
853 | } | |
854 | } | |
855 | else { | |
856 | description = "a fallback locale"; | |
857 | name = trial_locales[i]; | |
858 | } | |
859 | if (name && strNE(name, "")) { | |
860 | PerlIO_printf(Perl_error_log, | |
861 | "perl: warning: %s %s (\"%s\").\n", msg, description, name); | |
862 | } | |
863 | else { | |
864 | PerlIO_printf(Perl_error_log, | |
865 | "perl: warning: %s %s.\n", msg, description); | |
866 | } | |
867 | } | |
868 | } /* End of tried to fallback */ | |
98994639 HS |
869 | |
870 | #ifdef USE_LOCALE_CTYPE | |
871 | new_ctype(curctype); | |
872 | #endif /* USE_LOCALE_CTYPE */ | |
873 | ||
874 | #ifdef USE_LOCALE_COLLATE | |
875 | new_collate(curcoll); | |
876 | #endif /* USE_LOCALE_COLLATE */ | |
877 | ||
878 | #ifdef USE_LOCALE_NUMERIC | |
879 | new_numeric(curnum); | |
880 | #endif /* USE_LOCALE_NUMERIC */ | |
b310b053 | 881 | |
8ef6e574 | 882 | #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE) |
49c85077 KW |
883 | /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE |
884 | * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by | |
885 | * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the | |
886 | * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open | |
887 | * discipline. */ | |
888 | PL_utf8locale = is_cur_LC_category_utf8(LC_CTYPE); | |
889 | ||
a05d7ebb | 890 | /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO. |
fde18df1 JH |
891 | This is an alternative to using the -C command line switch |
892 | (the -C if present will override this). */ | |
893 | { | |
dd374669 | 894 | const char *p = PerlEnv_getenv("PERL_UNICODE"); |
a05d7ebb | 895 | PL_unicode = p ? parse_unicode_opts(&p) : 0; |
5a22a2bb NC |
896 | if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG) |
897 | PL_utf8cache = -1; | |
b310b053 | 898 | } |
ec71e770 | 899 | #endif |
b310b053 | 900 | |
98994639 | 901 | #ifdef USE_LOCALE_CTYPE |
43c5f42d | 902 | Safefree(curctype); |
98994639 HS |
903 | #endif /* USE_LOCALE_CTYPE */ |
904 | #ifdef USE_LOCALE_COLLATE | |
43c5f42d | 905 | Safefree(curcoll); |
98994639 HS |
906 | #endif /* USE_LOCALE_COLLATE */ |
907 | #ifdef USE_LOCALE_NUMERIC | |
43c5f42d | 908 | Safefree(curnum); |
98994639 | 909 | #endif /* USE_LOCALE_NUMERIC */ |
8ef6e574 KW |
910 | |
911 | #endif /* USE_LOCALE */ | |
912 | ||
98994639 HS |
913 | return ok; |
914 | } | |
915 | ||
49c85077 | 916 | |
98994639 HS |
917 | #ifdef USE_LOCALE_COLLATE |
918 | ||
919 | /* | |
920 | * mem_collxfrm() is a bit like strxfrm() but with two important | |
921 | * differences. First, it handles embedded NULs. Second, it allocates | |
922 | * a bit more memory than needed for the transformed data itself. | |
923 | * The real transformed data begins at offset sizeof(collationix). | |
924 | * Please see sv_collxfrm() to see how this is used. | |
925 | */ | |
166f8a29 | 926 | |
98994639 HS |
927 | char * |
928 | Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen) | |
929 | { | |
97aff369 | 930 | dVAR; |
98994639 HS |
931 | char *xbuf; |
932 | STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */ | |
933 | ||
7918f24d NC |
934 | PERL_ARGS_ASSERT_MEM_COLLXFRM; |
935 | ||
98994639 HS |
936 | /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */ |
937 | /* the +1 is for the terminating NUL. */ | |
938 | ||
939 | xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1; | |
a02a5408 | 940 | Newx(xbuf, xAlloc, char); |
98994639 HS |
941 | if (! xbuf) |
942 | goto bad; | |
943 | ||
944 | *(U32*)xbuf = PL_collation_ix; | |
945 | xout = sizeof(PL_collation_ix); | |
946 | for (xin = 0; xin < len; ) { | |
224e8ef5 | 947 | Size_t xused; |
98994639 HS |
948 | |
949 | for (;;) { | |
950 | xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout); | |
224e8ef5 | 951 | if (xused >= PERL_INT_MAX) |
98994639 | 952 | goto bad; |
eb160463 | 953 | if ((STRLEN)xused < xAlloc - xout) |
98994639 HS |
954 | break; |
955 | xAlloc = (2 * xAlloc) + 1; | |
956 | Renew(xbuf, xAlloc, char); | |
957 | if (! xbuf) | |
958 | goto bad; | |
959 | } | |
960 | ||
961 | xin += strlen(s + xin) + 1; | |
962 | xout += xused; | |
963 | ||
964 | /* Embedded NULs are understood but silently skipped | |
965 | * because they make no sense in locale collation. */ | |
966 | } | |
967 | ||
968 | xbuf[xout] = '\0'; | |
969 | *xlen = xout - sizeof(PL_collation_ix); | |
970 | return xbuf; | |
971 | ||
972 | bad: | |
973 | Safefree(xbuf); | |
974 | *xlen = 0; | |
975 | return NULL; | |
976 | } | |
977 | ||
978 | #endif /* USE_LOCALE_COLLATE */ | |
979 | ||
8ef6e574 KW |
980 | #ifdef USE_LOCALE |
981 | ||
637917bb | 982 | STATIC bool |
7d74bb61 KW |
983 | S_is_cur_LC_category_utf8(pTHX_ int category) |
984 | { | |
985 | /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE | |
986 | * otherwise. 'category' may not be LC_ALL. If the platform doesn't have | |
119ee68b KW |
987 | * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence |
988 | * could give the wrong result. It errs on the side of not being a UTF-8 | |
989 | * locale. */ | |
7d74bb61 KW |
990 | |
991 | char *save_input_locale = NULL; | |
7d74bb61 KW |
992 | STRLEN final_pos; |
993 | ||
8ef6e574 | 994 | #ifdef LC_ALL |
7d74bb61 | 995 | assert(category != LC_ALL); |
8ef6e574 | 996 | #endif |
7d74bb61 KW |
997 | |
998 | /* First dispose of the trivial cases */ | |
b07fffd1 | 999 | save_input_locale = setlocale(category, NULL); |
7d74bb61 | 1000 | if (! save_input_locale) { |
69014004 KW |
1001 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1002 | "Could not find current locale for category %d\n", | |
1003 | category)); | |
7d74bb61 KW |
1004 | return FALSE; /* XXX maybe should croak */ |
1005 | } | |
b07fffd1 | 1006 | save_input_locale = stdize_locale(savepv(save_input_locale)); |
7d74bb61 KW |
1007 | if ((*save_input_locale == 'C' && save_input_locale[1] == '\0') |
1008 | || strEQ(save_input_locale, "POSIX")) | |
1009 | { | |
69014004 KW |
1010 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1011 | "Current locale for category %d is %s\n", | |
1012 | category, save_input_locale)); | |
b07fffd1 | 1013 | Safefree(save_input_locale); |
7d74bb61 KW |
1014 | return FALSE; |
1015 | } | |
1016 | ||
1d958db2 KW |
1017 | #if defined(USE_LOCALE_CTYPE) \ |
1018 | && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET))) | |
7d74bb61 | 1019 | |
1d958db2 | 1020 | { /* Next try nl_langinfo or MB_CUR_MAX if available */ |
7d74bb61 KW |
1021 | |
1022 | char *save_ctype_locale = NULL; | |
119ee68b | 1023 | bool is_utf8; |
7d74bb61 | 1024 | |
119ee68b | 1025 | if (category != LC_CTYPE) { /* These work only on LC_CTYPE */ |
7d74bb61 KW |
1026 | |
1027 | /* Get the current LC_CTYPE locale */ | |
1028 | save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL))); | |
1029 | if (! save_ctype_locale) { | |
69014004 KW |
1030 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1031 | "Could not find current locale for LC_CTYPE\n")); | |
7d74bb61 KW |
1032 | goto cant_use_nllanginfo; |
1033 | } | |
1034 | ||
1035 | /* If LC_CTYPE and the desired category use the same locale, this | |
1036 | * means that finding the value for LC_CTYPE is the same as finding | |
1037 | * the value for the desired category. Otherwise, switch LC_CTYPE | |
1038 | * to the desired category's locale */ | |
1039 | if (strEQ(save_ctype_locale, save_input_locale)) { | |
1040 | Safefree(save_ctype_locale); | |
1041 | save_ctype_locale = NULL; | |
1042 | } | |
1043 | else if (! setlocale(LC_CTYPE, save_input_locale)) { | |
69014004 KW |
1044 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1045 | "Could not change LC_CTYPE locale to %s\n", | |
1046 | save_input_locale)); | |
7d74bb61 KW |
1047 | Safefree(save_ctype_locale); |
1048 | goto cant_use_nllanginfo; | |
1049 | } | |
1050 | } | |
1051 | ||
69014004 KW |
1052 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n", |
1053 | save_input_locale)); | |
1054 | ||
7d74bb61 | 1055 | /* Here the current LC_CTYPE is set to the locale of the category whose |
1d958db2 KW |
1056 | * information is desired. This means that nl_langinfo() and MB_CUR_MAX |
1057 | * should give the correct results */ | |
119ee68b | 1058 | |
1d958db2 KW |
1059 | # if defined(HAS_NL_LANGINFO) && defined(CODESET) |
1060 | { | |
1061 | char *codeset = savepv(nl_langinfo(CODESET)); | |
1062 | if (codeset && strNE(codeset, "")) { | |
119ee68b | 1063 | |
1d958db2 KW |
1064 | /* If we switched LC_CTYPE, switch back */ |
1065 | if (save_ctype_locale) { | |
1066 | setlocale(LC_CTYPE, save_ctype_locale); | |
1067 | Safefree(save_ctype_locale); | |
1068 | } | |
1069 | ||
1070 | is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8")) | |
1071 | || foldEQ(codeset, STR_WITH_LEN("UTF8")); | |
1072 | ||
69014004 KW |
1073 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1074 | "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n", | |
1075 | codeset, is_utf8)); | |
1d958db2 KW |
1076 | Safefree(codeset); |
1077 | Safefree(save_input_locale); | |
1078 | return is_utf8; | |
1079 | } | |
119ee68b KW |
1080 | } |
1081 | ||
1d958db2 KW |
1082 | # endif |
1083 | # ifdef MB_CUR_MAX | |
1084 | ||
1085 | /* Here, either we don't have nl_langinfo, or it didn't return a | |
1086 | * codeset. Try MB_CUR_MAX */ | |
1087 | ||
119ee68b KW |
1088 | /* Standard UTF-8 needs at least 4 bytes to represent the maximum |
1089 | * Unicode code point. Since UTF-8 is the only non-single byte | |
1090 | * encoding we handle, we just say any such encoding is UTF-8, and if | |
1091 | * turns out to be wrong, other things will fail */ | |
1092 | is_utf8 = MB_CUR_MAX >= 4; | |
1093 | ||
69014004 KW |
1094 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1095 | "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n", | |
1096 | (int) MB_CUR_MAX, is_utf8)); | |
1097 | ||
119ee68b KW |
1098 | Safefree(save_input_locale); |
1099 | ||
1100 | # ifdef HAS_MBTOWC | |
1101 | ||
1102 | /* ... But, most system that have MB_CUR_MAX will also have mbtowc(), | |
1103 | * since they are both in the C99 standard. We can feed a known byte | |
1104 | * string to the latter function, and check that it gives the expected | |
1105 | * result */ | |
1106 | if (is_utf8) { | |
1107 | wchar_t wc; | |
a6715020 | 1108 | GCC_DIAG_IGNORE(-Wunused-result); |
1d958db2 | 1109 | (void) mbtowc(&wc, NULL, 0); /* Reset any shift state */ |
a6715020 | 1110 | GCC_DIAG_RESTORE; |
69014004 | 1111 | errno = 0; |
119ee68b KW |
1112 | if (mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)) |
1113 | != strlen(HYPHEN_UTF8) | |
1114 | || wc != (wchar_t) 0x2010) | |
1115 | { | |
1116 | is_utf8 = FALSE; | |
69014004 KW |
1117 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", wc)); |
1118 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1119 | "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n", | |
1120 | mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno)); | |
119ee68b KW |
1121 | } |
1122 | } | |
119ee68b KW |
1123 | # endif |
1124 | ||
1d958db2 KW |
1125 | /* If we switched LC_CTYPE, switch back */ |
1126 | if (save_ctype_locale) { | |
1127 | setlocale(LC_CTYPE, save_ctype_locale); | |
1128 | Safefree(save_ctype_locale); | |
119ee68b | 1129 | } |
7d74bb61 | 1130 | |
1d958db2 | 1131 | return is_utf8; |
119ee68b | 1132 | # endif |
7d74bb61 | 1133 | } |
119ee68b | 1134 | |
7d74bb61 KW |
1135 | cant_use_nllanginfo: |
1136 | ||
1137 | #endif /* HAS_NL_LANGINFO etc */ | |
1138 | ||
1139 | /* nl_langinfo not available or failed somehow. Look at the locale name to | |
751e9426 | 1140 | * see if it matches qr/UTF -? 8 /ix */ |
7d74bb61 KW |
1141 | |
1142 | final_pos = strlen(save_input_locale) - 1; | |
751e9426 KW |
1143 | if (final_pos >= 3) { |
1144 | char *name = save_input_locale; | |
1145 | ||
1146 | /* Find next 'U' or 'u' and look from there */ | |
1147 | while ((name += strcspn(name, "Uu") + 1) | |
1148 | <= save_input_locale + final_pos - 2) | |
7d74bb61 | 1149 | { |
751e9426 KW |
1150 | if (toFOLD(*(name)) != 't' |
1151 | || toFOLD(*(name + 1)) != 'f') | |
1152 | { | |
1153 | continue; | |
1154 | } | |
1155 | name += 2; | |
1156 | if (*(name) == '-') { | |
1157 | if ((name > save_input_locale + final_pos - 1)) { | |
1158 | break; | |
1159 | } | |
1160 | name++; | |
1161 | } | |
1162 | if (*(name) == '8') { | |
1163 | Safefree(save_input_locale); | |
69014004 KW |
1164 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1165 | "Locale %s ends with UTF-8 in name\n", | |
1166 | save_input_locale)); | |
751e9426 KW |
1167 | return TRUE; |
1168 | } | |
7d74bb61 | 1169 | } |
69014004 KW |
1170 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1171 | "Locale %s doesn't end with UTF-8 in name\n", | |
1172 | save_input_locale)); | |
7d74bb61 KW |
1173 | } |
1174 | ||
1175 | #ifdef WIN32 | |
1176 | /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */ | |
1177 | if (final_pos >= 4 | |
1178 | && *(save_input_locale + final_pos - 0) == '1' | |
1179 | && *(save_input_locale + final_pos - 1) == '0' | |
1180 | && *(save_input_locale + final_pos - 2) == '0' | |
1181 | && *(save_input_locale + final_pos - 3) == '5' | |
1182 | && *(save_input_locale + final_pos - 4) == '6') | |
1183 | { | |
1184 | Safefree(save_input_locale); | |
69014004 KW |
1185 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1186 | "Locale %s ends with 10056 in name, is UTF-8 locale\n", | |
1187 | save_input_locale)); | |
7d74bb61 KW |
1188 | return TRUE; |
1189 | } | |
1190 | #endif | |
1191 | ||
fa9b773e KW |
1192 | /* Other common encodings are the ISO 8859 series, which aren't UTF-8 */ |
1193 | if (instr(save_input_locale, "8859")) { | |
1194 | Safefree(save_input_locale); | |
69014004 KW |
1195 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1196 | "Locale %s has 8859 in name, not UTF-8 locale\n", | |
1197 | save_input_locale)); | |
fa9b773e KW |
1198 | return FALSE; |
1199 | } | |
1200 | ||
1201 | #ifdef HAS_LOCALECONV | |
1202 | ||
1203 | # ifdef USE_LOCALE_MONETARY | |
1204 | ||
1205 | /* Here, there is nothing in the locale name to indicate whether the locale | |
1206 | * is UTF-8 or not. This "name", the return of setlocale(), is actually | |
1207 | * defined to be opaque, so we can't really rely on the absence of various | |
1208 | * substrings in the name to indicate its UTF-8ness. Look at the locale's | |
1209 | * currency symbol. Often that will be in the native script, and if the | |
1210 | * symbol isn't in UTF-8, we know that the locale isn't. If it is | |
1211 | * non-ASCII UTF-8, we infer that the locale is too. | |
1212 | * To do this, like above for LC_CTYPE, we first set LC_MONETARY to the | |
1213 | * locale of the desired category, if it isn't that locale already */ | |
1214 | ||
1215 | { | |
1216 | char *save_monetary_locale = NULL; | |
1217 | bool illegal_utf8 = FALSE; | |
1218 | bool only_ascii = FALSE; | |
1219 | const struct lconv* const lc = localeconv(); | |
1220 | ||
1221 | if (category != LC_MONETARY) { | |
1222 | ||
1223 | save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY, | |
1224 | NULL))); | |
1225 | if (! save_monetary_locale) { | |
69014004 KW |
1226 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1227 | "Could not find current locale for LC_MONETARY\n")); | |
fa9b773e KW |
1228 | goto cant_use_monetary; |
1229 | } | |
1230 | ||
1231 | if (strNE(save_monetary_locale, save_input_locale)) { | |
1232 | if (! setlocale(LC_MONETARY, save_input_locale)) { | |
69014004 KW |
1233 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1234 | "Could not change LC_MONETARY locale to %s\n", | |
1235 | save_input_locale)); | |
fa9b773e KW |
1236 | Safefree(save_monetary_locale); |
1237 | goto cant_use_monetary; | |
1238 | } | |
1239 | } | |
1240 | } | |
1241 | ||
1242 | /* Here the current LC_MONETARY is set to the locale of the category | |
1243 | * whose information is desired. */ | |
1244 | ||
1245 | if (lc && lc->currency_symbol) { | |
1246 | if (! is_utf8_string((U8 *) lc->currency_symbol, 0)) { | |
69014004 KW |
1247 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1248 | "Currency symbol for %s is not legal UTF-8\n", | |
1249 | save_input_locale)); | |
fa9b773e KW |
1250 | illegal_utf8 = TRUE; |
1251 | } | |
1252 | else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) { | |
69014004 | 1253 | 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 |
1254 | only_ascii = TRUE; |
1255 | } | |
1256 | } | |
1257 | ||
1258 | /* If we changed it, restore LC_MONETARY to its original locale */ | |
1259 | if (save_monetary_locale) { | |
1260 | setlocale(LC_MONETARY, save_monetary_locale); | |
1261 | Safefree(save_monetary_locale); | |
1262 | } | |
1263 | ||
1264 | Safefree(save_input_locale); | |
1265 | ||
1266 | /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; otherwise | |
1267 | * assume the locale is UTF-8 if and only if the symbol is non-ascii | |
1268 | * UTF-8. (We can't really tell if the locale is UTF-8 or not if the | |
1269 | * symbol is just a '$', so we err on the side of it not being UTF-8) | |
1270 | * */ | |
69014004 KW |
1271 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\tis_utf8=%d\n", (illegal_utf8) |
1272 | ? FALSE | |
1273 | : ! only_ascii)); | |
fa9b773e KW |
1274 | return (illegal_utf8) |
1275 | ? FALSE | |
1276 | : ! only_ascii; | |
1277 | ||
1278 | } | |
1279 | cant_use_monetary: | |
1280 | ||
1281 | # endif /* USE_LOCALE_MONETARY */ | |
1282 | #endif /* HAS_LOCALECONV */ | |
1283 | ||
1284 | #if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES) | |
1285 | ||
1286 | /* This code is ifdefd out because it was found to not be necessary in testing | |
1287 | * on our dromedary test machine, which has over 700 locales. There, looking | |
1288 | * at just the currency symbol gave essentially the same results as doing this | |
1289 | * extra work. Executing this also caused segfaults in miniperl. I left it in | |
1290 | * so as to avoid rewriting it if real-world experience indicates that | |
1291 | * dromedary is an outlier. Essentially, instead of returning abpve if we | |
1292 | * haven't found illegal utf8, we continue on and examine all the strerror() | |
1293 | * messages on the platform for utf8ness. If all are ASCII, we still don't | |
1294 | * know the answer; but otherwise we have a pretty good indication of the | |
1295 | * utf8ness. The reason this doesn't necessarily help much is that the | |
1296 | * messages may not have been translated into the locale. The currency symbol | |
1297 | * is much more likely to have been translated. The code below would need to | |
1298 | * be altered somewhat to just be a continuation of testing the currency | |
1299 | * symbol. */ | |
1300 | int e; | |
1301 | unsigned int failures = 0, non_ascii = 0; | |
1302 | char *save_messages_locale = NULL; | |
1303 | ||
1304 | /* Like above for LC_CTYPE, we set LC_MESSAGES to the locale of the | |
1305 | * desired category, if it isn't that locale already */ | |
1306 | ||
1307 | if (category != LC_MESSAGES) { | |
1308 | ||
1309 | save_messages_locale = stdize_locale(savepv(setlocale(LC_MESSAGES, | |
1310 | NULL))); | |
1311 | if (! save_messages_locale) { | |
1312 | goto cant_use_messages; | |
1313 | } | |
1314 | ||
1315 | if (strEQ(save_messages_locale, save_input_locale)) { | |
1316 | Safefree(save_input_locale); | |
1317 | } | |
1318 | else if (! setlocale(LC_MESSAGES, save_input_locale)) { | |
1319 | Safefree(save_messages_locale); | |
1320 | goto cant_use_messages; | |
1321 | } | |
1322 | } | |
1323 | ||
1324 | /* Here the current LC_MESSAGES is set to the locale of the category | |
1325 | * whose information is desired. Look through all the messages */ | |
1326 | ||
1327 | for (e = 0; | |
1328 | #ifdef HAS_SYS_ERRLIST | |
1329 | e <= sys_nerr | |
1330 | #endif | |
1331 | ; e++) | |
1332 | { | |
1333 | const U8* const errmsg = (U8 *) Strerror(e) ; | |
1334 | if (!errmsg) | |
1335 | break; | |
1336 | if (! is_utf8_string(errmsg, 0)) { | |
1337 | failures++; | |
1338 | break; | |
1339 | } | |
1340 | else if (! is_ascii_string(errmsg, 0)) { | |
1341 | non_ascii++; | |
1342 | } | |
1343 | } | |
1344 | ||
1345 | /* And, if we changed it, restore LC_MESSAGES to its original locale */ | |
1346 | if (save_messages_locale) { | |
1347 | setlocale(LC_MESSAGES, save_messages_locale); | |
1348 | Safefree(save_messages_locale); | |
1349 | } | |
1350 | ||
1351 | /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid, | |
1352 | * any non-ascii means it is one; otherwise we assume it isn't */ | |
1353 | return (failures) ? FALSE : non_ascii; | |
1354 | ||
1355 | } | |
1356 | cant_use_messages: | |
1357 | ||
1358 | #endif | |
1359 | ||
69014004 KW |
1360 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1361 | "Assuming locale %s is not a UTF-8 locale\n", | |
1362 | save_input_locale)); | |
fa9b773e | 1363 | Safefree(save_input_locale); |
7d74bb61 KW |
1364 | return FALSE; |
1365 | } | |
1366 | ||
8ef6e574 | 1367 | #endif |
7d74bb61 | 1368 | |
66610fdd RGS |
1369 | /* |
1370 | * Local variables: | |
1371 | * c-indentation-style: bsd | |
1372 | * c-basic-offset: 4 | |
14d04a33 | 1373 | * indent-tabs-mode: nil |
66610fdd RGS |
1374 | * End: |
1375 | * | |
14d04a33 | 1376 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 1377 | */ |