3 * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
4 * 2002, 2003, 2005, 2006, 2007, 2008 by Larry Wall and others
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.
12 * A Elbereth Gilthoniel,
13 * silivren penna míriel
14 * o menel aglar elenath!
15 * Na-chaered palan-díriel
16 * o galadhremmin ennorath,
17 * Fanuilos, le linnathon
18 * nef aear, si nef aearon!
20 * [p.238 of _The Lord of the Rings_, II/i: "Many Meetings"]
23 /* utility functions for handling locale-specific stuff like what
24 * character represents the decimal point.
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.
36 #define PERL_IN_LOCALE_C
40 # include <langinfo.h>
48 * Standardize the locale name from a string returned by 'setlocale', possibly
49 * modifying that string.
51 * The typical return value of setlocale() is either
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,
55 * in some unspecified order). This is not handled by this function.
57 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
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 '='
63 S_stdize_locale(pTHX_ char *locs)
65 const char * const s = strchr(locs, '=');
68 PERL_ARGS_ASSERT_STDIZE_LOCALE;
71 const char * const t = strchr(s, '.');
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);
85 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
93 Perl_set_numeric_radix(pTHX)
95 #ifdef USE_LOCALE_NUMERIC
96 # ifdef HAS_LOCALECONV
97 const struct lconv* const lc = localeconv();
99 if (lc && lc->decimal_point) {
100 if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) {
101 SvREFCNT_dec(PL_numeric_radix_sv);
102 PL_numeric_radix_sv = NULL;
105 if (PL_numeric_radix_sv)
106 sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
108 PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
109 if (! is_invariant_string((U8 *) lc->decimal_point, 0)
110 && is_utf8_string((U8 *) lc->decimal_point, 0)
111 && _is_cur_LC_category_utf8(LC_NUMERIC))
113 SvUTF8_on(PL_numeric_radix_sv);
118 PL_numeric_radix_sv = NULL;
120 DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is %s\n",
121 (PL_numeric_radix_sv)
125 # endif /* HAS_LOCALECONV */
126 #endif /* USE_LOCALE_NUMERIC */
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 */
139 #define isNAME_C_OR_POSIX(name) ((name) != NULL \
140 && ((*(name) == 'C' && (*(name + 1)) == '\0') \
141 || strEQ((name), "POSIX")))
144 Perl_new_numeric(pTHX_ const char *newnum)
146 #ifdef USE_LOCALE_NUMERIC
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.
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.
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
161 * This sets several interpreter-level variables:
162 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
163 * PL_numeric_local A boolean indicating if the toggled state is such
164 * that the current locale is the program's underlying
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
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
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() */
182 Safefree(PL_numeric_name);
183 PL_numeric_name = NULL;
184 PL_numeric_standard = TRUE;
185 PL_numeric_local = TRUE;
189 save_newnum = stdize_locale(savepv(newnum));
190 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
191 Safefree(PL_numeric_name);
192 PL_numeric_name = save_newnum;
195 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
196 PL_numeric_local = TRUE;
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();
206 PERL_UNUSED_ARG(newnum);
207 #endif /* USE_LOCALE_NUMERIC */
211 Perl_set_numeric_standard(pTHX)
213 #ifdef USE_LOCALE_NUMERIC
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) */
220 setlocale(LC_NUMERIC, "C");
221 PL_numeric_standard = TRUE;
222 PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
224 DEBUG_L(PerlIO_printf(Perl_debug_log,
225 "Underlying LC_NUMERIC locale now is C\n"));
227 #endif /* USE_LOCALE_NUMERIC */
231 Perl_set_numeric_local(pTHX)
233 #ifdef USE_LOCALE_NUMERIC
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) */
240 setlocale(LC_NUMERIC, PL_numeric_name);
241 PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
242 PL_numeric_local = TRUE;
244 DEBUG_L(PerlIO_printf(Perl_debug_log,
245 "Underlying LC_NUMERIC locale now is %s\n",
248 #endif /* USE_LOCALE_NUMERIC */
252 * Set up for a new ctype locale.
255 Perl_new_ctype(pTHX_ const char *newctype)
257 #ifdef USE_LOCALE_CTYPE
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.
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,
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() */
273 PERL_ARGS_ASSERT_NEW_CTYPE;
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;
282 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
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);
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
294 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
296 bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
298 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
300 unsigned int bad_count = 0; /* Count of bad characters */
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);
308 PL_fold_locale[i] = (U8) i;
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'))
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)))
327 if (bad_count) { /* Separate multiple entries with a
329 bad_chars_list[bad_count++] = ' ';
331 bad_chars_list[bad_count++] = '\'';
333 bad_chars_list[bad_count++] = (char) i;
336 bad_chars_list[bad_count++] = '\\';
338 bad_chars_list[bad_count++] = 'n';
342 bad_chars_list[bad_count++] = 't';
345 bad_chars_list[bad_count++] = '\'';
346 bad_chars_list[bad_count] = '\0';
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
355 if (check_for_problems && MB_CUR_MAX > 1
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"))
365 multi_byte_locale = TRUE;
369 if (bad_count || multi_byte_locale) {
370 PL_warn_locale = Perl_newSVpvf(aTHX_
371 "Locale '%s' may not work well.%s%s%s\n",
374 ? " Some characters in it are not recognized by"
378 ? "\nThe following characters (and maybe others)"
379 " may not have the same meaning as the Perl"
380 " program expects:\n"
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)) {
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);
398 setlocale(LC_CTYPE, "C");
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);
404 SvREFCNT_dec_NN(PL_warn_locale);
405 PL_warn_locale = NULL;
410 #endif /* USE_LOCALE_CTYPE */
411 PERL_ARGS_ASSERT_NEW_CTYPE;
412 PERL_UNUSED_ARG(newctype);
417 Perl__warn_problematic_locale()
420 #ifdef USE_LOCALE_CTYPE
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 */
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;
443 Perl_new_collate(pTHX_ const char *newcoll)
445 #ifdef USE_LOCALE_COLLATE
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.
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() */
456 if (PL_collation_name) {
458 Safefree(PL_collation_name);
459 PL_collation_name = NULL;
461 PL_collation_standard = TRUE;
462 PL_collxfrm_base = 0;
463 PL_collxfrm_mult = 2;
467 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
469 Safefree(PL_collation_name);
470 PL_collation_name = stdize_locale(savepv(newcoll));
471 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
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];
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;
481 if (mult < 1 && !(fa == 0 && fb == 0))
482 Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
484 PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
485 PL_collxfrm_mult = mult;
490 PERL_UNUSED_ARG(newcoll);
491 #endif /* USE_LOCALE_COLLATE */
497 Perl_my_setlocale(pTHX_ int category, const char* locale)
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. */
511 bool override_LC_ALL = FALSE;
514 if (locale && strEQ(locale, "")) {
516 locale = PerlEnv_getenv("LC_ALL");
522 override_LC_ALL = TRUE;
523 break; /* We already know its variable isn't set */
525 # ifdef USE_LOCALE_TIME
527 locale = PerlEnv_getenv("LC_TIME");
530 # ifdef USE_LOCALE_CTYPE
532 locale = PerlEnv_getenv("LC_CTYPE");
535 # ifdef USE_LOCALE_COLLATE
537 locale = PerlEnv_getenv("LC_COLLATE");
540 # ifdef USE_LOCALE_MONETARY
542 locale = PerlEnv_getenv("LC_MONETARY");
545 # ifdef USE_LOCALE_NUMERIC
547 locale = PerlEnv_getenv("LC_NUMERIC");
550 # ifdef USE_LOCALE_MESSAGES
552 locale = PerlEnv_getenv("LC_MESSAGES");
556 /* This is a category, like PAPER_SIZE that we don't
557 * know about; and so can't provide a wrapper. */
561 locale = PerlEnv_getenv("LANG");
571 result = setlocale(category, locale);
573 if (! override_LC_ALL) {
577 /* Here the input category was LC_ALL, and we have set it to what is in the
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) */
582 # ifdef USE_LOCALE_TIME
583 result = PerlEnv_getenv("LC_TIME");
584 if (result && strNE(result, "")) {
585 setlocale(LC_TIME, result);
588 # ifdef USE_LOCALE_CTYPE
589 result = PerlEnv_getenv("LC_CTYPE");
590 if (result && strNE(result, "")) {
591 setlocale(LC_CTYPE, result);
594 # ifdef USE_LOCALE_COLLATE
595 result = PerlEnv_getenv("LC_COLLATE");
596 if (result && strNE(result, "")) {
597 setlocale(LC_COLLATE, result);
600 # ifdef USE_LOCALE_MONETARY
601 result = PerlEnv_getenv("LC_MONETARY");
602 if (result && strNE(result, "")) {
603 setlocale(LC_MONETARY, result);
606 # ifdef USE_LOCALE_NUMERIC
607 result = PerlEnv_getenv("LC_NUMERIC");
608 if (result && strNE(result, "")) {
609 setlocale(LC_NUMERIC, result);
612 # ifdef USE_LOCALE_MESSAGES
613 result = PerlEnv_getenv("LC_MESSAGES");
614 if (result && strNE(result, "")) {
615 setlocale(LC_MESSAGES, result);
619 return setlocale(LC_ALL, NULL);
627 * Initialize locale awareness.
630 Perl_init_i18nl10n(pTHX_ int printwarn)
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
639 * 1 = set ok or not applicable,
640 * 0 = fallback to a locale of lower priority
641 * -1 = fallback to all locales failed, not even to the C locale
646 #if defined(USE_LOCALE)
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
655 #endif /* USE_LOCALE_NUMERIC */
657 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
660 /* NULL uses the existing already set up locale */
661 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
664 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
665 unsigned int trial_locales_count;
666 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
667 const char * const lang = savepv(PerlEnv_getenv("LANG"));
668 bool setlocale_failure = FALSE;
672 /* A later getenv() could zap this, so only use here */
673 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
675 const bool locwarn = (printwarn > 1
677 && (! bad_lang_use_once
679 /* disallow with "" or "0" */
681 && strNE("0", bad_lang_use_once)))));
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
688 #ifdef SYSTEM_DEFAULT_LOCALE
689 const char *system_default_locale = NULL;
692 #ifndef LOCALE_ENVIRON_REQUIRED
693 PERL_UNUSED_VAR(done);
697 * Ultrix setlocale(..., "") fails if there are no environment
698 * variables from which to get a locale name.
703 if (my_setlocale(LC_ALL, setlocale_init))
706 setlocale_failure = TRUE;
708 if (!setlocale_failure) {
709 # ifdef USE_LOCALE_CTYPE
712 my_setlocale(LC_CTYPE,
713 (!done && (lang || PerlEnv_getenv("LC_CTYPE")))
714 ? setlocale_init : NULL)))
715 setlocale_failure = TRUE;
717 curctype = savepv(curctype);
718 # endif /* USE_LOCALE_CTYPE */
719 # ifdef USE_LOCALE_COLLATE
722 my_setlocale(LC_COLLATE,
723 (!done && (lang || PerlEnv_getenv("LC_COLLATE")))
724 ? setlocale_init : NULL)))
725 setlocale_failure = TRUE;
727 curcoll = savepv(curcoll);
728 # endif /* USE_LOCALE_COLLATE */
729 # ifdef USE_LOCALE_NUMERIC
732 my_setlocale(LC_NUMERIC,
733 (!done && (lang || PerlEnv_getenv("LC_NUMERIC")))
734 ? setlocale_init : NULL)))
735 setlocale_failure = TRUE;
737 curnum = savepv(curnum);
738 # endif /* USE_LOCALE_NUMERIC */
739 # ifdef USE_LOCALE_MESSAGES
740 if (! my_setlocale(LC_MESSAGES,
741 (!done && (lang || PerlEnv_getenv("LC_MESSAGES")))
742 ? setlocale_init : NULL))
744 setlocale_failure = TRUE;
746 # endif /* USE_LOCALE_MESSAGES */
747 # ifdef USE_LOCALE_MONETARY
748 if (! my_setlocale(LC_MONETARY,
749 (!done && (lang || PerlEnv_getenv("LC_MONETARY")))
750 ? setlocale_init : NULL))
752 setlocale_failure = TRUE;
754 # endif /* USE_LOCALE_MONETARY */
759 #endif /* !LOCALE_ENVIRON_REQUIRED */
761 /* We try each locale in the list until we get one that works, or exhaust
762 * the list. Normally the loop is executed just once. But if setting the
763 * locale fails, inside the loop we add fallback trials to the array and so
764 * will execute the loop multiple times */
765 trial_locales[0] = setlocale_init;
766 trial_locales_count = 1;
767 for (i= 0; i < trial_locales_count; i++) {
768 const char * trial_locale = trial_locales[i];
772 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
773 * when i==0, but I (khw) don't think that behavior makes much
775 setlocale_failure = FALSE;
777 #ifdef SYSTEM_DEFAULT_LOCALE
779 /* On Windows machines, an entry of "" after the 0th means to use
780 * the system default locale, which we now proceed to get. */
781 if (strEQ(trial_locale, "")) {
784 /* Note that this may change the locale, but we are going to do
785 * that anyway just below */
786 system_default_locale = setlocale(LC_ALL, "");
788 /* Skip if invalid or it's already on the list of locales to
790 if (! system_default_locale) {
793 for (j = 0; j < trial_locales_count; j++) {
794 if (strEQ(system_default_locale, trial_locales[j])) {
799 trial_locale = system_default_locale;
802 #endif /* SYSTEM_DEFAULT_LOCALE */
806 if (! my_setlocale(LC_ALL, trial_locale)) {
807 setlocale_failure = TRUE;
810 /* Since LC_ALL succeeded, it should have changed all the other
811 * categories it can to its value; so we massage things so that the
812 * setlocales below just return their category's current values.
813 * This adequately handles the case in NetBSD where LC_COLLATE may
814 * not be defined for a locale, and setting it individually will
815 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
816 * the POSIX locale. */
821 if (!setlocale_failure) {
822 #ifdef USE_LOCALE_CTYPE
824 if (! (curctype = my_setlocale(LC_CTYPE, trial_locale)))
825 setlocale_failure = TRUE;
827 curctype = savepv(curctype);
828 #endif /* USE_LOCALE_CTYPE */
829 #ifdef USE_LOCALE_COLLATE
831 if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale)))
832 setlocale_failure = TRUE;
834 curcoll = savepv(curcoll);
835 #endif /* USE_LOCALE_COLLATE */
836 #ifdef USE_LOCALE_NUMERIC
838 if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale)))
839 setlocale_failure = TRUE;
841 curnum = savepv(curnum);
842 #endif /* USE_LOCALE_NUMERIC */
843 #ifdef USE_LOCALE_MESSAGES
844 if (! (my_setlocale(LC_MESSAGES, trial_locale)))
845 setlocale_failure = TRUE;
846 #endif /* USE_LOCALE_MESSAGES */
847 #ifdef USE_LOCALE_MONETARY
848 if (! (my_setlocale(LC_MONETARY, trial_locale)))
849 setlocale_failure = TRUE;
850 #endif /* USE_LOCALE_MONETARY */
852 if (! setlocale_failure) { /* Success */
857 /* Here, something failed; will need to try a fallback. */
863 if (locwarn) { /* Output failure info only on the first one */
866 PerlIO_printf(Perl_error_log,
867 "perl: warning: Setting locale failed.\n");
871 PerlIO_printf(Perl_error_log,
872 "perl: warning: Setting locale failed for the categories:\n\t");
873 # ifdef USE_LOCALE_CTYPE
875 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
876 # endif /* USE_LOCALE_CTYPE */
877 # ifdef USE_LOCALE_COLLATE
879 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
880 # endif /* USE_LOCALE_COLLATE */
881 # ifdef USE_LOCALE_NUMERIC
883 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
884 # endif /* USE_LOCALE_NUMERIC */
885 PerlIO_printf(Perl_error_log, "and possibly others\n");
889 PerlIO_printf(Perl_error_log,
890 "perl: warning: Please check that your locale settings:\n");
893 PerlIO_printf(Perl_error_log,
894 "\tLANGUAGE = %c%s%c,\n",
895 language ? '"' : '(',
896 language ? language : "unset",
897 language ? '"' : ')');
900 PerlIO_printf(Perl_error_log,
901 "\tLC_ALL = %c%s%c,\n",
903 lc_all ? lc_all : "unset",
906 #if defined(USE_ENVIRON_ARRAY)
909 for (e = environ; *e; e++) {
910 if (strnEQ(*e, "LC_", 3)
911 && strnNE(*e, "LC_ALL=", 7)
912 && (p = strchr(*e, '=')))
913 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
914 (int)(p - *e), *e, p + 1);
918 PerlIO_printf(Perl_error_log,
919 "\t(possibly more locale environment variables)\n");
922 PerlIO_printf(Perl_error_log,
925 lang ? lang : "unset",
928 PerlIO_printf(Perl_error_log,
929 " are supported and installed on your system.\n");
932 /* Calculate what fallback locales to try. We have avoided this
933 * until we have to, because failure is quite unlikely. This will
934 * usually change the upper bound of the loop we are in.
936 * Since the system's default way of setting the locale has not
937 * found one that works, We use Perl's defined ordering: LC_ALL,
938 * LANG, and the C locale. We don't try the same locale twice, so
939 * don't add to the list if already there. (On POSIX systems, the
940 * LC_ALL element will likely be a repeat of the 0th element "",
941 * but there's no harm done by doing it explicitly */
943 for (j = 0; j < trial_locales_count; j++) {
944 if (strEQ(lc_all, trial_locales[j])) {
948 trial_locales[trial_locales_count++] = lc_all;
953 for (j = 0; j < trial_locales_count; j++) {
954 if (strEQ(lang, trial_locales[j])) {
958 trial_locales[trial_locales_count++] = lang;
962 #if defined(WIN32) && defined(LC_ALL)
963 /* For Windows, we also try the system default locale before "C".
964 * (If there exists a Windows without LC_ALL we skip this because
965 * it gets too complicated. For those, the "C" is the next
966 * fallback possibility). The "" is the same as the 0th element of
967 * the array, but the code at the loop above knows to treat it
968 * differently when not the 0th */
969 trial_locales[trial_locales_count++] = "";
972 for (j = 0; j < trial_locales_count; j++) {
973 if (strEQ("C", trial_locales[j])) {
977 trial_locales[trial_locales_count++] = "C";
980 } /* end of first time through the loop */
986 } /* end of looping through the trial locales */
988 if (ok < 1) { /* If we tried to fallback */
990 if (! setlocale_failure) { /* fallback succeeded */
991 msg = "Falling back to";
993 else { /* fallback failed */
995 /* We dropped off the end of the loop, so have to decrement i to
996 * get back to the value the last time through */
1000 msg = "Failed to fall back to";
1002 /* To continue, we should use whatever values we've got */
1003 #ifdef USE_LOCALE_CTYPE
1005 curctype = savepv(setlocale(LC_CTYPE, NULL));
1006 #endif /* USE_LOCALE_CTYPE */
1007 #ifdef USE_LOCALE_COLLATE
1009 curcoll = savepv(setlocale(LC_COLLATE, NULL));
1010 #endif /* USE_LOCALE_COLLATE */
1011 #ifdef USE_LOCALE_NUMERIC
1013 curnum = savepv(setlocale(LC_NUMERIC, NULL));
1014 #endif /* USE_LOCALE_NUMERIC */
1018 const char * description;
1019 const char * name = "";
1020 if (strEQ(trial_locales[i], "C")) {
1021 description = "the standard locale";
1024 #ifdef SYSTEM_DEFAULT_LOCALE
1025 else if (strEQ(trial_locales[i], "")) {
1026 description = "the system default locale";
1027 if (system_default_locale) {
1028 name = system_default_locale;
1031 #endif /* SYSTEM_DEFAULT_LOCALE */
1033 description = "a fallback locale";
1034 name = trial_locales[i];
1036 if (name && strNE(name, "")) {
1037 PerlIO_printf(Perl_error_log,
1038 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1041 PerlIO_printf(Perl_error_log,
1042 "perl: warning: %s %s.\n", msg, description);
1045 } /* End of tried to fallback */
1047 #ifdef USE_LOCALE_CTYPE
1048 new_ctype(curctype);
1049 #endif /* USE_LOCALE_CTYPE */
1051 #ifdef USE_LOCALE_COLLATE
1052 new_collate(curcoll);
1053 #endif /* USE_LOCALE_COLLATE */
1055 #ifdef USE_LOCALE_NUMERIC
1056 new_numeric(curnum);
1057 #endif /* USE_LOCALE_NUMERIC */
1059 #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
1060 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1061 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1062 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1063 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1065 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
1067 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
1068 This is an alternative to using the -C command line switch
1069 (the -C if present will override this). */
1071 const char *p = PerlEnv_getenv("PERL_UNICODE");
1072 PL_unicode = p ? parse_unicode_opts(&p) : 0;
1073 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1078 #ifdef USE_LOCALE_CTYPE
1080 #endif /* USE_LOCALE_CTYPE */
1081 #ifdef USE_LOCALE_COLLATE
1083 #endif /* USE_LOCALE_COLLATE */
1084 #ifdef USE_LOCALE_NUMERIC
1086 #endif /* USE_LOCALE_NUMERIC */
1095 #else /* !USE_LOCALE */
1096 PERL_UNUSED_ARG(printwarn);
1097 #endif /* USE_LOCALE */
1103 #ifdef USE_LOCALE_COLLATE
1106 * mem_collxfrm() is a bit like strxfrm() but with two important
1107 * differences. First, it handles embedded NULs. Second, it allocates
1108 * a bit more memory than needed for the transformed data itself.
1109 * The real transformed data begins at offset sizeof(collationix).
1110 * Please see sv_collxfrm() to see how this is used.
1114 Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1117 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1119 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1121 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1122 /* the +1 is for the terminating NUL. */
1124 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
1125 Newx(xbuf, xAlloc, char);
1129 *(U32*)xbuf = PL_collation_ix;
1130 xout = sizeof(PL_collation_ix);
1131 for (xin = 0; xin < len; ) {
1135 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
1136 if (xused >= PERL_INT_MAX)
1138 if ((STRLEN)xused < xAlloc - xout)
1140 xAlloc = (2 * xAlloc) + 1;
1141 Renew(xbuf, xAlloc, char);
1146 xin += strlen(s + xin) + 1;
1149 /* Embedded NULs are understood but silently skipped
1150 * because they make no sense in locale collation. */
1154 *xlen = xout - sizeof(PL_collation_ix);
1163 #endif /* USE_LOCALE_COLLATE */
1168 Perl__is_cur_LC_category_utf8(pTHX_ int category)
1170 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1171 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
1172 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
1173 * could give the wrong result. The result will very likely be correct for
1174 * languages that have commonly used non-ASCII characters, but for notably
1175 * English, it comes down to if the locale's name ends in something like
1176 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
1178 char *save_input_locale = NULL;
1182 assert(category != LC_ALL);
1185 /* First dispose of the trivial cases */
1186 save_input_locale = setlocale(category, NULL);
1187 if (! save_input_locale) {
1188 DEBUG_L(PerlIO_printf(Perl_debug_log,
1189 "Could not find current locale for category %d\n",
1191 return FALSE; /* XXX maybe should croak */
1193 save_input_locale = stdize_locale(savepv(save_input_locale));
1194 if (isNAME_C_OR_POSIX(save_input_locale)) {
1195 DEBUG_L(PerlIO_printf(Perl_debug_log,
1196 "Current locale for category %d is %s\n",
1197 category, save_input_locale));
1198 Safefree(save_input_locale);
1202 #if defined(USE_LOCALE_CTYPE) \
1203 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
1205 { /* Next try nl_langinfo or MB_CUR_MAX if available */
1207 char *save_ctype_locale = NULL;
1210 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
1212 /* Get the current LC_CTYPE locale */
1213 save_ctype_locale = setlocale(LC_CTYPE, NULL);
1214 if (! save_ctype_locale) {
1215 DEBUG_L(PerlIO_printf(Perl_debug_log,
1216 "Could not find current locale for LC_CTYPE\n"));
1217 goto cant_use_nllanginfo;
1219 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
1221 /* If LC_CTYPE and the desired category use the same locale, this
1222 * means that finding the value for LC_CTYPE is the same as finding
1223 * the value for the desired category. Otherwise, switch LC_CTYPE
1224 * to the desired category's locale */
1225 if (strEQ(save_ctype_locale, save_input_locale)) {
1226 Safefree(save_ctype_locale);
1227 save_ctype_locale = NULL;
1229 else if (! setlocale(LC_CTYPE, save_input_locale)) {
1230 DEBUG_L(PerlIO_printf(Perl_debug_log,
1231 "Could not change LC_CTYPE locale to %s\n",
1232 save_input_locale));
1233 Safefree(save_ctype_locale);
1234 goto cant_use_nllanginfo;
1238 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1239 save_input_locale));
1241 /* Here the current LC_CTYPE is set to the locale of the category whose
1242 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1243 * should give the correct results */
1245 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
1247 char *codeset = nl_langinfo(CODESET);
1248 if (codeset && strNE(codeset, "")) {
1249 codeset = savepv(codeset);
1251 /* If we switched LC_CTYPE, switch back */
1252 if (save_ctype_locale) {
1253 setlocale(LC_CTYPE, save_ctype_locale);
1254 Safefree(save_ctype_locale);
1257 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1258 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1260 DEBUG_L(PerlIO_printf(Perl_debug_log,
1261 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1264 Safefree(save_input_locale);
1272 /* Here, either we don't have nl_langinfo, or it didn't return a
1273 * codeset. Try MB_CUR_MAX */
1275 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1276 * Unicode code point. Since UTF-8 is the only non-single byte
1277 * encoding we handle, we just say any such encoding is UTF-8, and if
1278 * turns out to be wrong, other things will fail */
1279 is_utf8 = MB_CUR_MAX >= 4;
1281 DEBUG_L(PerlIO_printf(Perl_debug_log,
1282 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1283 (int) MB_CUR_MAX, is_utf8));
1285 Safefree(save_input_locale);
1289 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1290 * since they are both in the C99 standard. We can feed a known byte
1291 * string to the latter function, and check that it gives the expected
1295 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
1297 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
1298 != strlen(HYPHEN_UTF8)
1299 || wc != (wchar_t) 0x2010)
1302 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
1303 DEBUG_L(PerlIO_printf(Perl_debug_log,
1304 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1305 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
1310 /* If we switched LC_CTYPE, switch back */
1311 if (save_ctype_locale) {
1312 setlocale(LC_CTYPE, save_ctype_locale);
1313 Safefree(save_ctype_locale);
1320 cant_use_nllanginfo:
1322 #else /* nl_langinfo should work if available, so don't bother compiling this
1323 fallback code. The final fallback of looking at the name is
1324 compiled, and will be executed if nl_langinfo fails */
1326 /* nl_langinfo not available or failed somehow. Next try looking at the
1327 * currency symbol to see if it disambiguates things. Often that will be
1328 * in the native script, and if the symbol isn't in UTF-8, we know that the
1329 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
1330 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1333 #ifdef HAS_LOCALECONV
1334 # ifdef USE_LOCALE_MONETARY
1336 char *save_monetary_locale = NULL;
1337 bool only_ascii = FALSE;
1338 bool is_utf8 = FALSE;
1341 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1342 * the desired category, if it isn't that locale already */
1344 if (category != LC_MONETARY) {
1346 save_monetary_locale = setlocale(LC_MONETARY, NULL);
1347 if (! save_monetary_locale) {
1348 DEBUG_L(PerlIO_printf(Perl_debug_log,
1349 "Could not find current locale for LC_MONETARY\n"));
1350 goto cant_use_monetary;
1352 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
1354 if (strEQ(save_monetary_locale, save_input_locale)) {
1355 Safefree(save_monetary_locale);
1356 save_monetary_locale = NULL;
1358 else if (! setlocale(LC_MONETARY, save_input_locale)) {
1359 DEBUG_L(PerlIO_printf(Perl_debug_log,
1360 "Could not change LC_MONETARY locale to %s\n",
1361 save_input_locale));
1362 Safefree(save_monetary_locale);
1363 goto cant_use_monetary;
1367 /* Here the current LC_MONETARY is set to the locale of the category
1368 * whose information is desired. */
1372 || ! lc->currency_symbol
1373 || is_invariant_string((U8 *) lc->currency_symbol, 0))
1375 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));
1379 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
1382 /* If we changed it, restore LC_MONETARY to its original locale */
1383 if (save_monetary_locale) {
1384 setlocale(LC_MONETARY, save_monetary_locale);
1385 Safefree(save_monetary_locale);
1390 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1391 * otherwise assume the locale is UTF-8 if and only if the symbol
1392 * is non-ascii UTF-8. */
1393 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1394 save_input_locale, is_utf8));
1395 Safefree(save_input_locale);
1401 # endif /* USE_LOCALE_MONETARY */
1402 #endif /* HAS_LOCALECONV */
1404 #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1406 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1407 * the names of the months and weekdays, timezone, and am/pm indicator */
1409 char *save_time_locale = NULL;
1411 bool is_dst = FALSE;
1415 char * formatted_time;
1418 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1419 * desired category, if it isn't that locale already */
1421 if (category != LC_TIME) {
1423 save_time_locale = setlocale(LC_TIME, NULL);
1424 if (! save_time_locale) {
1425 DEBUG_L(PerlIO_printf(Perl_debug_log,
1426 "Could not find current locale for LC_TIME\n"));
1429 save_time_locale = stdize_locale(savepv(save_time_locale));
1431 if (strEQ(save_time_locale, save_input_locale)) {
1432 Safefree(save_time_locale);
1433 save_time_locale = NULL;
1435 else if (! setlocale(LC_TIME, save_input_locale)) {
1436 DEBUG_L(PerlIO_printf(Perl_debug_log,
1437 "Could not change LC_TIME locale to %s\n",
1438 save_input_locale));
1439 Safefree(save_time_locale);
1444 /* Here the current LC_TIME is set to the locale of the category
1445 * whose information is desired. Look at all the days of the week and
1446 * month names, and the timezone and am/pm indicator for UTF-8 variant
1447 * characters. The first such a one found will tell us if the locale
1448 * is UTF-8 or not */
1450 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1451 formatted_time = my_strftime("%A %B %Z %p",
1452 0, 0, hour, dom, month, 112, 0, 0, is_dst);
1453 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
1455 /* Here, we didn't find a non-ASCII. Try the next time through
1456 * with the complemented dst and am/pm, and try with the next
1457 * weekday. After we have gotten all weekdays, try the next
1460 hour = (hour + 12) % 24;
1468 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1469 * false otherwise. But first, restore LC_TIME to its original
1470 * locale if we changed it */
1471 if (save_time_locale) {
1472 setlocale(LC_TIME, save_time_locale);
1473 Safefree(save_time_locale);
1476 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1478 is_utf8_string((U8 *) formatted_time, 0)));
1479 Safefree(save_input_locale);
1480 return is_utf8_string((U8 *) formatted_time, 0);
1483 /* Falling off the end of the loop indicates all the names were just
1484 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1485 * to its original locale */
1486 if (save_time_locale) {
1487 setlocale(LC_TIME, save_time_locale);
1488 Safefree(save_time_locale);
1490 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));
1496 #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
1498 /* This code is ifdefd out because it was found to not be necessary in testing
1499 * on our dromedary test machine, which has over 700 locales. There, this
1500 * added no value to looking at the currency symbol and the time strings. I
1501 * left it in so as to avoid rewriting it if real-world experience indicates
1502 * that dromedary is an outlier. Essentially, instead of returning abpve if we
1503 * haven't found illegal utf8, we continue on and examine all the strerror()
1504 * messages on the platform for utf8ness. If all are ASCII, we still don't
1505 * know the answer; but otherwise we have a pretty good indication of the
1506 * utf8ness. The reason this doesn't help much is that the messages may not
1507 * have been translated into the locale. The currency symbol and time strings
1508 * are much more likely to have been translated. */
1511 bool is_utf8 = FALSE;
1512 bool non_ascii = FALSE;
1513 char *save_messages_locale = NULL;
1514 const char * errmsg = NULL;
1516 /* Like above, we set LC_MESSAGES to the locale of the desired
1517 * category, if it isn't that locale already */
1519 if (category != LC_MESSAGES) {
1521 save_messages_locale = setlocale(LC_MESSAGES, NULL);
1522 if (! save_messages_locale) {
1523 DEBUG_L(PerlIO_printf(Perl_debug_log,
1524 "Could not find current locale for LC_MESSAGES\n"));
1525 goto cant_use_messages;
1527 save_messages_locale = stdize_locale(savepv(save_messages_locale));
1529 if (strEQ(save_messages_locale, save_input_locale)) {
1530 Safefree(save_messages_locale);
1531 save_messages_locale = NULL;
1533 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
1534 DEBUG_L(PerlIO_printf(Perl_debug_log,
1535 "Could not change LC_MESSAGES locale to %s\n",
1536 save_input_locale));
1537 Safefree(save_messages_locale);
1538 goto cant_use_messages;
1542 /* Here the current LC_MESSAGES is set to the locale of the category
1543 * whose information is desired. Look through all the messages. We
1544 * can't use Strerror() here because it may expand to code that
1545 * segfaults in miniperl */
1547 for (e = 0; e <= sys_nerr; e++) {
1549 errmsg = sys_errlist[e];
1550 if (errno || !errmsg) {
1553 errmsg = savepv(errmsg);
1554 if (! is_invariant_string((U8 *) errmsg, 0)) {
1556 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1562 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1563 if (save_messages_locale) {
1564 setlocale(LC_MESSAGES, save_messages_locale);
1565 Safefree(save_messages_locale);
1570 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1571 * any non-ascii means it is one; otherwise we assume it isn't */
1572 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1575 Safefree(save_input_locale);
1579 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));
1585 #endif /* the code that is compiled when no nl_langinfo */
1587 #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1589 /* As a last resort, look at the locale name to see if it matches
1590 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1591 * return of setlocale(), is actually defined to be opaque, so we can't
1592 * really rely on the absence of various substrings in the name to indicate
1593 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1594 * be a UTF-8 locale. Similarly for the other common names */
1596 final_pos = strlen(save_input_locale) - 1;
1597 if (final_pos >= 3) {
1598 char *name = save_input_locale;
1600 /* Find next 'U' or 'u' and look from there */
1601 while ((name += strcspn(name, "Uu") + 1)
1602 <= save_input_locale + final_pos - 2)
1604 if (!isALPHA_FOLD_NE(*name, 't')
1605 || isALPHA_FOLD_NE(*(name + 1), 'f'))
1610 if (*(name) == '-') {
1611 if ((name > save_input_locale + final_pos - 1)) {
1616 if (*(name) == '8') {
1617 DEBUG_L(PerlIO_printf(Perl_debug_log,
1618 "Locale %s ends with UTF-8 in name\n",
1619 save_input_locale));
1620 Safefree(save_input_locale);
1624 DEBUG_L(PerlIO_printf(Perl_debug_log,
1625 "Locale %s doesn't end with UTF-8 in name\n",
1626 save_input_locale));
1631 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1633 && *(save_input_locale + final_pos - 0) == '1'
1634 && *(save_input_locale + final_pos - 1) == '0'
1635 && *(save_input_locale + final_pos - 2) == '0'
1636 && *(save_input_locale + final_pos - 3) == '5'
1637 && *(save_input_locale + final_pos - 4) == '6')
1639 DEBUG_L(PerlIO_printf(Perl_debug_log,
1640 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1641 save_input_locale));
1642 Safefree(save_input_locale);
1647 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1648 * since we are about to return FALSE anyway, there is no point in doing
1649 * this extra work */
1651 if (instr(save_input_locale, "8859")) {
1652 DEBUG_L(PerlIO_printf(Perl_debug_log,
1653 "Locale %s has 8859 in name, not UTF-8 locale\n",
1654 save_input_locale));
1655 Safefree(save_input_locale);
1660 DEBUG_L(PerlIO_printf(Perl_debug_log,
1661 "Assuming locale %s is not a UTF-8 locale\n",
1662 save_input_locale));
1663 Safefree(save_input_locale);
1671 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1674 /* Internal function which returns if we are in the scope of a pragma that
1675 * enables the locale category 'category'. 'compiling' should indicate if
1676 * this is during the compilation phase (TRUE) or not (FALSE). */
1678 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1680 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1681 if (! categories || categories == &PL_sv_placeholder) {
1685 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1686 * a valid unsigned */
1687 assert(category >= -1);
1688 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1692 Perl_my_strerror(pTHX_ const int errnum) {
1694 /* Uses C locale for the error text unless within scope of 'use locale' for
1697 #ifdef USE_LOCALE_MESSAGES
1698 if (! IN_LC(LC_MESSAGES)) {
1699 char * save_locale = setlocale(LC_MESSAGES, NULL);
1700 if (! isNAME_C_OR_POSIX(save_locale)) {
1703 /* The next setlocale likely will zap this, so create a copy */
1704 save_locale = savepv(save_locale);
1706 setlocale(LC_MESSAGES, "C");
1708 /* This points to the static space in Strerror, with all its
1710 errstr = Strerror(errnum);
1712 setlocale(LC_MESSAGES, save_locale);
1713 Safefree(save_locale);
1719 return Strerror(errnum);
1724 =head1 Locale-related functions and macros
1726 =for apidoc sync_locale
1728 Changing the program's locale should be avoided by XS code. Nevertheless,
1729 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1730 happens, Perl needs to be told that the locale has changed. Use this function
1731 to do so, before returning to Perl.
1737 Perl_sync_locale(pTHX)
1740 #ifdef USE_LOCALE_CTYPE
1741 new_ctype(setlocale(LC_CTYPE, NULL));
1742 #endif /* USE_LOCALE_CTYPE */
1744 #ifdef USE_LOCALE_COLLATE
1745 new_collate(setlocale(LC_COLLATE, NULL));
1748 #ifdef USE_LOCALE_NUMERIC
1749 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1750 new_numeric(setlocale(LC_NUMERIC, NULL));
1751 #endif /* USE_LOCALE_NUMERIC */
1758 * ex: set ts=8 sts=4 sw=4 et: