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 code generally doesn't pay
27 * any 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
39 #include "perl_langinfo.h"
44 /* If the environment says to, we can output debugging information during
45 * initialization. This is done before option parsing, and before any thread
46 * creation, so can be a file-level static */
47 #if ! defined(DEBUGGING) || defined(PERL_GLOBAL_STRUCT)
48 # define debug_initialization 0
49 # define DEBUG_INITIALIZATION_set(v)
51 static bool debug_initialization = FALSE;
52 # define DEBUG_INITIALIZATION_set(v) (debug_initialization = v)
55 /* strlen() of a literal string constant. We might want this more general,
56 * but using it in just this file for now. A problem with more generality is
57 * the compiler warnings about comparing unlike signs */
58 #define STRLENs(s) (sizeof("" s "") - 1)
60 /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
61 * return of setlocale(), then this is extremely likely to be the C or POSIX
62 * locale. However, the output of setlocale() is documented to be opaque, but
63 * the odds are extremely small that it would return these two strings for some
64 * other locale. Note that VMS in these two locales includes many non-ASCII
65 * characters as controls and punctuation (below are hex bytes):
67 * punct: A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
68 * Oddly, none there are listed as alphas, though some represent alphabetics
69 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
70 #define isNAME_C_OR_POSIX(name) \
72 && (( *(name) == 'C' && (*(name + 1)) == '\0') \
73 || strEQ((name), "POSIX")))
77 /* This code keeps a LRU cache of the UTF-8ness of the locales it has so-far
78 * looked up. This is in the form of a C string: */
80 #define UTF8NESS_SEP "\v"
81 #define UTF8NESS_PREFIX "\f"
83 /* So, the string looks like:
85 * \vC\a0\vPOSIX\a0\vam_ET\a0\vaf_ZA.utf8\a1\ven_US.UTF-8\a1\0
87 * where the digit 0 after the \a indicates that the locale starting just
88 * after the preceding \v is not UTF-8, and the digit 1 mean it is. */
90 STATIC_ASSERT_DECL(STRLENs(UTF8NESS_SEP) == 1);
91 STATIC_ASSERT_DECL(STRLENs(UTF8NESS_PREFIX) == 1);
93 #define C_and_POSIX_utf8ness UTF8NESS_SEP "C" UTF8NESS_PREFIX "0" \
94 UTF8NESS_SEP "POSIX" UTF8NESS_PREFIX "0"
96 /* The cache is initialized to C_and_POSIX_utf8ness at start up. These are
97 * kept there always. The remining portion of the cache is LRU, with the
98 * oldest looked-up locale at the tail end */
101 S_stdize_locale(pTHX_ char *locs)
103 /* Standardize the locale name from a string returned by 'setlocale',
104 * possibly modifying that string.
106 * The typical return value of setlocale() is either
107 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
108 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
109 * (the space-separated values represent the various sublocales,
110 * in some unspecified order). This is not handled by this function.
112 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
113 * which is harmful for further use of the string in setlocale(). This
114 * function removes the trailing new line and everything up through the '='
117 const char * const s = strchr(locs, '=');
120 PERL_ARGS_ASSERT_STDIZE_LOCALE;
123 const char * const t = strchr(s, '.');
126 const char * const u = strchr(t, '\n');
127 if (u && (u[1] == 0)) {
128 const STRLEN len = u - s;
129 Move(s + 1, locs, len, char);
137 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
142 /* Two parallel arrays; first the locale categories Perl uses on this system;
143 * the second array is their names. These arrays are in mostly arbitrary
146 const int categories[] = {
148 # ifdef USE_LOCALE_NUMERIC
151 # ifdef USE_LOCALE_CTYPE
154 # ifdef USE_LOCALE_COLLATE
157 # ifdef USE_LOCALE_TIME
160 # ifdef USE_LOCALE_MESSAGES
163 # ifdef USE_LOCALE_MONETARY
166 # ifdef USE_LOCALE_ADDRESS
169 # ifdef USE_LOCALE_IDENTIFICATION
172 # ifdef USE_LOCALE_MEASUREMENT
175 # ifdef USE_LOCALE_PAPER
178 # ifdef USE_LOCALE_TELEPHONE
184 -1 /* Placeholder because C doesn't allow a
185 trailing comma, and it would get complicated
186 with all the #ifdef's */
189 /* The top-most real element is LC_ALL */
191 const char * category_names[] = {
193 # ifdef USE_LOCALE_NUMERIC
196 # ifdef USE_LOCALE_CTYPE
199 # ifdef USE_LOCALE_COLLATE
202 # ifdef USE_LOCALE_TIME
205 # ifdef USE_LOCALE_MESSAGES
208 # ifdef USE_LOCALE_MONETARY
211 # ifdef USE_LOCALE_ADDRESS
214 # ifdef USE_LOCALE_IDENTIFICATION
217 # ifdef USE_LOCALE_MEASUREMENT
220 # ifdef USE_LOCALE_PAPER
223 # ifdef USE_LOCALE_TELEPHONE
229 NULL /* Placeholder */
234 /* On systems with LC_ALL, it is kept in the highest index position. (-2
235 * to account for the final unused placeholder element.) */
236 # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 2)
240 /* On systems without LC_ALL, we pretend it is there, one beyond the real
241 * top element, hence in the unused placeholder element. */
242 # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 1)
246 /* Pretending there is an LC_ALL element just above allows us to avoid most
247 * special cases. Most loops through these arrays in the code below are
248 * written like 'for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++)'. They will work
249 * on either type of system. But the code must be written to not access the
250 * element at 'LC_ALL_INDEX' except on platforms that have it. This can be
251 * checked for at compile time by using the #define LC_ALL_INDEX which is only
252 * defined if we do have LC_ALL. */
255 S_category_name(const int category)
261 if (category == LC_ALL) {
267 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
268 if (category == categories[i]) {
269 return category_names[i];
274 const char suffix[] = " (unknown)";
276 Size_t length = sizeof(suffix) + 1;
285 /* Calculate the number of digits */
291 Newx(unknown, length, char);
292 my_snprintf(unknown, length, "%d%s", category, suffix);
298 /* Now create LC_foo_INDEX #defines for just those categories on this system */
299 # ifdef USE_LOCALE_NUMERIC
300 # define LC_NUMERIC_INDEX 0
301 # define _DUMMY_NUMERIC LC_NUMERIC_INDEX
303 # define _DUMMY_NUMERIC -1
305 # ifdef USE_LOCALE_CTYPE
306 # define LC_CTYPE_INDEX _DUMMY_NUMERIC + 1
307 # define _DUMMY_CTYPE LC_CTYPE_INDEX
309 # define _DUMMY_CTYPE _DUMMY_NUMERIC
311 # ifdef USE_LOCALE_COLLATE
312 # define LC_COLLATE_INDEX _DUMMY_CTYPE + 1
313 # define _DUMMY_COLLATE LC_COLLATE_INDEX
315 # define _DUMMY_COLLATE _DUMMY_COLLATE
317 # ifdef USE_LOCALE_TIME
318 # define LC_TIME_INDEX _DUMMY_COLLATE + 1
319 # define _DUMMY_TIME LC_TIME_INDEX
321 # define _DUMMY_TIME _DUMMY_COLLATE
323 # ifdef USE_LOCALE_MESSAGES
324 # define LC_MESSAGES_INDEX _DUMMY_TIME + 1
325 # define _DUMMY_MESSAGES LC_MESSAGES_INDEX
327 # define _DUMMY_MESSAGES _DUMMY_TIME
329 # ifdef USE_LOCALE_MONETARY
330 # define LC_MONETARY_INDEX _DUMMY_MESSAGES + 1
331 # define _DUMMY_MONETARY LC_MONETARY_INDEX
333 # define _DUMMY_MONETARY _DUMMY_MESSAGES
335 # ifdef USE_LOCALE_ADDRESS
336 # define LC_ADDRESS_INDEX _DUMMY_MONETARY + 1
337 # define _DUMMY_ADDRESS LC_ADDRESS_INDEX
339 # define _DUMMY_ADDRESS _DUMMY_MONETARY
341 # ifdef USE_LOCALE_IDENTIFICATION
342 # define LC_IDENTIFICATION_INDEX _DUMMY_ADDRESS + 1
343 # define _DUMMY_IDENTIFICATION LC_IDENTIFICATION_INDEX
345 # define _DUMMY_IDENTIFICATION _DUMMY_ADDRESS
347 # ifdef USE_LOCALE_MEASUREMENT
348 # define LC_MEASUREMENT_INDEX _DUMMY_IDENTIFICATION + 1
349 # define _DUMMY_MEASUREMENT LC_MEASUREMENT_INDEX
351 # define _DUMMY_MEASUREMENT _DUMMY_IDENTIFICATION
353 # ifdef USE_LOCALE_PAPER
354 # define LC_PAPER_INDEX _DUMMY_MEASUREMENT + 1
355 # define _DUMMY_PAPER LC_PAPER_INDEX
357 # define _DUMMY_PAPER _DUMMY_MEASUREMENT
359 # ifdef USE_LOCALE_TELEPHONE
360 # define LC_TELEPHONE_INDEX _DUMMY_PAPER + 1
361 # define _DUMMY_TELEPHONE LC_TELEPHONE_INDEX
363 # define _DUMMY_TELEPHONE _DUMMY_PAPER
366 # define LC_ALL_INDEX _DUMMY_TELEPHONE + 1
368 #endif /* ifdef USE_LOCALE */
370 /* Windows requres a customized base-level setlocale() */
372 # define my_setlocale(cat, locale) win32_setlocale(cat, locale)
374 # define my_setlocale(cat, locale) setlocale(cat, locale)
377 /* Just placeholders for now. "_c" is intended to be called when the category
378 * is a constant known at compile time; "_r", not known until run time */
379 # define do_setlocale_c(category, locale) my_setlocale(category, locale)
380 # define do_setlocale_r(category, locale) my_setlocale(category, locale)
383 S_set_numeric_radix(pTHX_ const bool use_locale)
385 /* If 'use_locale' is FALSE, set to use a dot for the radix character. If
386 * TRUE, use the radix character derived from the current locale */
388 #if defined(USE_LOCALE_NUMERIC) && ( defined(HAS_LOCALECONV) \
389 || defined(HAS_NL_LANGINFO))
391 /* We only set up the radix SV if we are to use a locale radix ... */
393 const char * radix = my_nl_langinfo(PERL_RADIXCHAR, FALSE);
394 /* FALSE => already in dest locale */
395 /* ... and the character being used isn't a dot */
396 if (strNE(radix, ".")) {
397 const U8 * first_variant;
399 if (PL_numeric_radix_sv) {
400 sv_setpv(PL_numeric_radix_sv, radix);
403 PL_numeric_radix_sv = newSVpv(radix, 0);
406 /* If there is a byte variant under UTF-8, and if the remainder of
407 * the string starting there is valid UTF-8, and we are in a UTF-8
408 * locale, then mark the radix as being in UTF-8 */
409 if ( ! is_utf8_invariant_string_loc(
410 (U8 *) SvPVX(PL_numeric_radix_sv),
411 SvCUR(PL_numeric_radix_sv),
413 && is_utf8_string(first_variant,
414 SvCUR(PL_numeric_radix_sv)
415 - ((char *) first_variant
416 - SvPVX(PL_numeric_radix_sv)))
417 && _is_cur_LC_category_utf8(LC_NUMERIC))
419 SvUTF8_on(PL_numeric_radix_sv);
425 SvREFCNT_dec(PL_numeric_radix_sv);
426 PL_numeric_radix_sv = NULL;
432 if (DEBUG_L_TEST || debug_initialization) {
433 PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
434 (PL_numeric_radix_sv)
435 ? SvPVX(PL_numeric_radix_sv)
437 (PL_numeric_radix_sv)
438 ? cBOOL(SvUTF8(PL_numeric_radix_sv))
443 #endif /* USE_LOCALE_NUMERIC and can find the radix char */
449 Perl_new_numeric(pTHX_ const char *newnum)
452 #ifndef USE_LOCALE_NUMERIC
454 PERL_UNUSED_ARG(newnum);
458 /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell
459 * core Perl this and that 'newnum' is the name of the new locale.
460 * It installs this locale as the current underlying default.
462 * The default locale and the C locale can be toggled between by use of the
463 * set_numeric_underlying() and set_numeric_standard() functions, which
464 * should probably not be called directly, but only via macros like
465 * SET_NUMERIC_STANDARD() in perl.h.
467 * The toggling is necessary mainly so that a non-dot radix decimal point
468 * character can be output, while allowing internal calculations to use a
471 * This sets several interpreter-level variables:
472 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
473 * PL_numeric_underlying A boolean indicating if the toggled state is such
474 * that the current locale is the program's underlying
476 * PL_numeric_standard An int indicating if the toggled state is such
477 * that the current locale is the C locale or
478 * indistinguishable from the C locale. If non-zero, it
479 * is in C; if > 1, it means it may not be toggled away
481 * PL_numeric_underlying_is_standard A bool kept by this function
482 * indicating that the underlying locale and the standard
483 * C locale are indistinguishable for the purposes of
484 * LC_NUMERIC. This happens when both of the above two
485 * variables are true at the same time. (Toggling is a
486 * no-op under these circumstances.) This variable is
487 * used to avoid having to recalculate.
488 * Any code changing the locale (outside this file) should use
489 * POSIX::setlocale, which calls this function. Therefore this function
490 * should be called directly only from this file and from
491 * POSIX::setlocale() */
496 Safefree(PL_numeric_name);
497 PL_numeric_name = NULL;
498 PL_numeric_standard = TRUE;
499 PL_numeric_underlying = TRUE;
500 PL_numeric_underlying_is_standard = TRUE;
504 save_newnum = stdize_locale(savepv(newnum));
505 PL_numeric_underlying = TRUE;
506 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
508 /* If its name isn't C nor POSIX, it could still be indistinguishable from
510 if (! PL_numeric_standard) {
511 PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(PERL_RADIXCHAR,
512 FALSE /* Don't toggle locale */ ))
513 && strEQ("", my_nl_langinfo(PERL_THOUSEP,
517 /* Save the new name if it isn't the same as the previous one, if any */
518 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
519 Safefree(PL_numeric_name);
520 PL_numeric_name = save_newnum;
523 Safefree(save_newnum);
526 PL_numeric_underlying_is_standard = PL_numeric_standard;
528 if (DEBUG_L_TEST || debug_initialization) {
529 PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name);
532 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
533 * have to worry about the radix being a non-dot. (Core operations that
534 * need the underlying locale change to it temporarily). */
535 set_numeric_standard();
537 #endif /* USE_LOCALE_NUMERIC */
542 Perl_set_numeric_standard(pTHX)
545 #ifdef USE_LOCALE_NUMERIC
547 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
548 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
549 * macro avoids calling this routine if toggling isn't necessary according
550 * to our records (which could be wrong if some XS code has changed the
551 * locale behind our back) */
553 do_setlocale_c(LC_NUMERIC, "C");
554 PL_numeric_standard = TRUE;
555 PL_numeric_underlying = PL_numeric_underlying_is_standard;
556 set_numeric_radix(0);
560 if (DEBUG_L_TEST || debug_initialization) {
561 PerlIO_printf(Perl_debug_log,
562 "LC_NUMERIC locale now is standard C\n");
566 #endif /* USE_LOCALE_NUMERIC */
571 Perl_set_numeric_underlying(pTHX)
574 #ifdef USE_LOCALE_NUMERIC
576 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
577 * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h
578 * instead of calling this directly. The macro avoids calling this routine
579 * if toggling isn't necessary according to our records (which could be
580 * wrong if some XS code has changed the locale behind our back) */
582 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
583 PL_numeric_standard = PL_numeric_underlying_is_standard;
584 PL_numeric_underlying = TRUE;
585 set_numeric_radix(! PL_numeric_standard);
589 if (DEBUG_L_TEST || debug_initialization) {
590 PerlIO_printf(Perl_debug_log,
591 "LC_NUMERIC locale now is %s\n",
596 #endif /* USE_LOCALE_NUMERIC */
601 * Set up for a new ctype locale.
604 S_new_ctype(pTHX_ const char *newctype)
607 #ifndef USE_LOCALE_CTYPE
609 PERL_ARGS_ASSERT_NEW_CTYPE;
610 PERL_UNUSED_ARG(newctype);
615 /* Called after each libc setlocale() call affecting LC_CTYPE, to tell
616 * core Perl this and that 'newctype' is the name of the new locale.
618 * This function sets up the folding arrays for all 256 bytes, assuming
619 * that tofold() is tolc() since fold case is not a concept in POSIX,
621 * Any code changing the locale (outside this file) should use
622 * POSIX::setlocale, which calls this function. Therefore this function
623 * should be called directly only from this file and from
624 * POSIX::setlocale() */
629 PERL_ARGS_ASSERT_NEW_CTYPE;
631 /* We will replace any bad locale warning with 1) nothing if the new one is
632 * ok; or 2) a new warning for the bad new locale */
633 if (PL_warn_locale) {
634 SvREFCNT_dec_NN(PL_warn_locale);
635 PL_warn_locale = NULL;
638 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
640 /* A UTF-8 locale gets standard rules. But note that code still has to
641 * handle this specially because of the three problematic code points */
642 if (PL_in_utf8_CTYPE_locale) {
643 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
646 /* Assume enough space for every character being bad. 4 spaces each
647 * for the 94 printable characters that are output like "'x' "; and 5
648 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
650 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
652 /* Don't check for problems if we are suppressing the warnings */
653 bool check_for_problems = ckWARN_d(WARN_LOCALE)
654 || UNLIKELY(DEBUG_L_TEST);
655 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
657 unsigned int bad_count = 0; /* Count of bad characters */
659 for (i = 0; i < 256; i++) {
661 PL_fold_locale[i] = (U8) tolower(i);
663 PL_fold_locale[i] = (U8) toupper(i);
665 PL_fold_locale[i] = (U8) i;
667 /* If checking for locale problems, see if the native ASCII-range
668 * printables plus \n and \t are in their expected categories in
669 * the new locale. If not, this could mean big trouble, upending
670 * Perl's and most programs' assumptions, like having a
671 * metacharacter with special meaning become a \w. Fortunately,
672 * it's very rare to find locales that aren't supersets of ASCII
673 * nowadays. It isn't a problem for most controls to be changed
674 * into something else; we check only \n and \t, though perhaps \r
675 * could be an issue as well. */
676 if ( check_for_problems
677 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
679 if ( cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC(i))
680 || cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i))
681 || cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i))
682 || cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i))
683 || cBOOL(islower(i)) != cBOOL(isLOWER_A(i))
684 || cBOOL(isprint(i)) != cBOOL(isPRINT_A(i))
685 || cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i))
686 || cBOOL(isspace(i)) != cBOOL(isSPACE_A(i))
687 || cBOOL(isupper(i)) != cBOOL(isUPPER_A(i))
688 || cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i))
689 || tolower(i) != (int) toLOWER_A(i)
690 || toupper(i) != (int) toUPPER_A(i)
691 || (i == '\n' && ! isCNTRL_LC(i)))
693 if (bad_count) { /* Separate multiple entries with a
695 bad_chars_list[bad_count++] = ' ';
697 bad_chars_list[bad_count++] = '\'';
699 bad_chars_list[bad_count++] = (char) i;
702 bad_chars_list[bad_count++] = '\\';
704 bad_chars_list[bad_count++] = 'n';
708 bad_chars_list[bad_count++] = 't';
711 bad_chars_list[bad_count++] = '\'';
712 bad_chars_list[bad_count] = '\0';
719 /* We only handle single-byte locales (outside of UTF-8 ones; so if
720 * this locale requires more than one byte, there are going to be
722 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
723 "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n",
724 __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX));
726 if (check_for_problems && MB_CUR_MAX > 1
728 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
729 * locale. Just assume that the implementation for them (plus
730 * for POSIX) is correct and the > 1 value is spurious. (Since
731 * these are specially handled to never be considered UTF-8
732 * locales, as long as this is the only problem, everything
733 * should work fine */
734 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
736 multi_byte_locale = TRUE;
741 if (bad_count || multi_byte_locale) {
742 PL_warn_locale = Perl_newSVpvf(aTHX_
743 "Locale '%s' may not work well.%s%s%s\n",
746 ? " Some characters in it are not recognized by"
750 ? "\nThe following characters (and maybe others)"
751 " may not have the same meaning as the Perl"
752 " program expects:\n"
758 /* If we are actually in the scope of the locale or are debugging,
759 * output the message now. If not in that scope, we save the
760 * message to be output at the first operation using this locale,
761 * if that actually happens. Most programs don't use locales, so
762 * they are immune to bad ones. */
763 if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) {
765 /* We have to save 'newctype' because the setlocale() just
766 * below may destroy it. The next setlocale() further down
767 * should restore it properly so that the intermediate change
768 * here is transparent to this function's caller */
769 const char * const badlocale = savepv(newctype);
771 do_setlocale_c(LC_CTYPE, "C");
773 /* The '0' below suppresses a bogus gcc compiler warning */
774 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
776 do_setlocale_c(LC_CTYPE, badlocale);
779 if (IN_LC(LC_CTYPE)) {
780 SvREFCNT_dec_NN(PL_warn_locale);
781 PL_warn_locale = NULL;
787 #endif /* USE_LOCALE_CTYPE */
792 Perl__warn_problematic_locale()
795 #ifdef USE_LOCALE_CTYPE
799 /* Internal-to-core function that outputs the message in PL_warn_locale,
800 * and then NULLS it. Should be called only through the macro
801 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
803 if (PL_warn_locale) {
804 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
805 SvPVX(PL_warn_locale),
806 0 /* dummy to avoid compiler warning */ );
807 SvREFCNT_dec_NN(PL_warn_locale);
808 PL_warn_locale = NULL;
816 S_new_collate(pTHX_ const char *newcoll)
819 #ifndef USE_LOCALE_COLLATE
821 PERL_UNUSED_ARG(newcoll);
826 /* Called after each libc setlocale() call affecting LC_COLLATE, to tell
827 * core Perl this and that 'newcoll' is the name of the new locale.
829 * The design of locale collation is that every locale change is given an
830 * index 'PL_collation_ix'. The first time a string particpates in an
831 * operation that requires collation while locale collation is active, it
832 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
833 * magic includes the collation index, and the transformation of the string
834 * by strxfrm(), q.v. That transformation is used when doing comparisons,
835 * instead of the string itself. If a string changes, the magic is
836 * cleared. The next time the locale changes, the index is incremented,
837 * and so we know during a comparison that the transformation is not
838 * necessarily still valid, and so is recomputed. Note that if the locale
839 * changes enough times, the index could wrap (a U32), and it is possible
840 * that a transformation would improperly be considered valid, leading to
844 if (PL_collation_name) {
846 Safefree(PL_collation_name);
847 PL_collation_name = NULL;
849 PL_collation_standard = TRUE;
850 is_standard_collation:
851 PL_collxfrm_base = 0;
852 PL_collxfrm_mult = 2;
853 PL_in_utf8_COLLATE_locale = FALSE;
854 PL_strxfrm_NUL_replacement = '\0';
855 PL_strxfrm_max_cp = 0;
859 /* If this is not the same locale as currently, set the new one up */
860 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
862 Safefree(PL_collation_name);
863 PL_collation_name = stdize_locale(savepv(newcoll));
864 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
865 if (PL_collation_standard) {
866 goto is_standard_collation;
869 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
870 PL_strxfrm_NUL_replacement = '\0';
871 PL_strxfrm_max_cp = 0;
873 /* A locale collation definition includes primary, secondary, tertiary,
874 * etc. weights for each character. To sort, the primary weights are
875 * used, and only if they compare equal, then the secondary weights are
876 * used, and only if they compare equal, then the tertiary, etc.
878 * strxfrm() works by taking the input string, say ABC, and creating an
879 * output transformed string consisting of first the primary weights,
880 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
881 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
882 * may not have weights at every level. In our example, let's say B
883 * doesn't have a tertiary weight, and A doesn't have a secondary
884 * weight. The constructed string is then going to be
885 * A¹B¹C¹ B²C² A³C³ ....
886 * This has the desired effect that strcmp() will look at the secondary
887 * or tertiary weights only if the strings compare equal at all higher
888 * priority weights. The spaces shown here, like in
890 * are not just for readability. In the general case, these must
891 * actually be bytes, which we will call here 'separator weights'; and
892 * they must be smaller than any other weight value, but since these
893 * are C strings, only the terminating one can be a NUL (some
894 * implementations may include a non-NUL separator weight just before
895 * the NUL). Implementations tend to reserve 01 for the separator
896 * weights. They are needed so that a shorter string's secondary
897 * weights won't be misconstrued as primary weights of a longer string,
898 * etc. By making them smaller than any other weight, the shorter
899 * string will sort first. (Actually, if all secondary weights are
900 * smaller than all primary ones, there is no need for a separator
901 * weight between those two levels, etc.)
903 * The length of the transformed string is roughly a linear function of
904 * the input string. It's not exactly linear because some characters
905 * don't have weights at all levels. When we call strxfrm() we have to
906 * allocate some memory to hold the transformed string. The
907 * calculations below try to find coefficients 'm' and 'b' for this
908 * locale so that m*x + b equals how much space we need, given the size
909 * of the input string in 'x'. If we calculate too small, we increase
910 * the size as needed, and call strxfrm() again, but it is better to
911 * get it right the first time to avoid wasted expensive string
912 * transformations. */
915 /* We use the string below to find how long the tranformation of it
916 * is. Almost all locales are supersets of ASCII, or at least the
917 * ASCII letters. We use all of them, half upper half lower,
918 * because if we used fewer, we might hit just the ones that are
919 * outliers in a particular locale. Most of the strings being
920 * collated will contain a preponderance of letters, and even if
921 * they are above-ASCII, they are likely to have the same number of
922 * weight levels as the ASCII ones. It turns out that digits tend
923 * to have fewer levels, and some punctuation has more, but those
924 * are relatively sparse in text, and khw believes this gives a
925 * reasonable result, but it could be changed if experience so
927 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
928 char * x_longer; /* Transformed 'longer' */
929 Size_t x_len_longer; /* Length of 'x_longer' */
931 char * x_shorter; /* We also transform a substring of 'longer' */
932 Size_t x_len_shorter;
934 /* _mem_collxfrm() is used get the transformation (though here we
935 * are interested only in its length). It is used because it has
936 * the intelligence to handle all cases, but to work, it needs some
937 * values of 'm' and 'b' to get it started. For the purposes of
938 * this calculation we use a very conservative estimate of 'm' and
939 * 'b'. This assumes a weight can be multiple bytes, enough to
940 * hold any UV on the platform, and there are 5 levels, 4 weight
941 * bytes, and a trailing NUL. */
942 PL_collxfrm_base = 5;
943 PL_collxfrm_mult = 5 * sizeof(UV);
945 /* Find out how long the transformation really is */
946 x_longer = _mem_collxfrm(longer,
950 /* We avoid converting to UTF-8 in the
951 * called function by telling it the
952 * string is in UTF-8 if the locale is a
953 * UTF-8 one. Since the string passed
954 * here is invariant under UTF-8, we can
955 * claim it's UTF-8 even though it isn't.
957 PL_in_utf8_COLLATE_locale);
960 /* Find out how long the transformation of a substring of 'longer'
961 * is. Together the lengths of these transformations are
962 * sufficient to calculate 'm' and 'b'. The substring is all of
963 * 'longer' except the first character. This minimizes the chances
964 * of being swayed by outliers */
965 x_shorter = _mem_collxfrm(longer + 1,
968 PL_in_utf8_COLLATE_locale);
971 /* If the results are nonsensical for this simple test, the whole
972 * locale definition is suspect. Mark it so that locale collation
973 * is not active at all for it. XXX Should we warn? */
974 if ( x_len_shorter == 0
976 || x_len_shorter >= x_len_longer)
978 PL_collxfrm_mult = 0;
979 PL_collxfrm_base = 0;
982 SSize_t base; /* Temporary */
984 /* We have both: m * strlen(longer) + b = x_len_longer
985 * m * strlen(shorter) + b = x_len_shorter;
986 * subtracting yields:
987 * m * (strlen(longer) - strlen(shorter))
988 * = x_len_longer - x_len_shorter
989 * But we have set things up so that 'shorter' is 1 byte smaller
990 * than 'longer'. Hence:
991 * m = x_len_longer - x_len_shorter
993 * But if something went wrong, make sure the multiplier is at
996 if (x_len_longer > x_len_shorter) {
997 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
1000 PL_collxfrm_mult = 1;
1005 * but in case something has gone wrong, make sure it is
1007 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
1012 /* Add 1 for the trailing NUL */
1013 PL_collxfrm_base = base + 1;
1018 if (DEBUG_L_TEST || debug_initialization) {
1019 PerlIO_printf(Perl_debug_log,
1020 "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
1022 " collate multipler=%zu, collate base=%zu\n",
1024 PL_in_utf8_COLLATE_locale,
1025 x_len_shorter, x_len_longer,
1026 PL_collxfrm_mult, PL_collxfrm_base);
1033 #endif /* USE_LOCALE_COLLATE */
1040 S_win32_setlocale(pTHX_ int category, const char* locale)
1042 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
1043 * difference between the two unless the input locale is "", which normally
1044 * means on Windows to get the machine default, which is set via the
1045 * computer's "Regional and Language Options" (or its current equivalent).
1046 * In POSIX, it instead means to find the locale from the user's
1047 * environment. This routine changes the Windows behavior to first look in
1048 * the environment, and, if anything is found, use that instead of going to
1049 * the machine default. If there is no environment override, the machine
1050 * default is used, by calling the real setlocale() with "".
1052 * The POSIX behavior is to use the LC_ALL variable if set; otherwise to
1053 * use the particular category's variable if set; otherwise to use the LANG
1056 bool override_LC_ALL = FALSE;
1060 if (locale && strEQ(locale, "")) {
1064 locale = PerlEnv_getenv("LC_ALL");
1066 if (category == LC_ALL) {
1067 override_LC_ALL = TRUE;
1073 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
1074 if (category == categories[i]) {
1075 locale = PerlEnv_getenv(category_names[i]);
1080 locale = PerlEnv_getenv("LANG");
1096 result = setlocale(category, locale);
1097 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
1098 setlocale_debug_string(category, locale, result)));
1100 if (! override_LC_ALL) {
1104 /* Here the input category was LC_ALL, and we have set it to what is in the
1105 * LANG variable or the system default if there is no LANG. But these have
1106 * lower priority than the other LC_foo variables, so override it for each
1107 * one that is set. (If they are set to "", it means to use the same thing
1108 * we just set LC_ALL to, so can skip) */
1110 for (i = 0; i < LC_ALL_INDEX; i++) {
1111 result = PerlEnv_getenv(category_names[i]);
1112 if (result && strNE(result, "")) {
1113 setlocale(categories[i], result);
1114 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1116 setlocale_debug_string(categories[i], result, "not captured")));
1120 result = setlocale(LC_ALL, NULL);
1121 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1123 setlocale_debug_string(LC_ALL, NULL, result)));
1131 Perl_setlocale(int category, const char * locale)
1133 /* This wraps POSIX::setlocale() */
1139 #ifdef USE_LOCALE_NUMERIC
1141 /* A NULL locale means only query what the current one is. We have the
1142 * LC_NUMERIC name saved, because we are normally switched into the C
1143 * locale for it. For an LC_ALL query, switch back to get the correct
1144 * results. All other categories don't require special handling */
1145 if (locale == NULL) {
1146 if (category == LC_NUMERIC) {
1147 return savepv(PL_numeric_name);
1152 else if (category == LC_ALL && ! PL_numeric_underlying) {
1154 SET_NUMERIC_UNDERLYING();
1163 /* Save retval since subsequent setlocale() calls may overwrite it. */
1164 retval = savepv(do_setlocale_r(category, locale));
1166 DEBUG_L(PerlIO_printf(Perl_debug_log,
1167 "%s:%d: %s\n", __FILE__, __LINE__,
1168 setlocale_debug_string(category, locale, retval)));
1170 /* Should never happen that a query would return an error, but be
1171 * sure and reset to C locale */
1173 SET_NUMERIC_STANDARD();
1179 /* If locale == NULL, we are just querying the state, but may have switched
1180 * to NUMERIC_UNDERLYING. Switch back before returning. */
1181 if (locale == NULL) {
1182 SET_NUMERIC_STANDARD();
1186 /* Now that have switched locales, we have to update our records to
1191 #ifdef USE_LOCALE_CTYPE
1198 #ifdef USE_LOCALE_COLLATE
1201 new_collate(retval);
1205 #ifdef USE_LOCALE_NUMERIC
1208 new_numeric(retval);
1216 /* LC_ALL updates all the things we care about. The values may not
1217 * be the same as 'retval', as the locale "" may have set things
1220 # ifdef USE_LOCALE_CTYPE
1222 newlocale = do_setlocale_c(LC_CTYPE, NULL);
1223 new_ctype(newlocale);
1225 # endif /* USE_LOCALE_CTYPE */
1226 # ifdef USE_LOCALE_COLLATE
1228 newlocale = do_setlocale_c(LC_COLLATE, NULL);
1229 new_collate(newlocale);
1232 # ifdef USE_LOCALE_NUMERIC
1234 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
1235 new_numeric(newlocale);
1237 # endif /* USE_LOCALE_NUMERIC */
1249 PERL_STATIC_INLINE const char *
1250 S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset)
1252 /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size',
1253 * growing it if necessary */
1255 const Size_t string_size = strlen(string) + offset + 1;
1257 PERL_ARGS_ASSERT_SAVE_TO_BUFFER;
1259 if (*buf_size == 0) {
1260 Newx(*buf, string_size, char);
1261 *buf_size = string_size;
1263 else if (string_size > *buf_size) {
1264 Renew(*buf, string_size, char);
1265 *buf_size = string_size;
1268 Copy(string, *buf + offset, string_size - offset, char);
1274 =head1 Locale-related functions and macros
1276 =for apidoc Perl_langinfo
1278 This is an (almost ª) drop-in replacement for the system C<L<nl_langinfo(3)>>,
1279 taking the same C<item> parameter values, and returning the same information.
1280 But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks
1281 of Perl's locale handling from your code, and can be used on systems that lack
1282 a native C<nl_langinfo>.
1290 It delivers the correct results for the C<RADIXCHAR> and C<THOUSESEP> items,
1291 without you having to write extra code. The reason for the extra code would be
1292 because these are from the C<LC_NUMERIC> locale category, which is normally
1293 kept set to the C locale by Perl, no matter what the underlying locale is
1294 supposed to be, and so to get the expected results, you have to temporarily
1295 toggle into the underlying locale, and later toggle back. (You could use
1296 plain C<nl_langinfo> and C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this
1297 but then you wouldn't get the other advantages of C<Perl_langinfo()>; not
1298 keeping C<LC_NUMERIC> in the C locale would break a lot of CPAN, which is
1299 expecting the radix (decimal point) character to be a dot.)
1303 Depending on C<item>, it works on systems that don't have C<nl_langinfo>, hence
1304 makes your code more portable. Of the fifty-some possible items specified by
1305 the POSIX 2008 standard,
1306 L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>,
1307 only two are completely unimplemented. It uses various techniques to recover
1308 the other items, including calling C<L<localeconv(3)>>, and C<L<strftime(3)>>,
1309 both of which are specified in C89, so should be always be available. Later
1310 C<strftime()> versions have additional capabilities; C<""> is returned for
1311 those not available on your system.
1313 The details for those items which may differ from what this emulation returns
1314 and what a native C<nl_langinfo()> would return are:
1322 Unimplemented, so returns C<"">.
1332 Only the values for English are returned. C<YESSTR> and C<NOSTR> have been
1333 removed from POSIX 2008, and are retained for backwards compatibility. Your
1334 platform's C<nl_langinfo> may not support them.
1338 Always evaluates to C<%x>, the locale's appropriate date representation.
1342 Always evaluates to C<%X>, the locale's appropriate time representation.
1346 Always evaluates to C<%c>, the locale's appropriate date and time
1351 The return may be incorrect for those rare locales where the currency symbol
1352 replaces the radix character.
1353 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1354 to work differently.
1358 Currently this gives the same results as Linux does.
1359 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1360 to work differently.
1366 =item C<ERA_D_T_FMT>
1370 These are derived by using C<strftime()>, and not all versions of that function
1371 know about them. C<""> is returned for these on such systems.
1375 When using C<Perl_langinfo> on systems that don't have a native
1376 C<nl_langinfo()>, you must
1378 #include "perl_langinfo.h"
1380 before the C<perl.h> C<#include>. You can replace your C<langinfo.h>
1381 C<#include> with this one. (Doing it this way keeps out the symbols that plain
1382 C<langinfo.h> imports into the namespace for code that doesn't need it.)
1384 You also should not use the bare C<langinfo.h> item names, but should preface
1385 them with C<PERL_>, so use C<PERL_RADIXCHAR> instead of plain C<RADIXCHAR>.
1386 The C<PERL_I<foo>> versions will also work for this function on systems that do
1387 have a native C<nl_langinfo>.
1391 It is thread-friendly, returning its result in a buffer that won't be
1392 overwritten by another thread, so you don't have to code for that possibility.
1393 The buffer can be overwritten by the next call to C<nl_langinfo> or
1394 C<Perl_langinfo> in the same thread.
1398 ª It returns S<C<const char *>>, whereas plain C<nl_langinfo()> returns S<C<char
1399 *>>, but you are (only by documentation) forbidden to write into the buffer.
1400 By declaring this C<const>, the compiler enforces this restriction. The extra
1401 C<const> is why this isn't an unequivocal drop-in replacement for
1406 The original impetus for C<Perl_langinfo()> was so that code that needs to
1407 find out the current currency symbol, floating point radix character, or digit
1408 grouping separator can use, on all systems, the simpler and more
1409 thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a
1410 pain to make thread-friendly. For other fields returned by C<localeconv>, it
1411 is better to use the methods given in L<perlcall> to call
1412 L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly.
1419 #ifdef HAS_NL_LANGINFO
1420 Perl_langinfo(const nl_item item)
1422 Perl_langinfo(const int item)
1425 return my_nl_langinfo(item, TRUE);
1429 #ifdef HAS_NL_LANGINFO
1430 S_my_nl_langinfo(const nl_item item, bool toggle)
1432 S_my_nl_langinfo(const int item, bool toggle)
1437 #if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */
1438 #if ! defined(HAS_POSIX_2008_LOCALE)
1440 /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC
1441 * for those items dependent on it. This must be copied to a buffer before
1442 * switching back, as some systems destroy the buffer when setlocale() is
1448 if ( ! PL_numeric_underlying
1449 && (item == PERL_RADIXCHAR || item == PERL_THOUSEP))
1451 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
1458 save_to_buffer(nl_langinfo(item), &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
1461 do_setlocale_c(LC_NUMERIC, "C");
1466 # else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */
1468 bool do_free = FALSE;
1469 locale_t cur = uselocale((locale_t) 0);
1471 if (cur == LC_GLOBAL_LOCALE) {
1472 cur = duplocale(LC_GLOBAL_LOCALE);
1477 cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur);
1481 save_to_buffer(nl_langinfo_l(item, cur),
1482 &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
1489 if (strEQ(PL_langinfo_buf, "")) {
1490 if (item == PERL_YESSTR) {
1493 if (item == PERL_NOSTR) {
1498 return PL_langinfo_buf;
1500 #else /* Below, emulate nl_langinfo as best we can */
1504 # ifdef HAS_LOCALECONV
1506 const struct lconv* lc;
1509 # ifdef HAS_STRFTIME
1512 bool return_format = FALSE; /* Return the %format, not the value */
1513 const char * format;
1517 /* We copy the results to a per-thread buffer, even if not
1518 * multi-threaded. This is in part to simplify this code, and partly
1519 * because we need a buffer anyway for strftime(), and partly because a
1520 * call of localeconv() could otherwise wipe out the buffer, and the
1521 * programmer would not be expecting this, as this is a nl_langinfo()
1522 * substitute after all, so s/he might be thinking their localeconv()
1523 * is safe until another localeconv() call. */
1527 const char * retval;
1529 /* These 2 are unimplemented */
1531 case PERL_ERA: /* For use with strftime() %E modifier */
1536 /* We use only an English set, since we don't know any more */
1537 case PERL_YESEXPR: return "^[+1yY]";
1538 case PERL_YESSTR: return "yes";
1539 case PERL_NOEXPR: return "^[-0nN]";
1540 case PERL_NOSTR: return "no";
1542 # ifdef HAS_LOCALECONV
1548 /* We don't bother with localeconv_l() because any system that
1549 * has it is likely to also have nl_langinfo() */
1553 || ! lc->currency_symbol
1554 || strEQ("", lc->currency_symbol))
1560 /* Leave the first spot empty to be filled in below */
1561 save_to_buffer(lc->currency_symbol, &PL_langinfo_buf,
1562 &PL_langinfo_bufsize, 1);
1563 if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, ""))
1564 { /* khw couldn't figure out how the localedef specifications
1565 would show that the $ should replace the radix; this is
1566 just a guess as to how it might work.*/
1567 *PL_langinfo_buf = '.';
1569 else if (lc->p_cs_precedes) {
1570 *PL_langinfo_buf = '-';
1573 *PL_langinfo_buf = '+';
1579 case PERL_RADIXCHAR:
1585 if (! PL_numeric_underlying) {
1586 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
1598 retval = (item == PERL_RADIXCHAR)
1600 : lc->thousands_sep;
1606 save_to_buffer(retval, &PL_langinfo_buf,
1607 &PL_langinfo_bufsize, 0);
1610 do_setlocale_c(LC_NUMERIC, "C");
1618 # ifdef HAS_STRFTIME
1620 /* These are defined by C89, so we assume that strftime supports
1621 * them, and so are returned unconditionally; they may not be what
1622 * the locale actually says, but should give good enough results
1623 * for someone using them as formats (as opposed to trying to parse
1624 * them to figure out what the locale says). The other format
1625 * items are actually tested to verify they work on the platform */
1626 case PERL_D_FMT: return "%x";
1627 case PERL_T_FMT: return "%X";
1628 case PERL_D_T_FMT: return "%c";
1630 /* These formats are only available in later strfmtime's */
1631 case PERL_ERA_D_FMT: case PERL_ERA_T_FMT: case PERL_ERA_D_T_FMT:
1632 case PERL_T_FMT_AMPM:
1634 /* The rest can be gotten from most versions of strftime(). */
1635 case PERL_ABDAY_1: case PERL_ABDAY_2: case PERL_ABDAY_3:
1636 case PERL_ABDAY_4: case PERL_ABDAY_5: case PERL_ABDAY_6:
1638 case PERL_ALT_DIGITS:
1639 case PERL_AM_STR: case PERL_PM_STR:
1640 case PERL_ABMON_1: case PERL_ABMON_2: case PERL_ABMON_3:
1641 case PERL_ABMON_4: case PERL_ABMON_5: case PERL_ABMON_6:
1642 case PERL_ABMON_7: case PERL_ABMON_8: case PERL_ABMON_9:
1643 case PERL_ABMON_10: case PERL_ABMON_11: case PERL_ABMON_12:
1644 case PERL_DAY_1: case PERL_DAY_2: case PERL_DAY_3: case PERL_DAY_4:
1645 case PERL_DAY_5: case PERL_DAY_6: case PERL_DAY_7:
1646 case PERL_MON_1: case PERL_MON_2: case PERL_MON_3: case PERL_MON_4:
1647 case PERL_MON_5: case PERL_MON_6: case PERL_MON_7: case PERL_MON_8:
1648 case PERL_MON_9: case PERL_MON_10: case PERL_MON_11:
1653 init_tm(&tm); /* Precaution against core dumps */
1657 tm.tm_year = 2017 - 1900;
1664 "panic: %s: %d: switch case: %d problem",
1665 __FILE__, __LINE__, item);
1666 NOT_REACHED; /* NOTREACHED */
1668 case PERL_PM_STR: tm.tm_hour = 18;
1673 case PERL_ABDAY_7: tm.tm_wday++;
1674 case PERL_ABDAY_6: tm.tm_wday++;
1675 case PERL_ABDAY_5: tm.tm_wday++;
1676 case PERL_ABDAY_4: tm.tm_wday++;
1677 case PERL_ABDAY_3: tm.tm_wday++;
1678 case PERL_ABDAY_2: tm.tm_wday++;
1683 case PERL_DAY_7: tm.tm_wday++;
1684 case PERL_DAY_6: tm.tm_wday++;
1685 case PERL_DAY_5: tm.tm_wday++;
1686 case PERL_DAY_4: tm.tm_wday++;
1687 case PERL_DAY_3: tm.tm_wday++;
1688 case PERL_DAY_2: tm.tm_wday++;
1693 case PERL_ABMON_12: tm.tm_mon++;
1694 case PERL_ABMON_11: tm.tm_mon++;
1695 case PERL_ABMON_10: tm.tm_mon++;
1696 case PERL_ABMON_9: tm.tm_mon++;
1697 case PERL_ABMON_8: tm.tm_mon++;
1698 case PERL_ABMON_7: tm.tm_mon++;
1699 case PERL_ABMON_6: tm.tm_mon++;
1700 case PERL_ABMON_5: tm.tm_mon++;
1701 case PERL_ABMON_4: tm.tm_mon++;
1702 case PERL_ABMON_3: tm.tm_mon++;
1703 case PERL_ABMON_2: tm.tm_mon++;
1708 case PERL_MON_12: tm.tm_mon++;
1709 case PERL_MON_11: tm.tm_mon++;
1710 case PERL_MON_10: tm.tm_mon++;
1711 case PERL_MON_9: tm.tm_mon++;
1712 case PERL_MON_8: tm.tm_mon++;
1713 case PERL_MON_7: tm.tm_mon++;
1714 case PERL_MON_6: tm.tm_mon++;
1715 case PERL_MON_5: tm.tm_mon++;
1716 case PERL_MON_4: tm.tm_mon++;
1717 case PERL_MON_3: tm.tm_mon++;
1718 case PERL_MON_2: tm.tm_mon++;
1723 case PERL_T_FMT_AMPM:
1725 return_format = TRUE;
1728 case PERL_ERA_D_FMT:
1730 return_format = TRUE;
1733 case PERL_ERA_T_FMT:
1735 return_format = TRUE;
1738 case PERL_ERA_D_T_FMT:
1740 return_format = TRUE;
1743 case PERL_ALT_DIGITS:
1745 format = "%Ow"; /* Find the alternate digit for 0 */
1749 /* We can't use my_strftime() because it doesn't look at
1751 while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize,
1754 /* A zero return means one of:
1755 * a) there wasn't enough space in PL_langinfo_buf
1756 * b) the format, like a plain %p, returns empty
1757 * c) it was an illegal format, though some
1758 * implementations of strftime will just return the
1759 * illegal format as a plain character sequence.
1761 * To quickly test for case 'b)', try again but precede
1762 * the format with a plain character. If that result is
1763 * still empty, the problem is either 'a)' or 'c)' */
1765 Size_t format_size = strlen(format) + 1;
1766 Size_t mod_size = format_size + 1;
1770 Newx(mod_format, mod_size, char);
1771 Newx(temp_result, PL_langinfo_bufsize, char);
1773 my_strlcpy(mod_format + 1, format, mod_size);
1774 len = strftime(temp_result,
1775 PL_langinfo_bufsize,
1777 Safefree(mod_format);
1778 Safefree(temp_result);
1780 /* If 'len' is non-zero, it means that we had a case like
1781 * %p which means the current locale doesn't use a.m. or
1782 * p.m., and that is valid */
1785 /* Here, still didn't work. If we get well beyond a
1786 * reasonable size, bail out to prevent an infinite
1789 if (PL_langinfo_bufsize > 100 * format_size) {
1790 *PL_langinfo_buf = '\0';
1793 /* Double the buffer size to retry; Add 1 in case
1794 * original was 0, so we aren't stuck at 0. */
1795 PL_langinfo_bufsize *= 2;
1796 PL_langinfo_bufsize++;
1797 Renew(PL_langinfo_buf, PL_langinfo_bufsize, char);
1805 /* Here, we got a result.
1807 * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the
1808 * alternate format for wday 0. If the value is the same as
1809 * the normal 0, there isn't an alternate, so clear the buffer.
1811 if ( item == PERL_ALT_DIGITS
1812 && strEQ(PL_langinfo_buf, "0"))
1814 *PL_langinfo_buf = '\0';
1817 /* ALT_DIGITS is problematic. Experiments on it showed that
1818 * strftime() did not always work properly when going from
1819 * alt-9 to alt-10. Only a few locales have this item defined,
1820 * and in all of them on Linux that khw was able to find,
1821 * nl_langinfo() merely returned the alt-0 character, possibly
1822 * doubled. Most Unicode digits are in blocks of 10
1823 * consecutive code points, so that is sufficient information
1824 * for those scripts, as we can infer alt-1, alt-2, .... But
1825 * for a Japanese locale, a CJK ideographic 0 is returned, and
1826 * the CJK digits are not in code point order, so you can't
1827 * really infer anything. The localedef for this locale did
1828 * specify the succeeding digits, so that strftime() works
1829 * properly on them, without needing to infer anything. But
1830 * the nl_langinfo() return did not give sufficient information
1831 * for the caller to understand what's going on. So until
1832 * there is evidence that it should work differently, this
1833 * returns the alt-0 string for ALT_DIGITS.
1835 * wday was chosen because its range is all a single digit.
1836 * Things like tm_sec have two digits as the minimum: '00' */
1840 /* If to return the format, not the value, overwrite the buffer
1841 * with it. But some strftime()s will keep the original format
1842 * if illegal, so change those to "" */
1843 if (return_format) {
1844 if (strEQ(PL_langinfo_buf, format)) {
1845 *PL_langinfo_buf = '\0';
1848 save_to_buffer(format, &PL_langinfo_buf,
1849 &PL_langinfo_bufsize, 0);
1860 return PL_langinfo_buf;
1867 * Initialize locale awareness.
1870 Perl_init_i18nl10n(pTHX_ int printwarn)
1874 * 0 if not to output warning when setup locale is bad
1875 * 1 if to output warning based on value of PERL_BADLANG
1876 * >1 if to output regardless of PERL_BADLANG
1879 * 1 = set ok or not applicable,
1880 * 0 = fallback to a locale of lower priority
1881 * -1 = fallback to all locales failed, not even to the C locale
1883 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
1884 * set, debugging information is output.
1886 * This looks more complicated than it is, mainly due to the #ifdefs.
1888 * We try to set LC_ALL to the value determined by the environment. If
1889 * there is no LC_ALL on this platform, we try the individual categories we
1890 * know about. If this works, we are done.
1892 * But if it doesn't work, we have to do something else. We search the
1893 * environment variables ourselves instead of relying on the system to do
1894 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
1895 * think there is one), and the ultimate fallback "C". This is all done in
1896 * the same loop as above to avoid duplicating code, but it makes things
1897 * more complex. The 'trial_locales' array is initialized with just one
1898 * element; it causes the behavior described in the paragraph above this to
1899 * happen. If that fails, we add elements to 'trial_locales', and do extra
1900 * loop iterations to cause the behavior described in this paragraph.
1902 * On Ultrix, the locale MUST come from the environment, so there is
1903 * preliminary code to set it. I (khw) am not sure that it is necessary,
1904 * and that this couldn't be folded into the loop, but barring any real
1905 * platforms to test on, it's staying as-is
1907 * A slight complication is that in embedded Perls, the locale may already
1908 * be set-up, and we don't want to get it from the normal environment
1909 * variables. This is handled by having a special environment variable
1910 * indicate we're in this situation. We simply set setlocale's 2nd
1911 * parameter to be a NULL instead of "". That indicates to setlocale that
1912 * it is not to change anything, but to return the current value,
1913 * effectively initializing perl's db to what the locale already is.
1915 * We play the same trick with NULL if a LC_ALL succeeds. We call
1916 * setlocale() on the individual categores with NULL to get their existing
1917 * values for our db, instead of trying to change them.
1924 PERL_UNUSED_ARG(printwarn);
1926 #else /* USE_LOCALE */
1929 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
1933 /* NULL uses the existing already set up locale */
1934 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
1937 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
1938 unsigned int trial_locales_count;
1939 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
1940 const char * const lang = savepv(PerlEnv_getenv("LANG"));
1941 bool setlocale_failure = FALSE;
1944 /* A later getenv() could zap this, so only use here */
1945 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
1947 const bool locwarn = (printwarn > 1
1949 && ( ! bad_lang_use_once
1951 /* disallow with "" or "0" */
1953 && strNE("0", bad_lang_use_once)))));
1955 /* setlocale() return vals; not copied so must be looked at immediately */
1956 const char * sl_result[NOMINAL_LC_ALL_INDEX + 1];
1958 /* current locale for given category; should have been copied so aren't
1960 const char * curlocales[NOMINAL_LC_ALL_INDEX + 1];
1964 /* In some systems you can find out the system default locale
1965 * and use that as the fallback locale. */
1966 # define SYSTEM_DEFAULT_LOCALE
1968 # ifdef SYSTEM_DEFAULT_LOCALE
1970 const char *system_default_locale = NULL;
1975 # define DEBUG_LOCALE_INIT(a,b,c)
1978 DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT")));
1980 # define DEBUG_LOCALE_INIT(category, locale, result) \
1982 if (debug_initialization) { \
1983 PerlIO_printf(Perl_debug_log, \
1985 __FILE__, __LINE__, \
1986 setlocale_debug_string(category, \
1992 /* Make sure the parallel arrays are properly set up */
1993 # ifdef USE_LOCALE_NUMERIC
1994 assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC);
1995 assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC"));
1997 # ifdef USE_LOCALE_CTYPE
1998 assert(categories[LC_CTYPE_INDEX] == LC_CTYPE);
1999 assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE"));
2001 # ifdef USE_LOCALE_COLLATE
2002 assert(categories[LC_COLLATE_INDEX] == LC_COLLATE);
2003 assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE"));
2005 # ifdef USE_LOCALE_TIME
2006 assert(categories[LC_TIME_INDEX] == LC_TIME);
2007 assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME"));
2009 # ifdef USE_LOCALE_MESSAGES
2010 assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES);
2011 assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES"));
2013 # ifdef USE_LOCALE_MONETARY
2014 assert(categories[LC_MONETARY_INDEX] == LC_MONETARY);
2015 assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY"));
2017 # ifdef USE_LOCALE_ADDRESS
2018 assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS);
2019 assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS"));
2021 # ifdef USE_LOCALE_IDENTIFICATION
2022 assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION);
2023 assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION"));
2025 # ifdef USE_LOCALE_MEASUREMENT
2026 assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT);
2027 assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT"));
2029 # ifdef USE_LOCALE_PAPER
2030 assert(categories[LC_PAPER_INDEX] == LC_PAPER);
2031 assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER"));
2033 # ifdef USE_LOCALE_TELEPHONE
2034 assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE);
2035 assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE"));
2038 assert(categories[LC_ALL_INDEX] == LC_ALL);
2039 assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL"));
2040 assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX);
2042 # endif /* DEBUGGING */
2044 /* Initialize the cache of the program's UTF-8ness for the always known
2045 * locales C and POSIX */
2046 my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness,
2047 sizeof(PL_locale_utf8ness));
2049 # ifdef LOCALE_ENVIRON_REQUIRED
2052 * Ultrix setlocale(..., "") fails if there are no environment
2053 * variables from which to get a locale name.
2057 # error Ultrix without LC_ALL not implemented
2063 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init);
2064 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]);
2065 if (sl_result[LC_ALL_INDEX])
2068 setlocale_failure = TRUE;
2070 if (! setlocale_failure) {
2071 const char * locale_param;
2072 for (i = 0; i < LC_ALL_INDEX; i++) {
2073 locale_param = (! done && (lang || PerlEnv_getenv(category_names[i])))
2076 sl_result[i] = do_setlocale_r(categories[i], locale_param);
2077 if (! sl_result[i]) {
2078 setlocale_failure = TRUE;
2080 DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]);
2085 # endif /* LC_ALL */
2086 # endif /* LOCALE_ENVIRON_REQUIRED */
2088 /* We try each locale in the list until we get one that works, or exhaust
2089 * the list. Normally the loop is executed just once. But if setting the
2090 * locale fails, inside the loop we add fallback trials to the array and so
2091 * will execute the loop multiple times */
2092 trial_locales[0] = setlocale_init;
2093 trial_locales_count = 1;
2095 for (i= 0; i < trial_locales_count; i++) {
2096 const char * trial_locale = trial_locales[i];
2100 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
2101 * when i==0, but I (khw) don't think that behavior makes much
2103 setlocale_failure = FALSE;
2105 # ifdef SYSTEM_DEFAULT_LOCALE
2106 # ifdef WIN32 /* Note that assumes Win32 has LC_ALL */
2108 /* On Windows machines, an entry of "" after the 0th means to use
2109 * the system default locale, which we now proceed to get. */
2110 if (strEQ(trial_locale, "")) {
2113 /* Note that this may change the locale, but we are going to do
2114 * that anyway just below */
2115 system_default_locale = do_setlocale_c(LC_ALL, "");
2116 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
2118 /* Skip if invalid or if it's already on the list of locales to
2120 if (! system_default_locale) {
2121 goto next_iteration;
2123 for (j = 0; j < trial_locales_count; j++) {
2124 if (strEQ(system_default_locale, trial_locales[j])) {
2125 goto next_iteration;
2129 trial_locale = system_default_locale;
2132 # error SYSTEM_DEFAULT_LOCALE only implemented for Win32
2134 # endif /* SYSTEM_DEFAULT_LOCALE */
2140 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale);
2141 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]);
2142 if (! sl_result[LC_ALL_INDEX]) {
2143 setlocale_failure = TRUE;
2146 /* Since LC_ALL succeeded, it should have changed all the other
2147 * categories it can to its value; so we massage things so that the
2148 * setlocales below just return their category's current values.
2149 * This adequately handles the case in NetBSD where LC_COLLATE may
2150 * not be defined for a locale, and setting it individually will
2151 * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to
2152 * the POSIX locale. */
2153 trial_locale = NULL;
2156 # endif /* LC_ALL */
2158 if (! setlocale_failure) {
2160 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2162 = savepv(do_setlocale_r(categories[j], trial_locale));
2163 if (! curlocales[j]) {
2164 setlocale_failure = TRUE;
2166 DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]);
2169 if (! setlocale_failure) { /* All succeeded */
2170 break; /* Exit trial_locales loop */
2174 /* Here, something failed; will need to try a fallback. */
2180 if (locwarn) { /* Output failure info only on the first one */
2184 PerlIO_printf(Perl_error_log,
2185 "perl: warning: Setting locale failed.\n");
2187 # else /* !LC_ALL */
2189 PerlIO_printf(Perl_error_log,
2190 "perl: warning: Setting locale failed for the categories:\n\t");
2192 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2193 if (! curlocales[j]) {
2194 PerlIO_printf(Perl_error_log, category_names[j]);
2197 Safefree(curlocales[j]);
2201 # endif /* LC_ALL */
2203 PerlIO_printf(Perl_error_log,
2204 "perl: warning: Please check that your locale settings:\n");
2208 PerlIO_printf(Perl_error_log,
2209 "\tLANGUAGE = %c%s%c,\n",
2210 language ? '"' : '(',
2211 language ? language : "unset",
2212 language ? '"' : ')');
2215 PerlIO_printf(Perl_error_log,
2216 "\tLC_ALL = %c%s%c,\n",
2218 lc_all ? lc_all : "unset",
2219 lc_all ? '"' : ')');
2221 # if defined(USE_ENVIRON_ARRAY)
2226 /* Look through the environment for any variables of the
2227 * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was
2228 * already handled above. These are assumed to be locale
2229 * settings. Output them and their values. */
2230 for (e = environ; *e; e++) {
2231 const STRLEN prefix_len = sizeof("LC_") - 1;
2234 if ( strBEGINs(*e, "LC_")
2235 && ! strBEGINs(*e, "LC_ALL=")
2236 && (uppers_len = strspn(*e + prefix_len,
2237 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"))
2238 && ((*e)[prefix_len + uppers_len] == '='))
2240 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
2241 (int) (prefix_len + uppers_len), *e,
2242 *e + prefix_len + uppers_len + 1);
2249 PerlIO_printf(Perl_error_log,
2250 "\t(possibly more locale environment variables)\n");
2254 PerlIO_printf(Perl_error_log,
2255 "\tLANG = %c%s%c\n",
2257 lang ? lang : "unset",
2260 PerlIO_printf(Perl_error_log,
2261 " are supported and installed on your system.\n");
2264 /* Calculate what fallback locales to try. We have avoided this
2265 * until we have to, because failure is quite unlikely. This will
2266 * usually change the upper bound of the loop we are in.
2268 * Since the system's default way of setting the locale has not
2269 * found one that works, We use Perl's defined ordering: LC_ALL,
2270 * LANG, and the C locale. We don't try the same locale twice, so
2271 * don't add to the list if already there. (On POSIX systems, the
2272 * LC_ALL element will likely be a repeat of the 0th element "",
2273 * but there's no harm done by doing it explicitly.
2275 * Note that this tries the LC_ALL environment variable even on
2276 * systems which have no LC_ALL locale setting. This may or may
2277 * not have been originally intentional, but there's no real need
2278 * to change the behavior. */
2280 for (j = 0; j < trial_locales_count; j++) {
2281 if (strEQ(lc_all, trial_locales[j])) {
2285 trial_locales[trial_locales_count++] = lc_all;
2290 for (j = 0; j < trial_locales_count; j++) {
2291 if (strEQ(lang, trial_locales[j])) {
2295 trial_locales[trial_locales_count++] = lang;
2299 # if defined(WIN32) && defined(LC_ALL)
2301 /* For Windows, we also try the system default locale before "C".
2302 * (If there exists a Windows without LC_ALL we skip this because
2303 * it gets too complicated. For those, the "C" is the next
2304 * fallback possibility). The "" is the same as the 0th element of
2305 * the array, but the code at the loop above knows to treat it
2306 * differently when not the 0th */
2307 trial_locales[trial_locales_count++] = "";
2311 for (j = 0; j < trial_locales_count; j++) {
2312 if (strEQ("C", trial_locales[j])) {
2316 trial_locales[trial_locales_count++] = "C";
2319 } /* end of first time through the loop */
2327 } /* end of looping through the trial locales */
2329 if (ok < 1) { /* If we tried to fallback */
2331 if (! setlocale_failure) { /* fallback succeeded */
2332 msg = "Falling back to";
2334 else { /* fallback failed */
2337 /* We dropped off the end of the loop, so have to decrement i to
2338 * get back to the value the last time through */
2342 msg = "Failed to fall back to";
2344 /* To continue, we should use whatever values we've got */
2346 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2347 Safefree(curlocales[j]);
2348 curlocales[j] = savepv(do_setlocale_r(categories[j], NULL));
2349 DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]);
2354 const char * description;
2355 const char * name = "";
2356 if (strEQ(trial_locales[i], "C")) {
2357 description = "the standard locale";
2361 # ifdef SYSTEM_DEFAULT_LOCALE
2363 else if (strEQ(trial_locales[i], "")) {
2364 description = "the system default locale";
2365 if (system_default_locale) {
2366 name = system_default_locale;
2370 # endif /* SYSTEM_DEFAULT_LOCALE */
2373 description = "a fallback locale";
2374 name = trial_locales[i];
2376 if (name && strNE(name, "")) {
2377 PerlIO_printf(Perl_error_log,
2378 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
2381 PerlIO_printf(Perl_error_log,
2382 "perl: warning: %s %s.\n", msg, description);
2385 } /* End of tried to fallback */
2387 /* Done with finding the locales; update our records */
2389 # ifdef USE_LOCALE_CTYPE
2391 new_ctype(curlocales[LC_CTYPE_INDEX]);
2394 # ifdef USE_LOCALE_COLLATE
2396 new_collate(curlocales[LC_COLLATE_INDEX]);
2399 # ifdef USE_LOCALE_NUMERIC
2401 new_numeric(curlocales[LC_NUMERIC_INDEX]);
2405 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
2407 # if defined(USE_ITHREADS)
2409 /* This caches whether each category's locale is UTF-8 or not. This
2410 * may involve changing the locale. It is ok to do this at
2411 * initialization time before any threads have started, but not later.
2412 * Caching means that if the program heeds our dictate not to change
2413 * locales in threaded applications, this data will remain valid, and
2414 * it may get queried without changing locales. If the environment is
2415 * such that all categories have the same locale, this isn't needed, as
2416 * the code will not change the locale; but this handles the uncommon
2417 * case where the environment has disparate locales for the categories
2419 (void) _is_cur_LC_category_utf8(categories[i]);
2423 Safefree(curlocales[i]);
2426 # if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
2428 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
2429 * locale is UTF-8. The call to new_ctype() just above has already
2430 * calculated the latter value and saved it in PL_in_utf8_CTYPE_locale. If
2431 * both PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE})
2432 * are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer on
2433 * STDIN, STDOUT, STDERR, _and_ the default open discipline. */
2434 PL_utf8locale = PL_in_utf8_CTYPE_locale;
2436 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
2437 This is an alternative to using the -C command line switch
2438 (the -C if present will override this). */
2440 const char *p = PerlEnv_getenv("PERL_UNICODE");
2441 PL_unicode = p ? parse_unicode_opts(&p) : 0;
2442 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
2456 #endif /* USE_LOCALE */
2459 /* So won't continue to output stuff */
2460 DEBUG_INITIALIZATION_set(FALSE);
2467 #ifdef USE_LOCALE_COLLATE
2470 Perl__mem_collxfrm(pTHX_ const char *input_string,
2471 STRLEN len, /* Length of 'input_string' */
2472 STRLEN *xlen, /* Set to length of returned string
2473 (not including the collation index
2475 bool utf8 /* Is the input in UTF-8? */
2479 /* _mem_collxfrm() is a bit like strxfrm() but with two important
2480 * differences. First, it handles embedded NULs. Second, it allocates a bit
2481 * more memory than needed for the transformed data itself. The real
2482 * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
2483 * the length of that, and doesn't include the collation index size.
2484 * Please see sv_collxfrm() to see how this is used. */
2486 #define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
2488 char * s = (char *) input_string;
2489 STRLEN s_strlen = strlen(input_string);
2491 STRLEN xAlloc; /* xalloc is a reserved word in VC */
2492 STRLEN length_in_chars;
2493 bool first_time = TRUE; /* Cleared after first loop iteration */
2495 PERL_ARGS_ASSERT__MEM_COLLXFRM;
2497 /* Must be NUL-terminated */
2498 assert(*(input_string + len) == '\0');
2500 /* If this locale has defective collation, skip */
2501 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
2502 DEBUG_L(PerlIO_printf(Perl_debug_log,
2503 "_mem_collxfrm: locale's collation is defective\n"));
2507 /* Replace any embedded NULs with the control that sorts before any others.
2508 * This will give as good as possible results on strings that don't
2509 * otherwise contain that character, but otherwise there may be
2510 * less-than-perfect results with that character and NUL. This is
2511 * unavoidable unless we replace strxfrm with our own implementation. */
2512 if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded
2516 STRLEN sans_nuls_len;
2517 int try_non_controls;
2518 char this_replacement_char[] = "?\0"; /* Room for a two-byte string,
2519 making sure 2nd byte is NUL.
2521 STRLEN this_replacement_len;
2523 /* If we don't know what non-NUL control character sorts lowest for
2524 * this locale, find it */
2525 if (PL_strxfrm_NUL_replacement == '\0') {
2527 char * cur_min_x = NULL; /* The min_char's xfrm, (except it also
2528 includes the collation index
2531 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
2533 /* Unlikely, but it may be that no control will work to replace
2534 * NUL, in which case we instead look for any character. Controls
2535 * are preferred because collation order is, in general, context
2536 * sensitive, with adjoining characters affecting the order, and
2537 * controls are less likely to have such interactions, allowing the
2538 * NUL-replacement to stand on its own. (Another way to look at it
2539 * is to imagine what would happen if the NUL were replaced by a
2540 * combining character; it wouldn't work out all that well.) */
2541 for (try_non_controls = 0;
2542 try_non_controls < 2;
2545 /* Look through all legal code points (NUL isn't) */
2546 for (j = 1; j < 256; j++) {
2547 char * x; /* j's xfrm plus collation index */
2548 STRLEN x_len; /* length of 'x' */
2549 STRLEN trial_len = 1;
2550 char cur_source[] = { '\0', '\0' };
2552 /* Skip non-controls the first time through the loop. The
2553 * controls in a UTF-8 locale are the L1 ones */
2554 if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
2561 /* Create a 1-char string of the current code point */
2562 cur_source[0] = (char) j;
2564 /* Then transform it */
2565 x = _mem_collxfrm(cur_source, trial_len, &x_len,
2566 0 /* The string is not in UTF-8 */);
2568 /* Ignore any character that didn't successfully transform.
2574 /* If this character's transformation is lower than
2575 * the current lowest, this one becomes the lowest */
2576 if ( cur_min_x == NULL
2577 || strLT(x + COLLXFRM_HDR_LEN,
2578 cur_min_x + COLLXFRM_HDR_LEN))
2580 PL_strxfrm_NUL_replacement = j;
2586 } /* end of loop through all 255 characters */
2588 /* Stop looking if found */
2593 /* Unlikely, but possible, if there aren't any controls that
2594 * work in the locale, repeat the loop, looking for any
2595 * character that works */
2596 DEBUG_L(PerlIO_printf(Perl_debug_log,
2597 "_mem_collxfrm: No control worked. Trying non-controls\n"));
2598 } /* End of loop to try first the controls, then any char */
2601 DEBUG_L(PerlIO_printf(Perl_debug_log,
2602 "_mem_collxfrm: Couldn't find any character to replace"
2603 " embedded NULs in locale %s with", PL_collation_name));
2607 DEBUG_L(PerlIO_printf(Perl_debug_log,
2608 "_mem_collxfrm: Replacing embedded NULs in locale %s with "
2609 "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement));
2611 Safefree(cur_min_x);
2612 } /* End of determining the character that is to replace NULs */
2614 /* If the replacement is variant under UTF-8, it must match the
2615 * UTF8-ness of the original */
2616 if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) {
2617 this_replacement_char[0] =
2618 UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement);
2619 this_replacement_char[1] =
2620 UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement);
2621 this_replacement_len = 2;
2624 this_replacement_char[0] = PL_strxfrm_NUL_replacement;
2625 /* this_replacement_char[1] = '\0' was done at initialization */
2626 this_replacement_len = 1;
2629 /* The worst case length for the replaced string would be if every
2630 * character in it is NUL. Multiply that by the length of each
2631 * replacement, and allow for a trailing NUL */
2632 sans_nuls_len = (len * this_replacement_len) + 1;
2633 Newx(sans_nuls, sans_nuls_len, char);
2636 /* Replace each NUL with the lowest collating control. Loop until have
2637 * exhausted all the NULs */
2638 while (s + s_strlen < e) {
2639 my_strlcat(sans_nuls, s, sans_nuls_len);
2641 /* Do the actual replacement */
2642 my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len);
2644 /* Move past the input NUL */
2646 s_strlen = strlen(s);
2649 /* And add anything that trails the final NUL */
2650 my_strlcat(sans_nuls, s, sans_nuls_len);
2652 /* Switch so below we transform this modified string */
2655 } /* End of replacing NULs */
2657 /* Make sure the UTF8ness of the string and locale match */
2658 if (utf8 != PL_in_utf8_COLLATE_locale) {
2659 const char * const t = s; /* Temporary so we can later find where the
2662 /* Here they don't match. Change the string's to be what the locale is
2665 if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
2666 s = (char *) bytes_to_utf8((const U8 *) s, &len);
2669 else { /* locale is not UTF-8; but input is; downgrade the input */
2671 s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
2673 /* If the downgrade was successful we are done, but if the input
2674 * contains things that require UTF-8 to represent, have to do
2675 * damage control ... */
2676 if (UNLIKELY(utf8)) {
2678 /* What we do is construct a non-UTF-8 string with
2679 * 1) the characters representable by a single byte converted
2680 * to be so (if necessary);
2681 * 2) and the rest converted to collate the same as the
2682 * highest collating representable character. That makes
2683 * them collate at the end. This is similar to how we
2684 * handle embedded NULs, but we use the highest collating
2685 * code point instead of the smallest. Like the NUL case,
2686 * this isn't perfect, but is the best we can reasonably
2687 * do. Every above-255 code point will sort the same as
2688 * the highest-sorting 0-255 code point. If that code
2689 * point can combine in a sequence with some other code
2690 * points for weight calculations, us changing something to
2691 * be it can adversely affect the results. But in most
2692 * cases, it should work reasonably. And note that this is
2693 * really an illegal situation: using code points above 255
2694 * on a locale where only 0-255 are valid. If two strings
2695 * sort entirely equal, then the sort order for the
2696 * above-255 code points will be in code point order. */
2700 /* If we haven't calculated the code point with the maximum
2701 * collating order for this locale, do so now */
2702 if (! PL_strxfrm_max_cp) {
2705 /* The current transformed string that collates the
2706 * highest (except it also includes the prefixed collation
2708 char * cur_max_x = NULL;
2710 /* Look through all legal code points (NUL isn't) */
2711 for (j = 1; j < 256; j++) {
2714 char cur_source[] = { '\0', '\0' };
2716 /* Create a 1-char string of the current code point */
2717 cur_source[0] = (char) j;
2719 /* Then transform it */
2720 x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
2722 /* If something went wrong (which it shouldn't), just
2723 * ignore this code point */
2728 /* If this character's transformation is higher than
2729 * the current highest, this one becomes the highest */
2730 if ( cur_max_x == NULL
2731 || strGT(x + COLLXFRM_HDR_LEN,
2732 cur_max_x + COLLXFRM_HDR_LEN))
2734 PL_strxfrm_max_cp = j;
2743 DEBUG_L(PerlIO_printf(Perl_debug_log,
2744 "_mem_collxfrm: Couldn't find any character to"
2745 " replace above-Latin1 chars in locale %s with",
2746 PL_collation_name));
2750 DEBUG_L(PerlIO_printf(Perl_debug_log,
2751 "_mem_collxfrm: highest 1-byte collating character"
2752 " in locale %s is 0x%02X\n",
2754 PL_strxfrm_max_cp));
2756 Safefree(cur_max_x);
2759 /* Here we know which legal code point collates the highest.
2760 * We are ready to construct the non-UTF-8 string. The length
2761 * will be at least 1 byte smaller than the input string
2762 * (because we changed at least one 2-byte character into a
2763 * single byte), but that is eaten up by the trailing NUL */
2769 char * e = (char *) t + len;
2771 for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
2773 if (UTF8_IS_INVARIANT(cur_char)) {
2776 else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) {
2777 s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
2779 else { /* Replace illegal cp with highest collating
2781 s[d++] = PL_strxfrm_max_cp;
2785 Renew(s, d, char); /* Free up unused space */
2790 /* Here, we have constructed a modified version of the input. It could
2791 * be that we already had a modified copy before we did this version.
2792 * If so, that copy is no longer needed */
2793 if (t != input_string) {
2798 length_in_chars = (utf8)
2799 ? utf8_length((U8 *) s, (U8 *) s + len)
2802 /* The first element in the output is the collation id, used by
2803 * sv_collxfrm(); then comes the space for the transformed string. The
2804 * equation should give us a good estimate as to how much is needed */
2805 xAlloc = COLLXFRM_HDR_LEN
2807 + (PL_collxfrm_mult * length_in_chars);
2808 Newx(xbuf, xAlloc, char);
2809 if (UNLIKELY(! xbuf)) {
2810 DEBUG_L(PerlIO_printf(Perl_debug_log,
2811 "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
2815 /* Store the collation id */
2816 *(U32*)xbuf = PL_collation_ix;
2818 /* Then the transformation of the input. We loop until successful, or we
2822 *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
2824 /* If the transformed string occupies less space than we told strxfrm()
2825 * was available, it means it successfully transformed the whole
2827 if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
2829 /* Some systems include a trailing NUL in the returned length.
2830 * Ignore it, using a loop in case multiple trailing NULs are
2833 && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
2838 /* If the first try didn't get it, it means our prediction was low.
2839 * Modify the coefficients so that we predict a larger value in any
2840 * future transformations */
2842 STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
2843 STRLEN computed_guess = PL_collxfrm_base
2844 + (PL_collxfrm_mult * length_in_chars);
2846 /* On zero-length input, just keep current slope instead of
2848 const STRLEN new_m = (length_in_chars != 0)
2849 ? needed / length_in_chars
2852 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2853 "%s: %d: initial size of %zu bytes for a length "
2854 "%zu string was insufficient, %zu needed\n",
2856 computed_guess, length_in_chars, needed));
2858 /* If slope increased, use it, but discard this result for
2859 * length 1 strings, as we can't be sure that it's a real slope
2861 if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
2865 STRLEN old_m = PL_collxfrm_mult;
2866 STRLEN old_b = PL_collxfrm_base;
2870 PL_collxfrm_mult = new_m;
2871 PL_collxfrm_base = 1; /* +1 For trailing NUL */
2872 computed_guess = PL_collxfrm_base
2873 + (PL_collxfrm_mult * length_in_chars);
2874 if (computed_guess < needed) {
2875 PL_collxfrm_base += needed - computed_guess;
2878 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2879 "%s: %d: slope is now %zu; was %zu, base "
2880 "is now %zu; was %zu\n",
2882 PL_collxfrm_mult, old_m,
2883 PL_collxfrm_base, old_b));
2885 else { /* Slope didn't change, but 'b' did */
2886 const STRLEN new_b = needed
2889 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2890 "%s: %d: base is now %zu; was %zu\n",
2892 new_b, PL_collxfrm_base));
2893 PL_collxfrm_base = new_b;
2900 if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
2901 DEBUG_L(PerlIO_printf(Perl_debug_log,
2902 "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
2903 *xlen, PERL_INT_MAX));
2907 /* A well-behaved strxfrm() returns exactly how much space it needs
2908 * (usually not including the trailing NUL) when it fails due to not
2909 * enough space being provided. Assume that this is the case unless
2910 * it's been proven otherwise */
2911 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
2912 xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
2914 else { /* Here, either:
2915 * 1) The strxfrm() has previously shown bad behavior; or
2916 * 2) It isn't the first time through the loop, which means
2917 * that the strxfrm() is now showing bad behavior, because
2918 * we gave it what it said was needed in the previous
2919 * iteration, and it came back saying it needed still more.
2920 * (Many versions of cygwin fit this. When the buffer size
2921 * isn't sufficient, they return the input size instead of
2922 * how much is needed.)
2923 * Increase the buffer size by a fixed percentage and try again.
2925 xAlloc += (xAlloc / 4) + 1;
2926 PL_strxfrm_is_behaved = FALSE;
2930 if (DEBUG_Lv_TEST || debug_initialization) {
2931 PerlIO_printf(Perl_debug_log,
2932 "_mem_collxfrm required more space than previously calculated"
2933 " for locale %s, trying again with new guess=%d+%zu\n",
2934 PL_collation_name, (int) COLLXFRM_HDR_LEN,
2935 xAlloc - COLLXFRM_HDR_LEN);
2942 Renew(xbuf, xAlloc, char);
2943 if (UNLIKELY(! xbuf)) {
2944 DEBUG_L(PerlIO_printf(Perl_debug_log,
2945 "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
2955 if (DEBUG_Lv_TEST || debug_initialization) {
2957 print_collxfrm_input_and_return(s, s + len, xlen, utf8);
2958 PerlIO_printf(Perl_debug_log, "Its xfrm is:");
2959 PerlIO_printf(Perl_debug_log, "%s\n",
2960 _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN,
2966 /* Free up unneeded space; retain ehough for trailing NUL */
2967 Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
2969 if (s != input_string) {
2977 if (s != input_string) {
2984 if (DEBUG_Lv_TEST || debug_initialization) {
2985 print_collxfrm_input_and_return(s, s + len, NULL, utf8);
2996 S_print_collxfrm_input_and_return(pTHX_
2997 const char * const s,
2998 const char * const e,
2999 const STRLEN * const xlen,
3003 PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
3005 PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ",
3006 (UV)PL_collation_ix);
3008 PerlIO_printf(Perl_debug_log, "%zu", *xlen);
3011 PerlIO_printf(Perl_debug_log, "NULL");
3013 PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
3015 print_bytes_for_locale(s, e, is_utf8);
3017 PerlIO_printf(Perl_debug_log, "'\n");
3021 S_print_bytes_for_locale(pTHX_
3022 const char * const s,
3023 const char * const e,
3027 bool prev_was_printable = TRUE;
3028 bool first_time = TRUE;
3030 PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE;
3034 ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
3037 if (! prev_was_printable) {
3038 PerlIO_printf(Perl_debug_log, " ");
3040 PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
3041 prev_was_printable = TRUE;
3045 PerlIO_printf(Perl_debug_log, " ");
3047 PerlIO_printf(Perl_debug_log, "%02" UVXf, cp);
3048 prev_was_printable = FALSE;
3050 t += (is_utf8) ? UTF8SKIP(t) : 1;
3055 # endif /* #ifdef DEBUGGING */
3056 #endif /* USE_LOCALE_COLLATE */
3061 Perl__is_cur_LC_category_utf8(pTHX_ int category)
3063 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
3064 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
3065 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
3066 * could give the wrong result. The result will very likely be correct for
3067 * languages that have commonly used non-ASCII characters, but for notably
3068 * English, it comes down to if the locale's name ends in something like
3069 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
3071 /* Name of current locale corresponding to the input category */
3072 const char *save_input_locale = NULL;
3074 bool is_utf8 = FALSE; /* The return value */
3077 /* The variables below are for the cache of previous lookups using this
3078 * function. The cache is a C string, described at the definition for
3079 * 'C_and_POSIX_utf8ness'.
3081 * The first part of the cache is fixed, for the C and POSIX locales. The
3082 * varying part starts just after them. */
3083 char * utf8ness_cache = PL_locale_utf8ness + STRLENs(C_and_POSIX_utf8ness);
3085 Size_t utf8ness_cache_size; /* Size of the varying portion */
3086 Size_t input_name_len; /* Length in bytes of save_input_locale */
3087 Size_t input_name_len_with_overhead; /* plus extra chars used to store
3088 the name in the cache */
3089 char * delimited; /* The name plus the delimiters used to store
3091 char * name_pos; /* position of 'delimited' in the cache, or 0
3097 assert(category != LC_ALL);
3101 /* Get the desired category's locale */
3102 save_input_locale = do_setlocale_r(category, NULL);
3103 if (! save_input_locale) {
3105 "panic: %s: %d: Could not find current locale for %s\n",
3106 __FILE__, __LINE__, category_name(category));
3109 save_input_locale = stdize_locale(savepv(save_input_locale));
3110 DEBUG_L(PerlIO_printf(Perl_debug_log,
3111 "Current locale for %s is %s\n",
3112 category_name(category), save_input_locale));
3114 input_name_len = strlen(save_input_locale);
3116 /* In our cache, each name is accompanied by two delimiters and a single
3118 input_name_len_with_overhead = input_name_len + 3;
3120 /* Allocate and populate space for a copy of the name surrounded by the
3122 Newx(delimited, input_name_len_with_overhead, char);
3123 delimited[0] = UTF8NESS_SEP[0];
3124 Copy(save_input_locale, delimited + 1, input_name_len, char);
3125 delimited[input_name_len+1] = UTF8NESS_PREFIX[0];
3126 delimited[input_name_len+2] = '\0';
3128 /* And see if that is in the cache */
3129 name_pos = instr(PL_locale_utf8ness, delimited);
3131 is_utf8 = *(name_pos + input_name_len_with_overhead - 1) - '0';
3135 if (DEBUG_Lv_TEST || debug_initialization) {
3136 PerlIO_printf(Perl_debug_log, "UTF8ness for locale %s=%d, \n",
3137 save_input_locale, is_utf8);
3142 /* And, if not already in that position, move it to the beginning of
3143 * the non-constant portion of the list, since it is the most recently
3144 * used. (We don't have to worry about overflow, since just moving
3145 * existing names around) */
3146 if (name_pos > utf8ness_cache) {
3147 Move(utf8ness_cache,
3148 utf8ness_cache + input_name_len_with_overhead,
3149 name_pos - utf8ness_cache, char);
3152 input_name_len_with_overhead - 1, char);
3153 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3156 Safefree(delimited);
3157 Safefree(save_input_locale);
3161 /* Here we don't have stored the utf8ness for the input locale. We have to
3164 # if defined(USE_LOCALE_CTYPE) \
3165 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
3167 { /* Next try nl_langinfo or MB_CUR_MAX if available */
3169 char *save_ctype_locale = NULL;
3171 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
3173 /* Get the current LC_CTYPE locale */
3174 save_ctype_locale = do_setlocale_c(LC_CTYPE, NULL);
3175 if (! save_ctype_locale) {
3176 DEBUG_L(PerlIO_printf(Perl_debug_log,
3177 "Could not find current locale for LC_CTYPE\n"));
3178 goto cant_use_nllanginfo;
3180 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
3182 /* If LC_CTYPE and the desired category use the same locale, this
3183 * means that finding the value for LC_CTYPE is the same as finding
3184 * the value for the desired category. Otherwise, switch LC_CTYPE
3185 * to the desired category's locale */
3186 if (strEQ(save_ctype_locale, save_input_locale)) {
3187 Safefree(save_ctype_locale);
3188 save_ctype_locale = NULL;
3190 else if (! do_setlocale_c(LC_CTYPE, save_input_locale)) {
3191 DEBUG_L(PerlIO_printf(Perl_debug_log,
3192 "Could not change LC_CTYPE locale to %s\n",
3193 save_input_locale));
3194 Safefree(save_ctype_locale);
3195 goto cant_use_nllanginfo;
3199 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
3200 save_input_locale));
3202 /* Here the current LC_CTYPE is set to the locale of the category whose
3203 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
3204 * should give the correct results */
3206 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
3208 { /* The task is easiest if the platform has this POSIX 2001 function */
3209 const char *codeset = my_nl_langinfo(PERL_CODESET, FALSE);
3210 /* FALSE => already in dest locale */
3212 DEBUG_L(PerlIO_printf(Perl_debug_log,
3213 "\tnllanginfo returned CODESET '%s'\n", codeset));
3215 if (codeset && strNE(codeset, "")) {
3216 /* If we switched LC_CTYPE, switch back */
3217 if (save_ctype_locale) {
3218 do_setlocale_c(LC_CTYPE, save_ctype_locale);
3219 Safefree(save_ctype_locale);
3222 /* If the implementation of foldEQ() somehow were
3223 * to change to not go byte-by-byte, this could
3224 * read past end of string, as only one length is
3225 * checked. But currently, a premature NUL will
3226 * compare false, and it will stop there */
3227 is_utf8 = cBOOL( foldEQ(codeset, STR_WITH_LEN("UTF-8"))
3228 || foldEQ(codeset, STR_WITH_LEN("UTF8")));
3230 DEBUG_L(PerlIO_printf(Perl_debug_log,
3231 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
3233 goto finish_and_return;
3240 /* Here, either we don't have nl_langinfo, or it didn't return a
3241 * codeset. Try MB_CUR_MAX */
3243 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
3244 * Unicode code point. Since UTF-8 is the only non-single byte
3245 * encoding we handle, we just say any such encoding is UTF-8, and if
3246 * turns out to be wrong, other things will fail */
3247 is_utf8 = (unsigned) MB_CUR_MAX >= STRLENs(MAX_UNICODE_UTF8);
3249 DEBUG_L(PerlIO_printf(Perl_debug_log,
3250 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
3251 (int) MB_CUR_MAX, is_utf8));
3253 Safefree(save_input_locale);
3257 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
3258 * since they are both in the C99 standard. We can feed a known byte
3259 * string to the latter function, and check that it gives the expected
3265 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
3267 len = mbtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8));
3270 if ( len != STRLENs(REPLACEMENT_CHARACTER_UTF8)
3271 || wc != (wchar_t) UNICODE_REPLACEMENT)
3274 DEBUG_L(PerlIO_printf(Perl_debug_log, "\replacement=U+%x\n",
3276 DEBUG_L(PerlIO_printf(Perl_debug_log,
3277 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
3284 /* If we switched LC_CTYPE, switch back */
3285 if (save_ctype_locale) {
3286 do_setlocale_c(LC_CTYPE, save_ctype_locale);
3287 Safefree(save_ctype_locale);
3290 goto finish_and_return;
3296 cant_use_nllanginfo:
3298 # else /* nl_langinfo should work if available, so don't bother compiling this
3299 fallback code. The final fallback of looking at the name is
3300 compiled, and will be executed if nl_langinfo fails */
3302 /* nl_langinfo not available or failed somehow. Next try looking at the
3303 * currency symbol to see if it disambiguates things. Often that will be
3304 * in the native script, and if the symbol isn't in UTF-8, we know that the
3305 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
3306 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
3309 # ifdef HAS_LOCALECONV
3310 # ifdef USE_LOCALE_MONETARY
3313 char *save_monetary_locale = NULL;
3314 bool only_ascii = FALSE;
3317 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
3318 * the desired category, if it isn't that locale already */
3320 if (category != LC_MONETARY) {
3322 save_monetary_locale = do_setlocale_c(LC_MONETARY, NULL);
3323 if (! save_monetary_locale) {
3324 DEBUG_L(PerlIO_printf(Perl_debug_log,
3325 "Could not find current locale for LC_MONETARY\n"));
3326 goto cant_use_monetary;
3328 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
3330 if (strEQ(save_monetary_locale, save_input_locale)) {
3331 Safefree(save_monetary_locale);
3332 save_monetary_locale = NULL;
3334 else if (! do_setlocale_c(LC_MONETARY, save_input_locale)) {
3335 DEBUG_L(PerlIO_printf(Perl_debug_log,
3336 "Could not change LC_MONETARY locale to %s\n",
3337 save_input_locale));
3338 Safefree(save_monetary_locale);
3339 goto cant_use_monetary;
3343 /* Here the current LC_MONETARY is set to the locale of the category
3344 * whose information is desired. */
3348 || ! lc->currency_symbol
3349 || is_utf8_invariant_string((U8 *) lc->currency_symbol, 0))
3351 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));
3355 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
3358 /* If we changed it, restore LC_MONETARY to its original locale */
3359 if (save_monetary_locale) {
3360 do_setlocale_c(LC_MONETARY, save_monetary_locale);
3361 Safefree(save_monetary_locale);
3366 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
3367 * otherwise assume the locale is UTF-8 if and only if the symbol
3368 * is non-ascii UTF-8. */
3369 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
3370 save_input_locale, is_utf8));
3371 goto finish_and_return;
3376 # endif /* USE_LOCALE_MONETARY */
3377 # endif /* HAS_LOCALECONV */
3379 # if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
3381 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
3382 * the names of the months and weekdays, timezone, and am/pm indicator */
3384 char *save_time_locale = NULL;
3386 bool is_dst = FALSE;
3390 char * formatted_time;
3393 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
3394 * desired category, if it isn't that locale already */
3396 if (category != LC_TIME) {
3398 save_time_locale = do_setlocale_c(LC_TIME, NULL);
3399 if (! save_time_locale) {
3400 DEBUG_L(PerlIO_printf(Perl_debug_log,
3401 "Could not find current locale for LC_TIME\n"));
3404 save_time_locale = stdize_locale(savepv(save_time_locale));
3406 if (strEQ(save_time_locale, save_input_locale)) {
3407 Safefree(save_time_locale);
3408 save_time_locale = NULL;
3410 else if (! do_setlocale_c(LC_TIME, save_input_locale)) {
3411 DEBUG_L(PerlIO_printf(Perl_debug_log,
3412 "Could not change LC_TIME locale to %s\n",
3413 save_input_locale));
3414 Safefree(save_time_locale);
3419 /* Here the current LC_TIME is set to the locale of the category
3420 * whose information is desired. Look at all the days of the week and
3421 * month names, and the timezone and am/pm indicator for UTF-8 variant
3422 * characters. The first such a one found will tell us if the locale
3423 * is UTF-8 or not */
3425 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
3426 formatted_time = my_strftime("%A %B %Z %p",
3427 0, 0, hour, dom, month, 2012 - 1900, 0, 0, is_dst);
3428 if ( ! formatted_time
3429 || is_utf8_invariant_string((U8 *) formatted_time, 0))
3432 /* Here, we didn't find a non-ASCII. Try the next time through
3433 * with the complemented dst and am/pm, and try with the next
3434 * weekday. After we have gotten all weekdays, try the next
3437 hour = (hour + 12) % 24;
3445 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
3446 * false otherwise. But first, restore LC_TIME to its original
3447 * locale if we changed it */
3448 if (save_time_locale) {
3449 do_setlocale_c(LC_TIME, save_time_locale);
3450 Safefree(save_time_locale);
3453 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
3455 is_utf8_string((U8 *) formatted_time, 0)));
3456 is_utf8 = is_utf8_string((U8 *) formatted_time, 0);
3457 goto finish_and_return;
3460 /* Falling off the end of the loop indicates all the names were just
3461 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
3462 * to its original locale */
3463 if (save_time_locale) {
3464 do_setlocale_c(LC_TIME, save_time_locale);
3465 Safefree(save_time_locale);
3467 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));
3473 # if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
3475 /* This code is ifdefd out because it was found to not be necessary in testing
3476 * on our dromedary test machine, which has over 700 locales. There, this
3477 * added no value to looking at the currency symbol and the time strings. I
3478 * left it in so as to avoid rewriting it if real-world experience indicates
3479 * that dromedary is an outlier. Essentially, instead of returning abpve if we
3480 * haven't found illegal utf8, we continue on and examine all the strerror()
3481 * messages on the platform for utf8ness. If all are ASCII, we still don't
3482 * know the answer; but otherwise we have a pretty good indication of the
3483 * utf8ness. The reason this doesn't help much is that the messages may not
3484 * have been translated into the locale. The currency symbol and time strings
3485 * are much more likely to have been translated. */
3488 bool non_ascii = FALSE;
3489 char *save_messages_locale = NULL;
3490 const char * errmsg = NULL;
3492 /* Like above, we set LC_MESSAGES to the locale of the desired
3493 * category, if it isn't that locale already */
3495 if (category != LC_MESSAGES) {
3497 save_messages_locale = do_setlocale_c(LC_MESSAGES, NULL);
3498 if (! save_messages_locale) {
3499 DEBUG_L(PerlIO_printf(Perl_debug_log,
3500 "Could not find current locale for LC_MESSAGES\n"));
3501 goto cant_use_messages;
3503 save_messages_locale = stdize_locale(savepv(save_messages_locale));
3505 if (strEQ(save_messages_locale, save_input_locale)) {
3506 Safefree(save_messages_locale);
3507 save_messages_locale = NULL;
3509 else if (! do_setlocale_c(LC_MESSAGES, save_input_locale)) {
3510 DEBUG_L(PerlIO_printf(Perl_debug_log,
3511 "Could not change LC_MESSAGES locale to %s\n",
3512 save_input_locale));
3513 Safefree(save_messages_locale);
3514 goto cant_use_messages;
3518 /* Here the current LC_MESSAGES is set to the locale of the category
3519 * whose information is desired. Look through all the messages. We
3520 * can't use Strerror() here because it may expand to code that
3521 * segfaults in miniperl */
3523 for (e = 0; e <= sys_nerr; e++) {
3525 errmsg = sys_errlist[e];
3526 if (errno || !errmsg) {
3529 errmsg = savepv(errmsg);
3530 if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
3532 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
3538 /* And, if we changed it, restore LC_MESSAGES to its original locale */
3539 if (save_messages_locale) {
3540 do_setlocale_c(LC_MESSAGES, save_messages_locale);
3541 Safefree(save_messages_locale);
3546 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
3547 * any non-ascii means it is one; otherwise we assume it isn't */
3548 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
3551 goto finish_and_return;
3554 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));
3559 # endif /* the code that is compiled when no nl_langinfo */
3561 # ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
3564 /* As a last resort, look at the locale name to see if it matches
3565 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
3566 * return of setlocale(), is actually defined to be opaque, so we can't
3567 * really rely on the absence of various substrings in the name to indicate
3568 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
3569 * be a UTF-8 locale. Similarly for the other common names */
3571 final_pos = strlen(save_input_locale) - 1;
3572 if (final_pos >= 3) {
3573 const char *name = save_input_locale;
3575 /* Find next 'U' or 'u' and look from there */
3576 while ((name += strcspn(name, "Uu") + 1)
3577 <= save_input_locale + final_pos - 2)
3579 if ( isALPHA_FOLD_NE(*name, 't')
3580 || isALPHA_FOLD_NE(*(name + 1), 'f'))
3585 if (*(name) == '-') {
3586 if ((name > save_input_locale + final_pos - 1)) {
3591 if (*(name) == '8') {
3592 DEBUG_L(PerlIO_printf(Perl_debug_log,
3593 "Locale %s ends with UTF-8 in name\n",
3594 save_input_locale));
3596 goto finish_and_return;
3599 DEBUG_L(PerlIO_printf(Perl_debug_log,
3600 "Locale %s doesn't end with UTF-8 in name\n",
3601 save_input_locale));
3607 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
3608 if (memENDs(save_input_locale, final_pos, "65001")) {
3609 DEBUG_L(PerlIO_printf(Perl_debug_log,
3610 "Locale %s ends with 65001 in name, is UTF-8 locale\n",
3611 save_input_locale));
3613 goto finish_and_return;
3618 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
3619 * since we are about to return FALSE anyway, there is no point in doing
3620 * this extra work */
3623 if (instr(save_input_locale, "8859")) {
3624 DEBUG_L(PerlIO_printf(Perl_debug_log,
3625 "Locale %s has 8859 in name, not UTF-8 locale\n",
3626 save_input_locale));
3628 goto finish_and_return;
3632 DEBUG_L(PerlIO_printf(Perl_debug_log,
3633 "Assuming locale %s is not a UTF-8 locale\n",
3634 save_input_locale));
3639 /* Cache this result so we don't have to go through all this next time. */
3640 utf8ness_cache_size = sizeof(PL_locale_utf8ness)
3641 - (utf8ness_cache - PL_locale_utf8ness);
3643 /* But we can't save it if it is too large for the total space available */
3644 if (LIKELY(input_name_len_with_overhead < utf8ness_cache_size)) {
3645 Size_t utf8ness_cache_len = strlen(utf8ness_cache);
3647 /* Here it can fit, but we may need to clear out the oldest cached
3648 * result(s) to do so. Check */
3649 if (utf8ness_cache_len + input_name_len_with_overhead
3650 >= utf8ness_cache_size)
3652 /* Here we have to clear something out to make room for this.
3653 * Start looking at the rightmost place where it could fit and find
3654 * the beginning of the entry that extends past that. */
3655 char * cutoff = (char *) my_memrchr(utf8ness_cache,
3658 - input_name_len_with_overhead);
3661 assert(cutoff >= utf8ness_cache);
3663 /* This and all subsequent entries must be removed */
3665 utf8ness_cache_len = strlen(utf8ness_cache);
3668 /* Make space for the new entry */
3669 Move(utf8ness_cache,
3670 utf8ness_cache + input_name_len_with_overhead,
3671 utf8ness_cache_len + 1 /* Incl. trailing NUL */, char);
3674 Copy(delimited, utf8ness_cache, input_name_len_with_overhead - 1, char);
3675 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3677 if ((PL_locale_utf8ness[strlen(PL_locale_utf8ness)-1]
3678 & (PERL_UINTMAX_T) ~1) != '0')
3681 "panic: %s: %d: Corrupt utf8ness_cache=%s\nlen=%u,"
3682 " inserted_name=%s, its_len=%u\n",
3684 PL_locale_utf8ness, strlen(PL_locale_utf8ness),
3685 delimited, input_name_len_with_overhead);
3691 if (DEBUG_Lv_TEST || debug_initialization) {
3692 PerlIO_printf(Perl_debug_log,
3693 "PL_locale_utf8ness is now %s; returning %d\n",
3694 PL_locale_utf8ness, is_utf8);
3699 Safefree(delimited);
3700 Safefree(save_input_locale);
3708 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
3711 /* Internal function which returns if we are in the scope of a pragma that
3712 * enables the locale category 'category'. 'compiling' should indicate if
3713 * this is during the compilation phase (TRUE) or not (FALSE). */
3715 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
3717 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
3718 if (! categories || categories == &PL_sv_placeholder) {
3722 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
3723 * a valid unsigned */
3724 assert(category >= -1);
3725 return cBOOL(SvUV(categories) & (1U << (category + 1)));
3729 Perl_my_strerror(pTHX_ const int errnum)
3731 /* Returns a mortalized copy of the text of the error message associated
3732 * with 'errnum'. It uses the current locale's text unless the platform
3733 * doesn't have the LC_MESSAGES category or we are not being called from
3734 * within the scope of 'use locale'. In the former case, it uses whatever
3735 * strerror returns; in the latter case it uses the text from the C locale.
3737 * The function just calls strerror(), but temporarily switches, if needed,
3738 * to the C locale */
3743 #ifndef USE_LOCALE_MESSAGES
3745 /* If platform doesn't have messages category, we don't do any switching to
3746 * the C locale; we just use whatever strerror() returns */
3748 errstr = savepv(Strerror(errnum));
3750 #else /* Has locale messages */
3752 const bool within_locale_scope = IN_LC(LC_MESSAGES);
3754 # if defined(HAS_POSIX_2008_LOCALE) && defined(HAS_STRERROR_L)
3756 /* This function is trivial if we don't have to worry about thread safety
3757 * and have strerror_l(), as it handles the switch of locales so we don't
3758 * have to deal with that. We don't have to worry about thread safety if
3759 * this is an unthreaded build, or if strerror_r() is also available. Both
3760 * it and strerror_l() are thread-safe. Plain strerror() isn't thread
3761 * safe. But on threaded builds when strerror_r() is available, the
3762 * apparent call to strerror() below is actually a macro that
3763 * behind-the-scenes calls strerror_r().
3766 # if ! defined(USE_ITHREADS) || defined(HAS_STRERROR_R)
3768 if (within_locale_scope) {
3769 errstr = savepv(strerror(errnum));
3772 errstr = savepv(strerror_l(errnum, PL_C_locale_obj));
3777 /* Here we have strerror_l(), but not strerror_r() and we are on a
3778 * threaded-build. We use strerror_l() for everything, constructing a
3779 * locale to pass to it if necessary */
3781 bool do_free = FALSE;
3782 locale_t locale_to_use;
3784 if (within_locale_scope) {
3785 locale_to_use = uselocale((locale_t) 0);
3786 if (locale_to_use == LC_GLOBAL_LOCALE) {
3787 locale_to_use = duplocale(LC_GLOBAL_LOCALE);
3791 else { /* Use C locale if not within 'use locale' scope */
3792 locale_to_use = PL_C_locale_obj;
3795 errstr = savepv(strerror_l(errnum, locale_to_use));
3798 freelocale(locale_to_use);
3802 # else /* Doesn't have strerror_l() */
3804 # ifdef USE_POSIX_2008_LOCALE
3806 locale_t save_locale = NULL;
3810 const char * save_locale = NULL;
3811 bool locale_is_C = FALSE;
3813 /* We have a critical section to prevent another thread from changing the
3814 * locale out from under us (or zapping the buffer returned from
3820 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3821 "my_strerror called with errnum %d\n", errnum));
3822 if (! within_locale_scope) {
3825 # ifdef USE_POSIX_2008_LOCALE /* Use the thread-safe locale functions */
3827 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3828 "Not within locale scope, about to call"
3829 " uselocale(0x%p)\n", PL_C_locale_obj));
3830 save_locale = uselocale(PL_C_locale_obj);
3831 if (! save_locale) {
3832 DEBUG_L(PerlIO_printf(Perl_debug_log,
3833 "uselocale failed, errno=%d\n", errno));
3836 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3837 "uselocale returned 0x%p\n", save_locale));
3840 # else /* Not thread-safe build */
3842 save_locale = do_setlocale_c(LC_MESSAGES, NULL);
3843 if (! save_locale) {
3844 DEBUG_L(PerlIO_printf(Perl_debug_log,
3845 "setlocale failed, errno=%d\n", errno));
3848 locale_is_C = isNAME_C_OR_POSIX(save_locale);
3850 /* Switch to the C locale if not already in it */
3851 if (! locale_is_C) {
3853 /* The setlocale() just below likely will zap 'save_locale', so
3855 save_locale = savepv(save_locale);
3856 do_setlocale_c(LC_MESSAGES, "C");
3862 } /* end of ! within_locale_scope */
3864 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s: %d: WITHIN locale scope\n",
3865 __FILE__, __LINE__));
3868 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3869 "Any locale change has been done; about to call Strerror\n"));
3870 errstr = savepv(Strerror(errnum));
3872 if (! within_locale_scope) {
3875 # ifdef USE_POSIX_2008_LOCALE
3877 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3878 "%s: %d: not within locale scope, restoring the locale\n",
3879 __FILE__, __LINE__));
3880 if (save_locale && ! uselocale(save_locale)) {
3881 DEBUG_L(PerlIO_printf(Perl_debug_log,
3882 "uselocale restore failed, errno=%d\n", errno));
3888 if (save_locale && ! locale_is_C) {
3889 if (! do_setlocale_c(LC_MESSAGES, save_locale)) {
3890 DEBUG_L(PerlIO_printf(Perl_debug_log,
3891 "setlocale restore failed, errno=%d\n", errno));
3893 Safefree(save_locale);
3900 # endif /* End of doesn't have strerror_l */
3901 #endif /* End of does have locale messages */
3905 if (DEBUG_Lv_TEST) {
3906 PerlIO_printf(Perl_debug_log, "Strerror returned; saving a copy: '");
3907 print_bytes_for_locale(errstr, errstr + strlen(errstr), 0);
3908 PerlIO_printf(Perl_debug_log, "'\n");
3919 =for apidoc sync_locale
3921 Changing the program's locale should be avoided by XS code. Nevertheless,
3922 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
3923 happens, Perl needs to be told that the locale has changed. Use this function
3924 to do so, before returning to Perl.
3930 Perl_sync_locale(pTHX)
3934 #ifdef USE_LOCALE_CTYPE
3936 newlocale = do_setlocale_c(LC_CTYPE, NULL);
3937 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3938 "%s:%d: %s\n", __FILE__, __LINE__,
3939 setlocale_debug_string(LC_CTYPE, NULL, newlocale)));
3940 new_ctype(newlocale);
3942 #endif /* USE_LOCALE_CTYPE */
3943 #ifdef USE_LOCALE_COLLATE
3945 newlocale = do_setlocale_c(LC_COLLATE, NULL);
3946 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3947 "%s:%d: %s\n", __FILE__, __LINE__,
3948 setlocale_debug_string(LC_COLLATE, NULL, newlocale)));
3949 new_collate(newlocale);
3952 #ifdef USE_LOCALE_NUMERIC
3954 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
3955 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3956 "%s:%d: %s\n", __FILE__, __LINE__,
3957 setlocale_debug_string(LC_NUMERIC, NULL, newlocale)));
3958 new_numeric(newlocale);
3960 #endif /* USE_LOCALE_NUMERIC */
3964 #if defined(DEBUGGING) && defined(USE_LOCALE)
3967 S_setlocale_debug_string(const int category, /* category number,
3969 const char* const locale, /* locale name */
3971 /* return value from setlocale() when attempting to
3972 * set 'category' to 'locale' */
3973 const char* const retval)
3975 /* Returns a pointer to a NUL-terminated string in static storage with
3976 * added text about the info passed in. This is not thread safe and will
3977 * be overwritten by the next call, so this should be used just to
3978 * formulate a string to immediately print or savepv() on. */
3980 /* initialise to a non-null value to keep it out of BSS and so keep
3981 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
3982 static char ret[128] = "If you can read this, thank your buggy C"
3983 " library strlcpy(), and change your hints file"
3986 my_strlcpy(ret, "setlocale(", sizeof(ret));
3987 my_strlcat(ret, category_name(category), sizeof(ret));
3988 my_strlcat(ret, ", ", sizeof(ret));
3991 my_strlcat(ret, "\"", sizeof(ret));
3992 my_strlcat(ret, locale, sizeof(ret));
3993 my_strlcat(ret, "\"", sizeof(ret));
3996 my_strlcat(ret, "NULL", sizeof(ret));
3999 my_strlcat(ret, ") returned ", sizeof(ret));
4002 my_strlcat(ret, "\"", sizeof(ret));
4003 my_strlcat(ret, retval, sizeof(ret));
4004 my_strlcat(ret, "\"", sizeof(ret));
4007 my_strlcat(ret, "NULL", sizeof(ret));
4010 assert(strlen(ret) < sizeof(ret));
4019 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob