This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
mktables: Under -annotate use dotted circle for combining marks
[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
56279a21 710 Safefree(curctype);
98994639 711 if (! (curctype =
b385bb4d 712 my_setlocale(LC_CTYPE,
98994639 713 (!done && (lang || PerlEnv_getenv("LC_CTYPE")))
ccd65d51 714 ? setlocale_init : NULL)))
98994639
HS
715 setlocale_failure = TRUE;
716 else
717 curctype = savepv(curctype);
b3e384bf
KW
718# endif /* USE_LOCALE_CTYPE */
719# ifdef USE_LOCALE_COLLATE
56279a21 720 Safefree(curcoll);
98994639 721 if (! (curcoll =
b385bb4d 722 my_setlocale(LC_COLLATE,
98994639 723 (!done && (lang || PerlEnv_getenv("LC_COLLATE")))
ccd65d51 724 ? setlocale_init : NULL)))
98994639
HS
725 setlocale_failure = TRUE;
726 else
727 curcoll = savepv(curcoll);
b3e384bf
KW
728# endif /* USE_LOCALE_COLLATE */
729# ifdef USE_LOCALE_NUMERIC
56279a21 730 Safefree(curnum);
98994639 731 if (! (curnum =
b385bb4d 732 my_setlocale(LC_NUMERIC,
98994639 733 (!done && (lang || PerlEnv_getenv("LC_NUMERIC")))
ccd65d51 734 ? setlocale_init : NULL)))
98994639
HS
735 setlocale_failure = TRUE;
736 else
737 curnum = savepv(curnum);
b3e384bf 738# endif /* USE_LOCALE_NUMERIC */
a782673d
KW
739# ifdef USE_LOCALE_MESSAGES
740 if (! my_setlocale(LC_MESSAGES,
741 (!done && (lang || PerlEnv_getenv("LC_MESSAGES")))
742 ? setlocale_init : NULL))
743 {
744 setlocale_failure = TRUE;
745 }
746# endif /* USE_LOCALE_MESSAGES */
c835d6be
KW
747# ifdef USE_LOCALE_MONETARY
748 if (! my_setlocale(LC_MONETARY,
749 (!done && (lang || PerlEnv_getenv("LC_MONETARY")))
750 ? setlocale_init : NULL))
751 {
752 setlocale_failure = TRUE;
753 }
754# endif /* USE_LOCALE_MONETARY */
98994639
HS
755 }
756
b3e384bf 757# endif /* LC_ALL */
98994639
HS
758
759#endif /* !LOCALE_ENVIRON_REQUIRED */
760
65ebb059
KW
761 /* We try each locale in the list until we get one that works, or exhaust
762 * the list */
c3fcd832
KW
763 trial_locales[0] = setlocale_init;
764 trial_locales_count = 1;
65ebb059
KW
765 for (i= 0; i < trial_locales_count; i++) {
766 const char * trial_locale = trial_locales[i];
767
768 if (i > 0) {
769
770 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
771 * when i==0, but I (khw) don't think that behavior makes much
772 * sense */
773 setlocale_failure = FALSE;
774
6bce99ee
JH
775#ifdef SYSTEM_DEFAULT_LOCALE
776# ifdef WIN32
65ebb059
KW
777 /* On Windows machines, an entry of "" after the 0th means to use
778 * the system default locale, which we now proceed to get. */
779 if (strEQ(trial_locale, "")) {
780 unsigned int j;
781
782 /* Note that this may change the locale, but we are going to do
783 * that anyway just below */
784 system_default_locale = setlocale(LC_ALL, "");
785
786 /* Skip if invalid or it's already on the list of locales to
787 * try */
788 if (! system_default_locale) {
789 goto next_iteration;
790 }
791 for (j = 0; j < trial_locales_count; j++) {
792 if (strEQ(system_default_locale, trial_locales[j])) {
793 goto next_iteration;
794 }
795 }
796
797 trial_locale = system_default_locale;
798 }
6bce99ee
JH
799# endif /* WIN32 */
800#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
801 }
802
98994639 803#ifdef LC_ALL
7cd8b568 804 if (! my_setlocale(LC_ALL, trial_locale)) {
49c85077 805 setlocale_failure = TRUE;
7cd8b568
KW
806 }
807 else {
808 /* Since LC_ALL succeeded, it should have changed all the other
809 * categories it can to its value; so we massage things so that the
810 * setlocales below just return their category's current values.
811 * This adequately handles the case in NetBSD where LC_COLLATE may
812 * not be defined for a locale, and setting it individually will
813 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
814 * the POSIX locale. */
815 trial_locale = NULL;
816 }
98994639
HS
817#endif /* LC_ALL */
818
49c85077 819 if (!setlocale_failure) {
98994639 820#ifdef USE_LOCALE_CTYPE
49c85077
KW
821 Safefree(curctype);
822 if (! (curctype = my_setlocale(LC_CTYPE, trial_locale)))
823 setlocale_failure = TRUE;
824 else
825 curctype = savepv(curctype);
98994639
HS
826#endif /* USE_LOCALE_CTYPE */
827#ifdef USE_LOCALE_COLLATE
49c85077
KW
828 Safefree(curcoll);
829 if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale)))
830 setlocale_failure = TRUE;
831 else
832 curcoll = savepv(curcoll);
98994639
HS
833#endif /* USE_LOCALE_COLLATE */
834#ifdef USE_LOCALE_NUMERIC
49c85077
KW
835 Safefree(curnum);
836 if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale)))
837 setlocale_failure = TRUE;
838 else
839 curnum = savepv(curnum);
98994639 840#endif /* USE_LOCALE_NUMERIC */
a782673d
KW
841#ifdef USE_LOCALE_MESSAGES
842 if (! (my_setlocale(LC_MESSAGES, trial_locale)))
843 setlocale_failure = TRUE;
844#endif /* USE_LOCALE_MESSAGES */
c835d6be
KW
845#ifdef USE_LOCALE_MONETARY
846 if (! (my_setlocale(LC_MONETARY, trial_locale)))
847 setlocale_failure = TRUE;
848#endif /* USE_LOCALE_MONETARY */
849
49c85077
KW
850 if (! setlocale_failure) { /* Success */
851 break;
852 }
65ebb059 853 }
98994639 854
49c85077
KW
855 /* Here, something failed; will need to try a fallback. */
856 ok = 0;
65ebb059 857
49c85077
KW
858 if (i == 0) {
859 unsigned int j;
98994639 860
65ebb059 861 if (locwarn) { /* Output failure info only on the first one */
98994639
HS
862#ifdef LC_ALL
863
49c85077
KW
864 PerlIO_printf(Perl_error_log,
865 "perl: warning: Setting locale failed.\n");
98994639
HS
866
867#else /* !LC_ALL */
868
49c85077
KW
869 PerlIO_printf(Perl_error_log,
870 "perl: warning: Setting locale failed for the categories:\n\t");
98994639 871#ifdef USE_LOCALE_CTYPE
49c85077
KW
872 if (! curctype)
873 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
98994639
HS
874#endif /* USE_LOCALE_CTYPE */
875#ifdef USE_LOCALE_COLLATE
49c85077
KW
876 if (! curcoll)
877 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
98994639
HS
878#endif /* USE_LOCALE_COLLATE */
879#ifdef USE_LOCALE_NUMERIC
49c85077
KW
880 if (! curnum)
881 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
98994639 882#endif /* USE_LOCALE_NUMERIC */
a782673d 883 PerlIO_printf(Perl_error_log, "and possibly others\n");
98994639
HS
884
885#endif /* LC_ALL */
886
49c85077
KW
887 PerlIO_printf(Perl_error_log,
888 "perl: warning: Please check that your locale settings:\n");
98994639
HS
889
890#ifdef __GLIBC__
49c85077
KW
891 PerlIO_printf(Perl_error_log,
892 "\tLANGUAGE = %c%s%c,\n",
893 language ? '"' : '(',
894 language ? language : "unset",
895 language ? '"' : ')');
98994639
HS
896#endif
897
49c85077
KW
898 PerlIO_printf(Perl_error_log,
899 "\tLC_ALL = %c%s%c,\n",
900 lc_all ? '"' : '(',
901 lc_all ? lc_all : "unset",
902 lc_all ? '"' : ')');
98994639
HS
903
904#if defined(USE_ENVIRON_ARRAY)
49c85077
KW
905 {
906 char **e;
907 for (e = environ; *e; e++) {
908 if (strnEQ(*e, "LC_", 3)
909 && strnNE(*e, "LC_ALL=", 7)
910 && (p = strchr(*e, '=')))
911 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
912 (int)(p - *e), *e, p + 1);
913 }
914 }
98994639 915#else
49c85077
KW
916 PerlIO_printf(Perl_error_log,
917 "\t(possibly more locale environment variables)\n");
98994639
HS
918#endif
919
49c85077
KW
920 PerlIO_printf(Perl_error_log,
921 "\tLANG = %c%s%c\n",
922 lang ? '"' : '(',
923 lang ? lang : "unset",
924 lang ? '"' : ')');
98994639 925
49c85077
KW
926 PerlIO_printf(Perl_error_log,
927 " are supported and installed on your system.\n");
928 }
98994639 929
65ebb059 930 /* Calculate what fallback locales to try. We have avoided this
f6bab5f6 931 * until we have to, because failure is quite unlikely. This will
65ebb059
KW
932 * usually change the upper bound of the loop we are in.
933 *
934 * Since the system's default way of setting the locale has not
935 * found one that works, We use Perl's defined ordering: LC_ALL,
936 * LANG, and the C locale. We don't try the same locale twice, so
937 * don't add to the list if already there. (On POSIX systems, the
938 * LC_ALL element will likely be a repeat of the 0th element "",
939 * but there's no harm done by doing it explicitly */
940 if (lc_all) {
941 for (j = 0; j < trial_locales_count; j++) {
942 if (strEQ(lc_all, trial_locales[j])) {
943 goto done_lc_all;
944 }
945 }
946 trial_locales[trial_locales_count++] = lc_all;
947 }
948 done_lc_all:
98994639 949
65ebb059
KW
950 if (lang) {
951 for (j = 0; j < trial_locales_count; j++) {
952 if (strEQ(lang, trial_locales[j])) {
953 goto done_lang;
954 }
955 }
956 trial_locales[trial_locales_count++] = lang;
957 }
958 done_lang:
959
960#if defined(WIN32) && defined(LC_ALL)
961 /* For Windows, we also try the system default locale before "C".
962 * (If there exists a Windows without LC_ALL we skip this because
963 * it gets too complicated. For those, the "C" is the next
964 * fallback possibility). The "" is the same as the 0th element of
965 * the array, but the code at the loop above knows to treat it
966 * differently when not the 0th */
967 trial_locales[trial_locales_count++] = "";
968#endif
969
970 for (j = 0; j < trial_locales_count; j++) {
971 if (strEQ("C", trial_locales[j])) {
972 goto done_C;
973 }
974 }
975 trial_locales[trial_locales_count++] = "C";
98994639 976
65ebb059
KW
977 done_C: ;
978 } /* end of first time through the loop */
98994639 979
65ebb059
KW
980#ifdef WIN32
981 next_iteration: ;
982#endif
983
984 } /* end of looping through the trial locales */
985
986 if (ok < 1) { /* If we tried to fallback */
987 const char* msg;
988 if (! setlocale_failure) { /* fallback succeeded */
989 msg = "Falling back to";
990 }
991 else { /* fallback failed */
98994639 992
65ebb059
KW
993 /* We dropped off the end of the loop, so have to decrement i to
994 * get back to the value the last time through */
995 i--;
98994639 996
65ebb059
KW
997 ok = -1;
998 msg = "Failed to fall back to";
999
1000 /* To continue, we should use whatever values we've got */
98994639 1001#ifdef USE_LOCALE_CTYPE
49c85077
KW
1002 Safefree(curctype);
1003 curctype = savepv(setlocale(LC_CTYPE, NULL));
98994639
HS
1004#endif /* USE_LOCALE_CTYPE */
1005#ifdef USE_LOCALE_COLLATE
49c85077
KW
1006 Safefree(curcoll);
1007 curcoll = savepv(setlocale(LC_COLLATE, NULL));
98994639
HS
1008#endif /* USE_LOCALE_COLLATE */
1009#ifdef USE_LOCALE_NUMERIC
49c85077
KW
1010 Safefree(curnum);
1011 curnum = savepv(setlocale(LC_NUMERIC, NULL));
98994639 1012#endif /* USE_LOCALE_NUMERIC */
65ebb059
KW
1013 }
1014
1015 if (locwarn) {
1016 const char * description;
1017 const char * name = "";
1018 if (strEQ(trial_locales[i], "C")) {
1019 description = "the standard locale";
1020 name = "C";
1021 }
6bce99ee 1022#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059
KW
1023 else if (strEQ(trial_locales[i], "")) {
1024 description = "the system default locale";
1025 if (system_default_locale) {
1026 name = system_default_locale;
1027 }
1028 }
6bce99ee 1029#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
1030 else {
1031 description = "a fallback locale";
1032 name = trial_locales[i];
1033 }
1034 if (name && strNE(name, "")) {
1035 PerlIO_printf(Perl_error_log,
1036 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1037 }
1038 else {
1039 PerlIO_printf(Perl_error_log,
1040 "perl: warning: %s %s.\n", msg, description);
1041 }
1042 }
1043 } /* End of tried to fallback */
98994639
HS
1044
1045#ifdef USE_LOCALE_CTYPE
1046 new_ctype(curctype);
1047#endif /* USE_LOCALE_CTYPE */
1048
1049#ifdef USE_LOCALE_COLLATE
1050 new_collate(curcoll);
1051#endif /* USE_LOCALE_COLLATE */
1052
1053#ifdef USE_LOCALE_NUMERIC
1054 new_numeric(curnum);
1055#endif /* USE_LOCALE_NUMERIC */
b310b053 1056
8ef6e574 1057#if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
49c85077
KW
1058 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1059 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1060 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1061 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1062 * discipline. */
c1284011 1063 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
49c85077 1064
a05d7ebb 1065 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
fde18df1
JH
1066 This is an alternative to using the -C command line switch
1067 (the -C if present will override this). */
1068 {
dd374669 1069 const char *p = PerlEnv_getenv("PERL_UNICODE");
a05d7ebb 1070 PL_unicode = p ? parse_unicode_opts(&p) : 0;
5a22a2bb
NC
1071 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1072 PL_utf8cache = -1;
b310b053 1073 }
ec71e770 1074#endif
b310b053 1075
98994639 1076#ifdef USE_LOCALE_CTYPE
43c5f42d 1077 Safefree(curctype);
98994639
HS
1078#endif /* USE_LOCALE_CTYPE */
1079#ifdef USE_LOCALE_COLLATE
43c5f42d 1080 Safefree(curcoll);
98994639
HS
1081#endif /* USE_LOCALE_COLLATE */
1082#ifdef USE_LOCALE_NUMERIC
43c5f42d 1083 Safefree(curnum);
98994639 1084#endif /* USE_LOCALE_NUMERIC */
8ef6e574 1085
175c4cf9
KW
1086#ifdef __GLIBC__
1087 Safefree(language);
1088#endif
1089
1090 Safefree(lc_all);
1091 Safefree(lang);
1092
e3305790
KW
1093#else /* !USE_LOCALE */
1094 PERL_UNUSED_ARG(printwarn);
1095#endif /* USE_LOCALE */
1096
98994639
HS
1097 return ok;
1098}
1099
49c85077 1100
98994639
HS
1101#ifdef USE_LOCALE_COLLATE
1102
1103/*
1104 * mem_collxfrm() is a bit like strxfrm() but with two important
1105 * differences. First, it handles embedded NULs. Second, it allocates
1106 * a bit more memory than needed for the transformed data itself.
1107 * The real transformed data begins at offset sizeof(collationix).
1108 * Please see sv_collxfrm() to see how this is used.
1109 */
166f8a29 1110
98994639
HS
1111char *
1112Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1113{
1114 char *xbuf;
1115 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1116
7918f24d
NC
1117 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1118
98994639
HS
1119 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1120 /* the +1 is for the terminating NUL. */
1121
1122 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
a02a5408 1123 Newx(xbuf, xAlloc, char);
98994639
HS
1124 if (! xbuf)
1125 goto bad;
1126
1127 *(U32*)xbuf = PL_collation_ix;
1128 xout = sizeof(PL_collation_ix);
1129 for (xin = 0; xin < len; ) {
224e8ef5 1130 Size_t xused;
98994639
HS
1131
1132 for (;;) {
1133 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
224e8ef5 1134 if (xused >= PERL_INT_MAX)
98994639 1135 goto bad;
eb160463 1136 if ((STRLEN)xused < xAlloc - xout)
98994639
HS
1137 break;
1138 xAlloc = (2 * xAlloc) + 1;
1139 Renew(xbuf, xAlloc, char);
1140 if (! xbuf)
1141 goto bad;
1142 }
1143
1144 xin += strlen(s + xin) + 1;
1145 xout += xused;
1146
1147 /* Embedded NULs are understood but silently skipped
1148 * because they make no sense in locale collation. */
1149 }
1150
1151 xbuf[xout] = '\0';
1152 *xlen = xout - sizeof(PL_collation_ix);
1153 return xbuf;
1154
1155 bad:
1156 Safefree(xbuf);
1157 *xlen = 0;
1158 return NULL;
1159}
1160
1161#endif /* USE_LOCALE_COLLATE */
1162
8ef6e574
KW
1163#ifdef USE_LOCALE
1164
c1284011
KW
1165bool
1166Perl__is_cur_LC_category_utf8(pTHX_ int category)
7d74bb61
KW
1167{
1168 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1169 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
119ee68b 1170 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
609548d2
KW
1171 * could give the wrong result. The result will very likely be correct for
1172 * languages that have commonly used non-ASCII characters, but for notably
1173 * English, it comes down to if the locale's name ends in something like
1174 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
7d74bb61
KW
1175
1176 char *save_input_locale = NULL;
7d74bb61
KW
1177 STRLEN final_pos;
1178
8ef6e574 1179#ifdef LC_ALL
7d74bb61 1180 assert(category != LC_ALL);
8ef6e574 1181#endif
7d74bb61
KW
1182
1183 /* First dispose of the trivial cases */
b07fffd1 1184 save_input_locale = setlocale(category, NULL);
7d74bb61 1185 if (! save_input_locale) {
69014004
KW
1186 DEBUG_L(PerlIO_printf(Perl_debug_log,
1187 "Could not find current locale for category %d\n",
1188 category));
7d74bb61
KW
1189 return FALSE; /* XXX maybe should croak */
1190 }
b07fffd1 1191 save_input_locale = stdize_locale(savepv(save_input_locale));
a39edc4c 1192 if (isNAME_C_OR_POSIX(save_input_locale)) {
69014004
KW
1193 DEBUG_L(PerlIO_printf(Perl_debug_log,
1194 "Current locale for category %d is %s\n",
1195 category, save_input_locale));
b07fffd1 1196 Safefree(save_input_locale);
7d74bb61
KW
1197 return FALSE;
1198 }
1199
1d958db2
KW
1200#if defined(USE_LOCALE_CTYPE) \
1201 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
7d74bb61 1202
1d958db2 1203 { /* Next try nl_langinfo or MB_CUR_MAX if available */
7d74bb61
KW
1204
1205 char *save_ctype_locale = NULL;
119ee68b 1206 bool is_utf8;
7d74bb61 1207
119ee68b 1208 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
7d74bb61
KW
1209
1210 /* Get the current LC_CTYPE locale */
4f72bb37 1211 save_ctype_locale = setlocale(LC_CTYPE, NULL);
7d74bb61 1212 if (! save_ctype_locale) {
69014004
KW
1213 DEBUG_L(PerlIO_printf(Perl_debug_log,
1214 "Could not find current locale for LC_CTYPE\n"));
7d74bb61
KW
1215 goto cant_use_nllanginfo;
1216 }
4f72bb37 1217 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
7d74bb61
KW
1218
1219 /* If LC_CTYPE and the desired category use the same locale, this
1220 * means that finding the value for LC_CTYPE is the same as finding
1221 * the value for the desired category. Otherwise, switch LC_CTYPE
1222 * to the desired category's locale */
1223 if (strEQ(save_ctype_locale, save_input_locale)) {
1224 Safefree(save_ctype_locale);
1225 save_ctype_locale = NULL;
1226 }
1227 else if (! setlocale(LC_CTYPE, save_input_locale)) {
69014004
KW
1228 DEBUG_L(PerlIO_printf(Perl_debug_log,
1229 "Could not change LC_CTYPE locale to %s\n",
1230 save_input_locale));
7d74bb61
KW
1231 Safefree(save_ctype_locale);
1232 goto cant_use_nllanginfo;
1233 }
1234 }
1235
69014004
KW
1236 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1237 save_input_locale));
1238
7d74bb61 1239 /* Here the current LC_CTYPE is set to the locale of the category whose
1d958db2
KW
1240 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1241 * should give the correct results */
119ee68b 1242
1d958db2
KW
1243# if defined(HAS_NL_LANGINFO) && defined(CODESET)
1244 {
4f72bb37 1245 char *codeset = nl_langinfo(CODESET);
1d958db2 1246 if (codeset && strNE(codeset, "")) {
4f72bb37 1247 codeset = savepv(codeset);
119ee68b 1248
1d958db2
KW
1249 /* If we switched LC_CTYPE, switch back */
1250 if (save_ctype_locale) {
1251 setlocale(LC_CTYPE, save_ctype_locale);
1252 Safefree(save_ctype_locale);
1253 }
1254
1255 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1256 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1257
69014004
KW
1258 DEBUG_L(PerlIO_printf(Perl_debug_log,
1259 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1260 codeset, is_utf8));
1d958db2
KW
1261 Safefree(codeset);
1262 Safefree(save_input_locale);
1263 return is_utf8;
1264 }
119ee68b
KW
1265 }
1266
1d958db2
KW
1267# endif
1268# ifdef MB_CUR_MAX
1269
1270 /* Here, either we don't have nl_langinfo, or it didn't return a
1271 * codeset. Try MB_CUR_MAX */
1272
119ee68b
KW
1273 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1274 * Unicode code point. Since UTF-8 is the only non-single byte
1275 * encoding we handle, we just say any such encoding is UTF-8, and if
1276 * turns out to be wrong, other things will fail */
1277 is_utf8 = MB_CUR_MAX >= 4;
1278
69014004
KW
1279 DEBUG_L(PerlIO_printf(Perl_debug_log,
1280 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1281 (int) MB_CUR_MAX, is_utf8));
1282
119ee68b
KW
1283 Safefree(save_input_locale);
1284
1285# ifdef HAS_MBTOWC
1286
1287 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1288 * since they are both in the C99 standard. We can feed a known byte
1289 * string to the latter function, and check that it gives the expected
1290 * result */
1291 if (is_utf8) {
1292 wchar_t wc;
856b881c 1293 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
69014004 1294 errno = 0;
f019f68f 1295 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
119ee68b
KW
1296 != strlen(HYPHEN_UTF8)
1297 || wc != (wchar_t) 0x2010)
1298 {
1299 is_utf8 = FALSE;
abdcbdb8 1300 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
69014004
KW
1301 DEBUG_L(PerlIO_printf(Perl_debug_log,
1302 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1303 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
119ee68b
KW
1304 }
1305 }
119ee68b
KW
1306# endif
1307
1d958db2
KW
1308 /* If we switched LC_CTYPE, switch back */
1309 if (save_ctype_locale) {
1310 setlocale(LC_CTYPE, save_ctype_locale);
1311 Safefree(save_ctype_locale);
119ee68b 1312 }
7d74bb61 1313
1d958db2 1314 return is_utf8;
119ee68b 1315# endif
7d74bb61 1316 }
119ee68b 1317
7d74bb61
KW
1318 cant_use_nllanginfo:
1319
0080c90a
KW
1320#else /* nl_langinfo should work if available, so don't bother compiling this
1321 fallback code. The final fallback of looking at the name is
1322 compiled, and will be executed if nl_langinfo fails */
7d74bb61 1323
97f4de96
KW
1324 /* nl_langinfo not available or failed somehow. Next try looking at the
1325 * currency symbol to see if it disambiguates things. Often that will be
1326 * in the native script, and if the symbol isn't in UTF-8, we know that the
1327 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
609548d2
KW
1328 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1329 * */
fa9b773e
KW
1330
1331#ifdef HAS_LOCALECONV
fa9b773e 1332# ifdef USE_LOCALE_MONETARY
fa9b773e
KW
1333 {
1334 char *save_monetary_locale = NULL;
fa9b773e 1335 bool only_ascii = FALSE;
13542a67
KW
1336 bool is_utf8 = FALSE;
1337 struct lconv* lc;
fa9b773e 1338
97f4de96
KW
1339 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1340 * the desired category, if it isn't that locale already */
1341
fa9b773e
KW
1342 if (category != LC_MONETARY) {
1343
4f72bb37 1344 save_monetary_locale = setlocale(LC_MONETARY, NULL);
fa9b773e 1345 if (! save_monetary_locale) {
69014004
KW
1346 DEBUG_L(PerlIO_printf(Perl_debug_log,
1347 "Could not find current locale for LC_MONETARY\n"));
fa9b773e
KW
1348 goto cant_use_monetary;
1349 }
4f72bb37 1350 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
fa9b773e 1351
13542a67
KW
1352 if (strEQ(save_monetary_locale, save_input_locale)) {
1353 Safefree(save_monetary_locale);
1354 save_monetary_locale = NULL;
1355 }
1356 else if (! setlocale(LC_MONETARY, save_input_locale)) {
59c234b4
KW
1357 DEBUG_L(PerlIO_printf(Perl_debug_log,
1358 "Could not change LC_MONETARY locale to %s\n",
1359 save_input_locale));
1360 Safefree(save_monetary_locale);
1361 goto cant_use_monetary;
fa9b773e
KW
1362 }
1363 }
1364
1365 /* Here the current LC_MONETARY is set to the locale of the category
1366 * whose information is desired. */
1367
13542a67
KW
1368 lc = localeconv();
1369 if (! lc
1370 || ! lc->currency_symbol
9f10db87 1371 || is_invariant_string((U8 *) lc->currency_symbol, 0))
13542a67
KW
1372 {
1373 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));
1374 only_ascii = TRUE;
1375 }
1376 else {
1377 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
fa9b773e
KW
1378 }
1379
1380 /* If we changed it, restore LC_MONETARY to its original locale */
1381 if (save_monetary_locale) {
1382 setlocale(LC_MONETARY, save_monetary_locale);
1383 Safefree(save_monetary_locale);
1384 }
1385
13542a67 1386 if (! only_ascii) {
fa9b773e 1387
59c234b4
KW
1388 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1389 * otherwise assume the locale is UTF-8 if and only if the symbol
1390 * is non-ascii UTF-8. */
1391 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1392 save_input_locale, is_utf8));
1393 Safefree(save_input_locale);
1394 return is_utf8;
13542a67 1395 }
fa9b773e
KW
1396 }
1397 cant_use_monetary:
1398
1399# endif /* USE_LOCALE_MONETARY */
1400#endif /* HAS_LOCALECONV */
1401
15f7e74e
KW
1402#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1403
1404/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1405 * the names of the months and weekdays, timezone, and am/pm indicator */
1406 {
1407 char *save_time_locale = NULL;
1408 int hour = 10;
1409 bool is_dst = FALSE;
1410 int dom = 1;
1411 int month = 0;
1412 int i;
1413 char * formatted_time;
1414
1415
1416 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1417 * desired category, if it isn't that locale already */
1418
1419 if (category != LC_TIME) {
1420
1421 save_time_locale = setlocale(LC_TIME, NULL);
1422 if (! save_time_locale) {
1423 DEBUG_L(PerlIO_printf(Perl_debug_log,
1424 "Could not find current locale for LC_TIME\n"));
1425 goto cant_use_time;
1426 }
1427 save_time_locale = stdize_locale(savepv(save_time_locale));
1428
1429 if (strEQ(save_time_locale, save_input_locale)) {
1430 Safefree(save_time_locale);
1431 save_time_locale = NULL;
1432 }
1433 else if (! setlocale(LC_TIME, save_input_locale)) {
1434 DEBUG_L(PerlIO_printf(Perl_debug_log,
1435 "Could not change LC_TIME locale to %s\n",
1436 save_input_locale));
1437 Safefree(save_time_locale);
1438 goto cant_use_time;
1439 }
1440 }
1441
1442 /* Here the current LC_TIME is set to the locale of the category
1443 * whose information is desired. Look at all the days of the week and
9f10db87 1444 * month names, and the timezone and am/pm indicator for UTF-8 variant
15f7e74e
KW
1445 * characters. The first such a one found will tell us if the locale
1446 * is UTF-8 or not */
1447
1448 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1449 formatted_time = my_strftime("%A %B %Z %p",
1450 0, 0, hour, dom, month, 112, 0, 0, is_dst);
9f10db87 1451 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
15f7e74e
KW
1452
1453 /* Here, we didn't find a non-ASCII. Try the next time through
1454 * with the complemented dst and am/pm, and try with the next
1455 * weekday. After we have gotten all weekdays, try the next
1456 * month */
1457 is_dst = ! is_dst;
1458 hour = (hour + 12) % 24;
1459 dom++;
1460 if (i > 6) {
1461 month++;
1462 }
1463 continue;
1464 }
1465
1466 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1467 * false otherwise. But first, restore LC_TIME to its original
1468 * locale if we changed it */
1469 if (save_time_locale) {
1470 setlocale(LC_TIME, save_time_locale);
1471 Safefree(save_time_locale);
1472 }
1473
1474 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1475 save_input_locale,
1476 is_utf8_string((U8 *) formatted_time, 0)));
1477 Safefree(save_input_locale);
1478 return is_utf8_string((U8 *) formatted_time, 0);
1479 }
1480
1481 /* Falling off the end of the loop indicates all the names were just
1482 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1483 * to its original locale */
1484 if (save_time_locale) {
1485 setlocale(LC_TIME, save_time_locale);
1486 Safefree(save_time_locale);
1487 }
1488 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));
1489 }
1490 cant_use_time:
1491
1492#endif
1493
1494#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
855aeb93
JH
1495
1496/* This code is ifdefd out because it was found to not be necessary in testing
5857e934
KW
1497 * on our dromedary test machine, which has over 700 locales. There, this
1498 * added no value to looking at the currency symbol and the time strings. I
1499 * left it in so as to avoid rewriting it if real-world experience indicates
1500 * that dromedary is an outlier. Essentially, instead of returning abpve if we
855aeb93
JH
1501 * haven't found illegal utf8, we continue on and examine all the strerror()
1502 * messages on the platform for utf8ness. If all are ASCII, we still don't
1503 * know the answer; but otherwise we have a pretty good indication of the
5857e934
KW
1504 * utf8ness. The reason this doesn't help much is that the messages may not
1505 * have been translated into the locale. The currency symbol and time strings
1506 * are much more likely to have been translated. */
1507 {
855aeb93 1508 int e;
5857e934
KW
1509 bool is_utf8 = FALSE;
1510 bool non_ascii = FALSE;
855aeb93 1511 char *save_messages_locale = NULL;
5857e934 1512 const char * errmsg = NULL;
855aeb93 1513
5857e934
KW
1514 /* Like above, we set LC_MESSAGES to the locale of the desired
1515 * category, if it isn't that locale already */
855aeb93
JH
1516
1517 if (category != LC_MESSAGES) {
1518
5857e934 1519 save_messages_locale = setlocale(LC_MESSAGES, NULL);
855aeb93 1520 if (! save_messages_locale) {
5857e934
KW
1521 DEBUG_L(PerlIO_printf(Perl_debug_log,
1522 "Could not find current locale for LC_MESSAGES\n"));
855aeb93
JH
1523 goto cant_use_messages;
1524 }
5857e934 1525 save_messages_locale = stdize_locale(savepv(save_messages_locale));
855aeb93
JH
1526
1527 if (strEQ(save_messages_locale, save_input_locale)) {
5857e934
KW
1528 Safefree(save_messages_locale);
1529 save_messages_locale = NULL;
855aeb93
JH
1530 }
1531 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
5857e934
KW
1532 DEBUG_L(PerlIO_printf(Perl_debug_log,
1533 "Could not change LC_MESSAGES locale to %s\n",
1534 save_input_locale));
855aeb93
JH
1535 Safefree(save_messages_locale);
1536 goto cant_use_messages;
1537 }
1538 }
1539
1540 /* Here the current LC_MESSAGES is set to the locale of the category
5857e934
KW
1541 * whose information is desired. Look through all the messages. We
1542 * can't use Strerror() here because it may expand to code that
1543 * segfaults in miniperl */
855aeb93 1544
5857e934
KW
1545 for (e = 0; e <= sys_nerr; e++) {
1546 errno = 0;
1547 errmsg = sys_errlist[e];
1548 if (errno || !errmsg) {
855aeb93
JH
1549 break;
1550 }
5857e934 1551 errmsg = savepv(errmsg);
9f10db87 1552 if (! is_invariant_string((U8 *) errmsg, 0)) {
5857e934
KW
1553 non_ascii = TRUE;
1554 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1555 break;
855aeb93
JH
1556 }
1557 }
5857e934 1558 Safefree(errmsg);
855aeb93
JH
1559
1560 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1561 if (save_messages_locale) {
1562 setlocale(LC_MESSAGES, save_messages_locale);
1563 Safefree(save_messages_locale);
1564 }
1565
5857e934
KW
1566 if (non_ascii) {
1567
1568 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1569 * any non-ascii means it is one; otherwise we assume it isn't */
1570 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1571 save_input_locale,
1572 is_utf8));
1573 Safefree(save_input_locale);
1574 return is_utf8;
1575 }
855aeb93 1576
5857e934 1577 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
1578 }
1579 cant_use_messages:
1580
1581#endif
fa9b773e 1582
0080c90a
KW
1583#endif /* the code that is compiled when no nl_langinfo */
1584
92c0a900
KW
1585#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1586 UTF-8 locale */
97f4de96
KW
1587 /* As a last resort, look at the locale name to see if it matches
1588 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1589 * return of setlocale(), is actually defined to be opaque, so we can't
1590 * really rely on the absence of various substrings in the name to indicate
1591 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1592 * be a UTF-8 locale. Similarly for the other common names */
1593
1594 final_pos = strlen(save_input_locale) - 1;
1595 if (final_pos >= 3) {
1596 char *name = save_input_locale;
1597
1598 /* Find next 'U' or 'u' and look from there */
1599 while ((name += strcspn(name, "Uu") + 1)
1600 <= save_input_locale + final_pos - 2)
1601 {
305b8651
KW
1602 if (!isALPHA_FOLD_NE(*name, 't')
1603 || isALPHA_FOLD_NE(*(name + 1), 'f'))
97f4de96
KW
1604 {
1605 continue;
1606 }
1607 name += 2;
1608 if (*(name) == '-') {
1609 if ((name > save_input_locale + final_pos - 1)) {
1610 break;
1611 }
1612 name++;
1613 }
1614 if (*(name) == '8') {
97f4de96
KW
1615 DEBUG_L(PerlIO_printf(Perl_debug_log,
1616 "Locale %s ends with UTF-8 in name\n",
1617 save_input_locale));
00c54b9c 1618 Safefree(save_input_locale);
97f4de96
KW
1619 return TRUE;
1620 }
1621 }
1622 DEBUG_L(PerlIO_printf(Perl_debug_log,
1623 "Locale %s doesn't end with UTF-8 in name\n",
1624 save_input_locale));
1625 }
92c0a900 1626#endif
97f4de96
KW
1627
1628#ifdef WIN32
1629 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1630 if (final_pos >= 4
1631 && *(save_input_locale + final_pos - 0) == '1'
1632 && *(save_input_locale + final_pos - 1) == '0'
1633 && *(save_input_locale + final_pos - 2) == '0'
1634 && *(save_input_locale + final_pos - 3) == '5'
1635 && *(save_input_locale + final_pos - 4) == '6')
1636 {
1637 DEBUG_L(PerlIO_printf(Perl_debug_log,
1638 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1639 save_input_locale));
1640 Safefree(save_input_locale);
1641 return TRUE;
1642 }
1643#endif
1644
1645 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1646 * since we are about to return FALSE anyway, there is no point in doing
1647 * this extra work */
1648#if 0
1649 if (instr(save_input_locale, "8859")) {
1650 DEBUG_L(PerlIO_printf(Perl_debug_log,
1651 "Locale %s has 8859 in name, not UTF-8 locale\n",
1652 save_input_locale));
1653 Safefree(save_input_locale);
1654 return FALSE;
1655 }
1656#endif
1657
69014004
KW
1658 DEBUG_L(PerlIO_printf(Perl_debug_log,
1659 "Assuming locale %s is not a UTF-8 locale\n",
1660 save_input_locale));
fa9b773e 1661 Safefree(save_input_locale);
7d74bb61
KW
1662 return FALSE;
1663}
1664
8ef6e574 1665#endif
7d74bb61 1666
d6ded950
KW
1667
1668bool
1669Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1670{
1a4f13e1 1671 dVAR;
d6ded950
KW
1672 /* Internal function which returns if we are in the scope of a pragma that
1673 * enables the locale category 'category'. 'compiling' should indicate if
1674 * this is during the compilation phase (TRUE) or not (FALSE). */
1675
1676 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1677
1678 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1679 if (! categories || categories == &PL_sv_placeholder) {
1680 return FALSE;
1681 }
1682
1683 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1684 * a valid unsigned */
1685 assert(category >= -1);
1686 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1687}
1688
2c6ee1a7
KW
1689char *
1690Perl_my_strerror(pTHX_ const int errnum) {
1691
1692 /* Uses C locale for the error text unless within scope of 'use locale' for
1693 * LC_MESSAGES */
1694
1695#ifdef USE_LOCALE_MESSAGES
8ee4b769 1696 if (! IN_LC(LC_MESSAGES)) {
2c6ee1a7 1697 char * save_locale = setlocale(LC_MESSAGES, NULL);
a39edc4c 1698 if (! isNAME_C_OR_POSIX(save_locale)) {
2c6ee1a7
KW
1699 char *errstr;
1700
1701 /* The next setlocale likely will zap this, so create a copy */
1702 save_locale = savepv(save_locale);
1703
1704 setlocale(LC_MESSAGES, "C");
1705
1706 /* This points to the static space in Strerror, with all its
1707 * limitations */
1708 errstr = Strerror(errnum);
1709
1710 setlocale(LC_MESSAGES, save_locale);
1711 Safefree(save_locale);
1712 return errstr;
1713 }
1714 }
1715#endif
1716
1717 return Strerror(errnum);
1718}
1719
66610fdd 1720/*
747c467a
KW
1721
1722=head1 Locale-related functions and macros
1723
1724=for apidoc sync_locale
1725
1726Changing the program's locale should be avoided by XS code. Nevertheless,
1727certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1728happens, Perl needs to be told that the locale has changed. Use this function
1729to do so, before returning to Perl.
1730
1731=cut
1732*/
1733
1734void
1735Perl_sync_locale(pTHX)
1736{
1737
1738#ifdef USE_LOCALE_CTYPE
1739 new_ctype(setlocale(LC_CTYPE, NULL));
1740#endif /* USE_LOCALE_CTYPE */
1741
1742#ifdef USE_LOCALE_COLLATE
1743 new_collate(setlocale(LC_COLLATE, NULL));
1744#endif
1745
1746#ifdef USE_LOCALE_NUMERIC
1747 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1748 new_numeric(setlocale(LC_NUMERIC, NULL));
1749#endif /* USE_LOCALE_NUMERIC */
1750
1751}
1752
1753
1754
1755/*
14d04a33 1756 * ex: set ts=8 sts=4 sw=4 et:
37442d52 1757 */