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 more 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().
464 * The design of locale collation is that every locale change is given an
465 * index 'PL_collation_ix'. The first time a string particpates in an
466 * operation that requires collation while locale collation is active, it
467 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
468 * magic includes the collation index, and the transformation of the string
469 * by strxfrm(), q.v. That transformation is used when doing comparisons,
470 * instead of the string itself. If a string changes, the magic is
471 * cleared. The next time the locale changes, the index is incremented,
472 * and so we know during a comparison that the transformation is not
473 * necessarily still valid, and so is recomputed. Note that if the locale
474 * changes enough times, the index could wrap (a U32), and it is possible
475 * that a transformation would improperly be considered valid, leading to
479 if (PL_collation_name) {
481 Safefree(PL_collation_name);
482 PL_collation_name = NULL;
484 PL_collation_standard = TRUE;
485 PL_collxfrm_base = 0;
486 PL_collxfrm_mult = 2;
490 /* If this is not the same locale as currently, set the new one up */
491 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
493 Safefree(PL_collation_name);
494 PL_collation_name = stdize_locale(savepv(newcoll));
495 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
498 /* A locale collation definition includes primary, secondary,
499 * tertiary, etc. weights for each character. To sort, the primary
500 * weights are used, and only if they compare equal, then the
501 * secondary weights are used, and only if they compare equal, then
502 * the tertiary, etc. strxfrm() works by taking the input string,
503 * say ABC, and creating an output string consisting of first the
504 * primary weights, A¹B¹C¹ followed by the secondary ones, A²B²C²;
505 * and then the tertiary, etc, yielding A¹B¹C¹A²B²C²A³B³C³....
506 * Some characters may not have weights at every level. In our
507 * example, let's say B doesn't have a tertiary weight, and A
508 * doesn't have a secondary weight. The constructed string is then
509 * going to be A¹B¹C¹B²C²A³C³.... This has the desired
510 * characteristics that strcmp() will look at the secondary or
511 * tertiary weights only if the strings compare equal at all higher
512 * priority weights. The length of the transformed string is
513 * roughly a linear function of the input string. It's not exactly
514 * linear because some characters don't have weights at all levels,
515 * and there are some complications, so there is often per-string
516 * overhead. When we call strxfrm() we have to allocate some
517 * memory to hold the transformed string. The calculations below
518 * try to find constants for this locale 'm' and 'b' so that m*x +
519 * b equals how much space we need given the size of the input
520 * string in 'x'. If we calculate too small, we increase the size
521 * as needed, and call strxfrm() again, but it is better to get it
522 * right the first time to avoid wasted expensive string
523 * transformations. */
524 /* 2: at most so many chars ('a', 'b'). */
525 /* 50: surely no system expands a char more. */
526 #define XFRMBUFSIZE (2 * 50)
527 char xbuf[XFRMBUFSIZE];
528 const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE);
529 const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE);
530 const SSize_t mult = fb - fa;
531 if (mult < 1 && !(fa == 0 && fb == 0))
532 Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
534 PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
535 PL_collxfrm_mult = mult;
540 PERL_UNUSED_ARG(newcoll);
541 #endif /* USE_LOCALE_COLLATE */
547 Perl_my_setlocale(pTHX_ int category, const char* locale)
549 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
550 * difference unless the input locale is "", which means on Windows to get
551 * the machine default, which is set via the computer's "Regional and
552 * Language Options" (or its current equivalent). In POSIX, it instead
553 * means to find the locale from the user's environment. This routine
554 * looks in the environment, and, if anything is found, uses that instead
555 * of going to the machine default. If there is no environment override,
556 * the machine default is used, as normal, by calling the real setlocale()
557 * with "". The POSIX behavior is to use the LC_ALL variable if set;
558 * otherwise to use the particular category's variable if set; otherwise to
559 * use the LANG variable. */
561 bool override_LC_ALL = FALSE;
564 if (locale && strEQ(locale, "")) {
566 locale = PerlEnv_getenv("LC_ALL");
572 override_LC_ALL = TRUE;
573 break; /* We already know its variable isn't set */
575 # ifdef USE_LOCALE_TIME
577 locale = PerlEnv_getenv("LC_TIME");
580 # ifdef USE_LOCALE_CTYPE
582 locale = PerlEnv_getenv("LC_CTYPE");
585 # ifdef USE_LOCALE_COLLATE
587 locale = PerlEnv_getenv("LC_COLLATE");
590 # ifdef USE_LOCALE_MONETARY
592 locale = PerlEnv_getenv("LC_MONETARY");
595 # ifdef USE_LOCALE_NUMERIC
597 locale = PerlEnv_getenv("LC_NUMERIC");
600 # ifdef USE_LOCALE_MESSAGES
602 locale = PerlEnv_getenv("LC_MESSAGES");
606 /* This is a category, like PAPER_SIZE that we don't
607 * know about; and so can't provide a wrapper. */
611 locale = PerlEnv_getenv("LANG");
621 result = setlocale(category, locale);
622 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
623 _setlocale_debug_string(category, locale, result)));
625 if (! override_LC_ALL) {
629 /* Here the input category was LC_ALL, and we have set it to what is in the
630 * LANG variable or the system default if there is no LANG. But these have
631 * lower priority than the other LC_foo variables, so override it for each
632 * one that is set. (If they are set to "", it means to use the same thing
633 * we just set LC_ALL to, so can skip) */
634 # ifdef USE_LOCALE_TIME
635 result = PerlEnv_getenv("LC_TIME");
636 if (result && strNE(result, "")) {
637 setlocale(LC_TIME, result);
638 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
640 _setlocale_debug_string(LC_TIME, result, "not captured")));
643 # ifdef USE_LOCALE_CTYPE
644 result = PerlEnv_getenv("LC_CTYPE");
645 if (result && strNE(result, "")) {
646 setlocale(LC_CTYPE, result);
647 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
649 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
652 # ifdef USE_LOCALE_COLLATE
653 result = PerlEnv_getenv("LC_COLLATE");
654 if (result && strNE(result, "")) {
655 setlocale(LC_COLLATE, result);
656 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
658 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
661 # ifdef USE_LOCALE_MONETARY
662 result = PerlEnv_getenv("LC_MONETARY");
663 if (result && strNE(result, "")) {
664 setlocale(LC_MONETARY, result);
665 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
667 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
670 # ifdef USE_LOCALE_NUMERIC
671 result = PerlEnv_getenv("LC_NUMERIC");
672 if (result && strNE(result, "")) {
673 setlocale(LC_NUMERIC, result);
674 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
676 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
679 # ifdef USE_LOCALE_MESSAGES
680 result = PerlEnv_getenv("LC_MESSAGES");
681 if (result && strNE(result, "")) {
682 setlocale(LC_MESSAGES, result);
683 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
685 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
689 result = setlocale(LC_ALL, NULL);
690 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
692 _setlocale_debug_string(LC_ALL, NULL, result)));
701 * Initialize locale awareness.
704 Perl_init_i18nl10n(pTHX_ int printwarn)
708 * 0 if not to output warning when setup locale is bad
709 * 1 if to output warning based on value of PERL_BADLANG
710 * >1 if to output regardless of PERL_BADLANG
713 * 1 = set ok or not applicable,
714 * 0 = fallback to a locale of lower priority
715 * -1 = fallback to all locales failed, not even to the C locale
717 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
718 * set, debugging information is output.
720 * This looks more complicated than it is, mainly due to the #ifdefs.
722 * We try to set LC_ALL to the value determined by the environment. If
723 * there is no LC_ALL on this platform, we try the individual categories we
724 * know about. If this works, we are done.
726 * But if it doesn't work, we have to do something else. We search the
727 * environment variables ourselves instead of relying on the system to do
728 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
729 * think there is one), and the ultimate fallback "C". This is all done in
730 * the same loop as above to avoid duplicating code, but it makes things
731 * more complex. After the original failure, we add the fallback
732 * possibilities to the list of locales to try, and iterate the loop
733 * through them all until one succeeds.
735 * On Ultrix, the locale MUST come from the environment, so there is
736 * preliminary code to set it. I (khw) am not sure that it is necessary,
737 * and that this couldn't be folded into the loop, but barring any real
738 * platforms to test on, it's staying as-is
740 * A slight complication is that in embedded Perls, the locale may already
741 * be set-up, and we don't want to get it from the normal environment
742 * variables. This is handled by having a special environment variable
743 * indicate we're in this situation. We simply set setlocale's 2nd
744 * parameter to be a NULL instead of "". That indicates to setlocale that
745 * it is not to change anything, but to return the current value,
746 * effectively initializing perl's db to what the locale already is.
748 * We play the same trick with NULL if a LC_ALL succeeds. We call
749 * setlocale() on the individual categores with NULL to get their existing
750 * values for our db, instead of trying to change them.
755 #if defined(USE_LOCALE)
756 #ifdef USE_LOCALE_CTYPE
757 char *curctype = NULL;
758 #endif /* USE_LOCALE_CTYPE */
759 #ifdef USE_LOCALE_COLLATE
760 char *curcoll = NULL;
761 #endif /* USE_LOCALE_COLLATE */
762 #ifdef USE_LOCALE_NUMERIC
764 #endif /* USE_LOCALE_NUMERIC */
766 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
769 /* NULL uses the existing already set up locale */
770 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
774 const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
777 # define DEBUG_LOCALE_INIT(category, locale, result) \
780 PerlIO_printf(Perl_debug_log, \
782 __FILE__, __LINE__, \
783 _setlocale_debug_string(category, \
789 # define DEBUG_LOCALE_INIT(a,b,c)
791 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
792 unsigned int trial_locales_count;
793 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
794 const char * const lang = savepv(PerlEnv_getenv("LANG"));
795 bool setlocale_failure = FALSE;
799 /* A later getenv() could zap this, so only use here */
800 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
802 const bool locwarn = (printwarn > 1
804 && (! bad_lang_use_once
806 /* disallow with "" or "0" */
808 && strNE("0", bad_lang_use_once)))));
810 char * sl_result; /* return from setlocale() */
813 /* In some systems you can find out the system default locale
814 * and use that as the fallback locale. */
815 # define SYSTEM_DEFAULT_LOCALE
817 #ifdef SYSTEM_DEFAULT_LOCALE
818 const char *system_default_locale = NULL;
821 #ifndef LOCALE_ENVIRON_REQUIRED
822 PERL_UNUSED_VAR(done);
823 PERL_UNUSED_VAR(locale_param);
827 * Ultrix setlocale(..., "") fails if there are no environment
828 * variables from which to get a locale name.
833 sl_result = my_setlocale(LC_ALL, setlocale_init);
834 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
838 setlocale_failure = TRUE;
840 if (! setlocale_failure) {
841 # ifdef USE_LOCALE_CTYPE
842 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
845 curctype = my_setlocale(LC_CTYPE, locale_param);
846 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
848 setlocale_failure = TRUE;
850 curctype = savepv(curctype);
851 # endif /* USE_LOCALE_CTYPE */
852 # ifdef USE_LOCALE_COLLATE
853 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
856 curcoll = my_setlocale(LC_COLLATE, locale_param);
857 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
859 setlocale_failure = TRUE;
861 curcoll = savepv(curcoll);
862 # endif /* USE_LOCALE_COLLATE */
863 # ifdef USE_LOCALE_NUMERIC
864 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
867 curnum = my_setlocale(LC_NUMERIC, locale_param);
868 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
870 setlocale_failure = TRUE;
872 curnum = savepv(curnum);
873 # endif /* USE_LOCALE_NUMERIC */
874 # ifdef USE_LOCALE_MESSAGES
875 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
878 sl_result = my_setlocale(LC_MESSAGES, locale_param);
879 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
881 setlocale_failure = TRUE;
883 # endif /* USE_LOCALE_MESSAGES */
884 # ifdef USE_LOCALE_MONETARY
885 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
888 sl_result = my_setlocale(LC_MONETARY, locale_param);
889 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
891 setlocale_failure = TRUE;
893 # endif /* USE_LOCALE_MONETARY */
898 #endif /* !LOCALE_ENVIRON_REQUIRED */
900 /* We try each locale in the list until we get one that works, or exhaust
901 * the list. Normally the loop is executed just once. But if setting the
902 * locale fails, inside the loop we add fallback trials to the array and so
903 * will execute the loop multiple times */
904 trial_locales[0] = setlocale_init;
905 trial_locales_count = 1;
906 for (i= 0; i < trial_locales_count; i++) {
907 const char * trial_locale = trial_locales[i];
911 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
912 * when i==0, but I (khw) don't think that behavior makes much
914 setlocale_failure = FALSE;
916 #ifdef SYSTEM_DEFAULT_LOCALE
918 /* On Windows machines, an entry of "" after the 0th means to use
919 * the system default locale, which we now proceed to get. */
920 if (strEQ(trial_locale, "")) {
923 /* Note that this may change the locale, but we are going to do
924 * that anyway just below */
925 system_default_locale = setlocale(LC_ALL, "");
926 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
928 /* Skip if invalid or it's already on the list of locales to
930 if (! system_default_locale) {
933 for (j = 0; j < trial_locales_count; j++) {
934 if (strEQ(system_default_locale, trial_locales[j])) {
939 trial_locale = system_default_locale;
942 #endif /* SYSTEM_DEFAULT_LOCALE */
946 sl_result = my_setlocale(LC_ALL, trial_locale);
947 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
949 setlocale_failure = TRUE;
952 /* Since LC_ALL succeeded, it should have changed all the other
953 * categories it can to its value; so we massage things so that the
954 * setlocales below just return their category's current values.
955 * This adequately handles the case in NetBSD where LC_COLLATE may
956 * not be defined for a locale, and setting it individually will
957 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
958 * the POSIX locale. */
963 if (!setlocale_failure) {
964 #ifdef USE_LOCALE_CTYPE
966 curctype = my_setlocale(LC_CTYPE, trial_locale);
967 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
969 setlocale_failure = TRUE;
971 curctype = savepv(curctype);
972 #endif /* USE_LOCALE_CTYPE */
973 #ifdef USE_LOCALE_COLLATE
975 curcoll = my_setlocale(LC_COLLATE, trial_locale);
976 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
978 setlocale_failure = TRUE;
980 curcoll = savepv(curcoll);
981 #endif /* USE_LOCALE_COLLATE */
982 #ifdef USE_LOCALE_NUMERIC
984 curnum = my_setlocale(LC_NUMERIC, trial_locale);
985 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
987 setlocale_failure = TRUE;
989 curnum = savepv(curnum);
990 #endif /* USE_LOCALE_NUMERIC */
991 #ifdef USE_LOCALE_MESSAGES
992 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
993 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
995 setlocale_failure = TRUE;
996 #endif /* USE_LOCALE_MESSAGES */
997 #ifdef USE_LOCALE_MONETARY
998 sl_result = my_setlocale(LC_MONETARY, trial_locale);
999 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
1001 setlocale_failure = TRUE;
1002 #endif /* USE_LOCALE_MONETARY */
1004 if (! setlocale_failure) { /* Success */
1009 /* Here, something failed; will need to try a fallback. */
1015 if (locwarn) { /* Output failure info only on the first one */
1018 PerlIO_printf(Perl_error_log,
1019 "perl: warning: Setting locale failed.\n");
1023 PerlIO_printf(Perl_error_log,
1024 "perl: warning: Setting locale failed for the categories:\n\t");
1025 # ifdef USE_LOCALE_CTYPE
1027 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
1028 # endif /* USE_LOCALE_CTYPE */
1029 # ifdef USE_LOCALE_COLLATE
1031 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
1032 # endif /* USE_LOCALE_COLLATE */
1033 # ifdef USE_LOCALE_NUMERIC
1035 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
1036 # endif /* USE_LOCALE_NUMERIC */
1037 PerlIO_printf(Perl_error_log, "and possibly others\n");
1041 PerlIO_printf(Perl_error_log,
1042 "perl: warning: Please check that your locale settings:\n");
1045 PerlIO_printf(Perl_error_log,
1046 "\tLANGUAGE = %c%s%c,\n",
1047 language ? '"' : '(',
1048 language ? language : "unset",
1049 language ? '"' : ')');
1052 PerlIO_printf(Perl_error_log,
1053 "\tLC_ALL = %c%s%c,\n",
1055 lc_all ? lc_all : "unset",
1056 lc_all ? '"' : ')');
1058 #if defined(USE_ENVIRON_ARRAY)
1061 for (e = environ; *e; e++) {
1062 if (strnEQ(*e, "LC_", 3)
1063 && strnNE(*e, "LC_ALL=", 7)
1064 && (p = strchr(*e, '=')))
1065 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1066 (int)(p - *e), *e, p + 1);
1070 PerlIO_printf(Perl_error_log,
1071 "\t(possibly more locale environment variables)\n");
1074 PerlIO_printf(Perl_error_log,
1075 "\tLANG = %c%s%c\n",
1077 lang ? lang : "unset",
1080 PerlIO_printf(Perl_error_log,
1081 " are supported and installed on your system.\n");
1084 /* Calculate what fallback locales to try. We have avoided this
1085 * until we have to, because failure is quite unlikely. This will
1086 * usually change the upper bound of the loop we are in.
1088 * Since the system's default way of setting the locale has not
1089 * found one that works, We use Perl's defined ordering: LC_ALL,
1090 * LANG, and the C locale. We don't try the same locale twice, so
1091 * don't add to the list if already there. (On POSIX systems, the
1092 * LC_ALL element will likely be a repeat of the 0th element "",
1093 * but there's no harm done by doing it explicitly.
1095 * Note that this tries the LC_ALL environment variable even on
1096 * systems which have no LC_ALL locale setting. This may or may
1097 * not have been originally intentional, but there's no real need
1098 * to change the behavior. */
1100 for (j = 0; j < trial_locales_count; j++) {
1101 if (strEQ(lc_all, trial_locales[j])) {
1105 trial_locales[trial_locales_count++] = lc_all;
1110 for (j = 0; j < trial_locales_count; j++) {
1111 if (strEQ(lang, trial_locales[j])) {
1115 trial_locales[trial_locales_count++] = lang;
1119 #if defined(WIN32) && defined(LC_ALL)
1120 /* For Windows, we also try the system default locale before "C".
1121 * (If there exists a Windows without LC_ALL we skip this because
1122 * it gets too complicated. For those, the "C" is the next
1123 * fallback possibility). The "" is the same as the 0th element of
1124 * the array, but the code at the loop above knows to treat it
1125 * differently when not the 0th */
1126 trial_locales[trial_locales_count++] = "";
1129 for (j = 0; j < trial_locales_count; j++) {
1130 if (strEQ("C", trial_locales[j])) {
1134 trial_locales[trial_locales_count++] = "C";
1137 } /* end of first time through the loop */
1143 } /* end of looping through the trial locales */
1145 if (ok < 1) { /* If we tried to fallback */
1147 if (! setlocale_failure) { /* fallback succeeded */
1148 msg = "Falling back to";
1150 else { /* fallback failed */
1152 /* We dropped off the end of the loop, so have to decrement i to
1153 * get back to the value the last time through */
1157 msg = "Failed to fall back to";
1159 /* To continue, we should use whatever values we've got */
1160 #ifdef USE_LOCALE_CTYPE
1162 curctype = savepv(setlocale(LC_CTYPE, NULL));
1163 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
1164 #endif /* USE_LOCALE_CTYPE */
1165 #ifdef USE_LOCALE_COLLATE
1167 curcoll = savepv(setlocale(LC_COLLATE, NULL));
1168 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
1169 #endif /* USE_LOCALE_COLLATE */
1170 #ifdef USE_LOCALE_NUMERIC
1172 curnum = savepv(setlocale(LC_NUMERIC, NULL));
1173 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
1174 #endif /* USE_LOCALE_NUMERIC */
1178 const char * description;
1179 const char * name = "";
1180 if (strEQ(trial_locales[i], "C")) {
1181 description = "the standard locale";
1184 #ifdef SYSTEM_DEFAULT_LOCALE
1185 else if (strEQ(trial_locales[i], "")) {
1186 description = "the system default locale";
1187 if (system_default_locale) {
1188 name = system_default_locale;
1191 #endif /* SYSTEM_DEFAULT_LOCALE */
1193 description = "a fallback locale";
1194 name = trial_locales[i];
1196 if (name && strNE(name, "")) {
1197 PerlIO_printf(Perl_error_log,
1198 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1201 PerlIO_printf(Perl_error_log,
1202 "perl: warning: %s %s.\n", msg, description);
1205 } /* End of tried to fallback */
1207 #ifdef USE_LOCALE_CTYPE
1208 new_ctype(curctype);
1209 #endif /* USE_LOCALE_CTYPE */
1211 #ifdef USE_LOCALE_COLLATE
1212 new_collate(curcoll);
1213 #endif /* USE_LOCALE_COLLATE */
1215 #ifdef USE_LOCALE_NUMERIC
1216 new_numeric(curnum);
1217 #endif /* USE_LOCALE_NUMERIC */
1219 #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
1220 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1221 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1222 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1223 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1225 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
1227 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
1228 This is an alternative to using the -C command line switch
1229 (the -C if present will override this). */
1231 const char *p = PerlEnv_getenv("PERL_UNICODE");
1232 PL_unicode = p ? parse_unicode_opts(&p) : 0;
1233 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1238 #ifdef USE_LOCALE_CTYPE
1240 #endif /* USE_LOCALE_CTYPE */
1241 #ifdef USE_LOCALE_COLLATE
1243 #endif /* USE_LOCALE_COLLATE */
1244 #ifdef USE_LOCALE_NUMERIC
1246 #endif /* USE_LOCALE_NUMERIC */
1255 #else /* !USE_LOCALE */
1256 PERL_UNUSED_ARG(printwarn);
1257 #endif /* USE_LOCALE */
1263 #ifdef USE_LOCALE_COLLATE
1266 * mem_collxfrm() is a bit like strxfrm() but with two important
1267 * differences. First, it handles embedded NULs. Second, it allocates
1268 * a bit more memory than needed for the transformed data itself.
1269 * The real transformed data begins at offset sizeof(collationix).
1270 * *xlen is set to the length of that, and doesn't include the collation index
1272 * Please see sv_collxfrm() to see how this is used.
1276 Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1279 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1281 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1283 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1284 /* the +1 is for the terminating NUL. */
1286 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
1287 Newx(xbuf, xAlloc, char);
1288 if (UNLIKELY(! xbuf))
1291 /* Store the collation id */
1292 *(U32*)xbuf = PL_collation_ix;
1293 xout = sizeof(PL_collation_ix);
1295 /* Then the transformation of the input. We loop until successful, or we
1297 for (xin = 0; xin < len; ) {
1301 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
1303 /* If the transformed string occupies less space than we told
1304 * strxfrm() was available, it means it successfully transformed
1305 * the whole string. */
1306 if ((STRLEN)xused < xAlloc - xout)
1309 if (UNLIKELY(xused >= PERL_INT_MAX))
1312 /* Otherwise it should be that the transformation stopped in the
1313 * middle because it ran out of space. Malloc more, and try again.
1315 xAlloc = (2 * xAlloc) + 1;
1316 Renew(xbuf, xAlloc, char);
1317 if (UNLIKELY(! xbuf))
1321 xin += strlen(s + xin) + 1;
1324 /* Embedded NULs are understood but silently skipped
1325 * because they make no sense in locale collation. */
1329 *xlen = xout - sizeof(PL_collation_ix);
1338 #endif /* USE_LOCALE_COLLATE */
1343 Perl__is_cur_LC_category_utf8(pTHX_ int category)
1345 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1346 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
1347 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
1348 * could give the wrong result. The result will very likely be correct for
1349 * languages that have commonly used non-ASCII characters, but for notably
1350 * English, it comes down to if the locale's name ends in something like
1351 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
1353 char *save_input_locale = NULL;
1357 assert(category != LC_ALL);
1360 /* First dispose of the trivial cases */
1361 save_input_locale = setlocale(category, NULL);
1362 if (! save_input_locale) {
1363 DEBUG_L(PerlIO_printf(Perl_debug_log,
1364 "Could not find current locale for category %d\n",
1366 return FALSE; /* XXX maybe should croak */
1368 save_input_locale = stdize_locale(savepv(save_input_locale));
1369 if (isNAME_C_OR_POSIX(save_input_locale)) {
1370 DEBUG_L(PerlIO_printf(Perl_debug_log,
1371 "Current locale for category %d is %s\n",
1372 category, save_input_locale));
1373 Safefree(save_input_locale);
1377 #if defined(USE_LOCALE_CTYPE) \
1378 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
1380 { /* Next try nl_langinfo or MB_CUR_MAX if available */
1382 char *save_ctype_locale = NULL;
1385 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
1387 /* Get the current LC_CTYPE locale */
1388 save_ctype_locale = setlocale(LC_CTYPE, NULL);
1389 if (! save_ctype_locale) {
1390 DEBUG_L(PerlIO_printf(Perl_debug_log,
1391 "Could not find current locale for LC_CTYPE\n"));
1392 goto cant_use_nllanginfo;
1394 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
1396 /* If LC_CTYPE and the desired category use the same locale, this
1397 * means that finding the value for LC_CTYPE is the same as finding
1398 * the value for the desired category. Otherwise, switch LC_CTYPE
1399 * to the desired category's locale */
1400 if (strEQ(save_ctype_locale, save_input_locale)) {
1401 Safefree(save_ctype_locale);
1402 save_ctype_locale = NULL;
1404 else if (! setlocale(LC_CTYPE, save_input_locale)) {
1405 DEBUG_L(PerlIO_printf(Perl_debug_log,
1406 "Could not change LC_CTYPE locale to %s\n",
1407 save_input_locale));
1408 Safefree(save_ctype_locale);
1409 goto cant_use_nllanginfo;
1413 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1414 save_input_locale));
1416 /* Here the current LC_CTYPE is set to the locale of the category whose
1417 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1418 * should give the correct results */
1420 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
1422 char *codeset = nl_langinfo(CODESET);
1423 if (codeset && strNE(codeset, "")) {
1424 codeset = savepv(codeset);
1426 /* If we switched LC_CTYPE, switch back */
1427 if (save_ctype_locale) {
1428 setlocale(LC_CTYPE, save_ctype_locale);
1429 Safefree(save_ctype_locale);
1432 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1433 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1435 DEBUG_L(PerlIO_printf(Perl_debug_log,
1436 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1439 Safefree(save_input_locale);
1447 /* Here, either we don't have nl_langinfo, or it didn't return a
1448 * codeset. Try MB_CUR_MAX */
1450 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1451 * Unicode code point. Since UTF-8 is the only non-single byte
1452 * encoding we handle, we just say any such encoding is UTF-8, and if
1453 * turns out to be wrong, other things will fail */
1454 is_utf8 = MB_CUR_MAX >= 4;
1456 DEBUG_L(PerlIO_printf(Perl_debug_log,
1457 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1458 (int) MB_CUR_MAX, is_utf8));
1460 Safefree(save_input_locale);
1464 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1465 * since they are both in the C99 standard. We can feed a known byte
1466 * string to the latter function, and check that it gives the expected
1470 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
1472 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
1473 != strlen(HYPHEN_UTF8)
1474 || wc != (wchar_t) 0x2010)
1477 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
1478 DEBUG_L(PerlIO_printf(Perl_debug_log,
1479 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1480 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
1485 /* If we switched LC_CTYPE, switch back */
1486 if (save_ctype_locale) {
1487 setlocale(LC_CTYPE, save_ctype_locale);
1488 Safefree(save_ctype_locale);
1495 cant_use_nllanginfo:
1497 #else /* nl_langinfo should work if available, so don't bother compiling this
1498 fallback code. The final fallback of looking at the name is
1499 compiled, and will be executed if nl_langinfo fails */
1501 /* nl_langinfo not available or failed somehow. Next try looking at the
1502 * currency symbol to see if it disambiguates things. Often that will be
1503 * in the native script, and if the symbol isn't in UTF-8, we know that the
1504 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
1505 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1508 #ifdef HAS_LOCALECONV
1509 # ifdef USE_LOCALE_MONETARY
1511 char *save_monetary_locale = NULL;
1512 bool only_ascii = FALSE;
1513 bool is_utf8 = FALSE;
1516 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1517 * the desired category, if it isn't that locale already */
1519 if (category != LC_MONETARY) {
1521 save_monetary_locale = setlocale(LC_MONETARY, NULL);
1522 if (! save_monetary_locale) {
1523 DEBUG_L(PerlIO_printf(Perl_debug_log,
1524 "Could not find current locale for LC_MONETARY\n"));
1525 goto cant_use_monetary;
1527 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
1529 if (strEQ(save_monetary_locale, save_input_locale)) {
1530 Safefree(save_monetary_locale);
1531 save_monetary_locale = NULL;
1533 else if (! setlocale(LC_MONETARY, save_input_locale)) {
1534 DEBUG_L(PerlIO_printf(Perl_debug_log,
1535 "Could not change LC_MONETARY locale to %s\n",
1536 save_input_locale));
1537 Safefree(save_monetary_locale);
1538 goto cant_use_monetary;
1542 /* Here the current LC_MONETARY is set to the locale of the category
1543 * whose information is desired. */
1547 || ! lc->currency_symbol
1548 || is_invariant_string((U8 *) lc->currency_symbol, 0))
1550 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));
1554 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
1557 /* If we changed it, restore LC_MONETARY to its original locale */
1558 if (save_monetary_locale) {
1559 setlocale(LC_MONETARY, save_monetary_locale);
1560 Safefree(save_monetary_locale);
1565 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1566 * otherwise assume the locale is UTF-8 if and only if the symbol
1567 * is non-ascii UTF-8. */
1568 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1569 save_input_locale, is_utf8));
1570 Safefree(save_input_locale);
1576 # endif /* USE_LOCALE_MONETARY */
1577 #endif /* HAS_LOCALECONV */
1579 #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1581 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1582 * the names of the months and weekdays, timezone, and am/pm indicator */
1584 char *save_time_locale = NULL;
1586 bool is_dst = FALSE;
1590 char * formatted_time;
1593 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1594 * desired category, if it isn't that locale already */
1596 if (category != LC_TIME) {
1598 save_time_locale = setlocale(LC_TIME, NULL);
1599 if (! save_time_locale) {
1600 DEBUG_L(PerlIO_printf(Perl_debug_log,
1601 "Could not find current locale for LC_TIME\n"));
1604 save_time_locale = stdize_locale(savepv(save_time_locale));
1606 if (strEQ(save_time_locale, save_input_locale)) {
1607 Safefree(save_time_locale);
1608 save_time_locale = NULL;
1610 else if (! setlocale(LC_TIME, save_input_locale)) {
1611 DEBUG_L(PerlIO_printf(Perl_debug_log,
1612 "Could not change LC_TIME locale to %s\n",
1613 save_input_locale));
1614 Safefree(save_time_locale);
1619 /* Here the current LC_TIME is set to the locale of the category
1620 * whose information is desired. Look at all the days of the week and
1621 * month names, and the timezone and am/pm indicator for UTF-8 variant
1622 * characters. The first such a one found will tell us if the locale
1623 * is UTF-8 or not */
1625 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1626 formatted_time = my_strftime("%A %B %Z %p",
1627 0, 0, hour, dom, month, 112, 0, 0, is_dst);
1628 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
1630 /* Here, we didn't find a non-ASCII. Try the next time through
1631 * with the complemented dst and am/pm, and try with the next
1632 * weekday. After we have gotten all weekdays, try the next
1635 hour = (hour + 12) % 24;
1643 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1644 * false otherwise. But first, restore LC_TIME to its original
1645 * locale if we changed it */
1646 if (save_time_locale) {
1647 setlocale(LC_TIME, save_time_locale);
1648 Safefree(save_time_locale);
1651 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1653 is_utf8_string((U8 *) formatted_time, 0)));
1654 Safefree(save_input_locale);
1655 return is_utf8_string((U8 *) formatted_time, 0);
1658 /* Falling off the end of the loop indicates all the names were just
1659 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1660 * to its original locale */
1661 if (save_time_locale) {
1662 setlocale(LC_TIME, save_time_locale);
1663 Safefree(save_time_locale);
1665 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));
1671 #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
1673 /* This code is ifdefd out because it was found to not be necessary in testing
1674 * on our dromedary test machine, which has over 700 locales. There, this
1675 * added no value to looking at the currency symbol and the time strings. I
1676 * left it in so as to avoid rewriting it if real-world experience indicates
1677 * that dromedary is an outlier. Essentially, instead of returning abpve if we
1678 * haven't found illegal utf8, we continue on and examine all the strerror()
1679 * messages on the platform for utf8ness. If all are ASCII, we still don't
1680 * know the answer; but otherwise we have a pretty good indication of the
1681 * utf8ness. The reason this doesn't help much is that the messages may not
1682 * have been translated into the locale. The currency symbol and time strings
1683 * are much more likely to have been translated. */
1686 bool is_utf8 = FALSE;
1687 bool non_ascii = FALSE;
1688 char *save_messages_locale = NULL;
1689 const char * errmsg = NULL;
1691 /* Like above, we set LC_MESSAGES to the locale of the desired
1692 * category, if it isn't that locale already */
1694 if (category != LC_MESSAGES) {
1696 save_messages_locale = setlocale(LC_MESSAGES, NULL);
1697 if (! save_messages_locale) {
1698 DEBUG_L(PerlIO_printf(Perl_debug_log,
1699 "Could not find current locale for LC_MESSAGES\n"));
1700 goto cant_use_messages;
1702 save_messages_locale = stdize_locale(savepv(save_messages_locale));
1704 if (strEQ(save_messages_locale, save_input_locale)) {
1705 Safefree(save_messages_locale);
1706 save_messages_locale = NULL;
1708 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
1709 DEBUG_L(PerlIO_printf(Perl_debug_log,
1710 "Could not change LC_MESSAGES locale to %s\n",
1711 save_input_locale));
1712 Safefree(save_messages_locale);
1713 goto cant_use_messages;
1717 /* Here the current LC_MESSAGES is set to the locale of the category
1718 * whose information is desired. Look through all the messages. We
1719 * can't use Strerror() here because it may expand to code that
1720 * segfaults in miniperl */
1722 for (e = 0; e <= sys_nerr; e++) {
1724 errmsg = sys_errlist[e];
1725 if (errno || !errmsg) {
1728 errmsg = savepv(errmsg);
1729 if (! is_invariant_string((U8 *) errmsg, 0)) {
1731 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1737 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1738 if (save_messages_locale) {
1739 setlocale(LC_MESSAGES, save_messages_locale);
1740 Safefree(save_messages_locale);
1745 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1746 * any non-ascii means it is one; otherwise we assume it isn't */
1747 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1750 Safefree(save_input_locale);
1754 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));
1760 #endif /* the code that is compiled when no nl_langinfo */
1762 #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1764 /* As a last resort, look at the locale name to see if it matches
1765 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1766 * return of setlocale(), is actually defined to be opaque, so we can't
1767 * really rely on the absence of various substrings in the name to indicate
1768 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1769 * be a UTF-8 locale. Similarly for the other common names */
1771 final_pos = strlen(save_input_locale) - 1;
1772 if (final_pos >= 3) {
1773 char *name = save_input_locale;
1775 /* Find next 'U' or 'u' and look from there */
1776 while ((name += strcspn(name, "Uu") + 1)
1777 <= save_input_locale + final_pos - 2)
1779 if (!isALPHA_FOLD_NE(*name, 't')
1780 || isALPHA_FOLD_NE(*(name + 1), 'f'))
1785 if (*(name) == '-') {
1786 if ((name > save_input_locale + final_pos - 1)) {
1791 if (*(name) == '8') {
1792 DEBUG_L(PerlIO_printf(Perl_debug_log,
1793 "Locale %s ends with UTF-8 in name\n",
1794 save_input_locale));
1795 Safefree(save_input_locale);
1799 DEBUG_L(PerlIO_printf(Perl_debug_log,
1800 "Locale %s doesn't end with UTF-8 in name\n",
1801 save_input_locale));
1806 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1808 && *(save_input_locale + final_pos - 0) == '1'
1809 && *(save_input_locale + final_pos - 1) == '0'
1810 && *(save_input_locale + final_pos - 2) == '0'
1811 && *(save_input_locale + final_pos - 3) == '5'
1812 && *(save_input_locale + final_pos - 4) == '6')
1814 DEBUG_L(PerlIO_printf(Perl_debug_log,
1815 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1816 save_input_locale));
1817 Safefree(save_input_locale);
1822 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1823 * since we are about to return FALSE anyway, there is no point in doing
1824 * this extra work */
1826 if (instr(save_input_locale, "8859")) {
1827 DEBUG_L(PerlIO_printf(Perl_debug_log,
1828 "Locale %s has 8859 in name, not UTF-8 locale\n",
1829 save_input_locale));
1830 Safefree(save_input_locale);
1835 DEBUG_L(PerlIO_printf(Perl_debug_log,
1836 "Assuming locale %s is not a UTF-8 locale\n",
1837 save_input_locale));
1838 Safefree(save_input_locale);
1846 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1849 /* Internal function which returns if we are in the scope of a pragma that
1850 * enables the locale category 'category'. 'compiling' should indicate if
1851 * this is during the compilation phase (TRUE) or not (FALSE). */
1853 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1855 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1856 if (! categories || categories == &PL_sv_placeholder) {
1860 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1861 * a valid unsigned */
1862 assert(category >= -1);
1863 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1867 Perl_my_strerror(pTHX_ const int errnum) {
1870 /* Uses C locale for the error text unless within scope of 'use locale' for
1873 #ifdef USE_LOCALE_MESSAGES
1874 if (! IN_LC(LC_MESSAGES)) {
1877 /* We have a critical section to prevent another thread from changing
1878 * the locale out from under us (or zapping the buffer returned from
1882 save_locale = setlocale(LC_MESSAGES, NULL);
1883 if (! isNAME_C_OR_POSIX(save_locale)) {
1886 /* The next setlocale likely will zap this, so create a copy */
1887 save_locale = savepv(save_locale);
1889 setlocale(LC_MESSAGES, "C");
1891 /* This points to the static space in Strerror, with all its
1893 errstr = Strerror(errnum);
1895 setlocale(LC_MESSAGES, save_locale);
1896 Safefree(save_locale);
1907 return Strerror(errnum);
1912 =head1 Locale-related functions and macros
1914 =for apidoc sync_locale
1916 Changing the program's locale should be avoided by XS code. Nevertheless,
1917 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1918 happens, Perl needs to be told that the locale has changed. Use this function
1919 to do so, before returning to Perl.
1925 Perl_sync_locale(pTHX)
1928 #ifdef USE_LOCALE_CTYPE
1929 new_ctype(setlocale(LC_CTYPE, NULL));
1930 #endif /* USE_LOCALE_CTYPE */
1932 #ifdef USE_LOCALE_COLLATE
1933 new_collate(setlocale(LC_COLLATE, NULL));
1936 #ifdef USE_LOCALE_NUMERIC
1937 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1938 new_numeric(setlocale(LC_NUMERIC, NULL));
1939 #endif /* USE_LOCALE_NUMERIC */
1943 #if defined(DEBUGGING) && defined(USE_LOCALE)
1946 Perl__setlocale_debug_string(const int category, /* category number,
1948 const char* const locale, /* locale name */
1950 /* return value from setlocale() when attempting to
1951 * set 'category' to 'locale' */
1952 const char* const retval)
1954 /* Returns a pointer to a NUL-terminated string in static storage with
1955 * added text about the info passed in. This is not thread safe and will
1956 * be overwritten by the next call, so this should be used just to
1957 * formulate a string to immediately print or savepv() on. */
1959 /* initialise to a non-null value to keep it out of BSS and so keep
1960 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
1961 static char ret[128] = "x";
1963 my_strlcpy(ret, "setlocale(", sizeof(ret));
1967 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
1971 my_strlcat(ret, "LC_ALL", sizeof(ret));
1976 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
1981 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
1986 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
1991 my_strlcat(ret, "LC_TIME", sizeof(ret));
1996 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
2001 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
2006 my_strlcat(ret, ", ", sizeof(ret));
2009 my_strlcat(ret, "\"", sizeof(ret));
2010 my_strlcat(ret, locale, sizeof(ret));
2011 my_strlcat(ret, "\"", sizeof(ret));
2014 my_strlcat(ret, "NULL", sizeof(ret));
2017 my_strlcat(ret, ") returned ", sizeof(ret));
2020 my_strlcat(ret, "\"", sizeof(ret));
2021 my_strlcat(ret, retval, sizeof(ret));
2022 my_strlcat(ret, "\"", sizeof(ret));
2025 my_strlcat(ret, "NULL", sizeof(ret));
2028 assert(strlen(ret) < sizeof(ret));
2037 * ex: set ts=8 sts=4 sw=4 et: