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. And, LC_MESSAGES is
33 * switched to the C locale for outputting the message unless within the scope
38 #define PERL_IN_LOCALE_C
42 # include <langinfo.h>
50 * Standardize the locale name from a string returned by 'setlocale', possibly
51 * modifying that string.
53 * The typical return value of setlocale() is either
54 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
55 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
56 * (the space-separated values represent the various sublocales,
57 * in some unspecified order). This is not handled by this function.
59 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
60 * which is harmful for further use of the string in setlocale(). This
61 * function removes the trailing new line and everything up through the '='
65 S_stdize_locale(pTHX_ char *locs)
67 const char * const s = strchr(locs, '=');
70 PERL_ARGS_ASSERT_STDIZE_LOCALE;
73 const char * const t = strchr(s, '.');
76 const char * const u = strchr(t, '\n');
77 if (u && (u[1] == 0)) {
78 const STRLEN len = u - s;
79 Move(s + 1, locs, len, char);
87 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
95 Perl_set_numeric_radix(pTHX)
97 #ifdef USE_LOCALE_NUMERIC
98 # ifdef HAS_LOCALECONV
99 const struct lconv* const lc = localeconv();
101 if (lc && lc->decimal_point) {
102 if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) {
103 SvREFCNT_dec(PL_numeric_radix_sv);
104 PL_numeric_radix_sv = NULL;
107 if (PL_numeric_radix_sv)
108 sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
110 PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
111 if (! is_invariant_string((U8 *) lc->decimal_point, 0)
112 && is_utf8_string((U8 *) lc->decimal_point, 0)
113 && _is_cur_LC_category_utf8(LC_NUMERIC))
115 SvUTF8_on(PL_numeric_radix_sv);
120 PL_numeric_radix_sv = NULL;
122 DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is %s, ?UTF-8=%d\n",
123 (PL_numeric_radix_sv)
124 ? SvPVX(PL_numeric_radix_sv)
126 (PL_numeric_radix_sv)
127 ? cBOOL(SvUTF8(PL_numeric_radix_sv))
130 # endif /* HAS_LOCALECONV */
131 #endif /* USE_LOCALE_NUMERIC */
134 /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
135 * return of setlocale(), then this is extremely likely to be the C or POSIX
136 * locale. However, the output of setlocale() is documented to be opaque, but
137 * the odds are extremely small that it would return these two strings for some
138 * other locale. Note that VMS in these two locales includes many non-ASCII
139 * characters as controls and punctuation (below are hex bytes):
140 * cntrl: 00-1F 7F 84-97 9B-9F
141 * 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
142 * Oddly, none there are listed as alphas, though some represent alphabetics
143 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
144 #define isNAME_C_OR_POSIX(name) ((name) != NULL \
145 && ((*(name) == 'C' && (*(name + 1)) == '\0') \
146 || strEQ((name), "POSIX")))
149 Perl_new_numeric(pTHX_ const char *newnum)
151 #ifdef USE_LOCALE_NUMERIC
153 /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell
154 * core Perl this and that 'newnum' is the name of the new locale.
155 * It installs this locale as the current underlying default.
157 * The default locale and the C locale can be toggled between by use of the
158 * set_numeric_local() and set_numeric_standard() functions, which should
159 * probably not be called directly, but only via macros like
160 * SET_NUMERIC_STANDARD() in perl.h.
162 * The toggling is necessary mainly so that a non-dot radix decimal point
163 * character can be output, while allowing internal calculations to use a
166 * This sets several interpreter-level variables:
167 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
168 * PL_numeric_local A boolean indicating if the toggled state is such
169 * that the current locale is the program's underlying
171 * PL_numeric_standard An int indicating if the toggled state is such
172 * that the current locale is the C locale. If non-zero,
173 * it is in C; if > 1, it means it may not be toggled away
175 * Note that both of the last two variables can be true at the same time,
176 * if the underlying locale is C. (Toggling is a no-op under these
179 * Any code changing the locale (outside this file) should use
180 * POSIX::setlocale, which calls this function. Therefore this function
181 * should be called directly only from this file and from
182 * POSIX::setlocale() */
187 Safefree(PL_numeric_name);
188 PL_numeric_name = NULL;
189 PL_numeric_standard = TRUE;
190 PL_numeric_local = TRUE;
194 save_newnum = stdize_locale(savepv(newnum));
196 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
197 PL_numeric_local = TRUE;
199 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
200 Safefree(PL_numeric_name);
201 PL_numeric_name = save_newnum;
204 Safefree(save_newnum);
207 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
208 * have to worry about the radix being a non-dot. (Core operations that
209 * need the underlying locale change to it temporarily). */
210 set_numeric_standard();
215 PERL_UNUSED_ARG(newnum);
216 #endif /* USE_LOCALE_NUMERIC */
220 Perl_set_numeric_standard(pTHX)
222 #ifdef USE_LOCALE_NUMERIC
223 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
224 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
225 * macro avoids calling this routine if toggling isn't necessary according
226 * to our records (which could be wrong if some XS code has changed the
227 * locale behind our back) */
229 setlocale(LC_NUMERIC, "C");
230 PL_numeric_standard = TRUE;
231 PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
233 DEBUG_L(PerlIO_printf(Perl_debug_log,
234 "Underlying LC_NUMERIC locale now is C\n"));
236 #endif /* USE_LOCALE_NUMERIC */
240 Perl_set_numeric_local(pTHX)
242 #ifdef USE_LOCALE_NUMERIC
243 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
244 * code should use the macros like SET_NUMERIC_LOCAL() in perl.h instead of
245 * calling this directly. The macro avoids calling this routine if
246 * toggling isn't necessary according to our records (which could be wrong
247 * if some XS code has changed the locale behind our back) */
249 setlocale(LC_NUMERIC, PL_numeric_name);
250 PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
251 PL_numeric_local = TRUE;
253 DEBUG_L(PerlIO_printf(Perl_debug_log,
254 "Underlying LC_NUMERIC locale now is %s\n",
257 #endif /* USE_LOCALE_NUMERIC */
261 * Set up for a new ctype locale.
264 Perl_new_ctype(pTHX_ const char *newctype)
266 #ifdef USE_LOCALE_CTYPE
268 /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell
269 * core Perl this and that 'newctype' is the name of the new locale.
271 * This function sets up the folding arrays for all 256 bytes, assuming
272 * that tofold() is tolc() since fold case is not a concept in POSIX,
274 * Any code changing the locale (outside this file) should use
275 * POSIX::setlocale, which calls this function. Therefore this function
276 * should be called directly only from this file and from
277 * POSIX::setlocale() */
282 PERL_ARGS_ASSERT_NEW_CTYPE;
284 /* We will replace any bad locale warning with 1) nothing if the new one is
285 * ok; or 2) a new warning for the bad new locale */
286 if (PL_warn_locale) {
287 SvREFCNT_dec_NN(PL_warn_locale);
288 PL_warn_locale = NULL;
291 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
293 /* A UTF-8 locale gets standard rules. But note that code still has to
294 * handle this specially because of the three problematic code points */
295 if (PL_in_utf8_CTYPE_locale) {
296 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
299 /* Assume enough space for every character being bad. 4 spaces each
300 * for the 94 printable characters that are output like "'x' "; and 5
301 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
303 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
305 bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
307 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
309 unsigned int bad_count = 0; /* Count of bad characters */
311 for (i = 0; i < 256; i++) {
312 if (isUPPER_LC((U8) i))
313 PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
314 else if (isLOWER_LC((U8) i))
315 PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
317 PL_fold_locale[i] = (U8) i;
319 /* If checking for locale problems, see if the native ASCII-range
320 * printables plus \n and \t are in their expected categories in
321 * the new locale. If not, this could mean big trouble, upending
322 * Perl's and most programs' assumptions, like having a
323 * metacharacter with special meaning become a \w. Fortunately,
324 * it's very rare to find locales that aren't supersets of ASCII
325 * nowadays. It isn't a problem for most controls to be changed
326 * into something else; we check only \n and \t, though perhaps \r
327 * could be an issue as well. */
328 if (check_for_problems
329 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
331 if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
332 || (isPUNCT_A(i) && ! isPUNCT_LC(i))
333 || (isBLANK_A(i) && ! isBLANK_LC(i))
334 || (i == '\n' && ! isCNTRL_LC(i)))
336 if (bad_count) { /* Separate multiple entries with a
338 bad_chars_list[bad_count++] = ' ';
340 bad_chars_list[bad_count++] = '\'';
342 bad_chars_list[bad_count++] = (char) i;
345 bad_chars_list[bad_count++] = '\\';
347 bad_chars_list[bad_count++] = 'n';
351 bad_chars_list[bad_count++] = 't';
354 bad_chars_list[bad_count++] = '\'';
355 bad_chars_list[bad_count] = '\0';
361 /* We only handle single-byte locales (outside of UTF-8 ones; so if
362 * this locale requires than one byte, there are going to be
364 if (check_for_problems && MB_CUR_MAX > 1
366 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
367 * locale. Just assume that the implementation for them (plus
368 * for POSIX) is correct and the > 1 value is spurious. (Since
369 * these are specially handled to never be considered UTF-8
370 * locales, as long as this is the only problem, everything
371 * should work fine */
372 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
374 multi_byte_locale = TRUE;
378 if (bad_count || multi_byte_locale) {
379 PL_warn_locale = Perl_newSVpvf(aTHX_
380 "Locale '%s' may not work well.%s%s%s\n",
383 ? " Some characters in it are not recognized by"
387 ? "\nThe following characters (and maybe others)"
388 " may not have the same meaning as the Perl"
389 " program expects:\n"
395 /* If we are actually in the scope of the locale, output the
396 * message now. Otherwise we save it to be output at the first
397 * operation using this locale, if that actually happens. Most
398 * programs don't use locales, so they are immune to bad ones */
399 if (IN_LC(LC_CTYPE)) {
401 /* We have to save 'newctype' because the setlocale() just
402 * below may destroy it. The next setlocale() further down
403 * should restore it properly so that the intermediate change
404 * here is transparent to this function's caller */
405 const char * const badlocale = savepv(newctype);
407 setlocale(LC_CTYPE, "C");
409 /* The '0' below suppresses a bogus gcc compiler warning */
410 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
411 setlocale(LC_CTYPE, badlocale);
413 SvREFCNT_dec_NN(PL_warn_locale);
414 PL_warn_locale = NULL;
419 #endif /* USE_LOCALE_CTYPE */
420 PERL_ARGS_ASSERT_NEW_CTYPE;
421 PERL_UNUSED_ARG(newctype);
426 Perl__warn_problematic_locale()
429 #ifdef USE_LOCALE_CTYPE
433 /* Internal-to-core function that outputs the message in PL_warn_locale,
434 * and then NULLS it. Should be called only through the macro
435 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
437 if (PL_warn_locale) {
438 /*GCC_DIAG_IGNORE(-Wformat-security); Didn't work */
439 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
440 SvPVX(PL_warn_locale),
441 0 /* dummy to avoid compiler warning */ );
442 /* GCC_DIAG_RESTORE; */
443 SvREFCNT_dec_NN(PL_warn_locale);
444 PL_warn_locale = NULL;
452 Perl_new_collate(pTHX_ const char *newcoll)
454 #ifdef USE_LOCALE_COLLATE
456 /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell
457 * core Perl this and that 'newcoll' is the name of the new locale.
459 * Any code changing the locale (outside this file) should use
460 * POSIX::setlocale, which calls this function. Therefore this function
461 * should be called directly only from this file and from
462 * POSIX::setlocale() */
465 if (PL_collation_name) {
467 Safefree(PL_collation_name);
468 PL_collation_name = NULL;
470 PL_collation_standard = TRUE;
471 PL_collxfrm_base = 0;
472 PL_collxfrm_mult = 2;
476 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
478 Safefree(PL_collation_name);
479 PL_collation_name = stdize_locale(savepv(newcoll));
480 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
483 /* 2: at most so many chars ('a', 'b'). */
484 /* 50: surely no system expands a char more. */
485 #define XFRMBUFSIZE (2 * 50)
486 char xbuf[XFRMBUFSIZE];
487 const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE);
488 const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE);
489 const SSize_t mult = fb - fa;
490 if (mult < 1 && !(fa == 0 && fb == 0))
491 Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
493 PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
494 PL_collxfrm_mult = mult;
499 PERL_UNUSED_ARG(newcoll);
500 #endif /* USE_LOCALE_COLLATE */
506 Perl_my_setlocale(pTHX_ int category, const char* locale)
508 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
509 * difference unless the input locale is "", which means on Windows to get
510 * the machine default, which is set via the computer's "Regional and
511 * Language Options" (or its current equivalent). In POSIX, it instead
512 * means to find the locale from the user's environment. This routine
513 * looks in the environment, and, if anything is found, uses that instead
514 * of going to the machine default. If there is no environment override,
515 * the machine default is used, as normal, by calling the real setlocale()
516 * with "". The POSIX behavior is to use the LC_ALL variable if set;
517 * otherwise to use the particular category's variable if set; otherwise to
518 * use the LANG variable. */
520 bool override_LC_ALL = FALSE;
523 if (locale && strEQ(locale, "")) {
525 locale = PerlEnv_getenv("LC_ALL");
531 override_LC_ALL = TRUE;
532 break; /* We already know its variable isn't set */
534 # ifdef USE_LOCALE_TIME
536 locale = PerlEnv_getenv("LC_TIME");
539 # ifdef USE_LOCALE_CTYPE
541 locale = PerlEnv_getenv("LC_CTYPE");
544 # ifdef USE_LOCALE_COLLATE
546 locale = PerlEnv_getenv("LC_COLLATE");
549 # ifdef USE_LOCALE_MONETARY
551 locale = PerlEnv_getenv("LC_MONETARY");
554 # ifdef USE_LOCALE_NUMERIC
556 locale = PerlEnv_getenv("LC_NUMERIC");
559 # ifdef USE_LOCALE_MESSAGES
561 locale = PerlEnv_getenv("LC_MESSAGES");
565 /* This is a category, like PAPER_SIZE that we don't
566 * know about; and so can't provide a wrapper. */
570 locale = PerlEnv_getenv("LANG");
580 result = setlocale(category, locale);
581 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
582 _setlocale_debug_string(category, locale, result)));
584 if (! override_LC_ALL) {
588 /* Here the input category was LC_ALL, and we have set it to what is in the
589 * LANG variable or the system default if there is no LANG. But these have
590 * lower priority than the other LC_foo variables, so override it for each
591 * one that is set. (If they are set to "", it means to use the same thing
592 * we just set LC_ALL to, so can skip) */
593 # ifdef USE_LOCALE_TIME
594 result = PerlEnv_getenv("LC_TIME");
595 if (result && strNE(result, "")) {
596 setlocale(LC_TIME, result);
597 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
599 _setlocale_debug_string(LC_TIME, result, "not captured")));
602 # ifdef USE_LOCALE_CTYPE
603 result = PerlEnv_getenv("LC_CTYPE");
604 if (result && strNE(result, "")) {
605 setlocale(LC_CTYPE, result);
606 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
608 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
611 # ifdef USE_LOCALE_COLLATE
612 result = PerlEnv_getenv("LC_COLLATE");
613 if (result && strNE(result, "")) {
614 setlocale(LC_COLLATE, result);
615 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
617 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
620 # ifdef USE_LOCALE_MONETARY
621 result = PerlEnv_getenv("LC_MONETARY");
622 if (result && strNE(result, "")) {
623 setlocale(LC_MONETARY, result);
624 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
626 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
629 # ifdef USE_LOCALE_NUMERIC
630 result = PerlEnv_getenv("LC_NUMERIC");
631 if (result && strNE(result, "")) {
632 setlocale(LC_NUMERIC, result);
633 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
635 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
638 # ifdef USE_LOCALE_MESSAGES
639 result = PerlEnv_getenv("LC_MESSAGES");
640 if (result && strNE(result, "")) {
641 setlocale(LC_MESSAGES, result);
642 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
644 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
648 result = setlocale(LC_ALL, NULL);
649 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
651 _setlocale_debug_string(LC_ALL, NULL, result)));
660 * Initialize locale awareness.
663 Perl_init_i18nl10n(pTHX_ int printwarn)
667 * 0 if not to output warning when setup locale is bad
668 * 1 if to output warning based on value of PERL_BADLANG
669 * >1 if to output regardless of PERL_BADLANG
672 * 1 = set ok or not applicable,
673 * 0 = fallback to a locale of lower priority
674 * -1 = fallback to all locales failed, not even to the C locale
676 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
677 * set, debugging information is output.
679 * This looks more complicated than it is, mainly due to the #ifdefs.
681 * We try to set LC_ALL to the value determined by the environment. If
682 * there is no LC_ALL on this platform, we try the individual categories we
683 * know about. If this works, we are done.
685 * But if it doesn't work, we have to do something else. We search the
686 * environment variables ourselves instead of relying on the system to do
687 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
688 * think there is one), and the ultimate fallback "C". This is all done in
689 * the same loop as above to avoid duplicating code, but it makes things
690 * more complex. After the original failure, we add the fallback
691 * possibilities to the list of locales to try, and iterate the loop
692 * through them all until one succeeds.
694 * On Ultrix, the locale MUST come from the environment, so there is
695 * preliminary code to set it. I (khw) am not sure that it is necessary,
696 * and that this couldn't be folded into the loop, but barring any real
697 * platforms to test on, it's staying as-is
699 * A slight complication is that in embedded Perls, the locale may already
700 * be set-up, and we don't want to get it from the normal environment
701 * variables. This is handled by having a special environment variable
702 * indicate we're in this situation. We simply set setlocale's 2nd
703 * parameter to be a NULL instead of "". That indicates to setlocale that
704 * it is not to change anything, but to return the current value,
705 * effectively initializing perl's db to what the locale already is.
707 * We play the same trick with NULL if a LC_ALL succeeds. We call
708 * setlocale() on the individual categores with NULL to get their existing
709 * values for our db, instead of trying to change them.
714 #if defined(USE_LOCALE)
715 #ifdef USE_LOCALE_CTYPE
716 char *curctype = NULL;
717 #endif /* USE_LOCALE_CTYPE */
718 #ifdef USE_LOCALE_COLLATE
719 char *curcoll = NULL;
720 #endif /* USE_LOCALE_COLLATE */
721 #ifdef USE_LOCALE_NUMERIC
723 #endif /* USE_LOCALE_NUMERIC */
725 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
728 /* NULL uses the existing already set up locale */
729 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
733 const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
736 # define DEBUG_LOCALE_INIT(category, locale, result) \
739 PerlIO_printf(Perl_debug_log, \
741 __FILE__, __LINE__, \
742 _setlocale_debug_string(category, \
748 # define DEBUG_LOCALE_INIT(a,b,c)
750 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
751 unsigned int trial_locales_count;
752 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
753 const char * const lang = savepv(PerlEnv_getenv("LANG"));
754 bool setlocale_failure = FALSE;
758 /* A later getenv() could zap this, so only use here */
759 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
761 const bool locwarn = (printwarn > 1
763 && (! bad_lang_use_once
765 /* disallow with "" or "0" */
767 && strNE("0", bad_lang_use_once)))));
769 char * sl_result; /* return from setlocale() */
772 /* In some systems you can find out the system default locale
773 * and use that as the fallback locale. */
774 # define SYSTEM_DEFAULT_LOCALE
776 #ifdef SYSTEM_DEFAULT_LOCALE
777 const char *system_default_locale = NULL;
780 #ifndef LOCALE_ENVIRON_REQUIRED
781 PERL_UNUSED_VAR(done);
782 PERL_UNUSED_VAR(locale_param);
786 * Ultrix setlocale(..., "") fails if there are no environment
787 * variables from which to get a locale name.
792 sl_result = my_setlocale(LC_ALL, setlocale_init);
793 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
797 setlocale_failure = TRUE;
799 if (! setlocale_failure) {
800 # ifdef USE_LOCALE_CTYPE
801 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
804 curctype = my_setlocale(LC_CTYPE, locale_param);
805 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
807 setlocale_failure = TRUE;
809 curctype = savepv(curctype);
810 # endif /* USE_LOCALE_CTYPE */
811 # ifdef USE_LOCALE_COLLATE
812 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
815 curcoll = my_setlocale(LC_COLLATE, locale_param);
816 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
818 setlocale_failure = TRUE;
820 curcoll = savepv(curcoll);
821 # endif /* USE_LOCALE_COLLATE */
822 # ifdef USE_LOCALE_NUMERIC
823 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
826 curnum = my_setlocale(LC_NUMERIC, locale_param);
827 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
829 setlocale_failure = TRUE;
831 curnum = savepv(curnum);
832 # endif /* USE_LOCALE_NUMERIC */
833 # ifdef USE_LOCALE_MESSAGES
834 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
837 sl_result = my_setlocale(LC_MESSAGES, locale_param);
838 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
840 setlocale_failure = TRUE;
842 # endif /* USE_LOCALE_MESSAGES */
843 # ifdef USE_LOCALE_MONETARY
844 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
847 sl_result = my_setlocale(LC_MONETARY, locale_param);
848 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
850 setlocale_failure = TRUE;
852 # endif /* USE_LOCALE_MONETARY */
857 #endif /* !LOCALE_ENVIRON_REQUIRED */
859 /* We try each locale in the list until we get one that works, or exhaust
860 * the list. Normally the loop is executed just once. But if setting the
861 * locale fails, inside the loop we add fallback trials to the array and so
862 * will execute the loop multiple times */
863 trial_locales[0] = setlocale_init;
864 trial_locales_count = 1;
865 for (i= 0; i < trial_locales_count; i++) {
866 const char * trial_locale = trial_locales[i];
870 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
871 * when i==0, but I (khw) don't think that behavior makes much
873 setlocale_failure = FALSE;
875 #ifdef SYSTEM_DEFAULT_LOCALE
877 /* On Windows machines, an entry of "" after the 0th means to use
878 * the system default locale, which we now proceed to get. */
879 if (strEQ(trial_locale, "")) {
882 /* Note that this may change the locale, but we are going to do
883 * that anyway just below */
884 system_default_locale = setlocale(LC_ALL, "");
885 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
887 /* Skip if invalid or it's already on the list of locales to
889 if (! system_default_locale) {
892 for (j = 0; j < trial_locales_count; j++) {
893 if (strEQ(system_default_locale, trial_locales[j])) {
898 trial_locale = system_default_locale;
901 #endif /* SYSTEM_DEFAULT_LOCALE */
905 sl_result = my_setlocale(LC_ALL, trial_locale);
906 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
908 setlocale_failure = TRUE;
911 /* Since LC_ALL succeeded, it should have changed all the other
912 * categories it can to its value; so we massage things so that the
913 * setlocales below just return their category's current values.
914 * This adequately handles the case in NetBSD where LC_COLLATE may
915 * not be defined for a locale, and setting it individually will
916 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
917 * the POSIX locale. */
922 if (!setlocale_failure) {
923 #ifdef USE_LOCALE_CTYPE
925 curctype = my_setlocale(LC_CTYPE, trial_locale);
926 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
928 setlocale_failure = TRUE;
930 curctype = savepv(curctype);
931 #endif /* USE_LOCALE_CTYPE */
932 #ifdef USE_LOCALE_COLLATE
934 curcoll = my_setlocale(LC_COLLATE, trial_locale);
935 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
937 setlocale_failure = TRUE;
939 curcoll = savepv(curcoll);
940 #endif /* USE_LOCALE_COLLATE */
941 #ifdef USE_LOCALE_NUMERIC
943 curnum = my_setlocale(LC_NUMERIC, trial_locale);
944 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
946 setlocale_failure = TRUE;
948 curnum = savepv(curnum);
949 #endif /* USE_LOCALE_NUMERIC */
950 #ifdef USE_LOCALE_MESSAGES
951 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
952 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
954 setlocale_failure = TRUE;
955 #endif /* USE_LOCALE_MESSAGES */
956 #ifdef USE_LOCALE_MONETARY
957 sl_result = my_setlocale(LC_MONETARY, trial_locale);
958 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
960 setlocale_failure = TRUE;
961 #endif /* USE_LOCALE_MONETARY */
963 if (! setlocale_failure) { /* Success */
968 /* Here, something failed; will need to try a fallback. */
974 if (locwarn) { /* Output failure info only on the first one */
977 PerlIO_printf(Perl_error_log,
978 "perl: warning: Setting locale failed.\n");
982 PerlIO_printf(Perl_error_log,
983 "perl: warning: Setting locale failed for the categories:\n\t");
984 # ifdef USE_LOCALE_CTYPE
986 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
987 # endif /* USE_LOCALE_CTYPE */
988 # ifdef USE_LOCALE_COLLATE
990 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
991 # endif /* USE_LOCALE_COLLATE */
992 # ifdef USE_LOCALE_NUMERIC
994 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
995 # endif /* USE_LOCALE_NUMERIC */
996 PerlIO_printf(Perl_error_log, "and possibly others\n");
1000 PerlIO_printf(Perl_error_log,
1001 "perl: warning: Please check that your locale settings:\n");
1004 PerlIO_printf(Perl_error_log,
1005 "\tLANGUAGE = %c%s%c,\n",
1006 language ? '"' : '(',
1007 language ? language : "unset",
1008 language ? '"' : ')');
1011 PerlIO_printf(Perl_error_log,
1012 "\tLC_ALL = %c%s%c,\n",
1014 lc_all ? lc_all : "unset",
1015 lc_all ? '"' : ')');
1017 #if defined(USE_ENVIRON_ARRAY)
1020 for (e = environ; *e; e++) {
1021 if (strnEQ(*e, "LC_", 3)
1022 && strnNE(*e, "LC_ALL=", 7)
1023 && (p = strchr(*e, '=')))
1024 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1025 (int)(p - *e), *e, p + 1);
1029 PerlIO_printf(Perl_error_log,
1030 "\t(possibly more locale environment variables)\n");
1033 PerlIO_printf(Perl_error_log,
1034 "\tLANG = %c%s%c\n",
1036 lang ? lang : "unset",
1039 PerlIO_printf(Perl_error_log,
1040 " are supported and installed on your system.\n");
1043 /* Calculate what fallback locales to try. We have avoided this
1044 * until we have to, because failure is quite unlikely. This will
1045 * usually change the upper bound of the loop we are in.
1047 * Since the system's default way of setting the locale has not
1048 * found one that works, We use Perl's defined ordering: LC_ALL,
1049 * LANG, and the C locale. We don't try the same locale twice, so
1050 * don't add to the list if already there. (On POSIX systems, the
1051 * LC_ALL element will likely be a repeat of the 0th element "",
1052 * but there's no harm done by doing it explicitly.
1054 * Note that this tries the LC_ALL environment variable even on
1055 * systems which have no LC_ALL locale setting. This may or may
1056 * not have been originally intentional, but there's no real need
1057 * to change the behavior. */
1059 for (j = 0; j < trial_locales_count; j++) {
1060 if (strEQ(lc_all, trial_locales[j])) {
1064 trial_locales[trial_locales_count++] = lc_all;
1069 for (j = 0; j < trial_locales_count; j++) {
1070 if (strEQ(lang, trial_locales[j])) {
1074 trial_locales[trial_locales_count++] = lang;
1078 #if defined(WIN32) && defined(LC_ALL)
1079 /* For Windows, we also try the system default locale before "C".
1080 * (If there exists a Windows without LC_ALL we skip this because
1081 * it gets too complicated. For those, the "C" is the next
1082 * fallback possibility). The "" is the same as the 0th element of
1083 * the array, but the code at the loop above knows to treat it
1084 * differently when not the 0th */
1085 trial_locales[trial_locales_count++] = "";
1088 for (j = 0; j < trial_locales_count; j++) {
1089 if (strEQ("C", trial_locales[j])) {
1093 trial_locales[trial_locales_count++] = "C";
1096 } /* end of first time through the loop */
1102 } /* end of looping through the trial locales */
1104 if (ok < 1) { /* If we tried to fallback */
1106 if (! setlocale_failure) { /* fallback succeeded */
1107 msg = "Falling back to";
1109 else { /* fallback failed */
1111 /* We dropped off the end of the loop, so have to decrement i to
1112 * get back to the value the last time through */
1116 msg = "Failed to fall back to";
1118 /* To continue, we should use whatever values we've got */
1119 #ifdef USE_LOCALE_CTYPE
1121 curctype = savepv(setlocale(LC_CTYPE, NULL));
1122 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
1123 #endif /* USE_LOCALE_CTYPE */
1124 #ifdef USE_LOCALE_COLLATE
1126 curcoll = savepv(setlocale(LC_COLLATE, NULL));
1127 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
1128 #endif /* USE_LOCALE_COLLATE */
1129 #ifdef USE_LOCALE_NUMERIC
1131 curnum = savepv(setlocale(LC_NUMERIC, NULL));
1132 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
1133 #endif /* USE_LOCALE_NUMERIC */
1137 const char * description;
1138 const char * name = "";
1139 if (strEQ(trial_locales[i], "C")) {
1140 description = "the standard locale";
1143 #ifdef SYSTEM_DEFAULT_LOCALE
1144 else if (strEQ(trial_locales[i], "")) {
1145 description = "the system default locale";
1146 if (system_default_locale) {
1147 name = system_default_locale;
1150 #endif /* SYSTEM_DEFAULT_LOCALE */
1152 description = "a fallback locale";
1153 name = trial_locales[i];
1155 if (name && strNE(name, "")) {
1156 PerlIO_printf(Perl_error_log,
1157 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1160 PerlIO_printf(Perl_error_log,
1161 "perl: warning: %s %s.\n", msg, description);
1164 } /* End of tried to fallback */
1166 #ifdef USE_LOCALE_CTYPE
1167 new_ctype(curctype);
1168 #endif /* USE_LOCALE_CTYPE */
1170 #ifdef USE_LOCALE_COLLATE
1171 new_collate(curcoll);
1172 #endif /* USE_LOCALE_COLLATE */
1174 #ifdef USE_LOCALE_NUMERIC
1175 new_numeric(curnum);
1176 #endif /* USE_LOCALE_NUMERIC */
1178 #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
1179 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1180 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1181 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1182 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1184 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
1186 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
1187 This is an alternative to using the -C command line switch
1188 (the -C if present will override this). */
1190 const char *p = PerlEnv_getenv("PERL_UNICODE");
1191 PL_unicode = p ? parse_unicode_opts(&p) : 0;
1192 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1197 #ifdef USE_LOCALE_CTYPE
1199 #endif /* USE_LOCALE_CTYPE */
1200 #ifdef USE_LOCALE_COLLATE
1202 #endif /* USE_LOCALE_COLLATE */
1203 #ifdef USE_LOCALE_NUMERIC
1205 #endif /* USE_LOCALE_NUMERIC */
1214 #else /* !USE_LOCALE */
1215 PERL_UNUSED_ARG(printwarn);
1216 #endif /* USE_LOCALE */
1222 #ifdef USE_LOCALE_COLLATE
1225 * mem_collxfrm() is a bit like strxfrm() but with two important
1226 * differences. First, it handles embedded NULs. Second, it allocates
1227 * a bit more memory than needed for the transformed data itself.
1228 * The real transformed data begins at offset sizeof(collationix).
1229 * Please see sv_collxfrm() to see how this is used.
1233 Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1236 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1238 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1240 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1241 /* the +1 is for the terminating NUL. */
1243 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
1244 Newx(xbuf, xAlloc, char);
1248 *(U32*)xbuf = PL_collation_ix;
1249 xout = sizeof(PL_collation_ix);
1250 for (xin = 0; xin < len; ) {
1254 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
1255 if (xused >= PERL_INT_MAX)
1257 if ((STRLEN)xused < xAlloc - xout)
1259 xAlloc = (2 * xAlloc) + 1;
1260 Renew(xbuf, xAlloc, char);
1265 xin += strlen(s + xin) + 1;
1268 /* Embedded NULs are understood but silently skipped
1269 * because they make no sense in locale collation. */
1273 *xlen = xout - sizeof(PL_collation_ix);
1282 #endif /* USE_LOCALE_COLLATE */
1287 Perl__is_cur_LC_category_utf8(pTHX_ int category)
1289 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1290 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
1291 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
1292 * could give the wrong result. The result will very likely be correct for
1293 * languages that have commonly used non-ASCII characters, but for notably
1294 * English, it comes down to if the locale's name ends in something like
1295 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
1297 char *save_input_locale = NULL;
1301 assert(category != LC_ALL);
1304 /* First dispose of the trivial cases */
1305 save_input_locale = setlocale(category, NULL);
1306 if (! save_input_locale) {
1307 DEBUG_L(PerlIO_printf(Perl_debug_log,
1308 "Could not find current locale for category %d\n",
1310 return FALSE; /* XXX maybe should croak */
1312 save_input_locale = stdize_locale(savepv(save_input_locale));
1313 if (isNAME_C_OR_POSIX(save_input_locale)) {
1314 DEBUG_L(PerlIO_printf(Perl_debug_log,
1315 "Current locale for category %d is %s\n",
1316 category, save_input_locale));
1317 Safefree(save_input_locale);
1321 #if defined(USE_LOCALE_CTYPE) \
1322 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
1324 { /* Next try nl_langinfo or MB_CUR_MAX if available */
1326 char *save_ctype_locale = NULL;
1329 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
1331 /* Get the current LC_CTYPE locale */
1332 save_ctype_locale = setlocale(LC_CTYPE, NULL);
1333 if (! save_ctype_locale) {
1334 DEBUG_L(PerlIO_printf(Perl_debug_log,
1335 "Could not find current locale for LC_CTYPE\n"));
1336 goto cant_use_nllanginfo;
1338 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
1340 /* If LC_CTYPE and the desired category use the same locale, this
1341 * means that finding the value for LC_CTYPE is the same as finding
1342 * the value for the desired category. Otherwise, switch LC_CTYPE
1343 * to the desired category's locale */
1344 if (strEQ(save_ctype_locale, save_input_locale)) {
1345 Safefree(save_ctype_locale);
1346 save_ctype_locale = NULL;
1348 else if (! setlocale(LC_CTYPE, save_input_locale)) {
1349 DEBUG_L(PerlIO_printf(Perl_debug_log,
1350 "Could not change LC_CTYPE locale to %s\n",
1351 save_input_locale));
1352 Safefree(save_ctype_locale);
1353 goto cant_use_nllanginfo;
1357 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1358 save_input_locale));
1360 /* Here the current LC_CTYPE is set to the locale of the category whose
1361 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1362 * should give the correct results */
1364 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
1366 char *codeset = nl_langinfo(CODESET);
1367 if (codeset && strNE(codeset, "")) {
1368 codeset = savepv(codeset);
1370 /* If we switched LC_CTYPE, switch back */
1371 if (save_ctype_locale) {
1372 setlocale(LC_CTYPE, save_ctype_locale);
1373 Safefree(save_ctype_locale);
1376 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1377 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1379 DEBUG_L(PerlIO_printf(Perl_debug_log,
1380 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1383 Safefree(save_input_locale);
1391 /* Here, either we don't have nl_langinfo, or it didn't return a
1392 * codeset. Try MB_CUR_MAX */
1394 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1395 * Unicode code point. Since UTF-8 is the only non-single byte
1396 * encoding we handle, we just say any such encoding is UTF-8, and if
1397 * turns out to be wrong, other things will fail */
1398 is_utf8 = MB_CUR_MAX >= 4;
1400 DEBUG_L(PerlIO_printf(Perl_debug_log,
1401 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1402 (int) MB_CUR_MAX, is_utf8));
1404 Safefree(save_input_locale);
1408 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1409 * since they are both in the C99 standard. We can feed a known byte
1410 * string to the latter function, and check that it gives the expected
1414 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
1416 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
1417 != strlen(HYPHEN_UTF8)
1418 || wc != (wchar_t) 0x2010)
1421 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
1422 DEBUG_L(PerlIO_printf(Perl_debug_log,
1423 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1424 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
1429 /* If we switched LC_CTYPE, switch back */
1430 if (save_ctype_locale) {
1431 setlocale(LC_CTYPE, save_ctype_locale);
1432 Safefree(save_ctype_locale);
1439 cant_use_nllanginfo:
1441 #else /* nl_langinfo should work if available, so don't bother compiling this
1442 fallback code. The final fallback of looking at the name is
1443 compiled, and will be executed if nl_langinfo fails */
1445 /* nl_langinfo not available or failed somehow. Next try looking at the
1446 * currency symbol to see if it disambiguates things. Often that will be
1447 * in the native script, and if the symbol isn't in UTF-8, we know that the
1448 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
1449 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1452 #ifdef HAS_LOCALECONV
1453 # ifdef USE_LOCALE_MONETARY
1455 char *save_monetary_locale = NULL;
1456 bool only_ascii = FALSE;
1457 bool is_utf8 = FALSE;
1460 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1461 * the desired category, if it isn't that locale already */
1463 if (category != LC_MONETARY) {
1465 save_monetary_locale = setlocale(LC_MONETARY, NULL);
1466 if (! save_monetary_locale) {
1467 DEBUG_L(PerlIO_printf(Perl_debug_log,
1468 "Could not find current locale for LC_MONETARY\n"));
1469 goto cant_use_monetary;
1471 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
1473 if (strEQ(save_monetary_locale, save_input_locale)) {
1474 Safefree(save_monetary_locale);
1475 save_monetary_locale = NULL;
1477 else if (! setlocale(LC_MONETARY, save_input_locale)) {
1478 DEBUG_L(PerlIO_printf(Perl_debug_log,
1479 "Could not change LC_MONETARY locale to %s\n",
1480 save_input_locale));
1481 Safefree(save_monetary_locale);
1482 goto cant_use_monetary;
1486 /* Here the current LC_MONETARY is set to the locale of the category
1487 * whose information is desired. */
1491 || ! lc->currency_symbol
1492 || is_invariant_string((U8 *) lc->currency_symbol, 0))
1494 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));
1498 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
1501 /* If we changed it, restore LC_MONETARY to its original locale */
1502 if (save_monetary_locale) {
1503 setlocale(LC_MONETARY, save_monetary_locale);
1504 Safefree(save_monetary_locale);
1509 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1510 * otherwise assume the locale is UTF-8 if and only if the symbol
1511 * is non-ascii UTF-8. */
1512 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1513 save_input_locale, is_utf8));
1514 Safefree(save_input_locale);
1520 # endif /* USE_LOCALE_MONETARY */
1521 #endif /* HAS_LOCALECONV */
1523 #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1525 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1526 * the names of the months and weekdays, timezone, and am/pm indicator */
1528 char *save_time_locale = NULL;
1530 bool is_dst = FALSE;
1534 char * formatted_time;
1537 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1538 * desired category, if it isn't that locale already */
1540 if (category != LC_TIME) {
1542 save_time_locale = setlocale(LC_TIME, NULL);
1543 if (! save_time_locale) {
1544 DEBUG_L(PerlIO_printf(Perl_debug_log,
1545 "Could not find current locale for LC_TIME\n"));
1548 save_time_locale = stdize_locale(savepv(save_time_locale));
1550 if (strEQ(save_time_locale, save_input_locale)) {
1551 Safefree(save_time_locale);
1552 save_time_locale = NULL;
1554 else if (! setlocale(LC_TIME, save_input_locale)) {
1555 DEBUG_L(PerlIO_printf(Perl_debug_log,
1556 "Could not change LC_TIME locale to %s\n",
1557 save_input_locale));
1558 Safefree(save_time_locale);
1563 /* Here the current LC_TIME is set to the locale of the category
1564 * whose information is desired. Look at all the days of the week and
1565 * month names, and the timezone and am/pm indicator for UTF-8 variant
1566 * characters. The first such a one found will tell us if the locale
1567 * is UTF-8 or not */
1569 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1570 formatted_time = my_strftime("%A %B %Z %p",
1571 0, 0, hour, dom, month, 112, 0, 0, is_dst);
1572 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
1574 /* Here, we didn't find a non-ASCII. Try the next time through
1575 * with the complemented dst and am/pm, and try with the next
1576 * weekday. After we have gotten all weekdays, try the next
1579 hour = (hour + 12) % 24;
1587 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1588 * false otherwise. But first, restore LC_TIME to its original
1589 * locale if we changed it */
1590 if (save_time_locale) {
1591 setlocale(LC_TIME, save_time_locale);
1592 Safefree(save_time_locale);
1595 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1597 is_utf8_string((U8 *) formatted_time, 0)));
1598 Safefree(save_input_locale);
1599 return is_utf8_string((U8 *) formatted_time, 0);
1602 /* Falling off the end of the loop indicates all the names were just
1603 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1604 * to its original locale */
1605 if (save_time_locale) {
1606 setlocale(LC_TIME, save_time_locale);
1607 Safefree(save_time_locale);
1609 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));
1615 #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
1617 /* This code is ifdefd out because it was found to not be necessary in testing
1618 * on our dromedary test machine, which has over 700 locales. There, this
1619 * added no value to looking at the currency symbol and the time strings. I
1620 * left it in so as to avoid rewriting it if real-world experience indicates
1621 * that dromedary is an outlier. Essentially, instead of returning abpve if we
1622 * haven't found illegal utf8, we continue on and examine all the strerror()
1623 * messages on the platform for utf8ness. If all are ASCII, we still don't
1624 * know the answer; but otherwise we have a pretty good indication of the
1625 * utf8ness. The reason this doesn't help much is that the messages may not
1626 * have been translated into the locale. The currency symbol and time strings
1627 * are much more likely to have been translated. */
1630 bool is_utf8 = FALSE;
1631 bool non_ascii = FALSE;
1632 char *save_messages_locale = NULL;
1633 const char * errmsg = NULL;
1635 /* Like above, we set LC_MESSAGES to the locale of the desired
1636 * category, if it isn't that locale already */
1638 if (category != LC_MESSAGES) {
1640 save_messages_locale = setlocale(LC_MESSAGES, NULL);
1641 if (! save_messages_locale) {
1642 DEBUG_L(PerlIO_printf(Perl_debug_log,
1643 "Could not find current locale for LC_MESSAGES\n"));
1644 goto cant_use_messages;
1646 save_messages_locale = stdize_locale(savepv(save_messages_locale));
1648 if (strEQ(save_messages_locale, save_input_locale)) {
1649 Safefree(save_messages_locale);
1650 save_messages_locale = NULL;
1652 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
1653 DEBUG_L(PerlIO_printf(Perl_debug_log,
1654 "Could not change LC_MESSAGES locale to %s\n",
1655 save_input_locale));
1656 Safefree(save_messages_locale);
1657 goto cant_use_messages;
1661 /* Here the current LC_MESSAGES is set to the locale of the category
1662 * whose information is desired. Look through all the messages. We
1663 * can't use Strerror() here because it may expand to code that
1664 * segfaults in miniperl */
1666 for (e = 0; e <= sys_nerr; e++) {
1668 errmsg = sys_errlist[e];
1669 if (errno || !errmsg) {
1672 errmsg = savepv(errmsg);
1673 if (! is_invariant_string((U8 *) errmsg, 0)) {
1675 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1681 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1682 if (save_messages_locale) {
1683 setlocale(LC_MESSAGES, save_messages_locale);
1684 Safefree(save_messages_locale);
1689 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1690 * any non-ascii means it is one; otherwise we assume it isn't */
1691 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1694 Safefree(save_input_locale);
1698 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));
1704 #endif /* the code that is compiled when no nl_langinfo */
1706 #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1708 /* As a last resort, look at the locale name to see if it matches
1709 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1710 * return of setlocale(), is actually defined to be opaque, so we can't
1711 * really rely on the absence of various substrings in the name to indicate
1712 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1713 * be a UTF-8 locale. Similarly for the other common names */
1715 final_pos = strlen(save_input_locale) - 1;
1716 if (final_pos >= 3) {
1717 char *name = save_input_locale;
1719 /* Find next 'U' or 'u' and look from there */
1720 while ((name += strcspn(name, "Uu") + 1)
1721 <= save_input_locale + final_pos - 2)
1723 if (!isALPHA_FOLD_NE(*name, 't')
1724 || isALPHA_FOLD_NE(*(name + 1), 'f'))
1729 if (*(name) == '-') {
1730 if ((name > save_input_locale + final_pos - 1)) {
1735 if (*(name) == '8') {
1736 DEBUG_L(PerlIO_printf(Perl_debug_log,
1737 "Locale %s ends with UTF-8 in name\n",
1738 save_input_locale));
1739 Safefree(save_input_locale);
1743 DEBUG_L(PerlIO_printf(Perl_debug_log,
1744 "Locale %s doesn't end with UTF-8 in name\n",
1745 save_input_locale));
1750 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1752 && *(save_input_locale + final_pos - 0) == '1'
1753 && *(save_input_locale + final_pos - 1) == '0'
1754 && *(save_input_locale + final_pos - 2) == '0'
1755 && *(save_input_locale + final_pos - 3) == '5'
1756 && *(save_input_locale + final_pos - 4) == '6')
1758 DEBUG_L(PerlIO_printf(Perl_debug_log,
1759 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1760 save_input_locale));
1761 Safefree(save_input_locale);
1766 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1767 * since we are about to return FALSE anyway, there is no point in doing
1768 * this extra work */
1770 if (instr(save_input_locale, "8859")) {
1771 DEBUG_L(PerlIO_printf(Perl_debug_log,
1772 "Locale %s has 8859 in name, not UTF-8 locale\n",
1773 save_input_locale));
1774 Safefree(save_input_locale);
1779 DEBUG_L(PerlIO_printf(Perl_debug_log,
1780 "Assuming locale %s is not a UTF-8 locale\n",
1781 save_input_locale));
1782 Safefree(save_input_locale);
1790 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1793 /* Internal function which returns if we are in the scope of a pragma that
1794 * enables the locale category 'category'. 'compiling' should indicate if
1795 * this is during the compilation phase (TRUE) or not (FALSE). */
1797 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1799 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1800 if (! categories || categories == &PL_sv_placeholder) {
1804 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1805 * a valid unsigned */
1806 assert(category >= -1);
1807 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1811 Perl_my_strerror(pTHX_ const int errnum) {
1813 /* Uses C locale for the error text unless within scope of 'use locale' for
1816 #ifdef USE_LOCALE_MESSAGES
1817 if (! IN_LC(LC_MESSAGES)) {
1818 char * save_locale = setlocale(LC_MESSAGES, NULL);
1819 if (! isNAME_C_OR_POSIX(save_locale)) {
1822 /* The next setlocale likely will zap this, so create a copy */
1823 save_locale = savepv(save_locale);
1825 setlocale(LC_MESSAGES, "C");
1827 /* This points to the static space in Strerror, with all its
1829 errstr = Strerror(errnum);
1831 setlocale(LC_MESSAGES, save_locale);
1832 Safefree(save_locale);
1838 return Strerror(errnum);
1843 =head1 Locale-related functions and macros
1845 =for apidoc sync_locale
1847 Changing the program's locale should be avoided by XS code. Nevertheless,
1848 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1849 happens, Perl needs to be told that the locale has changed. Use this function
1850 to do so, before returning to Perl.
1856 Perl_sync_locale(pTHX)
1859 #ifdef USE_LOCALE_CTYPE
1860 new_ctype(setlocale(LC_CTYPE, NULL));
1861 #endif /* USE_LOCALE_CTYPE */
1863 #ifdef USE_LOCALE_COLLATE
1864 new_collate(setlocale(LC_COLLATE, NULL));
1867 #ifdef USE_LOCALE_NUMERIC
1868 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1869 new_numeric(setlocale(LC_NUMERIC, NULL));
1870 #endif /* USE_LOCALE_NUMERIC */
1874 #if defined(DEBUGGING) && defined(USE_LOCALE)
1877 Perl__setlocale_debug_string(const int category, /* category number,
1879 const char* const locale, /* locale name */
1881 /* return value from setlocale() when attempting to
1882 * set 'category' to 'locale' */
1883 const char* const retval)
1885 /* Returns a pointer to a NUL-terminated string in static storage with
1886 * added text about the info passed in. This is not thread safe and will
1887 * be overwritten by the next call, so this should be used just to
1888 * formulate a string to immediately print or savepv() on. */
1890 /* initialise to a non-null value to keep it out of BSS and so keep
1891 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
1892 static char ret[128] = "x";
1894 my_strlcpy(ret, "setlocale(", sizeof(ret));
1898 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
1902 my_strlcat(ret, "LC_ALL", sizeof(ret));
1907 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
1912 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
1917 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
1922 my_strlcat(ret, "LC_TIME", sizeof(ret));
1927 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
1932 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
1937 my_strlcat(ret, ", ", sizeof(ret));
1940 my_strlcat(ret, "\"", sizeof(ret));
1941 my_strlcat(ret, locale, sizeof(ret));
1942 my_strlcat(ret, "\"", sizeof(ret));
1945 my_strlcat(ret, "NULL", sizeof(ret));
1948 my_strlcat(ret, ") returned ", sizeof(ret));
1951 my_strlcat(ret, "\"", sizeof(ret));
1952 my_strlcat(ret, retval, sizeof(ret));
1953 my_strlcat(ret, "\"", sizeof(ret));
1956 my_strlcat(ret, "NULL", sizeof(ret));
1959 assert(strlen(ret) < sizeof(ret));
1968 * ex: set ts=8 sts=4 sw=4 et: