This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
report missing authors to stderr
[perl5.git] / locale.c
CommitLineData
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 */
62STATIC char *
63S_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
92void
93Perl_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);
9f10db87 109 if (! is_invariant_string((U8 *) lc->decimal_point, 0)
28acfe03 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 143void
8772537c 144Perl_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
210void
211Perl_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
230void
231Perl_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 */
254void
8772537c 255Perl_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
215c5139
KW
275 /* We will replace any bad locale warning with 1) nothing if the new one is
276 * ok; or 2) a new warning for the bad new locale */
277 if (PL_warn_locale) {
278 SvREFCNT_dec_NN(PL_warn_locale);
279 PL_warn_locale = NULL;
280 }
281
c1284011 282 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
31f05a37
KW
283
284 /* A UTF-8 locale gets standard rules. But note that code still has to
285 * handle this specially because of the three problematic code points */
286 if (PL_in_utf8_CTYPE_locale) {
287 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
288 }
289 else {
8c6180a9
KW
290 /* Assume enough space for every character being bad. 4 spaces each
291 * for the 94 printable characters that are output like "'x' "; and 5
292 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
293 * NUL */
294 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
295
296 bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
297 no check */
298 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
299 to start */
300 unsigned int bad_count = 0; /* Count of bad characters */
301
baa60164
KW
302 for (i = 0; i < 256; i++) {
303 if (isUPPER_LC((U8) i))
304 PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
305 else if (isLOWER_LC((U8) i))
306 PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
307 else
308 PL_fold_locale[i] = (U8) i;
8c6180a9
KW
309
310 /* If checking for locale problems, see if the native ASCII-range
311 * printables plus \n and \t are in their expected categories in
312 * the new locale. If not, this could mean big trouble, upending
313 * Perl's and most programs' assumptions, like having a
314 * metacharacter with special meaning become a \w. Fortunately,
315 * it's very rare to find locales that aren't supersets of ASCII
316 * nowadays. It isn't a problem for most controls to be changed
317 * into something else; we check only \n and \t, though perhaps \r
318 * could be an issue as well. */
319 if (check_for_problems
320 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
321 {
322 if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
323 || (isPUNCT_A(i) && ! isPUNCT_LC(i))
324 || (isBLANK_A(i) && ! isBLANK_LC(i))
325 || (i == '\n' && ! isCNTRL_LC(i)))
326 {
327 if (bad_count) { /* Separate multiple entries with a
328 blank */
329 bad_chars_list[bad_count++] = ' ';
330 }
331 bad_chars_list[bad_count++] = '\'';
332 if (isPRINT_A(i)) {
333 bad_chars_list[bad_count++] = (char) i;
334 }
335 else {
336 bad_chars_list[bad_count++] = '\\';
337 if (i == '\n') {
338 bad_chars_list[bad_count++] = 'n';
339 }
340 else {
341 assert(i == '\t');
342 bad_chars_list[bad_count++] = 't';
343 }
344 }
345 bad_chars_list[bad_count++] = '\'';
346 bad_chars_list[bad_count] = '\0';
347 }
348 }
349 }
350
351#ifdef MB_CUR_MAX
352 /* We only handle single-byte locales (outside of UTF-8 ones; so if
353 * this locale requires than one byte, there are going to be
354 * problems. */
ba1a4362
KW
355 if (check_for_problems && MB_CUR_MAX > 1
356
357 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
358 * locale. Just assume that the implementation for them (plus
359 * for POSIX) is correct and the > 1 value is spurious. (Since
360 * these are specially handled to never be considered UTF-8
361 * locales, as long as this is the only problem, everything
362 * should work fine */
363 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
364 {
8c6180a9
KW
365 multi_byte_locale = TRUE;
366 }
367#endif
368
369 if (bad_count || multi_byte_locale) {
780fcc9f 370 PL_warn_locale = Perl_newSVpvf(aTHX_
8c6180a9 371 "Locale '%s' may not work well.%s%s%s\n",
780fcc9f 372 newctype,
8c6180a9
KW
373 (multi_byte_locale)
374 ? " Some characters in it are not recognized by"
375 " Perl."
376 : "",
377 (bad_count)
378 ? "\nThe following characters (and maybe others)"
379 " may not have the same meaning as the Perl"
380 " program expects:\n"
381 : "",
382 (bad_count)
383 ? bad_chars_list
384 : ""
385 );
780fcc9f
KW
386 /* If we are actually in the scope of the locale, output the
387 * message now. Otherwise we save it to be output at the first
388 * operation using this locale, if that actually happens. Most
389 * programs don't use locales, so they are immune to bad ones */
390 if (IN_LC(LC_CTYPE)) {
391
392 /* We have to save 'newctype' because the setlocale() just
393 * below may destroy it. The next setlocale() further down
394 * should restore it properly so that the intermediate change
395 * here is transparent to this function's caller */
396 const char * const badlocale = savepv(newctype);
397
398 setlocale(LC_CTYPE, "C");
399
400 /* The '0' below suppresses a bogus gcc compiler warning */
401 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
402 setlocale(LC_CTYPE, badlocale);
c0f3a893 403 Safefree(badlocale);
780fcc9f
KW
404 SvREFCNT_dec_NN(PL_warn_locale);
405 PL_warn_locale = NULL;
406 }
baa60164 407 }
31f05a37 408 }
98994639
HS
409
410#endif /* USE_LOCALE_CTYPE */
7918f24d 411 PERL_ARGS_ASSERT_NEW_CTYPE;
8772537c 412 PERL_UNUSED_ARG(newctype);
96a5add6 413 PERL_UNUSED_CONTEXT;
98994639
HS
414}
415
98994639 416void
2726666d
KW
417Perl__warn_problematic_locale()
418{
2726666d
KW
419
420#ifdef USE_LOCALE_CTYPE
421
5f04a188
KW
422 dTHX;
423
424 /* Internal-to-core function that outputs the message in PL_warn_locale,
425 * and then NULLS it. Should be called only through the macro
426 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
427
2726666d
KW
428 if (PL_warn_locale) {
429 /*GCC_DIAG_IGNORE(-Wformat-security); Didn't work */
430 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
431 SvPVX(PL_warn_locale),
432 0 /* dummy to avoid compiler warning */ );
433 /* GCC_DIAG_RESTORE; */
434 SvREFCNT_dec_NN(PL_warn_locale);
435 PL_warn_locale = NULL;
436 }
437
438#endif
439
440}
441
442void
8772537c 443Perl_new_collate(pTHX_ const char *newcoll)
98994639
HS
444{
445#ifdef USE_LOCALE_COLLATE
0d071d52
KW
446
447 /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell
448 * core Perl this and that 'newcoll' is the name of the new locale.
449 *
450 * Any code changing the locale (outside this file) should use
451 * POSIX::setlocale, which calls this function. Therefore this function
452 * should be called directly only from this file and from
453 * POSIX::setlocale() */
454
98994639
HS
455 if (! newcoll) {
456 if (PL_collation_name) {
457 ++PL_collation_ix;
458 Safefree(PL_collation_name);
459 PL_collation_name = NULL;
460 }
461 PL_collation_standard = TRUE;
462 PL_collxfrm_base = 0;
463 PL_collxfrm_mult = 2;
464 return;
465 }
466
467 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
468 ++PL_collation_ix;
469 Safefree(PL_collation_name);
470 PL_collation_name = stdize_locale(savepv(newcoll));
a39edc4c 471 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
98994639
HS
472
473 {
474 /* 2: at most so many chars ('a', 'b'). */
475 /* 50: surely no system expands a char more. */
476#define XFRMBUFSIZE (2 * 50)
477 char xbuf[XFRMBUFSIZE];
8772537c
AL
478 const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE);
479 const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE);
480 const SSize_t mult = fb - fa;
e0601d3c 481 if (mult < 1 && !(fa == 0 && fb == 0))
ad49ad39
NC
482 Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
483 (UV) fa, (UV) fb);
eb160463 484 PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
98994639
HS
485 PL_collxfrm_mult = mult;
486 }
487 }
488
f2ce9e1c
JH
489#else
490 PERL_UNUSED_ARG(newcoll);
98994639
HS
491#endif /* USE_LOCALE_COLLATE */
492}
493
b385bb4d
KW
494#ifdef WIN32
495
496char *
497Perl_my_setlocale(pTHX_ int category, const char* locale)
498{
499 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
500 * difference unless the input locale is "", which means on Windows to get
501 * the machine default, which is set via the computer's "Regional and
502 * Language Options" (or its current equivalent). In POSIX, it instead
503 * means to find the locale from the user's environment. This routine
504 * looks in the environment, and, if anything is found, uses that instead
505 * of going to the machine default. If there is no environment override,
506 * the machine default is used, as normal, by calling the real setlocale()
507 * with "". The POSIX behavior is to use the LC_ALL variable if set;
508 * otherwise to use the particular category's variable if set; otherwise to
509 * use the LANG variable. */
510
175c4cf9 511 bool override_LC_ALL = FALSE;
89f7b9aa
KW
512 char * result;
513
b385bb4d
KW
514 if (locale && strEQ(locale, "")) {
515# ifdef LC_ALL
516 locale = PerlEnv_getenv("LC_ALL");
517 if (! locale) {
518#endif
519 switch (category) {
520# ifdef LC_ALL
521 case LC_ALL:
481465ea 522 override_LC_ALL = TRUE;
b385bb4d
KW
523 break; /* We already know its variable isn't set */
524# endif
525# ifdef USE_LOCALE_TIME
526 case LC_TIME:
527 locale = PerlEnv_getenv("LC_TIME");
528 break;
529# endif
530# ifdef USE_LOCALE_CTYPE
531 case LC_CTYPE:
532 locale = PerlEnv_getenv("LC_CTYPE");
533 break;
534# endif
535# ifdef USE_LOCALE_COLLATE
536 case LC_COLLATE:
537 locale = PerlEnv_getenv("LC_COLLATE");
538 break;
539# endif
540# ifdef USE_LOCALE_MONETARY
541 case LC_MONETARY:
542 locale = PerlEnv_getenv("LC_MONETARY");
543 break;
544# endif
545# ifdef USE_LOCALE_NUMERIC
546 case LC_NUMERIC:
547 locale = PerlEnv_getenv("LC_NUMERIC");
548 break;
549# endif
550# ifdef USE_LOCALE_MESSAGES
551 case LC_MESSAGES:
552 locale = PerlEnv_getenv("LC_MESSAGES");
553 break;
554# endif
555 default:
556 /* This is a category, like PAPER_SIZE that we don't
557 * know about; and so can't provide a wrapper. */
558 break;
559 }
560 if (! locale) {
561 locale = PerlEnv_getenv("LANG");
481465ea 562 if (! locale) {
b385bb4d
KW
563 locale = "";
564 }
565 }
566# ifdef LC_ALL
567 }
568# endif
569 }
570
89f7b9aa
KW
571 result = setlocale(category, locale);
572
481465ea 573 if (! override_LC_ALL) {
89f7b9aa
KW
574 return result;
575 }
576
dfd77d7a 577 /* Here the input category was LC_ALL, and we have set it to what is in the
481465ea
KW
578 * LANG variable or the system default if there is no LANG. But these have
579 * lower priority than the other LC_foo variables, so override it for each
580 * one that is set. (If they are set to "", it means to use the same thing
581 * we just set LC_ALL to, so can skip) */
89f7b9aa
KW
582# ifdef USE_LOCALE_TIME
583 result = PerlEnv_getenv("LC_TIME");
730252b2 584 if (result && strNE(result, "")) {
89f7b9aa
KW
585 setlocale(LC_TIME, result);
586 }
587# endif
588# ifdef USE_LOCALE_CTYPE
589 result = PerlEnv_getenv("LC_CTYPE");
730252b2 590 if (result && strNE(result, "")) {
89f7b9aa
KW
591 setlocale(LC_CTYPE, result);
592 }
593# endif
594# ifdef USE_LOCALE_COLLATE
595 result = PerlEnv_getenv("LC_COLLATE");
730252b2 596 if (result && strNE(result, "")) {
89f7b9aa
KW
597 setlocale(LC_COLLATE, result);
598 }
599# endif
600# ifdef USE_LOCALE_MONETARY
601 result = PerlEnv_getenv("LC_MONETARY");
730252b2 602 if (result && strNE(result, "")) {
89f7b9aa
KW
603 setlocale(LC_MONETARY, result);
604 }
605# endif
606# ifdef USE_LOCALE_NUMERIC
607 result = PerlEnv_getenv("LC_NUMERIC");
730252b2 608 if (result && strNE(result, "")) {
89f7b9aa
KW
609 setlocale(LC_NUMERIC, result);
610 }
611# endif
612# ifdef USE_LOCALE_MESSAGES
613 result = PerlEnv_getenv("LC_MESSAGES");
730252b2 614 if (result && strNE(result, "")) {
89f7b9aa
KW
615 setlocale(LC_MESSAGES, result);
616 }
617# endif
618
619 return setlocale(LC_ALL, NULL);
620
b385bb4d
KW
621}
622
623#endif
624
625
98994639
HS
626/*
627 * Initialize locale awareness.
628 */
629int
630Perl_init_i18nl10n(pTHX_ int printwarn)
631{
0e92a118
KW
632 /* printwarn is
633 *
634 * 0 if not to output warning when setup locale is bad
635 * 1 if to output warning based on value of PERL_BADLANG
636 * >1 if to output regardless of PERL_BADLANG
637 *
638 * returns
98994639 639 * 1 = set ok or not applicable,
0e92a118
KW
640 * 0 = fallback to a locale of lower priority
641 * -1 = fallback to all locales failed, not even to the C locale
98994639
HS
642 */
643
0e92a118
KW
644 int ok = 1;
645
98994639 646#if defined(USE_LOCALE)
98994639
HS
647#ifdef USE_LOCALE_CTYPE
648 char *curctype = NULL;
649#endif /* USE_LOCALE_CTYPE */
650#ifdef USE_LOCALE_COLLATE
651 char *curcoll = NULL;
652#endif /* USE_LOCALE_COLLATE */
653#ifdef USE_LOCALE_NUMERIC
654 char *curnum = NULL;
655#endif /* USE_LOCALE_NUMERIC */
656#ifdef __GLIBC__
175c4cf9 657 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
98994639 658#endif
65ebb059 659
ccd65d51
KW
660 /* NULL uses the existing already set up locale */
661 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
662 ? NULL
663 : "";
c3fcd832
KW
664 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
665 unsigned int trial_locales_count;
175c4cf9
KW
666 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
667 const char * const lang = savepv(PerlEnv_getenv("LANG"));
98994639 668 bool setlocale_failure = FALSE;
65ebb059
KW
669 unsigned int i;
670 char *p;
175c4cf9
KW
671
672 /* A later getenv() could zap this, so only use here */
673 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
674
675 const bool locwarn = (printwarn > 1
676 || (printwarn
677 && (! bad_lang_use_once
22ff3130
HS
678 || (
679 /* disallow with "" or "0" */
680 *bad_lang_use_once
681 && strNE("0", bad_lang_use_once)))));
0e92a118 682 bool done = FALSE;
6bce99ee
JH
683#ifdef WIN32
684 /* In some systems you can find out the system default locale
685 * and use that as the fallback locale. */
686# define SYSTEM_DEFAULT_LOCALE
687#endif
688#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059 689 const char *system_default_locale = NULL;
6bce99ee 690#endif
98994639 691
0e92a118
KW
692#ifndef LOCALE_ENVIRON_REQUIRED
693 PERL_UNUSED_VAR(done);
694#else
98994639
HS
695
696 /*
697 * Ultrix setlocale(..., "") fails if there are no environment
698 * variables from which to get a locale name.
699 */
700
b3e384bf 701# ifdef LC_ALL
98994639 702 if (lang) {
b385bb4d 703 if (my_setlocale(LC_ALL, setlocale_init))
98994639
HS
704 done = TRUE;
705 else
706 setlocale_failure = TRUE;
707 }
708 if (!setlocale_failure) {
b3e384bf 709# ifdef USE_LOCALE_CTYPE
98994639 710 if (! (curctype =
b385bb4d 711 my_setlocale(LC_CTYPE,
98994639 712 (!done && (lang || PerlEnv_getenv("LC_CTYPE")))
ccd65d51 713 ? setlocale_init : NULL)))
98994639
HS
714 setlocale_failure = TRUE;
715 else
716 curctype = savepv(curctype);
b3e384bf
KW
717# endif /* USE_LOCALE_CTYPE */
718# ifdef USE_LOCALE_COLLATE
98994639 719 if (! (curcoll =
b385bb4d 720 my_setlocale(LC_COLLATE,
98994639 721 (!done && (lang || PerlEnv_getenv("LC_COLLATE")))
ccd65d51 722 ? setlocale_init : NULL)))
98994639
HS
723 setlocale_failure = TRUE;
724 else
725 curcoll = savepv(curcoll);
b3e384bf
KW
726# endif /* USE_LOCALE_COLLATE */
727# ifdef USE_LOCALE_NUMERIC
98994639 728 if (! (curnum =
b385bb4d 729 my_setlocale(LC_NUMERIC,
98994639 730 (!done && (lang || PerlEnv_getenv("LC_NUMERIC")))
ccd65d51 731 ? setlocale_init : NULL)))
98994639
HS
732 setlocale_failure = TRUE;
733 else
734 curnum = savepv(curnum);
b3e384bf 735# endif /* USE_LOCALE_NUMERIC */
a782673d
KW
736# ifdef USE_LOCALE_MESSAGES
737 if (! my_setlocale(LC_MESSAGES,
738 (!done && (lang || PerlEnv_getenv("LC_MESSAGES")))
739 ? setlocale_init : NULL))
740 {
741 setlocale_failure = TRUE;
742 }
743# endif /* USE_LOCALE_MESSAGES */
c835d6be
KW
744# ifdef USE_LOCALE_MONETARY
745 if (! my_setlocale(LC_MONETARY,
746 (!done && (lang || PerlEnv_getenv("LC_MONETARY")))
747 ? setlocale_init : NULL))
748 {
749 setlocale_failure = TRUE;
750 }
751# endif /* USE_LOCALE_MONETARY */
98994639
HS
752 }
753
b3e384bf 754# endif /* LC_ALL */
98994639
HS
755
756#endif /* !LOCALE_ENVIRON_REQUIRED */
757
65ebb059 758 /* We try each locale in the list until we get one that works, or exhaust
20a240df
KW
759 * the list. Normally the loop is executed just once. But if setting the
760 * locale fails, inside the loop we add fallback trials to the array and so
761 * will execute the loop multiple times */
c3fcd832
KW
762 trial_locales[0] = setlocale_init;
763 trial_locales_count = 1;
65ebb059
KW
764 for (i= 0; i < trial_locales_count; i++) {
765 const char * trial_locale = trial_locales[i];
766
767 if (i > 0) {
768
769 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
770 * when i==0, but I (khw) don't think that behavior makes much
771 * sense */
772 setlocale_failure = FALSE;
773
6bce99ee
JH
774#ifdef SYSTEM_DEFAULT_LOCALE
775# ifdef WIN32
65ebb059
KW
776 /* On Windows machines, an entry of "" after the 0th means to use
777 * the system default locale, which we now proceed to get. */
778 if (strEQ(trial_locale, "")) {
779 unsigned int j;
780
781 /* Note that this may change the locale, but we are going to do
782 * that anyway just below */
783 system_default_locale = setlocale(LC_ALL, "");
784
785 /* Skip if invalid or it's already on the list of locales to
786 * try */
787 if (! system_default_locale) {
788 goto next_iteration;
789 }
790 for (j = 0; j < trial_locales_count; j++) {
791 if (strEQ(system_default_locale, trial_locales[j])) {
792 goto next_iteration;
793 }
794 }
795
796 trial_locale = system_default_locale;
797 }
6bce99ee
JH
798# endif /* WIN32 */
799#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
800 }
801
98994639 802#ifdef LC_ALL
7cd8b568 803 if (! my_setlocale(LC_ALL, trial_locale)) {
49c85077 804 setlocale_failure = TRUE;
7cd8b568
KW
805 }
806 else {
807 /* Since LC_ALL succeeded, it should have changed all the other
808 * categories it can to its value; so we massage things so that the
809 * setlocales below just return their category's current values.
810 * This adequately handles the case in NetBSD where LC_COLLATE may
811 * not be defined for a locale, and setting it individually will
812 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
813 * the POSIX locale. */
814 trial_locale = NULL;
815 }
98994639
HS
816#endif /* LC_ALL */
817
49c85077 818 if (!setlocale_failure) {
98994639 819#ifdef USE_LOCALE_CTYPE
49c85077
KW
820 Safefree(curctype);
821 if (! (curctype = my_setlocale(LC_CTYPE, trial_locale)))
822 setlocale_failure = TRUE;
823 else
824 curctype = savepv(curctype);
98994639
HS
825#endif /* USE_LOCALE_CTYPE */
826#ifdef USE_LOCALE_COLLATE
49c85077
KW
827 Safefree(curcoll);
828 if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale)))
829 setlocale_failure = TRUE;
830 else
831 curcoll = savepv(curcoll);
98994639
HS
832#endif /* USE_LOCALE_COLLATE */
833#ifdef USE_LOCALE_NUMERIC
49c85077
KW
834 Safefree(curnum);
835 if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale)))
836 setlocale_failure = TRUE;
837 else
838 curnum = savepv(curnum);
98994639 839#endif /* USE_LOCALE_NUMERIC */
a782673d
KW
840#ifdef USE_LOCALE_MESSAGES
841 if (! (my_setlocale(LC_MESSAGES, trial_locale)))
842 setlocale_failure = TRUE;
843#endif /* USE_LOCALE_MESSAGES */
c835d6be
KW
844#ifdef USE_LOCALE_MONETARY
845 if (! (my_setlocale(LC_MONETARY, trial_locale)))
846 setlocale_failure = TRUE;
847#endif /* USE_LOCALE_MONETARY */
848
49c85077
KW
849 if (! setlocale_failure) { /* Success */
850 break;
851 }
65ebb059 852 }
98994639 853
49c85077
KW
854 /* Here, something failed; will need to try a fallback. */
855 ok = 0;
65ebb059 856
49c85077
KW
857 if (i == 0) {
858 unsigned int j;
98994639 859
65ebb059 860 if (locwarn) { /* Output failure info only on the first one */
98994639
HS
861#ifdef LC_ALL
862
49c85077
KW
863 PerlIO_printf(Perl_error_log,
864 "perl: warning: Setting locale failed.\n");
98994639
HS
865
866#else /* !LC_ALL */
867
49c85077
KW
868 PerlIO_printf(Perl_error_log,
869 "perl: warning: Setting locale failed for the categories:\n\t");
20a240df 870# ifdef USE_LOCALE_CTYPE
49c85077
KW
871 if (! curctype)
872 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
20a240df
KW
873# endif /* USE_LOCALE_CTYPE */
874# ifdef USE_LOCALE_COLLATE
49c85077
KW
875 if (! curcoll)
876 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
20a240df
KW
877# endif /* USE_LOCALE_COLLATE */
878# ifdef USE_LOCALE_NUMERIC
49c85077
KW
879 if (! curnum)
880 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
20a240df 881# endif /* USE_LOCALE_NUMERIC */
a782673d 882 PerlIO_printf(Perl_error_log, "and possibly others\n");
98994639
HS
883
884#endif /* LC_ALL */
885
49c85077
KW
886 PerlIO_printf(Perl_error_log,
887 "perl: warning: Please check that your locale settings:\n");
98994639
HS
888
889#ifdef __GLIBC__
49c85077
KW
890 PerlIO_printf(Perl_error_log,
891 "\tLANGUAGE = %c%s%c,\n",
892 language ? '"' : '(',
893 language ? language : "unset",
894 language ? '"' : ')');
98994639
HS
895#endif
896
49c85077
KW
897 PerlIO_printf(Perl_error_log,
898 "\tLC_ALL = %c%s%c,\n",
899 lc_all ? '"' : '(',
900 lc_all ? lc_all : "unset",
901 lc_all ? '"' : ')');
98994639
HS
902
903#if defined(USE_ENVIRON_ARRAY)
49c85077
KW
904 {
905 char **e;
906 for (e = environ; *e; e++) {
907 if (strnEQ(*e, "LC_", 3)
908 && strnNE(*e, "LC_ALL=", 7)
909 && (p = strchr(*e, '=')))
910 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
911 (int)(p - *e), *e, p + 1);
912 }
913 }
98994639 914#else
49c85077
KW
915 PerlIO_printf(Perl_error_log,
916 "\t(possibly more locale environment variables)\n");
98994639
HS
917#endif
918
49c85077
KW
919 PerlIO_printf(Perl_error_log,
920 "\tLANG = %c%s%c\n",
921 lang ? '"' : '(',
922 lang ? lang : "unset",
923 lang ? '"' : ')');
98994639 924
49c85077
KW
925 PerlIO_printf(Perl_error_log,
926 " are supported and installed on your system.\n");
927 }
98994639 928
65ebb059 929 /* Calculate what fallback locales to try. We have avoided this
f6bab5f6 930 * until we have to, because failure is quite unlikely. This will
65ebb059
KW
931 * usually change the upper bound of the loop we are in.
932 *
933 * Since the system's default way of setting the locale has not
934 * found one that works, We use Perl's defined ordering: LC_ALL,
935 * LANG, and the C locale. We don't try the same locale twice, so
936 * don't add to the list if already there. (On POSIX systems, the
937 * LC_ALL element will likely be a repeat of the 0th element "",
938 * but there's no harm done by doing it explicitly */
939 if (lc_all) {
940 for (j = 0; j < trial_locales_count; j++) {
941 if (strEQ(lc_all, trial_locales[j])) {
942 goto done_lc_all;
943 }
944 }
945 trial_locales[trial_locales_count++] = lc_all;
946 }
947 done_lc_all:
98994639 948
65ebb059
KW
949 if (lang) {
950 for (j = 0; j < trial_locales_count; j++) {
951 if (strEQ(lang, trial_locales[j])) {
952 goto done_lang;
953 }
954 }
955 trial_locales[trial_locales_count++] = lang;
956 }
957 done_lang:
958
959#if defined(WIN32) && defined(LC_ALL)
960 /* For Windows, we also try the system default locale before "C".
961 * (If there exists a Windows without LC_ALL we skip this because
962 * it gets too complicated. For those, the "C" is the next
963 * fallback possibility). The "" is the same as the 0th element of
964 * the array, but the code at the loop above knows to treat it
965 * differently when not the 0th */
966 trial_locales[trial_locales_count++] = "";
967#endif
968
969 for (j = 0; j < trial_locales_count; j++) {
970 if (strEQ("C", trial_locales[j])) {
971 goto done_C;
972 }
973 }
974 trial_locales[trial_locales_count++] = "C";
98994639 975
65ebb059
KW
976 done_C: ;
977 } /* end of first time through the loop */
98994639 978
65ebb059
KW
979#ifdef WIN32
980 next_iteration: ;
981#endif
982
983 } /* end of looping through the trial locales */
984
985 if (ok < 1) { /* If we tried to fallback */
986 const char* msg;
987 if (! setlocale_failure) { /* fallback succeeded */
988 msg = "Falling back to";
989 }
990 else { /* fallback failed */
98994639 991
65ebb059
KW
992 /* We dropped off the end of the loop, so have to decrement i to
993 * get back to the value the last time through */
994 i--;
98994639 995
65ebb059
KW
996 ok = -1;
997 msg = "Failed to fall back to";
998
999 /* To continue, we should use whatever values we've got */
98994639 1000#ifdef USE_LOCALE_CTYPE
49c85077
KW
1001 Safefree(curctype);
1002 curctype = savepv(setlocale(LC_CTYPE, NULL));
98994639
HS
1003#endif /* USE_LOCALE_CTYPE */
1004#ifdef USE_LOCALE_COLLATE
49c85077
KW
1005 Safefree(curcoll);
1006 curcoll = savepv(setlocale(LC_COLLATE, NULL));
98994639
HS
1007#endif /* USE_LOCALE_COLLATE */
1008#ifdef USE_LOCALE_NUMERIC
49c85077
KW
1009 Safefree(curnum);
1010 curnum = savepv(setlocale(LC_NUMERIC, NULL));
98994639 1011#endif /* USE_LOCALE_NUMERIC */
65ebb059
KW
1012 }
1013
1014 if (locwarn) {
1015 const char * description;
1016 const char * name = "";
1017 if (strEQ(trial_locales[i], "C")) {
1018 description = "the standard locale";
1019 name = "C";
1020 }
6bce99ee 1021#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059
KW
1022 else if (strEQ(trial_locales[i], "")) {
1023 description = "the system default locale";
1024 if (system_default_locale) {
1025 name = system_default_locale;
1026 }
1027 }
6bce99ee 1028#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
1029 else {
1030 description = "a fallback locale";
1031 name = trial_locales[i];
1032 }
1033 if (name && strNE(name, "")) {
1034 PerlIO_printf(Perl_error_log,
1035 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1036 }
1037 else {
1038 PerlIO_printf(Perl_error_log,
1039 "perl: warning: %s %s.\n", msg, description);
1040 }
1041 }
1042 } /* End of tried to fallback */
98994639
HS
1043
1044#ifdef USE_LOCALE_CTYPE
1045 new_ctype(curctype);
1046#endif /* USE_LOCALE_CTYPE */
1047
1048#ifdef USE_LOCALE_COLLATE
1049 new_collate(curcoll);
1050#endif /* USE_LOCALE_COLLATE */
1051
1052#ifdef USE_LOCALE_NUMERIC
1053 new_numeric(curnum);
1054#endif /* USE_LOCALE_NUMERIC */
b310b053 1055
8ef6e574 1056#if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
49c85077
KW
1057 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1058 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1059 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1060 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1061 * discipline. */
c1284011 1062 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
49c85077 1063
a05d7ebb 1064 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
fde18df1
JH
1065 This is an alternative to using the -C command line switch
1066 (the -C if present will override this). */
1067 {
dd374669 1068 const char *p = PerlEnv_getenv("PERL_UNICODE");
a05d7ebb 1069 PL_unicode = p ? parse_unicode_opts(&p) : 0;
5a22a2bb
NC
1070 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1071 PL_utf8cache = -1;
b310b053 1072 }
ec71e770 1073#endif
b310b053 1074
98994639 1075#ifdef USE_LOCALE_CTYPE
43c5f42d 1076 Safefree(curctype);
98994639
HS
1077#endif /* USE_LOCALE_CTYPE */
1078#ifdef USE_LOCALE_COLLATE
43c5f42d 1079 Safefree(curcoll);
98994639
HS
1080#endif /* USE_LOCALE_COLLATE */
1081#ifdef USE_LOCALE_NUMERIC
43c5f42d 1082 Safefree(curnum);
98994639 1083#endif /* USE_LOCALE_NUMERIC */
8ef6e574 1084
175c4cf9
KW
1085#ifdef __GLIBC__
1086 Safefree(language);
1087#endif
1088
1089 Safefree(lc_all);
1090 Safefree(lang);
1091
e3305790
KW
1092#else /* !USE_LOCALE */
1093 PERL_UNUSED_ARG(printwarn);
1094#endif /* USE_LOCALE */
1095
98994639
HS
1096 return ok;
1097}
1098
49c85077 1099
98994639
HS
1100#ifdef USE_LOCALE_COLLATE
1101
1102/*
1103 * mem_collxfrm() is a bit like strxfrm() but with two important
1104 * differences. First, it handles embedded NULs. Second, it allocates
1105 * a bit more memory than needed for the transformed data itself.
1106 * The real transformed data begins at offset sizeof(collationix).
1107 * Please see sv_collxfrm() to see how this is used.
1108 */
166f8a29 1109
98994639
HS
1110char *
1111Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1112{
1113 char *xbuf;
1114 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1115
7918f24d
NC
1116 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1117
98994639
HS
1118 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1119 /* the +1 is for the terminating NUL. */
1120
1121 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
a02a5408 1122 Newx(xbuf, xAlloc, char);
98994639
HS
1123 if (! xbuf)
1124 goto bad;
1125
1126 *(U32*)xbuf = PL_collation_ix;
1127 xout = sizeof(PL_collation_ix);
1128 for (xin = 0; xin < len; ) {
224e8ef5 1129 Size_t xused;
98994639
HS
1130
1131 for (;;) {
1132 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
224e8ef5 1133 if (xused >= PERL_INT_MAX)
98994639 1134 goto bad;
eb160463 1135 if ((STRLEN)xused < xAlloc - xout)
98994639
HS
1136 break;
1137 xAlloc = (2 * xAlloc) + 1;
1138 Renew(xbuf, xAlloc, char);
1139 if (! xbuf)
1140 goto bad;
1141 }
1142
1143 xin += strlen(s + xin) + 1;
1144 xout += xused;
1145
1146 /* Embedded NULs are understood but silently skipped
1147 * because they make no sense in locale collation. */
1148 }
1149
1150 xbuf[xout] = '\0';
1151 *xlen = xout - sizeof(PL_collation_ix);
1152 return xbuf;
1153
1154 bad:
1155 Safefree(xbuf);
1156 *xlen = 0;
1157 return NULL;
1158}
1159
1160#endif /* USE_LOCALE_COLLATE */
1161
8ef6e574
KW
1162#ifdef USE_LOCALE
1163
c1284011
KW
1164bool
1165Perl__is_cur_LC_category_utf8(pTHX_ int category)
7d74bb61
KW
1166{
1167 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1168 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
119ee68b 1169 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
609548d2
KW
1170 * could give the wrong result. The result will very likely be correct for
1171 * languages that have commonly used non-ASCII characters, but for notably
1172 * English, it comes down to if the locale's name ends in something like
1173 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
7d74bb61
KW
1174
1175 char *save_input_locale = NULL;
7d74bb61
KW
1176 STRLEN final_pos;
1177
8ef6e574 1178#ifdef LC_ALL
7d74bb61 1179 assert(category != LC_ALL);
8ef6e574 1180#endif
7d74bb61
KW
1181
1182 /* First dispose of the trivial cases */
b07fffd1 1183 save_input_locale = setlocale(category, NULL);
7d74bb61 1184 if (! save_input_locale) {
69014004
KW
1185 DEBUG_L(PerlIO_printf(Perl_debug_log,
1186 "Could not find current locale for category %d\n",
1187 category));
7d74bb61
KW
1188 return FALSE; /* XXX maybe should croak */
1189 }
b07fffd1 1190 save_input_locale = stdize_locale(savepv(save_input_locale));
a39edc4c 1191 if (isNAME_C_OR_POSIX(save_input_locale)) {
69014004
KW
1192 DEBUG_L(PerlIO_printf(Perl_debug_log,
1193 "Current locale for category %d is %s\n",
1194 category, save_input_locale));
b07fffd1 1195 Safefree(save_input_locale);
7d74bb61
KW
1196 return FALSE;
1197 }
1198
1d958db2
KW
1199#if defined(USE_LOCALE_CTYPE) \
1200 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
7d74bb61 1201
1d958db2 1202 { /* Next try nl_langinfo or MB_CUR_MAX if available */
7d74bb61
KW
1203
1204 char *save_ctype_locale = NULL;
119ee68b 1205 bool is_utf8;
7d74bb61 1206
119ee68b 1207 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
7d74bb61
KW
1208
1209 /* Get the current LC_CTYPE locale */
4f72bb37 1210 save_ctype_locale = setlocale(LC_CTYPE, NULL);
7d74bb61 1211 if (! save_ctype_locale) {
69014004
KW
1212 DEBUG_L(PerlIO_printf(Perl_debug_log,
1213 "Could not find current locale for LC_CTYPE\n"));
7d74bb61
KW
1214 goto cant_use_nllanginfo;
1215 }
4f72bb37 1216 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
7d74bb61
KW
1217
1218 /* If LC_CTYPE and the desired category use the same locale, this
1219 * means that finding the value for LC_CTYPE is the same as finding
1220 * the value for the desired category. Otherwise, switch LC_CTYPE
1221 * to the desired category's locale */
1222 if (strEQ(save_ctype_locale, save_input_locale)) {
1223 Safefree(save_ctype_locale);
1224 save_ctype_locale = NULL;
1225 }
1226 else if (! setlocale(LC_CTYPE, save_input_locale)) {
69014004
KW
1227 DEBUG_L(PerlIO_printf(Perl_debug_log,
1228 "Could not change LC_CTYPE locale to %s\n",
1229 save_input_locale));
7d74bb61
KW
1230 Safefree(save_ctype_locale);
1231 goto cant_use_nllanginfo;
1232 }
1233 }
1234
69014004
KW
1235 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1236 save_input_locale));
1237
7d74bb61 1238 /* Here the current LC_CTYPE is set to the locale of the category whose
1d958db2
KW
1239 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1240 * should give the correct results */
119ee68b 1241
1d958db2
KW
1242# if defined(HAS_NL_LANGINFO) && defined(CODESET)
1243 {
4f72bb37 1244 char *codeset = nl_langinfo(CODESET);
1d958db2 1245 if (codeset && strNE(codeset, "")) {
4f72bb37 1246 codeset = savepv(codeset);
119ee68b 1247
1d958db2
KW
1248 /* If we switched LC_CTYPE, switch back */
1249 if (save_ctype_locale) {
1250 setlocale(LC_CTYPE, save_ctype_locale);
1251 Safefree(save_ctype_locale);
1252 }
1253
1254 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1255 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1256
69014004
KW
1257 DEBUG_L(PerlIO_printf(Perl_debug_log,
1258 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1259 codeset, is_utf8));
1d958db2
KW
1260 Safefree(codeset);
1261 Safefree(save_input_locale);
1262 return is_utf8;
1263 }
119ee68b
KW
1264 }
1265
1d958db2
KW
1266# endif
1267# ifdef MB_CUR_MAX
1268
1269 /* Here, either we don't have nl_langinfo, or it didn't return a
1270 * codeset. Try MB_CUR_MAX */
1271
119ee68b
KW
1272 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1273 * Unicode code point. Since UTF-8 is the only non-single byte
1274 * encoding we handle, we just say any such encoding is UTF-8, and if
1275 * turns out to be wrong, other things will fail */
1276 is_utf8 = MB_CUR_MAX >= 4;
1277
69014004
KW
1278 DEBUG_L(PerlIO_printf(Perl_debug_log,
1279 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1280 (int) MB_CUR_MAX, is_utf8));
1281
119ee68b
KW
1282 Safefree(save_input_locale);
1283
1284# ifdef HAS_MBTOWC
1285
1286 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1287 * since they are both in the C99 standard. We can feed a known byte
1288 * string to the latter function, and check that it gives the expected
1289 * result */
1290 if (is_utf8) {
1291 wchar_t wc;
856b881c 1292 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
69014004 1293 errno = 0;
f019f68f 1294 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
119ee68b
KW
1295 != strlen(HYPHEN_UTF8)
1296 || wc != (wchar_t) 0x2010)
1297 {
1298 is_utf8 = FALSE;
abdcbdb8 1299 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
69014004
KW
1300 DEBUG_L(PerlIO_printf(Perl_debug_log,
1301 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1302 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
119ee68b
KW
1303 }
1304 }
119ee68b
KW
1305# endif
1306
1d958db2
KW
1307 /* If we switched LC_CTYPE, switch back */
1308 if (save_ctype_locale) {
1309 setlocale(LC_CTYPE, save_ctype_locale);
1310 Safefree(save_ctype_locale);
119ee68b 1311 }
7d74bb61 1312
1d958db2 1313 return is_utf8;
119ee68b 1314# endif
7d74bb61 1315 }
119ee68b 1316
7d74bb61
KW
1317 cant_use_nllanginfo:
1318
0080c90a
KW
1319#else /* nl_langinfo should work if available, so don't bother compiling this
1320 fallback code. The final fallback of looking at the name is
1321 compiled, and will be executed if nl_langinfo fails */
7d74bb61 1322
97f4de96
KW
1323 /* nl_langinfo not available or failed somehow. Next try looking at the
1324 * currency symbol to see if it disambiguates things. Often that will be
1325 * in the native script, and if the symbol isn't in UTF-8, we know that the
1326 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
609548d2
KW
1327 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1328 * */
fa9b773e
KW
1329
1330#ifdef HAS_LOCALECONV
fa9b773e 1331# ifdef USE_LOCALE_MONETARY
fa9b773e
KW
1332 {
1333 char *save_monetary_locale = NULL;
fa9b773e 1334 bool only_ascii = FALSE;
13542a67
KW
1335 bool is_utf8 = FALSE;
1336 struct lconv* lc;
fa9b773e 1337
97f4de96
KW
1338 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1339 * the desired category, if it isn't that locale already */
1340
fa9b773e
KW
1341 if (category != LC_MONETARY) {
1342
4f72bb37 1343 save_monetary_locale = setlocale(LC_MONETARY, NULL);
fa9b773e 1344 if (! save_monetary_locale) {
69014004
KW
1345 DEBUG_L(PerlIO_printf(Perl_debug_log,
1346 "Could not find current locale for LC_MONETARY\n"));
fa9b773e
KW
1347 goto cant_use_monetary;
1348 }
4f72bb37 1349 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
fa9b773e 1350
13542a67
KW
1351 if (strEQ(save_monetary_locale, save_input_locale)) {
1352 Safefree(save_monetary_locale);
1353 save_monetary_locale = NULL;
1354 }
1355 else if (! setlocale(LC_MONETARY, save_input_locale)) {
59c234b4
KW
1356 DEBUG_L(PerlIO_printf(Perl_debug_log,
1357 "Could not change LC_MONETARY locale to %s\n",
1358 save_input_locale));
1359 Safefree(save_monetary_locale);
1360 goto cant_use_monetary;
fa9b773e
KW
1361 }
1362 }
1363
1364 /* Here the current LC_MONETARY is set to the locale of the category
1365 * whose information is desired. */
1366
13542a67
KW
1367 lc = localeconv();
1368 if (! lc
1369 || ! lc->currency_symbol
9f10db87 1370 || is_invariant_string((U8 *) lc->currency_symbol, 0))
13542a67
KW
1371 {
1372 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));
1373 only_ascii = TRUE;
1374 }
1375 else {
1376 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
fa9b773e
KW
1377 }
1378
1379 /* If we changed it, restore LC_MONETARY to its original locale */
1380 if (save_monetary_locale) {
1381 setlocale(LC_MONETARY, save_monetary_locale);
1382 Safefree(save_monetary_locale);
1383 }
1384
13542a67 1385 if (! only_ascii) {
fa9b773e 1386
59c234b4
KW
1387 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1388 * otherwise assume the locale is UTF-8 if and only if the symbol
1389 * is non-ascii UTF-8. */
1390 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1391 save_input_locale, is_utf8));
1392 Safefree(save_input_locale);
1393 return is_utf8;
13542a67 1394 }
fa9b773e
KW
1395 }
1396 cant_use_monetary:
1397
1398# endif /* USE_LOCALE_MONETARY */
1399#endif /* HAS_LOCALECONV */
1400
15f7e74e
KW
1401#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1402
1403/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1404 * the names of the months and weekdays, timezone, and am/pm indicator */
1405 {
1406 char *save_time_locale = NULL;
1407 int hour = 10;
1408 bool is_dst = FALSE;
1409 int dom = 1;
1410 int month = 0;
1411 int i;
1412 char * formatted_time;
1413
1414
1415 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1416 * desired category, if it isn't that locale already */
1417
1418 if (category != LC_TIME) {
1419
1420 save_time_locale = setlocale(LC_TIME, NULL);
1421 if (! save_time_locale) {
1422 DEBUG_L(PerlIO_printf(Perl_debug_log,
1423 "Could not find current locale for LC_TIME\n"));
1424 goto cant_use_time;
1425 }
1426 save_time_locale = stdize_locale(savepv(save_time_locale));
1427
1428 if (strEQ(save_time_locale, save_input_locale)) {
1429 Safefree(save_time_locale);
1430 save_time_locale = NULL;
1431 }
1432 else if (! setlocale(LC_TIME, save_input_locale)) {
1433 DEBUG_L(PerlIO_printf(Perl_debug_log,
1434 "Could not change LC_TIME locale to %s\n",
1435 save_input_locale));
1436 Safefree(save_time_locale);
1437 goto cant_use_time;
1438 }
1439 }
1440
1441 /* Here the current LC_TIME is set to the locale of the category
1442 * whose information is desired. Look at all the days of the week and
9f10db87 1443 * month names, and the timezone and am/pm indicator for UTF-8 variant
15f7e74e
KW
1444 * characters. The first such a one found will tell us if the locale
1445 * is UTF-8 or not */
1446
1447 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1448 formatted_time = my_strftime("%A %B %Z %p",
1449 0, 0, hour, dom, month, 112, 0, 0, is_dst);
9f10db87 1450 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
15f7e74e
KW
1451
1452 /* Here, we didn't find a non-ASCII. Try the next time through
1453 * with the complemented dst and am/pm, and try with the next
1454 * weekday. After we have gotten all weekdays, try the next
1455 * month */
1456 is_dst = ! is_dst;
1457 hour = (hour + 12) % 24;
1458 dom++;
1459 if (i > 6) {
1460 month++;
1461 }
1462 continue;
1463 }
1464
1465 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1466 * false otherwise. But first, restore LC_TIME to its original
1467 * locale if we changed it */
1468 if (save_time_locale) {
1469 setlocale(LC_TIME, save_time_locale);
1470 Safefree(save_time_locale);
1471 }
1472
1473 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1474 save_input_locale,
1475 is_utf8_string((U8 *) formatted_time, 0)));
1476 Safefree(save_input_locale);
1477 return is_utf8_string((U8 *) formatted_time, 0);
1478 }
1479
1480 /* Falling off the end of the loop indicates all the names were just
1481 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1482 * to its original locale */
1483 if (save_time_locale) {
1484 setlocale(LC_TIME, save_time_locale);
1485 Safefree(save_time_locale);
1486 }
1487 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));
1488 }
1489 cant_use_time:
1490
1491#endif
1492
1493#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
855aeb93
JH
1494
1495/* This code is ifdefd out because it was found to not be necessary in testing
5857e934
KW
1496 * on our dromedary test machine, which has over 700 locales. There, this
1497 * added no value to looking at the currency symbol and the time strings. I
1498 * left it in so as to avoid rewriting it if real-world experience indicates
1499 * that dromedary is an outlier. Essentially, instead of returning abpve if we
855aeb93
JH
1500 * haven't found illegal utf8, we continue on and examine all the strerror()
1501 * messages on the platform for utf8ness. If all are ASCII, we still don't
1502 * know the answer; but otherwise we have a pretty good indication of the
5857e934
KW
1503 * utf8ness. The reason this doesn't help much is that the messages may not
1504 * have been translated into the locale. The currency symbol and time strings
1505 * are much more likely to have been translated. */
1506 {
855aeb93 1507 int e;
5857e934
KW
1508 bool is_utf8 = FALSE;
1509 bool non_ascii = FALSE;
855aeb93 1510 char *save_messages_locale = NULL;
5857e934 1511 const char * errmsg = NULL;
855aeb93 1512
5857e934
KW
1513 /* Like above, we set LC_MESSAGES to the locale of the desired
1514 * category, if it isn't that locale already */
855aeb93
JH
1515
1516 if (category != LC_MESSAGES) {
1517
5857e934 1518 save_messages_locale = setlocale(LC_MESSAGES, NULL);
855aeb93 1519 if (! save_messages_locale) {
5857e934
KW
1520 DEBUG_L(PerlIO_printf(Perl_debug_log,
1521 "Could not find current locale for LC_MESSAGES\n"));
855aeb93
JH
1522 goto cant_use_messages;
1523 }
5857e934 1524 save_messages_locale = stdize_locale(savepv(save_messages_locale));
855aeb93
JH
1525
1526 if (strEQ(save_messages_locale, save_input_locale)) {
5857e934
KW
1527 Safefree(save_messages_locale);
1528 save_messages_locale = NULL;
855aeb93
JH
1529 }
1530 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
5857e934
KW
1531 DEBUG_L(PerlIO_printf(Perl_debug_log,
1532 "Could not change LC_MESSAGES locale to %s\n",
1533 save_input_locale));
855aeb93
JH
1534 Safefree(save_messages_locale);
1535 goto cant_use_messages;
1536 }
1537 }
1538
1539 /* Here the current LC_MESSAGES is set to the locale of the category
5857e934
KW
1540 * whose information is desired. Look through all the messages. We
1541 * can't use Strerror() here because it may expand to code that
1542 * segfaults in miniperl */
855aeb93 1543
5857e934
KW
1544 for (e = 0; e <= sys_nerr; e++) {
1545 errno = 0;
1546 errmsg = sys_errlist[e];
1547 if (errno || !errmsg) {
855aeb93
JH
1548 break;
1549 }
5857e934 1550 errmsg = savepv(errmsg);
9f10db87 1551 if (! is_invariant_string((U8 *) errmsg, 0)) {
5857e934
KW
1552 non_ascii = TRUE;
1553 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1554 break;
855aeb93
JH
1555 }
1556 }
5857e934 1557 Safefree(errmsg);
855aeb93
JH
1558
1559 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1560 if (save_messages_locale) {
1561 setlocale(LC_MESSAGES, save_messages_locale);
1562 Safefree(save_messages_locale);
1563 }
1564
5857e934
KW
1565 if (non_ascii) {
1566
1567 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1568 * any non-ascii means it is one; otherwise we assume it isn't */
1569 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1570 save_input_locale,
1571 is_utf8));
1572 Safefree(save_input_locale);
1573 return is_utf8;
1574 }
855aeb93 1575
5857e934 1576 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
1577 }
1578 cant_use_messages:
1579
1580#endif
fa9b773e 1581
0080c90a
KW
1582#endif /* the code that is compiled when no nl_langinfo */
1583
92c0a900
KW
1584#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1585 UTF-8 locale */
97f4de96
KW
1586 /* As a last resort, look at the locale name to see if it matches
1587 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1588 * return of setlocale(), is actually defined to be opaque, so we can't
1589 * really rely on the absence of various substrings in the name to indicate
1590 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1591 * be a UTF-8 locale. Similarly for the other common names */
1592
1593 final_pos = strlen(save_input_locale) - 1;
1594 if (final_pos >= 3) {
1595 char *name = save_input_locale;
1596
1597 /* Find next 'U' or 'u' and look from there */
1598 while ((name += strcspn(name, "Uu") + 1)
1599 <= save_input_locale + final_pos - 2)
1600 {
305b8651
KW
1601 if (!isALPHA_FOLD_NE(*name, 't')
1602 || isALPHA_FOLD_NE(*(name + 1), 'f'))
97f4de96
KW
1603 {
1604 continue;
1605 }
1606 name += 2;
1607 if (*(name) == '-') {
1608 if ((name > save_input_locale + final_pos - 1)) {
1609 break;
1610 }
1611 name++;
1612 }
1613 if (*(name) == '8') {
97f4de96
KW
1614 DEBUG_L(PerlIO_printf(Perl_debug_log,
1615 "Locale %s ends with UTF-8 in name\n",
1616 save_input_locale));
00c54b9c 1617 Safefree(save_input_locale);
97f4de96
KW
1618 return TRUE;
1619 }
1620 }
1621 DEBUG_L(PerlIO_printf(Perl_debug_log,
1622 "Locale %s doesn't end with UTF-8 in name\n",
1623 save_input_locale));
1624 }
92c0a900 1625#endif
97f4de96
KW
1626
1627#ifdef WIN32
1628 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1629 if (final_pos >= 4
1630 && *(save_input_locale + final_pos - 0) == '1'
1631 && *(save_input_locale + final_pos - 1) == '0'
1632 && *(save_input_locale + final_pos - 2) == '0'
1633 && *(save_input_locale + final_pos - 3) == '5'
1634 && *(save_input_locale + final_pos - 4) == '6')
1635 {
1636 DEBUG_L(PerlIO_printf(Perl_debug_log,
1637 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1638 save_input_locale));
1639 Safefree(save_input_locale);
1640 return TRUE;
1641 }
1642#endif
1643
1644 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1645 * since we are about to return FALSE anyway, there is no point in doing
1646 * this extra work */
1647#if 0
1648 if (instr(save_input_locale, "8859")) {
1649 DEBUG_L(PerlIO_printf(Perl_debug_log,
1650 "Locale %s has 8859 in name, not UTF-8 locale\n",
1651 save_input_locale));
1652 Safefree(save_input_locale);
1653 return FALSE;
1654 }
1655#endif
1656
69014004
KW
1657 DEBUG_L(PerlIO_printf(Perl_debug_log,
1658 "Assuming locale %s is not a UTF-8 locale\n",
1659 save_input_locale));
fa9b773e 1660 Safefree(save_input_locale);
7d74bb61
KW
1661 return FALSE;
1662}
1663
8ef6e574 1664#endif
7d74bb61 1665
d6ded950
KW
1666
1667bool
1668Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1669{
1a4f13e1 1670 dVAR;
d6ded950
KW
1671 /* Internal function which returns if we are in the scope of a pragma that
1672 * enables the locale category 'category'. 'compiling' should indicate if
1673 * this is during the compilation phase (TRUE) or not (FALSE). */
1674
1675 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1676
1677 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1678 if (! categories || categories == &PL_sv_placeholder) {
1679 return FALSE;
1680 }
1681
1682 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1683 * a valid unsigned */
1684 assert(category >= -1);
1685 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1686}
1687
2c6ee1a7
KW
1688char *
1689Perl_my_strerror(pTHX_ const int errnum) {
1690
1691 /* Uses C locale for the error text unless within scope of 'use locale' for
1692 * LC_MESSAGES */
1693
1694#ifdef USE_LOCALE_MESSAGES
8ee4b769 1695 if (! IN_LC(LC_MESSAGES)) {
2c6ee1a7 1696 char * save_locale = setlocale(LC_MESSAGES, NULL);
a39edc4c 1697 if (! isNAME_C_OR_POSIX(save_locale)) {
2c6ee1a7
KW
1698 char *errstr;
1699
1700 /* The next setlocale likely will zap this, so create a copy */
1701 save_locale = savepv(save_locale);
1702
1703 setlocale(LC_MESSAGES, "C");
1704
1705 /* This points to the static space in Strerror, with all its
1706 * limitations */
1707 errstr = Strerror(errnum);
1708
1709 setlocale(LC_MESSAGES, save_locale);
1710 Safefree(save_locale);
1711 return errstr;
1712 }
1713 }
1714#endif
1715
1716 return Strerror(errnum);
1717}
1718
66610fdd 1719/*
747c467a
KW
1720
1721=head1 Locale-related functions and macros
1722
1723=for apidoc sync_locale
1724
1725Changing the program's locale should be avoided by XS code. Nevertheless,
1726certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1727happens, Perl needs to be told that the locale has changed. Use this function
1728to do so, before returning to Perl.
1729
1730=cut
1731*/
1732
1733void
1734Perl_sync_locale(pTHX)
1735{
1736
1737#ifdef USE_LOCALE_CTYPE
1738 new_ctype(setlocale(LC_CTYPE, NULL));
1739#endif /* USE_LOCALE_CTYPE */
1740
1741#ifdef USE_LOCALE_COLLATE
1742 new_collate(setlocale(LC_COLLATE, NULL));
1743#endif
1744
1745#ifdef USE_LOCALE_NUMERIC
1746 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1747 new_numeric(setlocale(LC_NUMERIC, NULL));
1748#endif /* USE_LOCALE_NUMERIC */
1749
1750}
1751
1752
1753
1754/*
14d04a33 1755 * ex: set ts=8 sts=4 sw=4 et:
37442d52 1756 */