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; |
98994639 HS |
120 | # endif /* HAS_LOCALECONV */ |
121 | #endif /* USE_LOCALE_NUMERIC */ | |
122 | } | |
123 | ||
98994639 | 124 | void |
8772537c | 125 | Perl_new_numeric(pTHX_ const char *newnum) |
98994639 HS |
126 | { |
127 | #ifdef USE_LOCALE_NUMERIC | |
0d071d52 KW |
128 | |
129 | /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell | |
130 | * core Perl this and that 'newnum' is the name of the new locale. | |
131 | * It installs this locale as the current underlying default. | |
132 | * | |
133 | * The default locale and the C locale can be toggled between by use of the | |
134 | * set_numeric_local() and set_numeric_standard() functions, which should | |
135 | * probably not be called directly, but only via macros like | |
136 | * SET_NUMERIC_STANDARD() in perl.h. | |
137 | * | |
138 | * The toggling is necessary mainly so that a non-dot radix decimal point | |
139 | * character can be output, while allowing internal calculations to use a | |
140 | * dot. | |
141 | * | |
142 | * This sets several interpreter-level variables: | |
143 | * PL_numeric_name The default locale's name: a copy of 'newnum' | |
144 | * PL_numeric_local A boolean indicating if the toggled state is such | |
145 | * that the current locale is the default locale | |
146 | * PL_numeric_standard A boolean indicating if the toggled state is such | |
147 | * that the current locale is the C locale | |
148 | * Note that both of the last two variables can be true at the same time, | |
149 | * if the underlying locale is C. (Toggling is a no-op under these | |
150 | * circumstances.) | |
151 | * | |
152 | * Any code changing the locale (outside this file) should use | |
153 | * POSIX::setlocale, which calls this function. Therefore this function | |
154 | * should be called directly only from this file and from | |
155 | * POSIX::setlocale() */ | |
156 | ||
b03f34cf | 157 | char *save_newnum; |
97aff369 | 158 | dVAR; |
98994639 HS |
159 | |
160 | if (! newnum) { | |
43c5f42d NC |
161 | Safefree(PL_numeric_name); |
162 | PL_numeric_name = NULL; | |
98994639 HS |
163 | PL_numeric_standard = TRUE; |
164 | PL_numeric_local = TRUE; | |
165 | return; | |
166 | } | |
167 | ||
b03f34cf KW |
168 | save_newnum = stdize_locale(savepv(newnum)); |
169 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { | |
98994639 | 170 | Safefree(PL_numeric_name); |
b03f34cf | 171 | PL_numeric_name = save_newnum; |
b03f34cf | 172 | } |
98994639 | 173 | |
e19f01cb KW |
174 | PL_numeric_standard = ((*save_newnum == 'C' && save_newnum[1] == '\0') |
175 | || strEQ(save_newnum, "POSIX")); | |
176 | PL_numeric_local = TRUE; | |
177 | set_numeric_radix(); | |
6959d69d | 178 | |
98994639 HS |
179 | #endif /* USE_LOCALE_NUMERIC */ |
180 | } | |
181 | ||
182 | void | |
183 | Perl_set_numeric_standard(pTHX) | |
184 | { | |
185 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 186 | dVAR; |
98994639 | 187 | |
0d071d52 KW |
188 | /* Toggle the LC_NUMERIC locale to C, if not already there. Probably |
189 | * should use the macros like SET_NUMERIC_STANDARD() in perl.h instead of | |
190 | * calling this directly. */ | |
191 | ||
98994639 HS |
192 | if (! PL_numeric_standard) { |
193 | setlocale(LC_NUMERIC, "C"); | |
194 | PL_numeric_standard = TRUE; | |
195 | PL_numeric_local = FALSE; | |
196 | set_numeric_radix(); | |
197 | } | |
198 | ||
199 | #endif /* USE_LOCALE_NUMERIC */ | |
200 | } | |
201 | ||
202 | void | |
203 | Perl_set_numeric_local(pTHX) | |
204 | { | |
205 | #ifdef USE_LOCALE_NUMERIC | |
97aff369 | 206 | dVAR; |
98994639 | 207 | |
0d071d52 KW |
208 | /* Toggle the LC_NUMERIC locale to the current underlying default, if not |
209 | * already there. Probably should use the macros like SET_NUMERIC_LOCAL() | |
210 | * in perl.h instead of calling this directly. */ | |
211 | ||
98994639 HS |
212 | if (! PL_numeric_local) { |
213 | setlocale(LC_NUMERIC, PL_numeric_name); | |
214 | PL_numeric_standard = FALSE; | |
215 | PL_numeric_local = TRUE; | |
216 | set_numeric_radix(); | |
217 | } | |
218 | ||
219 | #endif /* USE_LOCALE_NUMERIC */ | |
220 | } | |
221 | ||
222 | /* | |
223 | * Set up for a new ctype locale. | |
224 | */ | |
225 | void | |
8772537c | 226 | Perl_new_ctype(pTHX_ const char *newctype) |
98994639 HS |
227 | { |
228 | #ifdef USE_LOCALE_CTYPE | |
0d071d52 KW |
229 | |
230 | /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell | |
231 | * core Perl this and that 'newctype' is the name of the new locale. | |
232 | * | |
233 | * This function sets up the folding arrays for all 256 bytes, assuming | |
234 | * that tofold() is tolc() since fold case is not a concept in POSIX, | |
235 | * | |
236 | * Any code changing the locale (outside this file) should use | |
237 | * POSIX::setlocale, which calls this function. Therefore this function | |
238 | * should be called directly only from this file and from | |
239 | * POSIX::setlocale() */ | |
240 | ||
27da23d5 | 241 | dVAR; |
68067e4e | 242 | UV i; |
98994639 | 243 | |
7918f24d NC |
244 | PERL_ARGS_ASSERT_NEW_CTYPE; |
245 | ||
31f05a37 KW |
246 | PL_in_utf8_CTYPE_locale = is_cur_LC_category_utf8(LC_CTYPE); |
247 | ||
248 | /* A UTF-8 locale gets standard rules. But note that code still has to | |
249 | * handle this specially because of the three problematic code points */ | |
250 | if (PL_in_utf8_CTYPE_locale) { | |
251 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); | |
252 | } | |
253 | else { | |
254 | ||
77806dea KW |
255 | for (i = 0; i < 256; i++) { |
256 | if (isUPPER_LC((U8) i)) | |
31f05a37 | 257 | PL_fold_locale[i] = (U8) toLOWER_LC((U8) i); |
77806dea | 258 | else if (isLOWER_LC((U8) i)) |
31f05a37 | 259 | PL_fold_locale[i] = (U8) toUPPER_LC((U8) i); |
98994639 | 260 | else |
e7e76d1b | 261 | PL_fold_locale[i] = (U8) i; |
98994639 | 262 | } |
31f05a37 | 263 | } |
98994639 HS |
264 | |
265 | #endif /* USE_LOCALE_CTYPE */ | |
7918f24d | 266 | PERL_ARGS_ASSERT_NEW_CTYPE; |
8772537c | 267 | PERL_UNUSED_ARG(newctype); |
96a5add6 | 268 | PERL_UNUSED_CONTEXT; |
98994639 HS |
269 | } |
270 | ||
98994639 | 271 | void |
8772537c | 272 | Perl_new_collate(pTHX_ const char *newcoll) |
98994639 HS |
273 | { |
274 | #ifdef USE_LOCALE_COLLATE | |
0d071d52 KW |
275 | |
276 | /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell | |
277 | * core Perl this and that 'newcoll' is the name of the new locale. | |
278 | * | |
279 | * Any code changing the locale (outside this file) should use | |
280 | * POSIX::setlocale, which calls this function. Therefore this function | |
281 | * should be called directly only from this file and from | |
282 | * POSIX::setlocale() */ | |
283 | ||
97aff369 | 284 | dVAR; |
98994639 HS |
285 | |
286 | if (! newcoll) { | |
287 | if (PL_collation_name) { | |
288 | ++PL_collation_ix; | |
289 | Safefree(PL_collation_name); | |
290 | PL_collation_name = NULL; | |
291 | } | |
292 | PL_collation_standard = TRUE; | |
293 | PL_collxfrm_base = 0; | |
294 | PL_collxfrm_mult = 2; | |
295 | return; | |
296 | } | |
297 | ||
298 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { | |
299 | ++PL_collation_ix; | |
300 | Safefree(PL_collation_name); | |
301 | PL_collation_name = stdize_locale(savepv(newcoll)); | |
770526c1 NC |
302 | PL_collation_standard = ((*newcoll == 'C' && newcoll[1] == '\0') |
303 | || strEQ(newcoll, "POSIX")); | |
98994639 HS |
304 | |
305 | { | |
306 | /* 2: at most so many chars ('a', 'b'). */ | |
307 | /* 50: surely no system expands a char more. */ | |
308 | #define XFRMBUFSIZE (2 * 50) | |
309 | char xbuf[XFRMBUFSIZE]; | |
8772537c AL |
310 | const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE); |
311 | const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE); | |
312 | const SSize_t mult = fb - fa; | |
e0601d3c | 313 | if (mult < 1 && !(fa == 0 && fb == 0)) |
ad49ad39 NC |
314 | Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf, |
315 | (UV) fa, (UV) fb); | |
eb160463 | 316 | PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0; |
98994639 HS |
317 | PL_collxfrm_mult = mult; |
318 | } | |
319 | } | |
320 | ||
321 | #endif /* USE_LOCALE_COLLATE */ | |
322 | } | |
323 | ||
324 | /* | |
325 | * Initialize locale awareness. | |
326 | */ | |
327 | int | |
328 | Perl_init_i18nl10n(pTHX_ int printwarn) | |
329 | { | |
330 | int ok = 1; | |
331 | /* returns | |
332 | * 1 = set ok or not applicable, | |
333 | * 0 = fallback to C locale, | |
334 | * -1 = fallback to C locale failed | |
335 | */ | |
336 | ||
337 | #if defined(USE_LOCALE) | |
97aff369 | 338 | dVAR; |
98994639 HS |
339 | |
340 | #ifdef USE_LOCALE_CTYPE | |
341 | char *curctype = NULL; | |
342 | #endif /* USE_LOCALE_CTYPE */ | |
343 | #ifdef USE_LOCALE_COLLATE | |
344 | char *curcoll = NULL; | |
345 | #endif /* USE_LOCALE_COLLATE */ | |
346 | #ifdef USE_LOCALE_NUMERIC | |
347 | char *curnum = NULL; | |
348 | #endif /* USE_LOCALE_NUMERIC */ | |
349 | #ifdef __GLIBC__ | |
7452cf6a | 350 | char * const language = PerlEnv_getenv("LANGUAGE"); |
98994639 | 351 | #endif |
ccd65d51 KW |
352 | /* NULL uses the existing already set up locale */ |
353 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) | |
354 | ? NULL | |
355 | : ""; | |
7452cf6a AL |
356 | char * const lc_all = PerlEnv_getenv("LC_ALL"); |
357 | char * const lang = PerlEnv_getenv("LANG"); | |
98994639 HS |
358 | bool setlocale_failure = FALSE; |
359 | ||
360 | #ifdef LOCALE_ENVIRON_REQUIRED | |
361 | ||
362 | /* | |
363 | * Ultrix setlocale(..., "") fails if there are no environment | |
364 | * variables from which to get a locale name. | |
365 | */ | |
366 | ||
367 | bool done = FALSE; | |
368 | ||
b3e384bf | 369 | # ifdef LC_ALL |
98994639 | 370 | if (lang) { |
ccd65d51 | 371 | if (setlocale(LC_ALL, setlocale_init)) |
98994639 HS |
372 | done = TRUE; |
373 | else | |
374 | setlocale_failure = TRUE; | |
375 | } | |
376 | if (!setlocale_failure) { | |
b3e384bf | 377 | # ifdef USE_LOCALE_CTYPE |
56279a21 | 378 | Safefree(curctype); |
98994639 HS |
379 | if (! (curctype = |
380 | setlocale(LC_CTYPE, | |
381 | (!done && (lang || PerlEnv_getenv("LC_CTYPE"))) | |
ccd65d51 | 382 | ? setlocale_init : NULL))) |
98994639 HS |
383 | setlocale_failure = TRUE; |
384 | else | |
385 | curctype = savepv(curctype); | |
b3e384bf KW |
386 | # endif /* USE_LOCALE_CTYPE */ |
387 | # ifdef USE_LOCALE_COLLATE | |
56279a21 | 388 | Safefree(curcoll); |
98994639 HS |
389 | if (! (curcoll = |
390 | setlocale(LC_COLLATE, | |
391 | (!done && (lang || PerlEnv_getenv("LC_COLLATE"))) | |
ccd65d51 | 392 | ? setlocale_init : NULL))) |
98994639 HS |
393 | setlocale_failure = TRUE; |
394 | else | |
395 | curcoll = savepv(curcoll); | |
b3e384bf KW |
396 | # endif /* USE_LOCALE_COLLATE */ |
397 | # ifdef USE_LOCALE_NUMERIC | |
56279a21 | 398 | Safefree(curnum); |
98994639 HS |
399 | if (! (curnum = |
400 | setlocale(LC_NUMERIC, | |
401 | (!done && (lang || PerlEnv_getenv("LC_NUMERIC"))) | |
ccd65d51 | 402 | ? setlocale_init : NULL))) |
98994639 HS |
403 | setlocale_failure = TRUE; |
404 | else | |
405 | curnum = savepv(curnum); | |
b3e384bf | 406 | # endif /* USE_LOCALE_NUMERIC */ |
98994639 HS |
407 | } |
408 | ||
b3e384bf | 409 | # endif /* LC_ALL */ |
98994639 HS |
410 | |
411 | #endif /* !LOCALE_ENVIRON_REQUIRED */ | |
412 | ||
413 | #ifdef LC_ALL | |
ccd65d51 | 414 | if (! setlocale(LC_ALL, setlocale_init)) |
98994639 HS |
415 | setlocale_failure = TRUE; |
416 | #endif /* LC_ALL */ | |
417 | ||
418 | if (!setlocale_failure) { | |
419 | #ifdef USE_LOCALE_CTYPE | |
6cb43dbf | 420 | Safefree(curctype); |
ccd65d51 | 421 | if (! (curctype = setlocale(LC_CTYPE, setlocale_init))) |
98994639 HS |
422 | setlocale_failure = TRUE; |
423 | else | |
424 | curctype = savepv(curctype); | |
425 | #endif /* USE_LOCALE_CTYPE */ | |
426 | #ifdef USE_LOCALE_COLLATE | |
6cb43dbf | 427 | Safefree(curcoll); |
ccd65d51 | 428 | if (! (curcoll = setlocale(LC_COLLATE, setlocale_init))) |
98994639 HS |
429 | setlocale_failure = TRUE; |
430 | else | |
431 | curcoll = savepv(curcoll); | |
432 | #endif /* USE_LOCALE_COLLATE */ | |
433 | #ifdef USE_LOCALE_NUMERIC | |
6cb43dbf | 434 | Safefree(curnum); |
ccd65d51 | 435 | if (! (curnum = setlocale(LC_NUMERIC, setlocale_init))) |
98994639 HS |
436 | setlocale_failure = TRUE; |
437 | else | |
438 | curnum = savepv(curnum); | |
439 | #endif /* USE_LOCALE_NUMERIC */ | |
440 | } | |
441 | ||
442 | if (setlocale_failure) { | |
443 | char *p; | |
0bd48802 | 444 | const bool locwarn = (printwarn > 1 || |
98994639 HS |
445 | (printwarn && |
446 | (!(p = PerlEnv_getenv("PERL_BADLANG")) || atoi(p)))); | |
447 | ||
448 | if (locwarn) { | |
449 | #ifdef LC_ALL | |
450 | ||
451 | PerlIO_printf(Perl_error_log, | |
452 | "perl: warning: Setting locale failed.\n"); | |
453 | ||
454 | #else /* !LC_ALL */ | |
455 | ||
456 | PerlIO_printf(Perl_error_log, | |
457 | "perl: warning: Setting locale failed for the categories:\n\t"); | |
458 | #ifdef USE_LOCALE_CTYPE | |
459 | if (! curctype) | |
460 | PerlIO_printf(Perl_error_log, "LC_CTYPE "); | |
461 | #endif /* USE_LOCALE_CTYPE */ | |
462 | #ifdef USE_LOCALE_COLLATE | |
463 | if (! curcoll) | |
464 | PerlIO_printf(Perl_error_log, "LC_COLLATE "); | |
465 | #endif /* USE_LOCALE_COLLATE */ | |
466 | #ifdef USE_LOCALE_NUMERIC | |
467 | if (! curnum) | |
468 | PerlIO_printf(Perl_error_log, "LC_NUMERIC "); | |
469 | #endif /* USE_LOCALE_NUMERIC */ | |
470 | PerlIO_printf(Perl_error_log, "\n"); | |
471 | ||
472 | #endif /* LC_ALL */ | |
473 | ||
474 | PerlIO_printf(Perl_error_log, | |
475 | "perl: warning: Please check that your locale settings:\n"); | |
476 | ||
477 | #ifdef __GLIBC__ | |
478 | PerlIO_printf(Perl_error_log, | |
479 | "\tLANGUAGE = %c%s%c,\n", | |
480 | language ? '"' : '(', | |
481 | language ? language : "unset", | |
482 | language ? '"' : ')'); | |
483 | #endif | |
484 | ||
485 | PerlIO_printf(Perl_error_log, | |
486 | "\tLC_ALL = %c%s%c,\n", | |
487 | lc_all ? '"' : '(', | |
488 | lc_all ? lc_all : "unset", | |
489 | lc_all ? '"' : ')'); | |
490 | ||
491 | #if defined(USE_ENVIRON_ARRAY) | |
492 | { | |
493 | char **e; | |
494 | for (e = environ; *e; e++) { | |
495 | if (strnEQ(*e, "LC_", 3) | |
496 | && strnNE(*e, "LC_ALL=", 7) | |
497 | && (p = strchr(*e, '='))) | |
498 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", | |
499 | (int)(p - *e), *e, p + 1); | |
500 | } | |
501 | } | |
502 | #else | |
503 | PerlIO_printf(Perl_error_log, | |
504 | "\t(possibly more locale environment variables)\n"); | |
505 | #endif | |
506 | ||
507 | PerlIO_printf(Perl_error_log, | |
508 | "\tLANG = %c%s%c\n", | |
509 | lang ? '"' : '(', | |
510 | lang ? lang : "unset", | |
511 | lang ? '"' : ')'); | |
512 | ||
513 | PerlIO_printf(Perl_error_log, | |
514 | " are supported and installed on your system.\n"); | |
515 | } | |
516 | ||
517 | #ifdef LC_ALL | |
518 | ||
519 | if (setlocale(LC_ALL, "C")) { | |
520 | if (locwarn) | |
521 | PerlIO_printf(Perl_error_log, | |
522 | "perl: warning: Falling back to the standard locale (\"C\").\n"); | |
523 | ok = 0; | |
524 | } | |
525 | else { | |
526 | if (locwarn) | |
527 | PerlIO_printf(Perl_error_log, | |
528 | "perl: warning: Failed to fall back to the standard locale (\"C\").\n"); | |
529 | ok = -1; | |
530 | } | |
531 | ||
532 | #else /* ! LC_ALL */ | |
533 | ||
534 | if (0 | |
535 | #ifdef USE_LOCALE_CTYPE | |
536 | || !(curctype || setlocale(LC_CTYPE, "C")) | |
537 | #endif /* USE_LOCALE_CTYPE */ | |
538 | #ifdef USE_LOCALE_COLLATE | |
539 | || !(curcoll || setlocale(LC_COLLATE, "C")) | |
540 | #endif /* USE_LOCALE_COLLATE */ | |
541 | #ifdef USE_LOCALE_NUMERIC | |
542 | || !(curnum || setlocale(LC_NUMERIC, "C")) | |
543 | #endif /* USE_LOCALE_NUMERIC */ | |
544 | ) | |
545 | { | |
546 | if (locwarn) | |
547 | PerlIO_printf(Perl_error_log, | |
548 | "perl: warning: Cannot fall back to the standard locale (\"C\").\n"); | |
549 | ok = -1; | |
550 | } | |
551 | ||
552 | #endif /* ! LC_ALL */ | |
553 | ||
554 | #ifdef USE_LOCALE_CTYPE | |
56279a21 | 555 | Safefree(curctype); |
bd61b366 | 556 | curctype = savepv(setlocale(LC_CTYPE, NULL)); |
98994639 HS |
557 | #endif /* USE_LOCALE_CTYPE */ |
558 | #ifdef USE_LOCALE_COLLATE | |
56279a21 | 559 | Safefree(curcoll); |
bd61b366 | 560 | curcoll = savepv(setlocale(LC_COLLATE, NULL)); |
98994639 HS |
561 | #endif /* USE_LOCALE_COLLATE */ |
562 | #ifdef USE_LOCALE_NUMERIC | |
56279a21 | 563 | Safefree(curnum); |
bd61b366 | 564 | curnum = savepv(setlocale(LC_NUMERIC, NULL)); |
98994639 HS |
565 | #endif /* USE_LOCALE_NUMERIC */ |
566 | } | |
567 | else { | |
568 | ||
569 | #ifdef USE_LOCALE_CTYPE | |
570 | new_ctype(curctype); | |
571 | #endif /* USE_LOCALE_CTYPE */ | |
572 | ||
573 | #ifdef USE_LOCALE_COLLATE | |
574 | new_collate(curcoll); | |
575 | #endif /* USE_LOCALE_COLLATE */ | |
576 | ||
577 | #ifdef USE_LOCALE_NUMERIC | |
578 | new_numeric(curnum); | |
579 | #endif /* USE_LOCALE_NUMERIC */ | |
b310b053 | 580 | |
98994639 HS |
581 | } |
582 | ||
8ef6e574 | 583 | #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE) |
b310b053 | 584 | { |
fde18df1 | 585 | /* Set PL_utf8locale to TRUE if using PerlIO _and_ |
7d74bb61 | 586 | the current LC_CTYPE locale is UTF-8. |
8aa8f774 | 587 | If PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE}) |
a05d7ebb | 588 | are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer |
fde18df1 | 589 | on STDIN, STDOUT, STDERR, _and_ the default open discipline. |
085a54d9 | 590 | */ |
7d74bb61 | 591 | PL_utf8locale = is_cur_LC_category_utf8(LC_CTYPE); |
fde18df1 | 592 | } |
a05d7ebb | 593 | /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO. |
fde18df1 JH |
594 | This is an alternative to using the -C command line switch |
595 | (the -C if present will override this). */ | |
596 | { | |
dd374669 | 597 | const char *p = PerlEnv_getenv("PERL_UNICODE"); |
a05d7ebb | 598 | PL_unicode = p ? parse_unicode_opts(&p) : 0; |
5a22a2bb NC |
599 | if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG) |
600 | PL_utf8cache = -1; | |
b310b053 | 601 | } |
ec71e770 | 602 | #endif |
b310b053 | 603 | |
98994639 | 604 | #ifdef USE_LOCALE_CTYPE |
43c5f42d | 605 | Safefree(curctype); |
98994639 HS |
606 | #endif /* USE_LOCALE_CTYPE */ |
607 | #ifdef USE_LOCALE_COLLATE | |
43c5f42d | 608 | Safefree(curcoll); |
98994639 HS |
609 | #endif /* USE_LOCALE_COLLATE */ |
610 | #ifdef USE_LOCALE_NUMERIC | |
43c5f42d | 611 | Safefree(curnum); |
98994639 | 612 | #endif /* USE_LOCALE_NUMERIC */ |
8ef6e574 KW |
613 | |
614 | #endif /* USE_LOCALE */ | |
615 | ||
98994639 HS |
616 | return ok; |
617 | } | |
618 | ||
98994639 HS |
619 | #ifdef USE_LOCALE_COLLATE |
620 | ||
621 | /* | |
622 | * mem_collxfrm() is a bit like strxfrm() but with two important | |
623 | * differences. First, it handles embedded NULs. Second, it allocates | |
624 | * a bit more memory than needed for the transformed data itself. | |
625 | * The real transformed data begins at offset sizeof(collationix). | |
626 | * Please see sv_collxfrm() to see how this is used. | |
627 | */ | |
166f8a29 | 628 | |
98994639 HS |
629 | char * |
630 | Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen) | |
631 | { | |
97aff369 | 632 | dVAR; |
98994639 HS |
633 | char *xbuf; |
634 | STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */ | |
635 | ||
7918f24d NC |
636 | PERL_ARGS_ASSERT_MEM_COLLXFRM; |
637 | ||
98994639 HS |
638 | /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */ |
639 | /* the +1 is for the terminating NUL. */ | |
640 | ||
641 | xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1; | |
a02a5408 | 642 | Newx(xbuf, xAlloc, char); |
98994639 HS |
643 | if (! xbuf) |
644 | goto bad; | |
645 | ||
646 | *(U32*)xbuf = PL_collation_ix; | |
647 | xout = sizeof(PL_collation_ix); | |
648 | for (xin = 0; xin < len; ) { | |
224e8ef5 | 649 | Size_t xused; |
98994639 HS |
650 | |
651 | for (;;) { | |
652 | xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout); | |
224e8ef5 | 653 | if (xused >= PERL_INT_MAX) |
98994639 | 654 | goto bad; |
eb160463 | 655 | if ((STRLEN)xused < xAlloc - xout) |
98994639 HS |
656 | break; |
657 | xAlloc = (2 * xAlloc) + 1; | |
658 | Renew(xbuf, xAlloc, char); | |
659 | if (! xbuf) | |
660 | goto bad; | |
661 | } | |
662 | ||
663 | xin += strlen(s + xin) + 1; | |
664 | xout += xused; | |
665 | ||
666 | /* Embedded NULs are understood but silently skipped | |
667 | * because they make no sense in locale collation. */ | |
668 | } | |
669 | ||
670 | xbuf[xout] = '\0'; | |
671 | *xlen = xout - sizeof(PL_collation_ix); | |
672 | return xbuf; | |
673 | ||
674 | bad: | |
675 | Safefree(xbuf); | |
676 | *xlen = 0; | |
677 | return NULL; | |
678 | } | |
679 | ||
680 | #endif /* USE_LOCALE_COLLATE */ | |
681 | ||
8ef6e574 KW |
682 | #ifdef USE_LOCALE |
683 | ||
637917bb | 684 | STATIC bool |
7d74bb61 KW |
685 | S_is_cur_LC_category_utf8(pTHX_ int category) |
686 | { | |
687 | /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE | |
688 | * otherwise. 'category' may not be LC_ALL. If the platform doesn't have | |
119ee68b KW |
689 | * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence |
690 | * could give the wrong result. It errs on the side of not being a UTF-8 | |
691 | * locale. */ | |
7d74bb61 KW |
692 | |
693 | char *save_input_locale = NULL; | |
7d74bb61 KW |
694 | STRLEN final_pos; |
695 | ||
8ef6e574 | 696 | #ifdef LC_ALL |
7d74bb61 | 697 | assert(category != LC_ALL); |
8ef6e574 | 698 | #endif |
7d74bb61 KW |
699 | |
700 | /* First dispose of the trivial cases */ | |
b07fffd1 | 701 | save_input_locale = setlocale(category, NULL); |
7d74bb61 KW |
702 | if (! save_input_locale) { |
703 | return FALSE; /* XXX maybe should croak */ | |
704 | } | |
b07fffd1 | 705 | save_input_locale = stdize_locale(savepv(save_input_locale)); |
7d74bb61 KW |
706 | if ((*save_input_locale == 'C' && save_input_locale[1] == '\0') |
707 | || strEQ(save_input_locale, "POSIX")) | |
708 | { | |
b07fffd1 | 709 | Safefree(save_input_locale); |
7d74bb61 KW |
710 | return FALSE; |
711 | } | |
712 | ||
119ee68b | 713 | #if defined(USE_LOCALE_CTYPE) && (defined(MB_CUR_MAX) || defined(HAS_NL_LANGINFO) && defined(CODESET)) |
7d74bb61 | 714 | |
119ee68b | 715 | { /* Next try MB_CUR_MAX or nl_langinfo if available */ |
7d74bb61 KW |
716 | |
717 | char *save_ctype_locale = NULL; | |
119ee68b | 718 | bool is_utf8; |
7d74bb61 | 719 | |
119ee68b | 720 | if (category != LC_CTYPE) { /* These work only on LC_CTYPE */ |
7d74bb61 KW |
721 | |
722 | /* Get the current LC_CTYPE locale */ | |
723 | save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL))); | |
724 | if (! save_ctype_locale) { | |
725 | goto cant_use_nllanginfo; | |
726 | } | |
727 | ||
728 | /* If LC_CTYPE and the desired category use the same locale, this | |
729 | * means that finding the value for LC_CTYPE is the same as finding | |
730 | * the value for the desired category. Otherwise, switch LC_CTYPE | |
731 | * to the desired category's locale */ | |
732 | if (strEQ(save_ctype_locale, save_input_locale)) { | |
733 | Safefree(save_ctype_locale); | |
734 | save_ctype_locale = NULL; | |
735 | } | |
736 | else if (! setlocale(LC_CTYPE, save_input_locale)) { | |
737 | Safefree(save_ctype_locale); | |
738 | goto cant_use_nllanginfo; | |
739 | } | |
740 | } | |
741 | ||
742 | /* Here the current LC_CTYPE is set to the locale of the category whose | |
119ee68b KW |
743 | * information is desired. This means that MB_CUR_MAX and |
744 | * nl_langinfo() should give the correct results */ | |
745 | ||
746 | # ifdef MB_CUR_MAX | |
747 | ||
748 | /* If we switched LC_CTYPE, switch back */ | |
749 | if (save_ctype_locale) { | |
750 | setlocale(LC_CTYPE, save_ctype_locale); | |
751 | Safefree(save_ctype_locale); | |
752 | } | |
753 | ||
754 | /* Standard UTF-8 needs at least 4 bytes to represent the maximum | |
755 | * Unicode code point. Since UTF-8 is the only non-single byte | |
756 | * encoding we handle, we just say any such encoding is UTF-8, and if | |
757 | * turns out to be wrong, other things will fail */ | |
758 | is_utf8 = MB_CUR_MAX >= 4; | |
759 | ||
760 | Safefree(save_input_locale); | |
761 | ||
762 | # ifdef HAS_MBTOWC | |
763 | ||
764 | /* ... But, most system that have MB_CUR_MAX will also have mbtowc(), | |
765 | * since they are both in the C99 standard. We can feed a known byte | |
766 | * string to the latter function, and check that it gives the expected | |
767 | * result */ | |
768 | if (is_utf8) { | |
769 | wchar_t wc; | |
770 | if (mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)) | |
771 | != strlen(HYPHEN_UTF8) | |
772 | || wc != (wchar_t) 0x2010) | |
773 | { | |
774 | is_utf8 = FALSE; | |
775 | } | |
776 | } | |
777 | ||
778 | # endif | |
779 | ||
780 | return is_utf8; | |
781 | #else | |
782 | # if defined(HAS_NL_LANGINFO) && defined(CODESET) | |
783 | { | |
648f6ebd KW |
784 | char *codeset = savepv(nl_langinfo(CODESET)); |
785 | if (codeset) { | |
7d74bb61 | 786 | |
648f6ebd KW |
787 | /* If we switched LC_CTYPE, switch back */ |
788 | if (save_ctype_locale) { | |
789 | setlocale(LC_CTYPE, save_ctype_locale); | |
790 | Safefree(save_ctype_locale); | |
791 | } | |
7d74bb61 | 792 | |
648f6ebd KW |
793 | is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8")) |
794 | || foldEQ(codeset, STR_WITH_LEN("UTF8")); | |
7d74bb61 | 795 | |
648f6ebd KW |
796 | Safefree(codeset); |
797 | Safefree(save_input_locale); | |
798 | return is_utf8; | |
799 | } | |
119ee68b | 800 | } |
7d74bb61 | 801 | |
119ee68b KW |
802 | # endif |
803 | # endif | |
7d74bb61 | 804 | } |
119ee68b | 805 | |
7d74bb61 KW |
806 | cant_use_nllanginfo: |
807 | ||
808 | #endif /* HAS_NL_LANGINFO etc */ | |
809 | ||
810 | /* nl_langinfo not available or failed somehow. Look at the locale name to | |
751e9426 | 811 | * see if it matches qr/UTF -? 8 /ix */ |
7d74bb61 KW |
812 | |
813 | final_pos = strlen(save_input_locale) - 1; | |
751e9426 KW |
814 | if (final_pos >= 3) { |
815 | char *name = save_input_locale; | |
816 | ||
817 | /* Find next 'U' or 'u' and look from there */ | |
818 | while ((name += strcspn(name, "Uu") + 1) | |
819 | <= save_input_locale + final_pos - 2) | |
7d74bb61 | 820 | { |
751e9426 KW |
821 | if (toFOLD(*(name)) != 't' |
822 | || toFOLD(*(name + 1)) != 'f') | |
823 | { | |
824 | continue; | |
825 | } | |
826 | name += 2; | |
827 | if (*(name) == '-') { | |
828 | if ((name > save_input_locale + final_pos - 1)) { | |
829 | break; | |
830 | } | |
831 | name++; | |
832 | } | |
833 | if (*(name) == '8') { | |
834 | Safefree(save_input_locale); | |
835 | return TRUE; | |
836 | } | |
7d74bb61 KW |
837 | } |
838 | } | |
839 | ||
840 | #ifdef WIN32 | |
841 | /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */ | |
842 | if (final_pos >= 4 | |
843 | && *(save_input_locale + final_pos - 0) == '1' | |
844 | && *(save_input_locale + final_pos - 1) == '0' | |
845 | && *(save_input_locale + final_pos - 2) == '0' | |
846 | && *(save_input_locale + final_pos - 3) == '5' | |
847 | && *(save_input_locale + final_pos - 4) == '6') | |
848 | { | |
849 | Safefree(save_input_locale); | |
850 | return TRUE; | |
851 | } | |
852 | #endif | |
853 | ||
fa9b773e KW |
854 | /* Other common encodings are the ISO 8859 series, which aren't UTF-8 */ |
855 | if (instr(save_input_locale, "8859")) { | |
856 | Safefree(save_input_locale); | |
857 | return FALSE; | |
858 | } | |
859 | ||
860 | #ifdef HAS_LOCALECONV | |
861 | ||
862 | # ifdef USE_LOCALE_MONETARY | |
863 | ||
864 | /* Here, there is nothing in the locale name to indicate whether the locale | |
865 | * is UTF-8 or not. This "name", the return of setlocale(), is actually | |
866 | * defined to be opaque, so we can't really rely on the absence of various | |
867 | * substrings in the name to indicate its UTF-8ness. Look at the locale's | |
868 | * currency symbol. Often that will be in the native script, and if the | |
869 | * symbol isn't in UTF-8, we know that the locale isn't. If it is | |
870 | * non-ASCII UTF-8, we infer that the locale is too. | |
871 | * To do this, like above for LC_CTYPE, we first set LC_MONETARY to the | |
872 | * locale of the desired category, if it isn't that locale already */ | |
873 | ||
874 | { | |
875 | char *save_monetary_locale = NULL; | |
876 | bool illegal_utf8 = FALSE; | |
877 | bool only_ascii = FALSE; | |
878 | const struct lconv* const lc = localeconv(); | |
879 | ||
880 | if (category != LC_MONETARY) { | |
881 | ||
882 | save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY, | |
883 | NULL))); | |
884 | if (! save_monetary_locale) { | |
885 | goto cant_use_monetary; | |
886 | } | |
887 | ||
888 | if (strNE(save_monetary_locale, save_input_locale)) { | |
889 | if (! setlocale(LC_MONETARY, save_input_locale)) { | |
890 | Safefree(save_monetary_locale); | |
891 | goto cant_use_monetary; | |
892 | } | |
893 | } | |
894 | } | |
895 | ||
896 | /* Here the current LC_MONETARY is set to the locale of the category | |
897 | * whose information is desired. */ | |
898 | ||
899 | if (lc && lc->currency_symbol) { | |
900 | if (! is_utf8_string((U8 *) lc->currency_symbol, 0)) { | |
901 | illegal_utf8 = TRUE; | |
902 | } | |
903 | else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) { | |
904 | only_ascii = TRUE; | |
905 | } | |
906 | } | |
907 | ||
908 | /* If we changed it, restore LC_MONETARY to its original locale */ | |
909 | if (save_monetary_locale) { | |
910 | setlocale(LC_MONETARY, save_monetary_locale); | |
911 | Safefree(save_monetary_locale); | |
912 | } | |
913 | ||
914 | Safefree(save_input_locale); | |
915 | ||
916 | /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; otherwise | |
917 | * assume the locale is UTF-8 if and only if the symbol is non-ascii | |
918 | * UTF-8. (We can't really tell if the locale is UTF-8 or not if the | |
919 | * symbol is just a '$', so we err on the side of it not being UTF-8) | |
920 | * */ | |
921 | return (illegal_utf8) | |
922 | ? FALSE | |
923 | : ! only_ascii; | |
924 | ||
925 | } | |
926 | cant_use_monetary: | |
927 | ||
928 | # endif /* USE_LOCALE_MONETARY */ | |
929 | #endif /* HAS_LOCALECONV */ | |
930 | ||
931 | #if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES) | |
932 | ||
933 | /* This code is ifdefd out because it was found to not be necessary in testing | |
934 | * on our dromedary test machine, which has over 700 locales. There, looking | |
935 | * at just the currency symbol gave essentially the same results as doing this | |
936 | * extra work. Executing this also caused segfaults in miniperl. I left it in | |
937 | * so as to avoid rewriting it if real-world experience indicates that | |
938 | * dromedary is an outlier. Essentially, instead of returning abpve if we | |
939 | * haven't found illegal utf8, we continue on and examine all the strerror() | |
940 | * messages on the platform for utf8ness. If all are ASCII, we still don't | |
941 | * know the answer; but otherwise we have a pretty good indication of the | |
942 | * utf8ness. The reason this doesn't necessarily help much is that the | |
943 | * messages may not have been translated into the locale. The currency symbol | |
944 | * is much more likely to have been translated. The code below would need to | |
945 | * be altered somewhat to just be a continuation of testing the currency | |
946 | * symbol. */ | |
947 | int e; | |
948 | unsigned int failures = 0, non_ascii = 0; | |
949 | char *save_messages_locale = NULL; | |
950 | ||
951 | /* Like above for LC_CTYPE, we set LC_MESSAGES to the locale of the | |
952 | * desired category, if it isn't that locale already */ | |
953 | ||
954 | if (category != LC_MESSAGES) { | |
955 | ||
956 | save_messages_locale = stdize_locale(savepv(setlocale(LC_MESSAGES, | |
957 | NULL))); | |
958 | if (! save_messages_locale) { | |
959 | goto cant_use_messages; | |
960 | } | |
961 | ||
962 | if (strEQ(save_messages_locale, save_input_locale)) { | |
963 | Safefree(save_input_locale); | |
964 | } | |
965 | else if (! setlocale(LC_MESSAGES, save_input_locale)) { | |
966 | Safefree(save_messages_locale); | |
967 | goto cant_use_messages; | |
968 | } | |
969 | } | |
970 | ||
971 | /* Here the current LC_MESSAGES is set to the locale of the category | |
972 | * whose information is desired. Look through all the messages */ | |
973 | ||
974 | for (e = 0; | |
975 | #ifdef HAS_SYS_ERRLIST | |
976 | e <= sys_nerr | |
977 | #endif | |
978 | ; e++) | |
979 | { | |
980 | const U8* const errmsg = (U8 *) Strerror(e) ; | |
981 | if (!errmsg) | |
982 | break; | |
983 | if (! is_utf8_string(errmsg, 0)) { | |
984 | failures++; | |
985 | break; | |
986 | } | |
987 | else if (! is_ascii_string(errmsg, 0)) { | |
988 | non_ascii++; | |
989 | } | |
990 | } | |
991 | ||
992 | /* And, if we changed it, restore LC_MESSAGES to its original locale */ | |
993 | if (save_messages_locale) { | |
994 | setlocale(LC_MESSAGES, save_messages_locale); | |
995 | Safefree(save_messages_locale); | |
996 | } | |
997 | ||
998 | /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid, | |
999 | * any non-ascii means it is one; otherwise we assume it isn't */ | |
1000 | return (failures) ? FALSE : non_ascii; | |
1001 | ||
1002 | } | |
1003 | cant_use_messages: | |
1004 | ||
1005 | #endif | |
1006 | ||
1007 | Safefree(save_input_locale); | |
7d74bb61 KW |
1008 | return FALSE; |
1009 | } | |
1010 | ||
8ef6e574 | 1011 | #endif |
7d74bb61 | 1012 | |
66610fdd RGS |
1013 | /* |
1014 | * Local variables: | |
1015 | * c-indentation-style: bsd | |
1016 | * c-basic-offset: 4 | |
14d04a33 | 1017 | * indent-tabs-mode: nil |
66610fdd RGS |
1018 | * End: |
1019 | * | |
14d04a33 | 1020 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 1021 | */ |