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);
572 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
573 _setlocale_debug_string(category, locale, result)));
575 if (! override_LC_ALL) {
579 /* Here the input category was LC_ALL, and we have set it to what is in the
580 * LANG variable or the system default if there is no LANG. But these have
581 * lower priority than the other LC_foo variables, so override it for each
582 * one that is set. (If they are set to "", it means to use the same thing
583 * we just set LC_ALL to, so can skip) */
584 # ifdef USE_LOCALE_TIME
585 result = PerlEnv_getenv("LC_TIME");
586 if (result && strNE(result, "")) {
587 setlocale(LC_TIME, result);
588 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
590 _setlocale_debug_string(LC_TIME, result, "not captured")));
593 # ifdef USE_LOCALE_CTYPE
594 result = PerlEnv_getenv("LC_CTYPE");
595 if (result && strNE(result, "")) {
596 setlocale(LC_CTYPE, result);
597 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
599 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
602 # ifdef USE_LOCALE_COLLATE
603 result = PerlEnv_getenv("LC_COLLATE");
604 if (result && strNE(result, "")) {
605 setlocale(LC_COLLATE, result);
606 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
608 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
611 # ifdef USE_LOCALE_MONETARY
612 result = PerlEnv_getenv("LC_MONETARY");
613 if (result && strNE(result, "")) {
614 setlocale(LC_MONETARY, result);
615 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
617 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
620 # ifdef USE_LOCALE_NUMERIC
621 result = PerlEnv_getenv("LC_NUMERIC");
622 if (result && strNE(result, "")) {
623 setlocale(LC_NUMERIC, result);
624 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
626 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
629 # ifdef USE_LOCALE_MESSAGES
630 result = PerlEnv_getenv("LC_MESSAGES");
631 if (result && strNE(result, "")) {
632 setlocale(LC_MESSAGES, result);
633 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
635 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
639 result = setlocale(LC_ALL, NULL);
640 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
642 _setlocale_debug_string(LC_ALL, NULL, result)));
651 * Initialize locale awareness.
654 Perl_init_i18nl10n(pTHX_ int printwarn)
658 * 0 if not to output warning when setup locale is bad
659 * 1 if to output warning based on value of PERL_BADLANG
660 * >1 if to output regardless of PERL_BADLANG
663 * 1 = set ok or not applicable,
664 * 0 = fallback to a locale of lower priority
665 * -1 = fallback to all locales failed, not even to the C locale
667 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
668 * set, debugging information is output.
670 * This looks more complicated than it is, mainly due to the #ifdefs.
672 * We try to set LC_ALL to the value determined by the environment. If
673 * there is no LC_ALL on this platform, we try the individual categories we
674 * know about. If this works, we are done.
676 * But if it doesn't work, we have to do something else. We search the
677 * environment variables ourselves instead of relying on the system to do
678 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
679 * think there is one), and the ultimate fallback "C". This is all done in
680 * the same loop as above to avoid duplicating code, but it makes things
681 * more complex. After the original failure, we add the fallback
682 * possibilities to the list of locales to try, and iterate the loop
683 * through them all until one succeeds.
685 * On Ultrix, the locale MUST come from the environment, so there is
686 * preliminary code to set it. I (khw) am not sure that it is necessary,
687 * and that this couldn't be folded into the loop, but barring any real
688 * platforms to test on, it's staying as-is
690 * A slight complication is that in embedded Perls, the locale may already
691 * be set-up, and we don't want to get it from the normal environment
692 * variables. This is handled by having a special environment variable
693 * indicate we're in this situation. We simply set setlocale's 2nd
694 * parameter to be a NULL instead of "". That indicates to setlocale that
695 * it is not to change anything, but to return the current value,
696 * effectively initializing perl's db to what the locale already is.
698 * We play the same trick with NULL if a LC_ALL succeeds. We call
699 * setlocale() on the individual categores with NULL to get their existing
700 * values for our db, instead of trying to change them.
705 #if defined(USE_LOCALE)
706 #ifdef USE_LOCALE_CTYPE
707 char *curctype = NULL;
708 #endif /* USE_LOCALE_CTYPE */
709 #ifdef USE_LOCALE_COLLATE
710 char *curcoll = NULL;
711 #endif /* USE_LOCALE_COLLATE */
712 #ifdef USE_LOCALE_NUMERIC
714 #endif /* USE_LOCALE_NUMERIC */
716 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
719 /* NULL uses the existing already set up locale */
720 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
724 const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
727 # define DEBUG_LOCALE_INIT(category, locale, result) \
730 PerlIO_printf(Perl_debug_log, \
732 __FILE__, __LINE__, \
733 _setlocale_debug_string(category, \
739 # define DEBUG_LOCALE_INIT(a,b,c)
741 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
742 unsigned int trial_locales_count;
743 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
744 const char * const lang = savepv(PerlEnv_getenv("LANG"));
745 bool setlocale_failure = FALSE;
749 /* A later getenv() could zap this, so only use here */
750 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
752 const bool locwarn = (printwarn > 1
754 && (! bad_lang_use_once
756 /* disallow with "" or "0" */
758 && strNE("0", bad_lang_use_once)))));
760 char * sl_result; /* return from setlocale() */
763 /* In some systems you can find out the system default locale
764 * and use that as the fallback locale. */
765 # define SYSTEM_DEFAULT_LOCALE
767 #ifdef SYSTEM_DEFAULT_LOCALE
768 const char *system_default_locale = NULL;
771 #ifndef LOCALE_ENVIRON_REQUIRED
772 PERL_UNUSED_VAR(done);
773 PERL_UNUSED_VAR(locale_param);
777 * Ultrix setlocale(..., "") fails if there are no environment
778 * variables from which to get a locale name.
783 sl_result = my_setlocale(LC_ALL, setlocale_init);
784 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
788 setlocale_failure = TRUE;
790 if (! setlocale_failure) {
791 # ifdef USE_LOCALE_CTYPE
792 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
795 curctype = my_setlocale(LC_CTYPE, locale_param);
796 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
798 setlocale_failure = TRUE;
800 curctype = savepv(curctype);
801 # endif /* USE_LOCALE_CTYPE */
802 # ifdef USE_LOCALE_COLLATE
803 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
806 curcoll = my_setlocale(LC_COLLATE, locale_param);
807 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
809 setlocale_failure = TRUE;
811 curcoll = savepv(curcoll);
812 # endif /* USE_LOCALE_COLLATE */
813 # ifdef USE_LOCALE_NUMERIC
814 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
817 curnum = my_setlocale(LC_NUMERIC, locale_param);
818 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
820 setlocale_failure = TRUE;
822 curnum = savepv(curnum);
823 # endif /* USE_LOCALE_NUMERIC */
824 # ifdef USE_LOCALE_MESSAGES
825 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
828 sl_result = my_setlocale(LC_MESSAGES, locale_param);
829 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
831 setlocale_failure = TRUE;
833 # endif /* USE_LOCALE_MESSAGES */
834 # ifdef USE_LOCALE_MONETARY
835 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
838 sl_result = my_setlocale(LC_MONETARY, locale_param);
839 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
841 setlocale_failure = TRUE;
843 # endif /* USE_LOCALE_MONETARY */
848 #endif /* !LOCALE_ENVIRON_REQUIRED */
850 /* We try each locale in the list until we get one that works, or exhaust
851 * the list. Normally the loop is executed just once. But if setting the
852 * locale fails, inside the loop we add fallback trials to the array and so
853 * will execute the loop multiple times */
854 trial_locales[0] = setlocale_init;
855 trial_locales_count = 1;
856 for (i= 0; i < trial_locales_count; i++) {
857 const char * trial_locale = trial_locales[i];
861 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
862 * when i==0, but I (khw) don't think that behavior makes much
864 setlocale_failure = FALSE;
866 #ifdef SYSTEM_DEFAULT_LOCALE
868 /* On Windows machines, an entry of "" after the 0th means to use
869 * the system default locale, which we now proceed to get. */
870 if (strEQ(trial_locale, "")) {
873 /* Note that this may change the locale, but we are going to do
874 * that anyway just below */
875 system_default_locale = setlocale(LC_ALL, "");
876 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
878 /* Skip if invalid or it's already on the list of locales to
880 if (! system_default_locale) {
883 for (j = 0; j < trial_locales_count; j++) {
884 if (strEQ(system_default_locale, trial_locales[j])) {
889 trial_locale = system_default_locale;
892 #endif /* SYSTEM_DEFAULT_LOCALE */
896 sl_result = my_setlocale(LC_ALL, trial_locale);
897 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
899 setlocale_failure = TRUE;
902 /* Since LC_ALL succeeded, it should have changed all the other
903 * categories it can to its value; so we massage things so that the
904 * setlocales below just return their category's current values.
905 * This adequately handles the case in NetBSD where LC_COLLATE may
906 * not be defined for a locale, and setting it individually will
907 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
908 * the POSIX locale. */
913 if (!setlocale_failure) {
914 #ifdef USE_LOCALE_CTYPE
916 curctype = my_setlocale(LC_CTYPE, trial_locale);
917 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
919 setlocale_failure = TRUE;
921 curctype = savepv(curctype);
922 #endif /* USE_LOCALE_CTYPE */
923 #ifdef USE_LOCALE_COLLATE
925 curcoll = my_setlocale(LC_COLLATE, trial_locale);
926 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
928 setlocale_failure = TRUE;
930 curcoll = savepv(curcoll);
931 #endif /* USE_LOCALE_COLLATE */
932 #ifdef USE_LOCALE_NUMERIC
934 curnum = my_setlocale(LC_NUMERIC, trial_locale);
935 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
937 setlocale_failure = TRUE;
939 curnum = savepv(curnum);
940 #endif /* USE_LOCALE_NUMERIC */
941 #ifdef USE_LOCALE_MESSAGES
942 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
943 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
945 setlocale_failure = TRUE;
946 #endif /* USE_LOCALE_MESSAGES */
947 #ifdef USE_LOCALE_MONETARY
948 sl_result = my_setlocale(LC_MONETARY, trial_locale);
949 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
951 setlocale_failure = TRUE;
952 #endif /* USE_LOCALE_MONETARY */
954 if (! setlocale_failure) { /* Success */
959 /* Here, something failed; will need to try a fallback. */
965 if (locwarn) { /* Output failure info only on the first one */
968 PerlIO_printf(Perl_error_log,
969 "perl: warning: Setting locale failed.\n");
973 PerlIO_printf(Perl_error_log,
974 "perl: warning: Setting locale failed for the categories:\n\t");
975 # ifdef USE_LOCALE_CTYPE
977 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
978 # endif /* USE_LOCALE_CTYPE */
979 # ifdef USE_LOCALE_COLLATE
981 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
982 # endif /* USE_LOCALE_COLLATE */
983 # ifdef USE_LOCALE_NUMERIC
985 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
986 # endif /* USE_LOCALE_NUMERIC */
987 PerlIO_printf(Perl_error_log, "and possibly others\n");
991 PerlIO_printf(Perl_error_log,
992 "perl: warning: Please check that your locale settings:\n");
995 PerlIO_printf(Perl_error_log,
996 "\tLANGUAGE = %c%s%c,\n",
997 language ? '"' : '(',
998 language ? language : "unset",
999 language ? '"' : ')');
1002 PerlIO_printf(Perl_error_log,
1003 "\tLC_ALL = %c%s%c,\n",
1005 lc_all ? lc_all : "unset",
1006 lc_all ? '"' : ')');
1008 #if defined(USE_ENVIRON_ARRAY)
1011 for (e = environ; *e; e++) {
1012 if (strnEQ(*e, "LC_", 3)
1013 && strnNE(*e, "LC_ALL=", 7)
1014 && (p = strchr(*e, '=')))
1015 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1016 (int)(p - *e), *e, p + 1);
1020 PerlIO_printf(Perl_error_log,
1021 "\t(possibly more locale environment variables)\n");
1024 PerlIO_printf(Perl_error_log,
1025 "\tLANG = %c%s%c\n",
1027 lang ? lang : "unset",
1030 PerlIO_printf(Perl_error_log,
1031 " are supported and installed on your system.\n");
1034 /* Calculate what fallback locales to try. We have avoided this
1035 * until we have to, because failure is quite unlikely. This will
1036 * usually change the upper bound of the loop we are in.
1038 * Since the system's default way of setting the locale has not
1039 * found one that works, We use Perl's defined ordering: LC_ALL,
1040 * LANG, and the C locale. We don't try the same locale twice, so
1041 * don't add to the list if already there. (On POSIX systems, the
1042 * LC_ALL element will likely be a repeat of the 0th element "",
1043 * but there's no harm done by doing it explicitly.
1045 * Note that this tries the LC_ALL environment variable even on
1046 * systems which have no LC_ALL locale setting. This may or may
1047 * not have been originally intentional, but there's no real need
1048 * to change the behavior. */
1050 for (j = 0; j < trial_locales_count; j++) {
1051 if (strEQ(lc_all, trial_locales[j])) {
1055 trial_locales[trial_locales_count++] = lc_all;
1060 for (j = 0; j < trial_locales_count; j++) {
1061 if (strEQ(lang, trial_locales[j])) {
1065 trial_locales[trial_locales_count++] = lang;
1069 #if defined(WIN32) && defined(LC_ALL)
1070 /* For Windows, we also try the system default locale before "C".
1071 * (If there exists a Windows without LC_ALL we skip this because
1072 * it gets too complicated. For those, the "C" is the next
1073 * fallback possibility). The "" is the same as the 0th element of
1074 * the array, but the code at the loop above knows to treat it
1075 * differently when not the 0th */
1076 trial_locales[trial_locales_count++] = "";
1079 for (j = 0; j < trial_locales_count; j++) {
1080 if (strEQ("C", trial_locales[j])) {
1084 trial_locales[trial_locales_count++] = "C";
1087 } /* end of first time through the loop */
1093 } /* end of looping through the trial locales */
1095 if (ok < 1) { /* If we tried to fallback */
1097 if (! setlocale_failure) { /* fallback succeeded */
1098 msg = "Falling back to";
1100 else { /* fallback failed */
1102 /* We dropped off the end of the loop, so have to decrement i to
1103 * get back to the value the last time through */
1107 msg = "Failed to fall back to";
1109 /* To continue, we should use whatever values we've got */
1110 #ifdef USE_LOCALE_CTYPE
1112 curctype = savepv(setlocale(LC_CTYPE, NULL));
1113 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
1114 #endif /* USE_LOCALE_CTYPE */
1115 #ifdef USE_LOCALE_COLLATE
1117 curcoll = savepv(setlocale(LC_COLLATE, NULL));
1118 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
1119 #endif /* USE_LOCALE_COLLATE */
1120 #ifdef USE_LOCALE_NUMERIC
1122 curnum = savepv(setlocale(LC_NUMERIC, NULL));
1123 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
1124 #endif /* USE_LOCALE_NUMERIC */
1128 const char * description;
1129 const char * name = "";
1130 if (strEQ(trial_locales[i], "C")) {
1131 description = "the standard locale";
1134 #ifdef SYSTEM_DEFAULT_LOCALE
1135 else if (strEQ(trial_locales[i], "")) {
1136 description = "the system default locale";
1137 if (system_default_locale) {
1138 name = system_default_locale;
1141 #endif /* SYSTEM_DEFAULT_LOCALE */
1143 description = "a fallback locale";
1144 name = trial_locales[i];
1146 if (name && strNE(name, "")) {
1147 PerlIO_printf(Perl_error_log,
1148 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1151 PerlIO_printf(Perl_error_log,
1152 "perl: warning: %s %s.\n", msg, description);
1155 } /* End of tried to fallback */
1157 #ifdef USE_LOCALE_CTYPE
1158 new_ctype(curctype);
1159 #endif /* USE_LOCALE_CTYPE */
1161 #ifdef USE_LOCALE_COLLATE
1162 new_collate(curcoll);
1163 #endif /* USE_LOCALE_COLLATE */
1165 #ifdef USE_LOCALE_NUMERIC
1166 new_numeric(curnum);
1167 #endif /* USE_LOCALE_NUMERIC */
1169 #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
1170 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1171 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1172 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1173 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1175 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
1177 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
1178 This is an alternative to using the -C command line switch
1179 (the -C if present will override this). */
1181 const char *p = PerlEnv_getenv("PERL_UNICODE");
1182 PL_unicode = p ? parse_unicode_opts(&p) : 0;
1183 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1188 #ifdef USE_LOCALE_CTYPE
1190 #endif /* USE_LOCALE_CTYPE */
1191 #ifdef USE_LOCALE_COLLATE
1193 #endif /* USE_LOCALE_COLLATE */
1194 #ifdef USE_LOCALE_NUMERIC
1196 #endif /* USE_LOCALE_NUMERIC */
1205 #else /* !USE_LOCALE */
1206 PERL_UNUSED_ARG(printwarn);
1207 #endif /* USE_LOCALE */
1213 #ifdef USE_LOCALE_COLLATE
1216 * mem_collxfrm() is a bit like strxfrm() but with two important
1217 * differences. First, it handles embedded NULs. Second, it allocates
1218 * a bit more memory than needed for the transformed data itself.
1219 * The real transformed data begins at offset sizeof(collationix).
1220 * Please see sv_collxfrm() to see how this is used.
1224 Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1227 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1229 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1231 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1232 /* the +1 is for the terminating NUL. */
1234 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
1235 Newx(xbuf, xAlloc, char);
1239 *(U32*)xbuf = PL_collation_ix;
1240 xout = sizeof(PL_collation_ix);
1241 for (xin = 0; xin < len; ) {
1245 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
1246 if (xused >= PERL_INT_MAX)
1248 if ((STRLEN)xused < xAlloc - xout)
1250 xAlloc = (2 * xAlloc) + 1;
1251 Renew(xbuf, xAlloc, char);
1256 xin += strlen(s + xin) + 1;
1259 /* Embedded NULs are understood but silently skipped
1260 * because they make no sense in locale collation. */
1264 *xlen = xout - sizeof(PL_collation_ix);
1273 #endif /* USE_LOCALE_COLLATE */
1278 Perl__is_cur_LC_category_utf8(pTHX_ int category)
1280 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1281 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
1282 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
1283 * could give the wrong result. The result will very likely be correct for
1284 * languages that have commonly used non-ASCII characters, but for notably
1285 * English, it comes down to if the locale's name ends in something like
1286 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
1288 char *save_input_locale = NULL;
1292 assert(category != LC_ALL);
1295 /* First dispose of the trivial cases */
1296 save_input_locale = setlocale(category, NULL);
1297 if (! save_input_locale) {
1298 DEBUG_L(PerlIO_printf(Perl_debug_log,
1299 "Could not find current locale for category %d\n",
1301 return FALSE; /* XXX maybe should croak */
1303 save_input_locale = stdize_locale(savepv(save_input_locale));
1304 if (isNAME_C_OR_POSIX(save_input_locale)) {
1305 DEBUG_L(PerlIO_printf(Perl_debug_log,
1306 "Current locale for category %d is %s\n",
1307 category, save_input_locale));
1308 Safefree(save_input_locale);
1312 #if defined(USE_LOCALE_CTYPE) \
1313 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
1315 { /* Next try nl_langinfo or MB_CUR_MAX if available */
1317 char *save_ctype_locale = NULL;
1320 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
1322 /* Get the current LC_CTYPE locale */
1323 save_ctype_locale = setlocale(LC_CTYPE, NULL);
1324 if (! save_ctype_locale) {
1325 DEBUG_L(PerlIO_printf(Perl_debug_log,
1326 "Could not find current locale for LC_CTYPE\n"));
1327 goto cant_use_nllanginfo;
1329 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
1331 /* If LC_CTYPE and the desired category use the same locale, this
1332 * means that finding the value for LC_CTYPE is the same as finding
1333 * the value for the desired category. Otherwise, switch LC_CTYPE
1334 * to the desired category's locale */
1335 if (strEQ(save_ctype_locale, save_input_locale)) {
1336 Safefree(save_ctype_locale);
1337 save_ctype_locale = NULL;
1339 else if (! setlocale(LC_CTYPE, save_input_locale)) {
1340 DEBUG_L(PerlIO_printf(Perl_debug_log,
1341 "Could not change LC_CTYPE locale to %s\n",
1342 save_input_locale));
1343 Safefree(save_ctype_locale);
1344 goto cant_use_nllanginfo;
1348 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1349 save_input_locale));
1351 /* Here the current LC_CTYPE is set to the locale of the category whose
1352 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1353 * should give the correct results */
1355 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
1357 char *codeset = nl_langinfo(CODESET);
1358 if (codeset && strNE(codeset, "")) {
1359 codeset = savepv(codeset);
1361 /* If we switched LC_CTYPE, switch back */
1362 if (save_ctype_locale) {
1363 setlocale(LC_CTYPE, save_ctype_locale);
1364 Safefree(save_ctype_locale);
1367 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1368 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1370 DEBUG_L(PerlIO_printf(Perl_debug_log,
1371 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1374 Safefree(save_input_locale);
1382 /* Here, either we don't have nl_langinfo, or it didn't return a
1383 * codeset. Try MB_CUR_MAX */
1385 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1386 * Unicode code point. Since UTF-8 is the only non-single byte
1387 * encoding we handle, we just say any such encoding is UTF-8, and if
1388 * turns out to be wrong, other things will fail */
1389 is_utf8 = MB_CUR_MAX >= 4;
1391 DEBUG_L(PerlIO_printf(Perl_debug_log,
1392 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1393 (int) MB_CUR_MAX, is_utf8));
1395 Safefree(save_input_locale);
1399 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1400 * since they are both in the C99 standard. We can feed a known byte
1401 * string to the latter function, and check that it gives the expected
1405 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
1407 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
1408 != strlen(HYPHEN_UTF8)
1409 || wc != (wchar_t) 0x2010)
1412 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
1413 DEBUG_L(PerlIO_printf(Perl_debug_log,
1414 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1415 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
1420 /* If we switched LC_CTYPE, switch back */
1421 if (save_ctype_locale) {
1422 setlocale(LC_CTYPE, save_ctype_locale);
1423 Safefree(save_ctype_locale);
1430 cant_use_nllanginfo:
1432 #else /* nl_langinfo should work if available, so don't bother compiling this
1433 fallback code. The final fallback of looking at the name is
1434 compiled, and will be executed if nl_langinfo fails */
1436 /* nl_langinfo not available or failed somehow. Next try looking at the
1437 * currency symbol to see if it disambiguates things. Often that will be
1438 * in the native script, and if the symbol isn't in UTF-8, we know that the
1439 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
1440 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1443 #ifdef HAS_LOCALECONV
1444 # ifdef USE_LOCALE_MONETARY
1446 char *save_monetary_locale = NULL;
1447 bool only_ascii = FALSE;
1448 bool is_utf8 = FALSE;
1451 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1452 * the desired category, if it isn't that locale already */
1454 if (category != LC_MONETARY) {
1456 save_monetary_locale = setlocale(LC_MONETARY, NULL);
1457 if (! save_monetary_locale) {
1458 DEBUG_L(PerlIO_printf(Perl_debug_log,
1459 "Could not find current locale for LC_MONETARY\n"));
1460 goto cant_use_monetary;
1462 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
1464 if (strEQ(save_monetary_locale, save_input_locale)) {
1465 Safefree(save_monetary_locale);
1466 save_monetary_locale = NULL;
1468 else if (! setlocale(LC_MONETARY, save_input_locale)) {
1469 DEBUG_L(PerlIO_printf(Perl_debug_log,
1470 "Could not change LC_MONETARY locale to %s\n",
1471 save_input_locale));
1472 Safefree(save_monetary_locale);
1473 goto cant_use_monetary;
1477 /* Here the current LC_MONETARY is set to the locale of the category
1478 * whose information is desired. */
1482 || ! lc->currency_symbol
1483 || is_invariant_string((U8 *) lc->currency_symbol, 0))
1485 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));
1489 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
1492 /* If we changed it, restore LC_MONETARY to its original locale */
1493 if (save_monetary_locale) {
1494 setlocale(LC_MONETARY, save_monetary_locale);
1495 Safefree(save_monetary_locale);
1500 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1501 * otherwise assume the locale is UTF-8 if and only if the symbol
1502 * is non-ascii UTF-8. */
1503 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1504 save_input_locale, is_utf8));
1505 Safefree(save_input_locale);
1511 # endif /* USE_LOCALE_MONETARY */
1512 #endif /* HAS_LOCALECONV */
1514 #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1516 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1517 * the names of the months and weekdays, timezone, and am/pm indicator */
1519 char *save_time_locale = NULL;
1521 bool is_dst = FALSE;
1525 char * formatted_time;
1528 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1529 * desired category, if it isn't that locale already */
1531 if (category != LC_TIME) {
1533 save_time_locale = setlocale(LC_TIME, NULL);
1534 if (! save_time_locale) {
1535 DEBUG_L(PerlIO_printf(Perl_debug_log,
1536 "Could not find current locale for LC_TIME\n"));
1539 save_time_locale = stdize_locale(savepv(save_time_locale));
1541 if (strEQ(save_time_locale, save_input_locale)) {
1542 Safefree(save_time_locale);
1543 save_time_locale = NULL;
1545 else if (! setlocale(LC_TIME, save_input_locale)) {
1546 DEBUG_L(PerlIO_printf(Perl_debug_log,
1547 "Could not change LC_TIME locale to %s\n",
1548 save_input_locale));
1549 Safefree(save_time_locale);
1554 /* Here the current LC_TIME is set to the locale of the category
1555 * whose information is desired. Look at all the days of the week and
1556 * month names, and the timezone and am/pm indicator for UTF-8 variant
1557 * characters. The first such a one found will tell us if the locale
1558 * is UTF-8 or not */
1560 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1561 formatted_time = my_strftime("%A %B %Z %p",
1562 0, 0, hour, dom, month, 112, 0, 0, is_dst);
1563 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
1565 /* Here, we didn't find a non-ASCII. Try the next time through
1566 * with the complemented dst and am/pm, and try with the next
1567 * weekday. After we have gotten all weekdays, try the next
1570 hour = (hour + 12) % 24;
1578 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1579 * false otherwise. But first, restore LC_TIME to its original
1580 * locale if we changed it */
1581 if (save_time_locale) {
1582 setlocale(LC_TIME, save_time_locale);
1583 Safefree(save_time_locale);
1586 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1588 is_utf8_string((U8 *) formatted_time, 0)));
1589 Safefree(save_input_locale);
1590 return is_utf8_string((U8 *) formatted_time, 0);
1593 /* Falling off the end of the loop indicates all the names were just
1594 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1595 * to its original locale */
1596 if (save_time_locale) {
1597 setlocale(LC_TIME, save_time_locale);
1598 Safefree(save_time_locale);
1600 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));
1606 #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
1608 /* This code is ifdefd out because it was found to not be necessary in testing
1609 * on our dromedary test machine, which has over 700 locales. There, this
1610 * added no value to looking at the currency symbol and the time strings. I
1611 * left it in so as to avoid rewriting it if real-world experience indicates
1612 * that dromedary is an outlier. Essentially, instead of returning abpve if we
1613 * haven't found illegal utf8, we continue on and examine all the strerror()
1614 * messages on the platform for utf8ness. If all are ASCII, we still don't
1615 * know the answer; but otherwise we have a pretty good indication of the
1616 * utf8ness. The reason this doesn't help much is that the messages may not
1617 * have been translated into the locale. The currency symbol and time strings
1618 * are much more likely to have been translated. */
1621 bool is_utf8 = FALSE;
1622 bool non_ascii = FALSE;
1623 char *save_messages_locale = NULL;
1624 const char * errmsg = NULL;
1626 /* Like above, we set LC_MESSAGES to the locale of the desired
1627 * category, if it isn't that locale already */
1629 if (category != LC_MESSAGES) {
1631 save_messages_locale = setlocale(LC_MESSAGES, NULL);
1632 if (! save_messages_locale) {
1633 DEBUG_L(PerlIO_printf(Perl_debug_log,
1634 "Could not find current locale for LC_MESSAGES\n"));
1635 goto cant_use_messages;
1637 save_messages_locale = stdize_locale(savepv(save_messages_locale));
1639 if (strEQ(save_messages_locale, save_input_locale)) {
1640 Safefree(save_messages_locale);
1641 save_messages_locale = NULL;
1643 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
1644 DEBUG_L(PerlIO_printf(Perl_debug_log,
1645 "Could not change LC_MESSAGES locale to %s\n",
1646 save_input_locale));
1647 Safefree(save_messages_locale);
1648 goto cant_use_messages;
1652 /* Here the current LC_MESSAGES is set to the locale of the category
1653 * whose information is desired. Look through all the messages. We
1654 * can't use Strerror() here because it may expand to code that
1655 * segfaults in miniperl */
1657 for (e = 0; e <= sys_nerr; e++) {
1659 errmsg = sys_errlist[e];
1660 if (errno || !errmsg) {
1663 errmsg = savepv(errmsg);
1664 if (! is_invariant_string((U8 *) errmsg, 0)) {
1666 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1672 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1673 if (save_messages_locale) {
1674 setlocale(LC_MESSAGES, save_messages_locale);
1675 Safefree(save_messages_locale);
1680 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1681 * any non-ascii means it is one; otherwise we assume it isn't */
1682 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1685 Safefree(save_input_locale);
1689 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));
1695 #endif /* the code that is compiled when no nl_langinfo */
1697 #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1699 /* As a last resort, look at the locale name to see if it matches
1700 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1701 * return of setlocale(), is actually defined to be opaque, so we can't
1702 * really rely on the absence of various substrings in the name to indicate
1703 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1704 * be a UTF-8 locale. Similarly for the other common names */
1706 final_pos = strlen(save_input_locale) - 1;
1707 if (final_pos >= 3) {
1708 char *name = save_input_locale;
1710 /* Find next 'U' or 'u' and look from there */
1711 while ((name += strcspn(name, "Uu") + 1)
1712 <= save_input_locale + final_pos - 2)
1714 if (!isALPHA_FOLD_NE(*name, 't')
1715 || isALPHA_FOLD_NE(*(name + 1), 'f'))
1720 if (*(name) == '-') {
1721 if ((name > save_input_locale + final_pos - 1)) {
1726 if (*(name) == '8') {
1727 DEBUG_L(PerlIO_printf(Perl_debug_log,
1728 "Locale %s ends with UTF-8 in name\n",
1729 save_input_locale));
1730 Safefree(save_input_locale);
1734 DEBUG_L(PerlIO_printf(Perl_debug_log,
1735 "Locale %s doesn't end with UTF-8 in name\n",
1736 save_input_locale));
1741 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1743 && *(save_input_locale + final_pos - 0) == '1'
1744 && *(save_input_locale + final_pos - 1) == '0'
1745 && *(save_input_locale + final_pos - 2) == '0'
1746 && *(save_input_locale + final_pos - 3) == '5'
1747 && *(save_input_locale + final_pos - 4) == '6')
1749 DEBUG_L(PerlIO_printf(Perl_debug_log,
1750 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1751 save_input_locale));
1752 Safefree(save_input_locale);
1757 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1758 * since we are about to return FALSE anyway, there is no point in doing
1759 * this extra work */
1761 if (instr(save_input_locale, "8859")) {
1762 DEBUG_L(PerlIO_printf(Perl_debug_log,
1763 "Locale %s has 8859 in name, not UTF-8 locale\n",
1764 save_input_locale));
1765 Safefree(save_input_locale);
1770 DEBUG_L(PerlIO_printf(Perl_debug_log,
1771 "Assuming locale %s is not a UTF-8 locale\n",
1772 save_input_locale));
1773 Safefree(save_input_locale);
1781 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1784 /* Internal function which returns if we are in the scope of a pragma that
1785 * enables the locale category 'category'. 'compiling' should indicate if
1786 * this is during the compilation phase (TRUE) or not (FALSE). */
1788 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1790 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1791 if (! categories || categories == &PL_sv_placeholder) {
1795 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1796 * a valid unsigned */
1797 assert(category >= -1);
1798 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1802 Perl_my_strerror(pTHX_ const int errnum) {
1804 /* Uses C locale for the error text unless within scope of 'use locale' for
1807 #ifdef USE_LOCALE_MESSAGES
1808 if (! IN_LC(LC_MESSAGES)) {
1809 char * save_locale = setlocale(LC_MESSAGES, NULL);
1810 if (! isNAME_C_OR_POSIX(save_locale)) {
1813 /* The next setlocale likely will zap this, so create a copy */
1814 save_locale = savepv(save_locale);
1816 setlocale(LC_MESSAGES, "C");
1818 /* This points to the static space in Strerror, with all its
1820 errstr = Strerror(errnum);
1822 setlocale(LC_MESSAGES, save_locale);
1823 Safefree(save_locale);
1829 return Strerror(errnum);
1834 =head1 Locale-related functions and macros
1836 =for apidoc sync_locale
1838 Changing the program's locale should be avoided by XS code. Nevertheless,
1839 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1840 happens, Perl needs to be told that the locale has changed. Use this function
1841 to do so, before returning to Perl.
1847 Perl_sync_locale(pTHX)
1850 #ifdef USE_LOCALE_CTYPE
1851 new_ctype(setlocale(LC_CTYPE, NULL));
1852 #endif /* USE_LOCALE_CTYPE */
1854 #ifdef USE_LOCALE_COLLATE
1855 new_collate(setlocale(LC_COLLATE, NULL));
1858 #ifdef USE_LOCALE_NUMERIC
1859 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1860 new_numeric(setlocale(LC_NUMERIC, NULL));
1861 #endif /* USE_LOCALE_NUMERIC */
1865 #if defined(DEBUGGING) && defined(USE_LOCALE)
1868 Perl__setlocale_debug_string(const int category, /* category number,
1870 const char* const locale, /* locale name */
1872 /* return value from setlocale() when attempting to
1873 * set 'category' to 'locale' */
1874 const char* const retval)
1876 /* Returns a pointer to a NUL-terminated string in static storage with
1877 * added text about the info passed in. This is not thread safe and will
1878 * be overwritten by the next call, so this should be used just to
1879 * formulate a string to immediately print or savepv() on. */
1881 /* initialise to a non-null value to keep it out of BSS and so keep
1882 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
1883 static char ret[128] = "x";
1885 my_strlcpy(ret, "setlocale(", sizeof(ret));
1889 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
1893 my_strlcat(ret, "LC_ALL", sizeof(ret));
1898 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
1903 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
1908 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
1913 my_strlcat(ret, "LC_TIME", sizeof(ret));
1918 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
1923 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
1928 my_strlcat(ret, ", ", sizeof(ret));
1931 my_strlcat(ret, "\"", sizeof(ret));
1932 my_strlcat(ret, locale, sizeof(ret));
1933 my_strlcat(ret, "\"", sizeof(ret));
1936 my_strlcat(ret, "NULL", sizeof(ret));
1939 my_strlcat(ret, ") returned ", sizeof(ret));
1942 my_strlcat(ret, "\"", sizeof(ret));
1943 my_strlcat(ret, retval, sizeof(ret));
1944 my_strlcat(ret, "\"", sizeof(ret));
1947 my_strlcat(ret, "NULL", sizeof(ret));
1950 assert(strlen(ret) < sizeof(ret));
1959 * ex: set ts=8 sts=4 sw=4 et: