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 | |
96 | # ifdef HAS_LOCALECONV | |
7452cf6a | 97 | const struct lconv* const lc = localeconv(); |
98994639 | 98 | |
98994639 HS |
99 | if (lc && lc->decimal_point) { |
100 | if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) { | |
101 | SvREFCNT_dec(PL_numeric_radix_sv); | |
a0714e2c | 102 | PL_numeric_radix_sv = NULL; |
98994639 HS |
103 | } |
104 | else { | |
105 | if (PL_numeric_radix_sv) | |
106 | sv_setpv(PL_numeric_radix_sv, lc->decimal_point); | |
107 | else | |
108 | PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0); | |
28acfe03 KW |
109 | if (! is_ascii_string((U8 *) lc->decimal_point, 0) |
110 | && is_utf8_string((U8 *) lc->decimal_point, 0) | |
c1284011 | 111 | && _is_cur_LC_category_utf8(LC_NUMERIC)) |
28acfe03 KW |
112 | { |
113 | SvUTF8_on(PL_numeric_radix_sv); | |
114 | } | |
98994639 HS |
115 | } |
116 | } | |
117 | else | |
a0714e2c | 118 | PL_numeric_radix_sv = NULL; |
69014004 KW |
119 | |
120 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is %s\n", | |
121 | (PL_numeric_radix_sv) | |
122 | ? lc->decimal_point | |
123 | : "NULL")); | |
124 | ||
98994639 HS |
125 | # endif /* HAS_LOCALECONV */ |
126 | #endif /* USE_LOCALE_NUMERIC */ | |
127 | } | |
128 | ||
a39edc4c KW |
129 | /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the |
130 | * return of setlocale(), then this is extremely likely to be the C or POSIX | |
131 | * locale. However, the output of setlocale() is documented to be opaque, but | |
132 | * the odds are extremely small that it would return these two strings for some | |
133 | * other locale. Note that VMS in these two locales includes many non-ASCII | |
134 | * characters as controls and punctuation (below are hex bytes): | |
135 | * cntrl: 00-1F 7F 84-97 9B-9F | |
136 | * punct: 21-2F 3A-40 5B-60 7B-7E A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD | |
137 | * Oddly, none there are listed as alphas, though some represent alphabetics | |
138 | * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */ | |
98b630b3 KW |
139 | #define isNAME_C_OR_POSIX(name) ((name) != NULL \ |
140 | && ((*(name) == 'C' && (*(name + 1)) == '\0') \ | |
a39edc4c KW |
141 | || strEQ((name), "POSIX"))) |
142 | ||
98994639 | 143 | void |
8772537c | 144 | Perl_new_numeric(pTHX_ const char *newnum) |
98994639 HS |
145 | { |
146 | #ifdef USE_LOCALE_NUMERIC | |
0d071d52 KW |
147 | |
148 | /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell | |
149 | * core Perl this and that 'newnum' is the name of the new locale. | |
150 | * It installs this locale as the current underlying default. | |
151 | * | |
152 | * The default locale and the C locale can be toggled between by use of the | |
153 | * set_numeric_local() and set_numeric_standard() functions, which should | |
154 | * probably not be called directly, but only via macros like | |
155 | * SET_NUMERIC_STANDARD() in perl.h. | |
156 | * | |
157 | * The toggling is necessary mainly so that a non-dot radix decimal point | |
158 | * character can be output, while allowing internal calculations to use a | |
159 | * dot. | |
160 | * | |
161 | * This sets several interpreter-level variables: | |
bb304765 | 162 | * PL_numeric_name The underlying locale's name: a copy of 'newnum' |
0d071d52 | 163 | * PL_numeric_local A boolean indicating if the toggled state is such |
7738054c KW |
164 | * that the current locale is the program's underlying |
165 | * locale | |
166 | * PL_numeric_standard An int indicating if the toggled state is such | |
167 | * that the current locale is the C locale. If non-zero, | |
168 | * it is in C; if > 1, it means it may not be toggled away | |
169 | * from C. | |
0d071d52 KW |
170 | * Note that both of the last two variables can be true at the same time, |
171 | * if the underlying locale is C. (Toggling is a no-op under these | |
172 | * circumstances.) | |
173 | * | |
174 | * Any code changing the locale (outside this file) should use | |
175 | * POSIX::setlocale, which calls this function. Therefore this function | |
176 | * should be called directly only from this file and from | |
177 | * POSIX::setlocale() */ | |
178 | ||
b03f34cf | 179 | char *save_newnum; |
98994639 HS |
180 | |
181 | if (! newnum) { | |
43c5f42d NC |
182 | Safefree(PL_numeric_name); |
183 | PL_numeric_name = NULL; | |
98994639 HS |
184 | PL_numeric_standard = TRUE; |
185 | PL_numeric_local = TRUE; | |
186 | return; | |
187 | } | |
188 | ||
b03f34cf KW |
189 | save_newnum = stdize_locale(savepv(newnum)); |
190 | if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) { | |
98994639 | 191 | Safefree(PL_numeric_name); |
b03f34cf | 192 | PL_numeric_name = save_newnum; |
b03f34cf | 193 | } |
98994639 | 194 | |
a39edc4c | 195 | PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum); |
e19f01cb | 196 | PL_numeric_local = TRUE; |
4c28b29c KW |
197 | |
198 | /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't | |
199 | * have to worry about the radix being a non-dot. (Core operations that | |
200 | * need the underlying locale change to it temporarily). */ | |
201 | set_numeric_standard(); | |
202 | ||
e19f01cb | 203 | set_numeric_radix(); |
6959d69d | 204 | |
f2ce9e1c JH |
205 | #else |
206 | PERL_UNUSED_ARG(newnum); | |
98994639 HS |
207 | #endif /* USE_LOCALE_NUMERIC */ |
208 | } | |
209 | ||
210 | void | |
211 | Perl_set_numeric_standard(pTHX) | |
212 | { | |
213 | #ifdef USE_LOCALE_NUMERIC | |
28c1bf33 KW |
214 | /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like |
215 | * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The | |
216 | * macro avoids calling this routine if toggling isn't necessary according | |
217 | * to our records (which could be wrong if some XS code has changed the | |
218 | * locale behind our back) */ | |
0d071d52 | 219 | |
a9b8c0d8 KW |
220 | setlocale(LC_NUMERIC, "C"); |
221 | PL_numeric_standard = TRUE; | |
222 | PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name); | |
223 | set_numeric_radix(); | |
69014004 KW |
224 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
225 | "Underlying LC_NUMERIC locale now is C\n")); | |
98994639 HS |
226 | |
227 | #endif /* USE_LOCALE_NUMERIC */ | |
228 | } | |
229 | ||
230 | void | |
231 | Perl_set_numeric_local(pTHX) | |
232 | { | |
233 | #ifdef USE_LOCALE_NUMERIC | |
28c1bf33 KW |
234 | /* Toggle the LC_NUMERIC locale to the current underlying default. Most |
235 | * code should use the macros like SET_NUMERIC_LOCAL() in perl.h instead of | |
236 | * calling this directly. The macro avoids calling this routine if | |
237 | * toggling isn't necessary according to our records (which could be wrong | |
238 | * if some XS code has changed the locale behind our back) */ | |
a9b8c0d8 KW |
239 | |
240 | setlocale(LC_NUMERIC, PL_numeric_name); | |
241 | PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name); | |
242 | PL_numeric_local = TRUE; | |
243 | set_numeric_radix(); | |
69014004 KW |
244 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
245 | "Underlying LC_NUMERIC locale now is %s\n", | |
246 | PL_numeric_name)); | |
98994639 HS |
247 | |
248 | #endif /* USE_LOCALE_NUMERIC */ | |
249 | } | |
250 | ||
251 | /* | |
252 | * Set up for a new ctype locale. | |
253 | */ | |
254 | void | |
8772537c | 255 | Perl_new_ctype(pTHX_ const char *newctype) |
98994639 HS |
256 | { |
257 | #ifdef USE_LOCALE_CTYPE | |
0d071d52 KW |
258 | |
259 | /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell | |
260 | * core Perl this and that 'newctype' is the name of the new locale. | |
261 | * | |
262 | * This function sets up the folding arrays for all 256 bytes, assuming | |
263 | * that tofold() is tolc() since fold case is not a concept in POSIX, | |
264 | * | |
265 | * Any code changing the locale (outside this file) should use | |
266 | * POSIX::setlocale, which calls this function. Therefore this function | |
267 | * should be called directly only from this file and from | |
268 | * POSIX::setlocale() */ | |
269 | ||
27da23d5 | 270 | dVAR; |
68067e4e | 271 | UV i; |
98994639 | 272 | |
7918f24d NC |
273 | PERL_ARGS_ASSERT_NEW_CTYPE; |
274 | ||
c1284011 | 275 | PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE); |
31f05a37 KW |
276 | |
277 | /* A UTF-8 locale gets standard rules. But note that code still has to | |
278 | * handle this specially because of the three problematic code points */ | |
279 | if (PL_in_utf8_CTYPE_locale) { | |
280 | Copy(PL_fold_latin1, PL_fold_locale, 256, U8); | |
281 | } | |
282 | else { | |
8c6180a9 KW |
283 | /* Assume enough space for every character being bad. 4 spaces each |
284 | * for the 94 printable characters that are output like "'x' "; and 5 | |
285 | * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating | |
286 | * NUL */ | |
287 | char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ]; | |
288 | ||
289 | bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means | |
290 | no check */ | |
291 | bool multi_byte_locale = FALSE; /* Assume is a single-byte locale | |
292 | to start */ | |
293 | unsigned int bad_count = 0; /* Count of bad characters */ | |
294 | ||
baa60164 KW |
295 | for (i = 0; i < 256; i++) { |
296 | if (isUPPER_LC((U8) i)) | |
297 | PL_fold_locale[i] = (U8) toLOWER_LC((U8) i); | |
298 | else if (isLOWER_LC((U8) i)) | |
299 | PL_fold_locale[i] = (U8) toUPPER_LC((U8) i); | |
300 | else | |
301 | PL_fold_locale[i] = (U8) i; | |
8c6180a9 KW |
302 | |
303 | /* If checking for locale problems, see if the native ASCII-range | |
304 | * printables plus \n and \t are in their expected categories in | |
305 | * the new locale. If not, this could mean big trouble, upending | |
306 | * Perl's and most programs' assumptions, like having a | |
307 | * metacharacter with special meaning become a \w. Fortunately, | |
308 | * it's very rare to find locales that aren't supersets of ASCII | |
309 | * nowadays. It isn't a problem for most controls to be changed | |
310 | * into something else; we check only \n and \t, though perhaps \r | |
311 | * could be an issue as well. */ | |
312 | if (check_for_problems | |
313 | && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n')) | |
314 | { | |
315 | if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i)) | |
316 | || (isPUNCT_A(i) && ! isPUNCT_LC(i)) | |
317 | || (isBLANK_A(i) && ! isBLANK_LC(i)) | |
318 | || (i == '\n' && ! isCNTRL_LC(i))) | |
319 | { | |
320 | if (bad_count) { /* Separate multiple entries with a | |
321 | blank */ | |
322 | bad_chars_list[bad_count++] = ' '; | |
323 | } | |
324 | bad_chars_list[bad_count++] = '\''; | |
325 | if (isPRINT_A(i)) { | |
326 | bad_chars_list[bad_count++] = (char) i; | |
327 | } | |
328 | else { | |
329 | bad_chars_list[bad_count++] = '\\'; | |
330 | if (i == '\n') { | |
331 | bad_chars_list[bad_count++] = 'n'; | |
332 | } | |
333 | else { | |
334 | assert(i == '\t'); | |
335 | bad_chars_list[bad_count++] = 't'; | |
336 | } | |
337 | } | |
338 | bad_chars_list[bad_count++] = '\''; | |
339 | bad_chars_list[bad_count] = '\0'; | |
340 | } | |
341 | } | |
342 | } | |
343 | ||
344 | #ifdef MB_CUR_MAX | |
345 | /* We only handle single-byte locales (outside of UTF-8 ones; so if | |
346 | * this locale requires than one byte, there are going to be | |
347 | * problems. */ | |
348 | if (check_for_problems && MB_CUR_MAX > 1) { | |
349 | multi_byte_locale = TRUE; | |
350 | } | |
351 | #endif | |
352 | ||
353 | if (bad_count || multi_byte_locale) { | |
354 | setlocale(LC_CTYPE, "C"); | |
355 | Perl_warner(aTHX_ packWARN(WARN_LOCALE), | |
356 | "Locale '%s' may not work well.%s%s%s\n", | |
357 | newctype, | |
358 | (multi_byte_locale) | |
359 | ? " Some characters in it are not recognized by" | |
360 | " Perl." | |
361 | : "", | |
362 | (bad_count) | |
363 | ? "\nThe following characters (and maybe others)" | |
364 | " may not have the same meaning as the Perl" | |
365 | " program expects:\n" | |
366 | : "", | |
367 | (bad_count) | |
368 | ? bad_chars_list | |
369 | : "" | |
370 | ); | |
371 | setlocale(LC_CTYPE, newctype); | |
baa60164 | 372 | } |
31f05a37 | 373 | } |
98994639 HS |
374 | |
375 | #endif /* USE_LOCALE_CTYPE */ | |
7918f24d | 376 | PERL_ARGS_ASSERT_NEW_CTYPE; |
8772537c | 377 | PERL_UNUSED_ARG(newctype); |
96a5add6 | 378 | PERL_UNUSED_CONTEXT; |
98994639 HS |
379 | } |
380 | ||
98994639 | 381 | void |
8772537c | 382 | Perl_new_collate(pTHX_ const char *newcoll) |
98994639 HS |
383 | { |
384 | #ifdef USE_LOCALE_COLLATE | |
0d071d52 KW |
385 | |
386 | /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell | |
387 | * core Perl this and that 'newcoll' is the name of the new locale. | |
388 | * | |
389 | * Any code changing the locale (outside this file) should use | |
390 | * POSIX::setlocale, which calls this function. Therefore this function | |
391 | * should be called directly only from this file and from | |
392 | * POSIX::setlocale() */ | |
393 | ||
98994639 HS |
394 | if (! newcoll) { |
395 | if (PL_collation_name) { | |
396 | ++PL_collation_ix; | |
397 | Safefree(PL_collation_name); | |
398 | PL_collation_name = NULL; | |
399 | } | |
400 | PL_collation_standard = TRUE; | |
401 | PL_collxfrm_base = 0; | |
402 | PL_collxfrm_mult = 2; | |
403 | return; | |
404 | } | |
405 | ||
406 | if (! PL_collation_name || strNE(PL_collation_name, newcoll)) { | |
407 | ++PL_collation_ix; | |
408 | Safefree(PL_collation_name); | |
409 | PL_collation_name = stdize_locale(savepv(newcoll)); | |
a39edc4c | 410 | PL_collation_standard = isNAME_C_OR_POSIX(newcoll); |
98994639 HS |
411 | |
412 | { | |
413 | /* 2: at most so many chars ('a', 'b'). */ | |
414 | /* 50: surely no system expands a char more. */ | |
415 | #define XFRMBUFSIZE (2 * 50) | |
416 | char xbuf[XFRMBUFSIZE]; | |
8772537c AL |
417 | const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE); |
418 | const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE); | |
419 | const SSize_t mult = fb - fa; | |
e0601d3c | 420 | if (mult < 1 && !(fa == 0 && fb == 0)) |
ad49ad39 NC |
421 | Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf, |
422 | (UV) fa, (UV) fb); | |
eb160463 | 423 | PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0; |
98994639 HS |
424 | PL_collxfrm_mult = mult; |
425 | } | |
426 | } | |
427 | ||
f2ce9e1c JH |
428 | #else |
429 | PERL_UNUSED_ARG(newcoll); | |
98994639 HS |
430 | #endif /* USE_LOCALE_COLLATE */ |
431 | } | |
432 | ||
b385bb4d KW |
433 | #ifdef WIN32 |
434 | ||
435 | char * | |
436 | Perl_my_setlocale(pTHX_ int category, const char* locale) | |
437 | { | |
438 | /* This, for Windows, emulates POSIX setlocale() behavior. There is no | |
439 | * difference unless the input locale is "", which means on Windows to get | |
440 | * the machine default, which is set via the computer's "Regional and | |
441 | * Language Options" (or its current equivalent). In POSIX, it instead | |
442 | * means to find the locale from the user's environment. This routine | |
443 | * looks in the environment, and, if anything is found, uses that instead | |
444 | * of going to the machine default. If there is no environment override, | |
445 | * the machine default is used, as normal, by calling the real setlocale() | |
446 | * with "". The POSIX behavior is to use the LC_ALL variable if set; | |
447 | * otherwise to use the particular category's variable if set; otherwise to | |
448 | * use the LANG variable. */ | |
449 | ||
481465ea | 450 | bool override_LC_ALL = 0; |
89f7b9aa KW |
451 | char * result; |
452 | ||
b385bb4d KW |
453 | if (locale && strEQ(locale, "")) { |
454 | # ifdef LC_ALL | |
455 | locale = PerlEnv_getenv("LC_ALL"); | |
456 | if (! locale) { | |
457 | #endif | |
458 | switch (category) { | |
459 | # ifdef LC_ALL | |
460 | case LC_ALL: | |
481465ea | 461 | override_LC_ALL = TRUE; |
b385bb4d KW |
462 | break; /* We already know its variable isn't set */ |
463 | # endif | |
464 | # ifdef USE_LOCALE_TIME | |
465 | case LC_TIME: | |
466 | locale = PerlEnv_getenv("LC_TIME"); | |
467 | break; | |
468 | # endif | |
469 | # ifdef USE_LOCALE_CTYPE | |
470 | case LC_CTYPE: | |
471 | locale = PerlEnv_getenv("LC_CTYPE"); | |
472 | break; | |
473 | # endif | |
474 | # ifdef USE_LOCALE_COLLATE | |
475 | case LC_COLLATE: | |
476 | locale = PerlEnv_getenv("LC_COLLATE"); | |
477 | break; | |
478 | # endif | |
479 | # ifdef USE_LOCALE_MONETARY | |
480 | case LC_MONETARY: | |
481 | locale = PerlEnv_getenv("LC_MONETARY"); | |
482 | break; | |
483 | # endif | |
484 | # ifdef USE_LOCALE_NUMERIC | |
485 | case LC_NUMERIC: | |
486 | locale = PerlEnv_getenv("LC_NUMERIC"); | |
487 | break; | |
488 | # endif | |
489 | # ifdef USE_LOCALE_MESSAGES | |
490 | case LC_MESSAGES: | |
491 | locale = PerlEnv_getenv("LC_MESSAGES"); | |
492 | break; | |
493 | # endif | |
494 | default: | |
495 | /* This is a category, like PAPER_SIZE that we don't | |
496 | * know about; and so can't provide a wrapper. */ | |
497 | break; | |
498 | } | |
499 | if (! locale) { | |
500 | locale = PerlEnv_getenv("LANG"); | |
481465ea | 501 | if (! locale) { |
b385bb4d KW |
502 | locale = ""; |
503 | } | |
504 | } | |
505 | # ifdef LC_ALL | |
506 | } | |
507 | # endif | |
508 | } | |
509 | ||
89f7b9aa KW |
510 | result = setlocale(category, locale); |
511 | ||
481465ea | 512 | if (! override_LC_ALL) { |
89f7b9aa KW |
513 | return result; |
514 | } | |
515 | ||
516 | /* Here the input locale was LC_ALL, and we have set it to what is in the | |
481465ea KW |
517 | * LANG variable or the system default if there is no LANG. But these have |
518 | * lower priority than the other LC_foo variables, so override it for each | |
519 | * one that is set. (If they are set to "", it means to use the same thing | |
520 | * we just set LC_ALL to, so can skip) */ | |
89f7b9aa KW |
521 | # ifdef USE_LOCALE_TIME |
522 | result = PerlEnv_getenv("LC_TIME"); | |
730252b2 | 523 | if (result && strNE(result, "")) { |
89f7b9aa KW |
524 | setlocale(LC_TIME, result); |
525 | } | |
526 | # endif | |
527 | # ifdef USE_LOCALE_CTYPE | |
528 | result = PerlEnv_getenv("LC_CTYPE"); | |
730252b2 | 529 | if (result && strNE(result, "")) { |
89f7b9aa KW |
530 | setlocale(LC_CTYPE, result); |
531 | } | |
532 | # endif | |
533 | # ifdef USE_LOCALE_COLLATE | |
534 | result = PerlEnv_getenv("LC_COLLATE"); | |
730252b2 | 535 | if (result && strNE(result, "")) { |
89f7b9aa KW |
536 | setlocale(LC_COLLATE, result); |
537 | } | |
538 | # endif | |
539 | # ifdef USE_LOCALE_MONETARY | |
540 | result = PerlEnv_getenv("LC_MONETARY"); | |
730252b2 | 541 | if (result && strNE(result, "")) { |
89f7b9aa KW |
542 | setlocale(LC_MONETARY, result); |
543 | } | |
544 | # endif | |
545 | # ifdef USE_LOCALE_NUMERIC | |
546 | result = PerlEnv_getenv("LC_NUMERIC"); | |
730252b2 | 547 | if (result && strNE(result, "")) { |
89f7b9aa KW |
548 | setlocale(LC_NUMERIC, result); |
549 | } | |
550 | # endif | |
551 | # ifdef USE_LOCALE_MESSAGES | |
552 | result = PerlEnv_getenv("LC_MESSAGES"); | |
730252b2 | 553 | if (result && strNE(result, "")) { |
89f7b9aa KW |
554 | setlocale(LC_MESSAGES, result); |
555 | } | |
556 | # endif | |
557 | ||
558 | return setlocale(LC_ALL, NULL); | |
559 | ||
b385bb4d KW |
560 | } |
561 | ||
562 | #endif | |
563 | ||
564 | ||
98994639 HS |
565 | /* |
566 | * Initialize locale awareness. | |
567 | */ | |
568 | int | |
569 | Perl_init_i18nl10n(pTHX_ int printwarn) | |
570 | { | |
0e92a118 KW |
571 | /* printwarn is |
572 | * | |
573 | * 0 if not to output warning when setup locale is bad | |
574 | * 1 if to output warning based on value of PERL_BADLANG | |
575 | * >1 if to output regardless of PERL_BADLANG | |
576 | * | |
577 | * returns | |
98994639 | 578 | * 1 = set ok or not applicable, |
0e92a118 KW |
579 | * 0 = fallback to a locale of lower priority |
580 | * -1 = fallback to all locales failed, not even to the C locale | |
98994639 HS |
581 | */ |
582 | ||
0e92a118 KW |
583 | int ok = 1; |
584 | ||
98994639 | 585 | #if defined(USE_LOCALE) |
98994639 HS |
586 | #ifdef USE_LOCALE_CTYPE |
587 | char *curctype = NULL; | |
588 | #endif /* USE_LOCALE_CTYPE */ | |
589 | #ifdef USE_LOCALE_COLLATE | |
590 | char *curcoll = NULL; | |
591 | #endif /* USE_LOCALE_COLLATE */ | |
592 | #ifdef USE_LOCALE_NUMERIC | |
593 | char *curnum = NULL; | |
594 | #endif /* USE_LOCALE_NUMERIC */ | |
595 | #ifdef __GLIBC__ | |
7452cf6a | 596 | char * const language = PerlEnv_getenv("LANGUAGE"); |
98994639 | 597 | #endif |
65ebb059 | 598 | |
ccd65d51 KW |
599 | /* NULL uses the existing already set up locale */ |
600 | const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT")) | |
601 | ? NULL | |
602 | : ""; | |
c3fcd832 KW |
603 | const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */ |
604 | unsigned int trial_locales_count; | |
7452cf6a AL |
605 | char * const lc_all = PerlEnv_getenv("LC_ALL"); |
606 | char * const lang = PerlEnv_getenv("LANG"); | |
98994639 | 607 | bool setlocale_failure = FALSE; |
65ebb059 KW |
608 | unsigned int i; |
609 | char *p; | |
610 | const bool locwarn = (printwarn > 1 || | |
611 | (printwarn && | |
96e440d2 JH |
612 | (!(p = PerlEnv_getenv("PERL_BADLANG")) || |
613 | grok_atou(p, NULL)))); | |
0e92a118 | 614 | bool done = FALSE; |
6bce99ee JH |
615 | #ifdef WIN32 |
616 | /* In some systems you can find out the system default locale | |
617 | * and use that as the fallback locale. */ | |
618 | # define SYSTEM_DEFAULT_LOCALE | |
619 | #endif | |
620 | #ifdef SYSTEM_DEFAULT_LOCALE | |
65ebb059 | 621 | const char *system_default_locale = NULL; |
6bce99ee | 622 | #endif |
98994639 | 623 | |
0e92a118 KW |
624 | #ifndef LOCALE_ENVIRON_REQUIRED |
625 | PERL_UNUSED_VAR(done); | |
626 | #else | |
98994639 HS |
627 | |
628 | /* | |
629 | * Ultrix setlocale(..., "") fails if there are no environment | |
630 | * variables from which to get a locale name. | |
631 | */ | |
632 | ||
b3e384bf | 633 | # ifdef LC_ALL |
98994639 | 634 | if (lang) { |
b385bb4d | 635 | if (my_setlocale(LC_ALL, setlocale_init)) |
98994639 HS |
636 | done = TRUE; |
637 | else | |
638 | setlocale_failure = TRUE; | |
639 | } | |
640 | if (!setlocale_failure) { | |
b3e384bf | 641 | # ifdef USE_LOCALE_CTYPE |
56279a21 | 642 | Safefree(curctype); |
98994639 | 643 | if (! (curctype = |
b385bb4d | 644 | my_setlocale(LC_CTYPE, |
98994639 | 645 | (!done && (lang || PerlEnv_getenv("LC_CTYPE"))) |
ccd65d51 | 646 | ? setlocale_init : NULL))) |
98994639 HS |
647 | setlocale_failure = TRUE; |
648 | else | |
649 | curctype = savepv(curctype); | |
b3e384bf KW |
650 | # endif /* USE_LOCALE_CTYPE */ |
651 | # ifdef USE_LOCALE_COLLATE | |
56279a21 | 652 | Safefree(curcoll); |
98994639 | 653 | if (! (curcoll = |
b385bb4d | 654 | my_setlocale(LC_COLLATE, |
98994639 | 655 | (!done && (lang || PerlEnv_getenv("LC_COLLATE"))) |
ccd65d51 | 656 | ? setlocale_init : NULL))) |
98994639 HS |
657 | setlocale_failure = TRUE; |
658 | else | |
659 | curcoll = savepv(curcoll); | |
b3e384bf KW |
660 | # endif /* USE_LOCALE_COLLATE */ |
661 | # ifdef USE_LOCALE_NUMERIC | |
56279a21 | 662 | Safefree(curnum); |
98994639 | 663 | if (! (curnum = |
b385bb4d | 664 | my_setlocale(LC_NUMERIC, |
98994639 | 665 | (!done && (lang || PerlEnv_getenv("LC_NUMERIC"))) |
ccd65d51 | 666 | ? setlocale_init : NULL))) |
98994639 HS |
667 | setlocale_failure = TRUE; |
668 | else | |
669 | curnum = savepv(curnum); | |
b3e384bf | 670 | # endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
671 | # ifdef USE_LOCALE_MESSAGES |
672 | if (! my_setlocale(LC_MESSAGES, | |
673 | (!done && (lang || PerlEnv_getenv("LC_MESSAGES"))) | |
674 | ? setlocale_init : NULL)) | |
675 | { | |
676 | setlocale_failure = TRUE; | |
677 | } | |
678 | # endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
679 | # ifdef USE_LOCALE_MONETARY |
680 | if (! my_setlocale(LC_MONETARY, | |
681 | (!done && (lang || PerlEnv_getenv("LC_MONETARY"))) | |
682 | ? setlocale_init : NULL)) | |
683 | { | |
684 | setlocale_failure = TRUE; | |
685 | } | |
686 | # endif /* USE_LOCALE_MONETARY */ | |
98994639 HS |
687 | } |
688 | ||
b3e384bf | 689 | # endif /* LC_ALL */ |
98994639 HS |
690 | |
691 | #endif /* !LOCALE_ENVIRON_REQUIRED */ | |
692 | ||
65ebb059 KW |
693 | /* We try each locale in the list until we get one that works, or exhaust |
694 | * the list */ | |
c3fcd832 KW |
695 | trial_locales[0] = setlocale_init; |
696 | trial_locales_count = 1; | |
65ebb059 KW |
697 | for (i= 0; i < trial_locales_count; i++) { |
698 | const char * trial_locale = trial_locales[i]; | |
699 | ||
700 | if (i > 0) { | |
701 | ||
702 | /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED | |
703 | * when i==0, but I (khw) don't think that behavior makes much | |
704 | * sense */ | |
705 | setlocale_failure = FALSE; | |
706 | ||
6bce99ee JH |
707 | #ifdef SYSTEM_DEFAULT_LOCALE |
708 | # ifdef WIN32 | |
65ebb059 KW |
709 | /* On Windows machines, an entry of "" after the 0th means to use |
710 | * the system default locale, which we now proceed to get. */ | |
711 | if (strEQ(trial_locale, "")) { | |
712 | unsigned int j; | |
713 | ||
714 | /* Note that this may change the locale, but we are going to do | |
715 | * that anyway just below */ | |
716 | system_default_locale = setlocale(LC_ALL, ""); | |
717 | ||
718 | /* Skip if invalid or it's already on the list of locales to | |
719 | * try */ | |
720 | if (! system_default_locale) { | |
721 | goto next_iteration; | |
722 | } | |
723 | for (j = 0; j < trial_locales_count; j++) { | |
724 | if (strEQ(system_default_locale, trial_locales[j])) { | |
725 | goto next_iteration; | |
726 | } | |
727 | } | |
728 | ||
729 | trial_locale = system_default_locale; | |
730 | } | |
6bce99ee JH |
731 | # endif /* WIN32 */ |
732 | #endif /* SYSTEM_DEFAULT_LOCALE */ | |
65ebb059 KW |
733 | } |
734 | ||
98994639 | 735 | #ifdef LC_ALL |
7cd8b568 | 736 | if (! my_setlocale(LC_ALL, trial_locale)) { |
49c85077 | 737 | setlocale_failure = TRUE; |
7cd8b568 KW |
738 | } |
739 | else { | |
740 | /* Since LC_ALL succeeded, it should have changed all the other | |
741 | * categories it can to its value; so we massage things so that the | |
742 | * setlocales below just return their category's current values. | |
743 | * This adequately handles the case in NetBSD where LC_COLLATE may | |
744 | * not be defined for a locale, and setting it individually will | |
745 | * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to | |
746 | * the POSIX locale. */ | |
747 | trial_locale = NULL; | |
748 | } | |
98994639 HS |
749 | #endif /* LC_ALL */ |
750 | ||
49c85077 | 751 | if (!setlocale_failure) { |
98994639 | 752 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
753 | Safefree(curctype); |
754 | if (! (curctype = my_setlocale(LC_CTYPE, trial_locale))) | |
755 | setlocale_failure = TRUE; | |
756 | else | |
757 | curctype = savepv(curctype); | |
98994639 HS |
758 | #endif /* USE_LOCALE_CTYPE */ |
759 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
760 | Safefree(curcoll); |
761 | if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale))) | |
762 | setlocale_failure = TRUE; | |
763 | else | |
764 | curcoll = savepv(curcoll); | |
98994639 HS |
765 | #endif /* USE_LOCALE_COLLATE */ |
766 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
767 | Safefree(curnum); |
768 | if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale))) | |
769 | setlocale_failure = TRUE; | |
770 | else | |
771 | curnum = savepv(curnum); | |
98994639 | 772 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d KW |
773 | #ifdef USE_LOCALE_MESSAGES |
774 | if (! (my_setlocale(LC_MESSAGES, trial_locale))) | |
775 | setlocale_failure = TRUE; | |
776 | #endif /* USE_LOCALE_MESSAGES */ | |
c835d6be KW |
777 | #ifdef USE_LOCALE_MONETARY |
778 | if (! (my_setlocale(LC_MONETARY, trial_locale))) | |
779 | setlocale_failure = TRUE; | |
780 | #endif /* USE_LOCALE_MONETARY */ | |
781 | ||
49c85077 KW |
782 | if (! setlocale_failure) { /* Success */ |
783 | break; | |
784 | } | |
65ebb059 | 785 | } |
98994639 | 786 | |
49c85077 KW |
787 | /* Here, something failed; will need to try a fallback. */ |
788 | ok = 0; | |
65ebb059 | 789 | |
49c85077 KW |
790 | if (i == 0) { |
791 | unsigned int j; | |
98994639 | 792 | |
65ebb059 | 793 | if (locwarn) { /* Output failure info only on the first one */ |
98994639 HS |
794 | #ifdef LC_ALL |
795 | ||
49c85077 KW |
796 | PerlIO_printf(Perl_error_log, |
797 | "perl: warning: Setting locale failed.\n"); | |
98994639 HS |
798 | |
799 | #else /* !LC_ALL */ | |
800 | ||
49c85077 KW |
801 | PerlIO_printf(Perl_error_log, |
802 | "perl: warning: Setting locale failed for the categories:\n\t"); | |
98994639 | 803 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
804 | if (! curctype) |
805 | PerlIO_printf(Perl_error_log, "LC_CTYPE "); | |
98994639 HS |
806 | #endif /* USE_LOCALE_CTYPE */ |
807 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
808 | if (! curcoll) |
809 | PerlIO_printf(Perl_error_log, "LC_COLLATE "); | |
98994639 HS |
810 | #endif /* USE_LOCALE_COLLATE */ |
811 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
812 | if (! curnum) |
813 | PerlIO_printf(Perl_error_log, "LC_NUMERIC "); | |
98994639 | 814 | #endif /* USE_LOCALE_NUMERIC */ |
a782673d | 815 | PerlIO_printf(Perl_error_log, "and possibly others\n"); |
98994639 HS |
816 | |
817 | #endif /* LC_ALL */ | |
818 | ||
49c85077 KW |
819 | PerlIO_printf(Perl_error_log, |
820 | "perl: warning: Please check that your locale settings:\n"); | |
98994639 HS |
821 | |
822 | #ifdef __GLIBC__ | |
49c85077 KW |
823 | PerlIO_printf(Perl_error_log, |
824 | "\tLANGUAGE = %c%s%c,\n", | |
825 | language ? '"' : '(', | |
826 | language ? language : "unset", | |
827 | language ? '"' : ')'); | |
98994639 HS |
828 | #endif |
829 | ||
49c85077 KW |
830 | PerlIO_printf(Perl_error_log, |
831 | "\tLC_ALL = %c%s%c,\n", | |
832 | lc_all ? '"' : '(', | |
833 | lc_all ? lc_all : "unset", | |
834 | lc_all ? '"' : ')'); | |
98994639 HS |
835 | |
836 | #if defined(USE_ENVIRON_ARRAY) | |
49c85077 KW |
837 | { |
838 | char **e; | |
839 | for (e = environ; *e; e++) { | |
840 | if (strnEQ(*e, "LC_", 3) | |
841 | && strnNE(*e, "LC_ALL=", 7) | |
842 | && (p = strchr(*e, '='))) | |
843 | PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n", | |
844 | (int)(p - *e), *e, p + 1); | |
845 | } | |
846 | } | |
98994639 | 847 | #else |
49c85077 KW |
848 | PerlIO_printf(Perl_error_log, |
849 | "\t(possibly more locale environment variables)\n"); | |
98994639 HS |
850 | #endif |
851 | ||
49c85077 KW |
852 | PerlIO_printf(Perl_error_log, |
853 | "\tLANG = %c%s%c\n", | |
854 | lang ? '"' : '(', | |
855 | lang ? lang : "unset", | |
856 | lang ? '"' : ')'); | |
98994639 | 857 | |
49c85077 KW |
858 | PerlIO_printf(Perl_error_log, |
859 | " are supported and installed on your system.\n"); | |
860 | } | |
98994639 | 861 | |
65ebb059 KW |
862 | /* Calculate what fallback locales to try. We have avoided this |
863 | * until we have to, becuase failure is quite unlikely. This will | |
864 | * usually change the upper bound of the loop we are in. | |
865 | * | |
866 | * Since the system's default way of setting the locale has not | |
867 | * found one that works, We use Perl's defined ordering: LC_ALL, | |
868 | * LANG, and the C locale. We don't try the same locale twice, so | |
869 | * don't add to the list if already there. (On POSIX systems, the | |
870 | * LC_ALL element will likely be a repeat of the 0th element "", | |
871 | * but there's no harm done by doing it explicitly */ | |
872 | if (lc_all) { | |
873 | for (j = 0; j < trial_locales_count; j++) { | |
874 | if (strEQ(lc_all, trial_locales[j])) { | |
875 | goto done_lc_all; | |
876 | } | |
877 | } | |
878 | trial_locales[trial_locales_count++] = lc_all; | |
879 | } | |
880 | done_lc_all: | |
98994639 | 881 | |
65ebb059 KW |
882 | if (lang) { |
883 | for (j = 0; j < trial_locales_count; j++) { | |
884 | if (strEQ(lang, trial_locales[j])) { | |
885 | goto done_lang; | |
886 | } | |
887 | } | |
888 | trial_locales[trial_locales_count++] = lang; | |
889 | } | |
890 | done_lang: | |
891 | ||
892 | #if defined(WIN32) && defined(LC_ALL) | |
893 | /* For Windows, we also try the system default locale before "C". | |
894 | * (If there exists a Windows without LC_ALL we skip this because | |
895 | * it gets too complicated. For those, the "C" is the next | |
896 | * fallback possibility). The "" is the same as the 0th element of | |
897 | * the array, but the code at the loop above knows to treat it | |
898 | * differently when not the 0th */ | |
899 | trial_locales[trial_locales_count++] = ""; | |
900 | #endif | |
901 | ||
902 | for (j = 0; j < trial_locales_count; j++) { | |
903 | if (strEQ("C", trial_locales[j])) { | |
904 | goto done_C; | |
905 | } | |
906 | } | |
907 | trial_locales[trial_locales_count++] = "C"; | |
98994639 | 908 | |
65ebb059 KW |
909 | done_C: ; |
910 | } /* end of first time through the loop */ | |
98994639 | 911 | |
65ebb059 KW |
912 | #ifdef WIN32 |
913 | next_iteration: ; | |
914 | #endif | |
915 | ||
916 | } /* end of looping through the trial locales */ | |
917 | ||
918 | if (ok < 1) { /* If we tried to fallback */ | |
919 | const char* msg; | |
920 | if (! setlocale_failure) { /* fallback succeeded */ | |
921 | msg = "Falling back to"; | |
922 | } | |
923 | else { /* fallback failed */ | |
98994639 | 924 | |
65ebb059 KW |
925 | /* We dropped off the end of the loop, so have to decrement i to |
926 | * get back to the value the last time through */ | |
927 | i--; | |
98994639 | 928 | |
65ebb059 KW |
929 | ok = -1; |
930 | msg = "Failed to fall back to"; | |
931 | ||
932 | /* To continue, we should use whatever values we've got */ | |
98994639 | 933 | #ifdef USE_LOCALE_CTYPE |
49c85077 KW |
934 | Safefree(curctype); |
935 | curctype = savepv(setlocale(LC_CTYPE, NULL)); | |
98994639 HS |
936 | #endif /* USE_LOCALE_CTYPE */ |
937 | #ifdef USE_LOCALE_COLLATE | |
49c85077 KW |
938 | Safefree(curcoll); |
939 | curcoll = savepv(setlocale(LC_COLLATE, NULL)); | |
98994639 HS |
940 | #endif /* USE_LOCALE_COLLATE */ |
941 | #ifdef USE_LOCALE_NUMERIC | |
49c85077 KW |
942 | Safefree(curnum); |
943 | curnum = savepv(setlocale(LC_NUMERIC, NULL)); | |
98994639 | 944 | #endif /* USE_LOCALE_NUMERIC */ |
65ebb059 KW |
945 | } |
946 | ||
947 | if (locwarn) { | |
948 | const char * description; | |
949 | const char * name = ""; | |
950 | if (strEQ(trial_locales[i], "C")) { | |
951 | description = "the standard locale"; | |
952 | name = "C"; | |
953 | } | |
6bce99ee | 954 | #ifdef SYSTEM_DEFAULT_LOCALE |
65ebb059 KW |
955 | else if (strEQ(trial_locales[i], "")) { |
956 | description = "the system default locale"; | |
957 | if (system_default_locale) { | |
958 | name = system_default_locale; | |
959 | } | |
960 | } | |
6bce99ee | 961 | #endif /* SYSTEM_DEFAULT_LOCALE */ |
65ebb059 KW |
962 | else { |
963 | description = "a fallback locale"; | |
964 | name = trial_locales[i]; | |
965 | } | |
966 | if (name && strNE(name, "")) { | |
967 | PerlIO_printf(Perl_error_log, | |
968 | "perl: warning: %s %s (\"%s\").\n", msg, description, name); | |
969 | } | |
970 | else { | |
971 | PerlIO_printf(Perl_error_log, | |
972 | "perl: warning: %s %s.\n", msg, description); | |
973 | } | |
974 | } | |
975 | } /* End of tried to fallback */ | |
98994639 HS |
976 | |
977 | #ifdef USE_LOCALE_CTYPE | |
978 | new_ctype(curctype); | |
979 | #endif /* USE_LOCALE_CTYPE */ | |
980 | ||
981 | #ifdef USE_LOCALE_COLLATE | |
982 | new_collate(curcoll); | |
983 | #endif /* USE_LOCALE_COLLATE */ | |
984 | ||
985 | #ifdef USE_LOCALE_NUMERIC | |
986 | new_numeric(curnum); | |
987 | #endif /* USE_LOCALE_NUMERIC */ | |
b310b053 | 988 | |
8ef6e574 | 989 | #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE) |
49c85077 KW |
990 | /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE |
991 | * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by | |
992 | * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the | |
993 | * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open | |
994 | * discipline. */ | |
c1284011 | 995 | PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE); |
49c85077 | 996 | |
a05d7ebb | 997 | /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO. |
fde18df1 JH |
998 | This is an alternative to using the -C command line switch |
999 | (the -C if present will override this). */ | |
1000 | { | |
dd374669 | 1001 | const char *p = PerlEnv_getenv("PERL_UNICODE"); |
a05d7ebb | 1002 | PL_unicode = p ? parse_unicode_opts(&p) : 0; |
5a22a2bb NC |
1003 | if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG) |
1004 | PL_utf8cache = -1; | |
b310b053 | 1005 | } |
ec71e770 | 1006 | #endif |
b310b053 | 1007 | |
98994639 | 1008 | #ifdef USE_LOCALE_CTYPE |
43c5f42d | 1009 | Safefree(curctype); |
98994639 HS |
1010 | #endif /* USE_LOCALE_CTYPE */ |
1011 | #ifdef USE_LOCALE_COLLATE | |
43c5f42d | 1012 | Safefree(curcoll); |
98994639 HS |
1013 | #endif /* USE_LOCALE_COLLATE */ |
1014 | #ifdef USE_LOCALE_NUMERIC | |
43c5f42d | 1015 | Safefree(curnum); |
98994639 | 1016 | #endif /* USE_LOCALE_NUMERIC */ |
8ef6e574 | 1017 | |
f2ce9e1c JH |
1018 | #else /* !USE_LOCALE */ |
1019 | PERL_UNUSED_ARG(printwarn); | |
8ef6e574 KW |
1020 | #endif /* USE_LOCALE */ |
1021 | ||
98994639 HS |
1022 | return ok; |
1023 | } | |
1024 | ||
49c85077 | 1025 | |
98994639 HS |
1026 | #ifdef USE_LOCALE_COLLATE |
1027 | ||
1028 | /* | |
1029 | * mem_collxfrm() is a bit like strxfrm() but with two important | |
1030 | * differences. First, it handles embedded NULs. Second, it allocates | |
1031 | * a bit more memory than needed for the transformed data itself. | |
1032 | * The real transformed data begins at offset sizeof(collationix). | |
1033 | * Please see sv_collxfrm() to see how this is used. | |
1034 | */ | |
166f8a29 | 1035 | |
98994639 HS |
1036 | char * |
1037 | Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen) | |
1038 | { | |
1039 | char *xbuf; | |
1040 | STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */ | |
1041 | ||
7918f24d NC |
1042 | PERL_ARGS_ASSERT_MEM_COLLXFRM; |
1043 | ||
98994639 HS |
1044 | /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */ |
1045 | /* the +1 is for the terminating NUL. */ | |
1046 | ||
1047 | xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1; | |
a02a5408 | 1048 | Newx(xbuf, xAlloc, char); |
98994639 HS |
1049 | if (! xbuf) |
1050 | goto bad; | |
1051 | ||
1052 | *(U32*)xbuf = PL_collation_ix; | |
1053 | xout = sizeof(PL_collation_ix); | |
1054 | for (xin = 0; xin < len; ) { | |
224e8ef5 | 1055 | Size_t xused; |
98994639 HS |
1056 | |
1057 | for (;;) { | |
1058 | xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout); | |
224e8ef5 | 1059 | if (xused >= PERL_INT_MAX) |
98994639 | 1060 | goto bad; |
eb160463 | 1061 | if ((STRLEN)xused < xAlloc - xout) |
98994639 HS |
1062 | break; |
1063 | xAlloc = (2 * xAlloc) + 1; | |
1064 | Renew(xbuf, xAlloc, char); | |
1065 | if (! xbuf) | |
1066 | goto bad; | |
1067 | } | |
1068 | ||
1069 | xin += strlen(s + xin) + 1; | |
1070 | xout += xused; | |
1071 | ||
1072 | /* Embedded NULs are understood but silently skipped | |
1073 | * because they make no sense in locale collation. */ | |
1074 | } | |
1075 | ||
1076 | xbuf[xout] = '\0'; | |
1077 | *xlen = xout - sizeof(PL_collation_ix); | |
1078 | return xbuf; | |
1079 | ||
1080 | bad: | |
1081 | Safefree(xbuf); | |
1082 | *xlen = 0; | |
1083 | return NULL; | |
1084 | } | |
1085 | ||
1086 | #endif /* USE_LOCALE_COLLATE */ | |
1087 | ||
8ef6e574 KW |
1088 | #ifdef USE_LOCALE |
1089 | ||
c1284011 KW |
1090 | bool |
1091 | Perl__is_cur_LC_category_utf8(pTHX_ int category) | |
7d74bb61 KW |
1092 | { |
1093 | /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE | |
1094 | * otherwise. 'category' may not be LC_ALL. If the platform doesn't have | |
119ee68b | 1095 | * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence |
609548d2 KW |
1096 | * could give the wrong result. The result will very likely be correct for |
1097 | * languages that have commonly used non-ASCII characters, but for notably | |
1098 | * English, it comes down to if the locale's name ends in something like | |
1099 | * "UTF-8". It errs on the side of not being a UTF-8 locale. */ | |
7d74bb61 KW |
1100 | |
1101 | char *save_input_locale = NULL; | |
7d74bb61 KW |
1102 | STRLEN final_pos; |
1103 | ||
8ef6e574 | 1104 | #ifdef LC_ALL |
7d74bb61 | 1105 | assert(category != LC_ALL); |
8ef6e574 | 1106 | #endif |
7d74bb61 KW |
1107 | |
1108 | /* First dispose of the trivial cases */ | |
b07fffd1 | 1109 | save_input_locale = setlocale(category, NULL); |
7d74bb61 | 1110 | if (! save_input_locale) { |
69014004 KW |
1111 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1112 | "Could not find current locale for category %d\n", | |
1113 | category)); | |
7d74bb61 KW |
1114 | return FALSE; /* XXX maybe should croak */ |
1115 | } | |
b07fffd1 | 1116 | save_input_locale = stdize_locale(savepv(save_input_locale)); |
a39edc4c | 1117 | if (isNAME_C_OR_POSIX(save_input_locale)) { |
69014004 KW |
1118 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1119 | "Current locale for category %d is %s\n", | |
1120 | category, save_input_locale)); | |
b07fffd1 | 1121 | Safefree(save_input_locale); |
7d74bb61 KW |
1122 | return FALSE; |
1123 | } | |
1124 | ||
1d958db2 KW |
1125 | #if defined(USE_LOCALE_CTYPE) \ |
1126 | && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET))) | |
7d74bb61 | 1127 | |
1d958db2 | 1128 | { /* Next try nl_langinfo or MB_CUR_MAX if available */ |
7d74bb61 KW |
1129 | |
1130 | char *save_ctype_locale = NULL; | |
119ee68b | 1131 | bool is_utf8; |
7d74bb61 | 1132 | |
119ee68b | 1133 | if (category != LC_CTYPE) { /* These work only on LC_CTYPE */ |
7d74bb61 KW |
1134 | |
1135 | /* Get the current LC_CTYPE locale */ | |
4f72bb37 | 1136 | save_ctype_locale = setlocale(LC_CTYPE, NULL); |
7d74bb61 | 1137 | if (! save_ctype_locale) { |
69014004 KW |
1138 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1139 | "Could not find current locale for LC_CTYPE\n")); | |
7d74bb61 KW |
1140 | goto cant_use_nllanginfo; |
1141 | } | |
4f72bb37 | 1142 | save_ctype_locale = stdize_locale(savepv(save_ctype_locale)); |
7d74bb61 KW |
1143 | |
1144 | /* If LC_CTYPE and the desired category use the same locale, this | |
1145 | * means that finding the value for LC_CTYPE is the same as finding | |
1146 | * the value for the desired category. Otherwise, switch LC_CTYPE | |
1147 | * to the desired category's locale */ | |
1148 | if (strEQ(save_ctype_locale, save_input_locale)) { | |
1149 | Safefree(save_ctype_locale); | |
1150 | save_ctype_locale = NULL; | |
1151 | } | |
1152 | else if (! setlocale(LC_CTYPE, save_input_locale)) { | |
69014004 KW |
1153 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1154 | "Could not change LC_CTYPE locale to %s\n", | |
1155 | save_input_locale)); | |
7d74bb61 KW |
1156 | Safefree(save_ctype_locale); |
1157 | goto cant_use_nllanginfo; | |
1158 | } | |
1159 | } | |
1160 | ||
69014004 KW |
1161 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n", |
1162 | save_input_locale)); | |
1163 | ||
7d74bb61 | 1164 | /* Here the current LC_CTYPE is set to the locale of the category whose |
1d958db2 KW |
1165 | * information is desired. This means that nl_langinfo() and MB_CUR_MAX |
1166 | * should give the correct results */ | |
119ee68b | 1167 | |
1d958db2 KW |
1168 | # if defined(HAS_NL_LANGINFO) && defined(CODESET) |
1169 | { | |
4f72bb37 | 1170 | char *codeset = nl_langinfo(CODESET); |
1d958db2 | 1171 | if (codeset && strNE(codeset, "")) { |
4f72bb37 | 1172 | codeset = savepv(codeset); |
119ee68b | 1173 | |
1d958db2 KW |
1174 | /* If we switched LC_CTYPE, switch back */ |
1175 | if (save_ctype_locale) { | |
1176 | setlocale(LC_CTYPE, save_ctype_locale); | |
1177 | Safefree(save_ctype_locale); | |
1178 | } | |
1179 | ||
1180 | is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8")) | |
1181 | || foldEQ(codeset, STR_WITH_LEN("UTF8")); | |
1182 | ||
69014004 KW |
1183 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1184 | "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n", | |
1185 | codeset, is_utf8)); | |
1d958db2 KW |
1186 | Safefree(codeset); |
1187 | Safefree(save_input_locale); | |
1188 | return is_utf8; | |
1189 | } | |
119ee68b KW |
1190 | } |
1191 | ||
1d958db2 KW |
1192 | # endif |
1193 | # ifdef MB_CUR_MAX | |
1194 | ||
1195 | /* Here, either we don't have nl_langinfo, or it didn't return a | |
1196 | * codeset. Try MB_CUR_MAX */ | |
1197 | ||
119ee68b KW |
1198 | /* Standard UTF-8 needs at least 4 bytes to represent the maximum |
1199 | * Unicode code point. Since UTF-8 is the only non-single byte | |
1200 | * encoding we handle, we just say any such encoding is UTF-8, and if | |
1201 | * turns out to be wrong, other things will fail */ | |
1202 | is_utf8 = MB_CUR_MAX >= 4; | |
1203 | ||
69014004 KW |
1204 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1205 | "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n", | |
1206 | (int) MB_CUR_MAX, is_utf8)); | |
1207 | ||
119ee68b KW |
1208 | Safefree(save_input_locale); |
1209 | ||
1210 | # ifdef HAS_MBTOWC | |
1211 | ||
1212 | /* ... But, most system that have MB_CUR_MAX will also have mbtowc(), | |
1213 | * since they are both in the C99 standard. We can feed a known byte | |
1214 | * string to the latter function, and check that it gives the expected | |
1215 | * result */ | |
1216 | if (is_utf8) { | |
1217 | wchar_t wc; | |
856b881c | 1218 | PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */ |
69014004 | 1219 | errno = 0; |
f019f68f | 1220 | if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)) |
119ee68b KW |
1221 | != strlen(HYPHEN_UTF8) |
1222 | || wc != (wchar_t) 0x2010) | |
1223 | { | |
1224 | is_utf8 = FALSE; | |
abdcbdb8 | 1225 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc)); |
69014004 KW |
1226 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1227 | "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n", | |
1228 | mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno)); | |
119ee68b KW |
1229 | } |
1230 | } | |
119ee68b KW |
1231 | # endif |
1232 | ||
1d958db2 KW |
1233 | /* If we switched LC_CTYPE, switch back */ |
1234 | if (save_ctype_locale) { | |
1235 | setlocale(LC_CTYPE, save_ctype_locale); | |
1236 | Safefree(save_ctype_locale); | |
119ee68b | 1237 | } |
7d74bb61 | 1238 | |
1d958db2 | 1239 | return is_utf8; |
119ee68b | 1240 | # endif |
7d74bb61 | 1241 | } |
119ee68b | 1242 | |
7d74bb61 KW |
1243 | cant_use_nllanginfo: |
1244 | ||
0080c90a KW |
1245 | #else /* nl_langinfo should work if available, so don't bother compiling this |
1246 | fallback code. The final fallback of looking at the name is | |
1247 | compiled, and will be executed if nl_langinfo fails */ | |
7d74bb61 | 1248 | |
97f4de96 KW |
1249 | /* nl_langinfo not available or failed somehow. Next try looking at the |
1250 | * currency symbol to see if it disambiguates things. Often that will be | |
1251 | * in the native script, and if the symbol isn't in UTF-8, we know that the | |
1252 | * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is | |
609548d2 KW |
1253 | * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small |
1254 | * */ | |
fa9b773e KW |
1255 | |
1256 | #ifdef HAS_LOCALECONV | |
fa9b773e | 1257 | # ifdef USE_LOCALE_MONETARY |
fa9b773e KW |
1258 | { |
1259 | char *save_monetary_locale = NULL; | |
fa9b773e | 1260 | bool only_ascii = FALSE; |
13542a67 KW |
1261 | bool is_utf8 = FALSE; |
1262 | struct lconv* lc; | |
fa9b773e | 1263 | |
97f4de96 KW |
1264 | /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of |
1265 | * the desired category, if it isn't that locale already */ | |
1266 | ||
fa9b773e KW |
1267 | if (category != LC_MONETARY) { |
1268 | ||
4f72bb37 | 1269 | save_monetary_locale = setlocale(LC_MONETARY, NULL); |
fa9b773e | 1270 | if (! save_monetary_locale) { |
69014004 KW |
1271 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1272 | "Could not find current locale for LC_MONETARY\n")); | |
fa9b773e KW |
1273 | goto cant_use_monetary; |
1274 | } | |
4f72bb37 | 1275 | save_monetary_locale = stdize_locale(savepv(save_monetary_locale)); |
fa9b773e | 1276 | |
13542a67 KW |
1277 | if (strEQ(save_monetary_locale, save_input_locale)) { |
1278 | Safefree(save_monetary_locale); | |
1279 | save_monetary_locale = NULL; | |
1280 | } | |
1281 | else if (! setlocale(LC_MONETARY, save_input_locale)) { | |
59c234b4 KW |
1282 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1283 | "Could not change LC_MONETARY locale to %s\n", | |
1284 | save_input_locale)); | |
1285 | Safefree(save_monetary_locale); | |
1286 | goto cant_use_monetary; | |
fa9b773e KW |
1287 | } |
1288 | } | |
1289 | ||
1290 | /* Here the current LC_MONETARY is set to the locale of the category | |
1291 | * whose information is desired. */ | |
1292 | ||
13542a67 KW |
1293 | lc = localeconv(); |
1294 | if (! lc | |
1295 | || ! lc->currency_symbol | |
1296 | || is_ascii_string((U8 *) lc->currency_symbol, 0)) | |
1297 | { | |
1298 | DEBUG_L(PerlIO_printf(Perl_debug_log, "Couldn't get currency symbol for %s, or contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); | |
1299 | only_ascii = TRUE; | |
1300 | } | |
1301 | else { | |
1302 | is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0); | |
fa9b773e KW |
1303 | } |
1304 | ||
1305 | /* If we changed it, restore LC_MONETARY to its original locale */ | |
1306 | if (save_monetary_locale) { | |
1307 | setlocale(LC_MONETARY, save_monetary_locale); | |
1308 | Safefree(save_monetary_locale); | |
1309 | } | |
1310 | ||
13542a67 | 1311 | if (! only_ascii) { |
fa9b773e | 1312 | |
59c234b4 KW |
1313 | /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; |
1314 | * otherwise assume the locale is UTF-8 if and only if the symbol | |
1315 | * is non-ascii UTF-8. */ | |
1316 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n", | |
1317 | save_input_locale, is_utf8)); | |
1318 | Safefree(save_input_locale); | |
1319 | return is_utf8; | |
13542a67 | 1320 | } |
fa9b773e KW |
1321 | } |
1322 | cant_use_monetary: | |
1323 | ||
1324 | # endif /* USE_LOCALE_MONETARY */ | |
1325 | #endif /* HAS_LOCALECONV */ | |
1326 | ||
15f7e74e KW |
1327 | #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME) |
1328 | ||
1329 | /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try | |
1330 | * the names of the months and weekdays, timezone, and am/pm indicator */ | |
1331 | { | |
1332 | char *save_time_locale = NULL; | |
1333 | int hour = 10; | |
1334 | bool is_dst = FALSE; | |
1335 | int dom = 1; | |
1336 | int month = 0; | |
1337 | int i; | |
1338 | char * formatted_time; | |
1339 | ||
1340 | ||
1341 | /* Like above for LC_MONETARY, we set LC_TIME to the locale of the | |
1342 | * desired category, if it isn't that locale already */ | |
1343 | ||
1344 | if (category != LC_TIME) { | |
1345 | ||
1346 | save_time_locale = setlocale(LC_TIME, NULL); | |
1347 | if (! save_time_locale) { | |
1348 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1349 | "Could not find current locale for LC_TIME\n")); | |
1350 | goto cant_use_time; | |
1351 | } | |
1352 | save_time_locale = stdize_locale(savepv(save_time_locale)); | |
1353 | ||
1354 | if (strEQ(save_time_locale, save_input_locale)) { | |
1355 | Safefree(save_time_locale); | |
1356 | save_time_locale = NULL; | |
1357 | } | |
1358 | else if (! setlocale(LC_TIME, save_input_locale)) { | |
1359 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1360 | "Could not change LC_TIME locale to %s\n", | |
1361 | save_input_locale)); | |
1362 | Safefree(save_time_locale); | |
1363 | goto cant_use_time; | |
1364 | } | |
1365 | } | |
1366 | ||
1367 | /* Here the current LC_TIME is set to the locale of the category | |
1368 | * whose information is desired. Look at all the days of the week and | |
1369 | * month names, and the timezone and am/pm indicator for non-ASCII | |
1370 | * characters. The first such a one found will tell us if the locale | |
1371 | * is UTF-8 or not */ | |
1372 | ||
1373 | for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */ | |
1374 | formatted_time = my_strftime("%A %B %Z %p", | |
1375 | 0, 0, hour, dom, month, 112, 0, 0, is_dst); | |
1376 | if (! formatted_time || is_ascii_string((U8 *) formatted_time, 0)) { | |
1377 | ||
1378 | /* Here, we didn't find a non-ASCII. Try the next time through | |
1379 | * with the complemented dst and am/pm, and try with the next | |
1380 | * weekday. After we have gotten all weekdays, try the next | |
1381 | * month */ | |
1382 | is_dst = ! is_dst; | |
1383 | hour = (hour + 12) % 24; | |
1384 | dom++; | |
1385 | if (i > 6) { | |
1386 | month++; | |
1387 | } | |
1388 | continue; | |
1389 | } | |
1390 | ||
1391 | /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8; | |
1392 | * false otherwise. But first, restore LC_TIME to its original | |
1393 | * locale if we changed it */ | |
1394 | if (save_time_locale) { | |
1395 | setlocale(LC_TIME, save_time_locale); | |
1396 | Safefree(save_time_locale); | |
1397 | } | |
1398 | ||
1399 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n", | |
1400 | save_input_locale, | |
1401 | is_utf8_string((U8 *) formatted_time, 0))); | |
1402 | Safefree(save_input_locale); | |
1403 | return is_utf8_string((U8 *) formatted_time, 0); | |
1404 | } | |
1405 | ||
1406 | /* Falling off the end of the loop indicates all the names were just | |
1407 | * ASCII. Go on to the next test. If we changed it, restore LC_TIME | |
1408 | * to its original locale */ | |
1409 | if (save_time_locale) { | |
1410 | setlocale(LC_TIME, save_time_locale); | |
1411 | Safefree(save_time_locale); | |
1412 | } | |
1413 | DEBUG_L(PerlIO_printf(Perl_debug_log, "All time-related words for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); | |
1414 | } | |
1415 | cant_use_time: | |
1416 | ||
1417 | #endif | |
1418 | ||
1419 | #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST) | |
855aeb93 JH |
1420 | |
1421 | /* This code is ifdefd out because it was found to not be necessary in testing | |
5857e934 KW |
1422 | * on our dromedary test machine, which has over 700 locales. There, this |
1423 | * added no value to looking at the currency symbol and the time strings. I | |
1424 | * left it in so as to avoid rewriting it if real-world experience indicates | |
1425 | * that dromedary is an outlier. Essentially, instead of returning abpve if we | |
855aeb93 JH |
1426 | * haven't found illegal utf8, we continue on and examine all the strerror() |
1427 | * messages on the platform for utf8ness. If all are ASCII, we still don't | |
1428 | * know the answer; but otherwise we have a pretty good indication of the | |
5857e934 KW |
1429 | * utf8ness. The reason this doesn't help much is that the messages may not |
1430 | * have been translated into the locale. The currency symbol and time strings | |
1431 | * are much more likely to have been translated. */ | |
1432 | { | |
855aeb93 | 1433 | int e; |
5857e934 KW |
1434 | bool is_utf8 = FALSE; |
1435 | bool non_ascii = FALSE; | |
855aeb93 | 1436 | char *save_messages_locale = NULL; |
5857e934 | 1437 | const char * errmsg = NULL; |
855aeb93 | 1438 | |
5857e934 KW |
1439 | /* Like above, we set LC_MESSAGES to the locale of the desired |
1440 | * category, if it isn't that locale already */ | |
855aeb93 JH |
1441 | |
1442 | if (category != LC_MESSAGES) { | |
1443 | ||
5857e934 | 1444 | save_messages_locale = setlocale(LC_MESSAGES, NULL); |
855aeb93 | 1445 | if (! save_messages_locale) { |
5857e934 KW |
1446 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1447 | "Could not find current locale for LC_MESSAGES\n")); | |
855aeb93 JH |
1448 | goto cant_use_messages; |
1449 | } | |
5857e934 | 1450 | save_messages_locale = stdize_locale(savepv(save_messages_locale)); |
855aeb93 JH |
1451 | |
1452 | if (strEQ(save_messages_locale, save_input_locale)) { | |
5857e934 KW |
1453 | Safefree(save_messages_locale); |
1454 | save_messages_locale = NULL; | |
855aeb93 JH |
1455 | } |
1456 | else if (! setlocale(LC_MESSAGES, save_input_locale)) { | |
5857e934 KW |
1457 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1458 | "Could not change LC_MESSAGES locale to %s\n", | |
1459 | save_input_locale)); | |
855aeb93 JH |
1460 | Safefree(save_messages_locale); |
1461 | goto cant_use_messages; | |
1462 | } | |
1463 | } | |
1464 | ||
1465 | /* Here the current LC_MESSAGES is set to the locale of the category | |
5857e934 KW |
1466 | * whose information is desired. Look through all the messages. We |
1467 | * can't use Strerror() here because it may expand to code that | |
1468 | * segfaults in miniperl */ | |
855aeb93 | 1469 | |
5857e934 KW |
1470 | for (e = 0; e <= sys_nerr; e++) { |
1471 | errno = 0; | |
1472 | errmsg = sys_errlist[e]; | |
1473 | if (errno || !errmsg) { | |
855aeb93 JH |
1474 | break; |
1475 | } | |
5857e934 KW |
1476 | errmsg = savepv(errmsg); |
1477 | if (! is_ascii_string((U8 *) errmsg, 0)) { | |
1478 | non_ascii = TRUE; | |
1479 | is_utf8 = is_utf8_string((U8 *) errmsg, 0); | |
1480 | break; | |
855aeb93 JH |
1481 | } |
1482 | } | |
5857e934 | 1483 | Safefree(errmsg); |
855aeb93 JH |
1484 | |
1485 | /* And, if we changed it, restore LC_MESSAGES to its original locale */ | |
1486 | if (save_messages_locale) { | |
1487 | setlocale(LC_MESSAGES, save_messages_locale); | |
1488 | Safefree(save_messages_locale); | |
1489 | } | |
1490 | ||
5857e934 KW |
1491 | if (non_ascii) { |
1492 | ||
1493 | /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid, | |
1494 | * any non-ascii means it is one; otherwise we assume it isn't */ | |
1495 | DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n", | |
1496 | save_input_locale, | |
1497 | is_utf8)); | |
1498 | Safefree(save_input_locale); | |
1499 | return is_utf8; | |
1500 | } | |
855aeb93 | 1501 | |
5857e934 | 1502 | DEBUG_L(PerlIO_printf(Perl_debug_log, "All error messages for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale)); |
855aeb93 JH |
1503 | } |
1504 | cant_use_messages: | |
1505 | ||
1506 | #endif | |
fa9b773e | 1507 | |
0080c90a KW |
1508 | #endif /* the code that is compiled when no nl_langinfo */ |
1509 | ||
92c0a900 KW |
1510 | #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a |
1511 | UTF-8 locale */ | |
97f4de96 KW |
1512 | /* As a last resort, look at the locale name to see if it matches |
1513 | * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the | |
1514 | * return of setlocale(), is actually defined to be opaque, so we can't | |
1515 | * really rely on the absence of various substrings in the name to indicate | |
1516 | * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to | |
1517 | * be a UTF-8 locale. Similarly for the other common names */ | |
1518 | ||
1519 | final_pos = strlen(save_input_locale) - 1; | |
1520 | if (final_pos >= 3) { | |
1521 | char *name = save_input_locale; | |
1522 | ||
1523 | /* Find next 'U' or 'u' and look from there */ | |
1524 | while ((name += strcspn(name, "Uu") + 1) | |
1525 | <= save_input_locale + final_pos - 2) | |
1526 | { | |
305b8651 KW |
1527 | if (!isALPHA_FOLD_NE(*name, 't') |
1528 | || isALPHA_FOLD_NE(*(name + 1), 'f')) | |
97f4de96 KW |
1529 | { |
1530 | continue; | |
1531 | } | |
1532 | name += 2; | |
1533 | if (*(name) == '-') { | |
1534 | if ((name > save_input_locale + final_pos - 1)) { | |
1535 | break; | |
1536 | } | |
1537 | name++; | |
1538 | } | |
1539 | if (*(name) == '8') { | |
97f4de96 KW |
1540 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1541 | "Locale %s ends with UTF-8 in name\n", | |
1542 | save_input_locale)); | |
00c54b9c | 1543 | Safefree(save_input_locale); |
97f4de96 KW |
1544 | return TRUE; |
1545 | } | |
1546 | } | |
1547 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1548 | "Locale %s doesn't end with UTF-8 in name\n", | |
1549 | save_input_locale)); | |
1550 | } | |
92c0a900 | 1551 | #endif |
97f4de96 KW |
1552 | |
1553 | #ifdef WIN32 | |
1554 | /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */ | |
1555 | if (final_pos >= 4 | |
1556 | && *(save_input_locale + final_pos - 0) == '1' | |
1557 | && *(save_input_locale + final_pos - 1) == '0' | |
1558 | && *(save_input_locale + final_pos - 2) == '0' | |
1559 | && *(save_input_locale + final_pos - 3) == '5' | |
1560 | && *(save_input_locale + final_pos - 4) == '6') | |
1561 | { | |
1562 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1563 | "Locale %s ends with 10056 in name, is UTF-8 locale\n", | |
1564 | save_input_locale)); | |
1565 | Safefree(save_input_locale); | |
1566 | return TRUE; | |
1567 | } | |
1568 | #endif | |
1569 | ||
1570 | /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But | |
1571 | * since we are about to return FALSE anyway, there is no point in doing | |
1572 | * this extra work */ | |
1573 | #if 0 | |
1574 | if (instr(save_input_locale, "8859")) { | |
1575 | DEBUG_L(PerlIO_printf(Perl_debug_log, | |
1576 | "Locale %s has 8859 in name, not UTF-8 locale\n", | |
1577 | save_input_locale)); | |
1578 | Safefree(save_input_locale); | |
1579 | return FALSE; | |
1580 | } | |
1581 | #endif | |
1582 | ||
69014004 KW |
1583 | DEBUG_L(PerlIO_printf(Perl_debug_log, |
1584 | "Assuming locale %s is not a UTF-8 locale\n", | |
1585 | save_input_locale)); | |
fa9b773e | 1586 | Safefree(save_input_locale); |
7d74bb61 KW |
1587 | return FALSE; |
1588 | } | |
1589 | ||
8ef6e574 | 1590 | #endif |
7d74bb61 | 1591 | |
d6ded950 KW |
1592 | |
1593 | bool | |
1594 | Perl__is_in_locale_category(pTHX_ const bool compiling, const int category) | |
1595 | { | |
1a4f13e1 | 1596 | dVAR; |
d6ded950 KW |
1597 | /* Internal function which returns if we are in the scope of a pragma that |
1598 | * enables the locale category 'category'. 'compiling' should indicate if | |
1599 | * this is during the compilation phase (TRUE) or not (FALSE). */ | |
1600 | ||
1601 | const COP * const cop = (compiling) ? &PL_compiling : PL_curcop; | |
1602 | ||
1603 | SV *categories = cop_hints_fetch_pvs(cop, "locale", 0); | |
1604 | if (! categories || categories == &PL_sv_placeholder) { | |
1605 | return FALSE; | |
1606 | } | |
1607 | ||
1608 | /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get | |
1609 | * a valid unsigned */ | |
1610 | assert(category >= -1); | |
1611 | return cBOOL(SvUV(categories) & (1U << (category + 1))); | |
1612 | } | |
1613 | ||
2c6ee1a7 KW |
1614 | char * |
1615 | Perl_my_strerror(pTHX_ const int errnum) { | |
1616 | ||
1617 | /* Uses C locale for the error text unless within scope of 'use locale' for | |
1618 | * LC_MESSAGES */ | |
1619 | ||
1620 | #ifdef USE_LOCALE_MESSAGES | |
8ee4b769 | 1621 | if (! IN_LC(LC_MESSAGES)) { |
2c6ee1a7 | 1622 | char * save_locale = setlocale(LC_MESSAGES, NULL); |
a39edc4c | 1623 | if (! isNAME_C_OR_POSIX(save_locale)) { |
2c6ee1a7 KW |
1624 | char *errstr; |
1625 | ||
1626 | /* The next setlocale likely will zap this, so create a copy */ | |
1627 | save_locale = savepv(save_locale); | |
1628 | ||
1629 | setlocale(LC_MESSAGES, "C"); | |
1630 | ||
1631 | /* This points to the static space in Strerror, with all its | |
1632 | * limitations */ | |
1633 | errstr = Strerror(errnum); | |
1634 | ||
1635 | setlocale(LC_MESSAGES, save_locale); | |
1636 | Safefree(save_locale); | |
1637 | return errstr; | |
1638 | } | |
1639 | } | |
1640 | #endif | |
1641 | ||
1642 | return Strerror(errnum); | |
1643 | } | |
1644 | ||
66610fdd | 1645 | /* |
747c467a KW |
1646 | |
1647 | =head1 Locale-related functions and macros | |
1648 | ||
1649 | =for apidoc sync_locale | |
1650 | ||
1651 | Changing the program's locale should be avoided by XS code. Nevertheless, | |
1652 | certain non-Perl libraries called from XS, such as C<Gtk> do so. When this | |
1653 | happens, Perl needs to be told that the locale has changed. Use this function | |
1654 | to do so, before returning to Perl. | |
1655 | ||
1656 | =cut | |
1657 | */ | |
1658 | ||
1659 | void | |
1660 | Perl_sync_locale(pTHX) | |
1661 | { | |
1662 | ||
1663 | #ifdef USE_LOCALE_CTYPE | |
1664 | new_ctype(setlocale(LC_CTYPE, NULL)); | |
1665 | #endif /* USE_LOCALE_CTYPE */ | |
1666 | ||
1667 | #ifdef USE_LOCALE_COLLATE | |
1668 | new_collate(setlocale(LC_COLLATE, NULL)); | |
1669 | #endif | |
1670 | ||
1671 | #ifdef USE_LOCALE_NUMERIC | |
1672 | set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */ | |
1673 | new_numeric(setlocale(LC_NUMERIC, NULL)); | |
1674 | #endif /* USE_LOCALE_NUMERIC */ | |
1675 | ||
1676 | } | |
1677 | ||
1678 | ||
1679 | ||
1680 | /* | |
66610fdd RGS |
1681 | * Local variables: |
1682 | * c-indentation-style: bsd | |
1683 | * c-basic-offset: 4 | |
14d04a33 | 1684 | * indent-tabs-mode: nil |
66610fdd RGS |
1685 | * End: |
1686 | * | |
14d04a33 | 1687 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 1688 | */ |