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 const char * radix = (use_locale)
392 ? my_nl_langinfo(PERL_RADIXCHAR, FALSE)
393 /* FALSE => already in dest locale */
396 sv_setpv(PL_numeric_radix_sv, radix);
398 /* If this is valid UTF-8 that isn't totally ASCII, and we are in
399 * a UTF-8 locale, then mark the radix as being in UTF-8 */
400 if (is_utf8_non_invariant_string((U8 *) SvPVX(PL_numeric_radix_sv),
401 SvCUR(PL_numeric_radix_sv))
402 && _is_cur_LC_category_utf8(LC_NUMERIC))
404 SvUTF8_on(PL_numeric_radix_sv);
409 if (DEBUG_L_TEST || debug_initialization) {
410 PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
411 SvPVX(PL_numeric_radix_sv),
412 cBOOL(SvUTF8(PL_numeric_radix_sv)));
416 #endif /* USE_LOCALE_NUMERIC and can find the radix char */
422 Perl_new_numeric(pTHX_ const char *newnum)
425 #ifndef USE_LOCALE_NUMERIC
427 PERL_UNUSED_ARG(newnum);
431 /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell
432 * core Perl this and that 'newnum' is the name of the new locale.
433 * It installs this locale as the current underlying default.
435 * The default locale and the C locale can be toggled between by use of the
436 * set_numeric_underlying() and set_numeric_standard() functions, which
437 * should probably not be called directly, but only via macros like
438 * SET_NUMERIC_STANDARD() in perl.h.
440 * The toggling is necessary mainly so that a non-dot radix decimal point
441 * character can be output, while allowing internal calculations to use a
444 * This sets several interpreter-level variables:
445 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
446 * PL_numeric_underlying A boolean indicating if the toggled state is such
447 * that the current locale is the program's underlying
449 * PL_numeric_standard An int indicating if the toggled state is such
450 * that the current locale is the C locale or
451 * indistinguishable from the C locale. If non-zero, it
452 * is in C; if > 1, it means it may not be toggled away
454 * PL_numeric_underlying_is_standard A bool kept by this function
455 * indicating that the underlying locale and the standard
456 * C locale are indistinguishable for the purposes of
457 * LC_NUMERIC. This happens when both of the above two
458 * variables are true at the same time. (Toggling is a
459 * no-op under these circumstances.) This variable is
460 * used to avoid having to recalculate.
461 * Any code changing the locale (outside this file) should use
462 * POSIX::setlocale, which calls this function. Therefore this function
463 * should be called directly only from this file and from
464 * POSIX::setlocale() */
469 Safefree(PL_numeric_name);
470 PL_numeric_name = NULL;
471 PL_numeric_standard = TRUE;
472 PL_numeric_underlying = TRUE;
473 PL_numeric_underlying_is_standard = TRUE;
477 save_newnum = stdize_locale(savepv(newnum));
478 PL_numeric_underlying = TRUE;
479 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
481 /* If its name isn't C nor POSIX, it could still be indistinguishable from
483 if (! PL_numeric_standard) {
484 PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(PERL_RADIXCHAR,
485 FALSE /* Don't toggle locale */ ))
486 && strEQ("", my_nl_langinfo(PERL_THOUSEP,
490 /* Save the new name if it isn't the same as the previous one, if any */
491 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
492 Safefree(PL_numeric_name);
493 PL_numeric_name = save_newnum;
496 Safefree(save_newnum);
499 PL_numeric_underlying_is_standard = PL_numeric_standard;
501 # if defined(HAS_NEWLOCALE) && ! defined(NO_POSIX_2008_LOCALE)
503 PL_underlying_numeric_obj = newlocale(LC_NUMERIC_MASK,
505 PL_underlying_numeric_obj);
509 if (DEBUG_L_TEST || debug_initialization) {
510 PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name);
513 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
514 * have to worry about the radix being a non-dot. (Core operations that
515 * need the underlying locale change to it temporarily). */
516 set_numeric_standard();
518 #endif /* USE_LOCALE_NUMERIC */
523 Perl_set_numeric_standard(pTHX)
526 #ifdef USE_LOCALE_NUMERIC
528 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
529 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
530 * macro avoids calling this routine if toggling isn't necessary according
531 * to our records (which could be wrong if some XS code has changed the
532 * locale behind our back) */
534 do_setlocale_c(LC_NUMERIC, "C");
535 PL_numeric_standard = TRUE;
536 PL_numeric_underlying = PL_numeric_underlying_is_standard;
537 set_numeric_radix(0);
541 if (DEBUG_L_TEST || debug_initialization) {
542 PerlIO_printf(Perl_debug_log,
543 "LC_NUMERIC locale now is standard C\n");
547 #endif /* USE_LOCALE_NUMERIC */
552 Perl_set_numeric_underlying(pTHX)
555 #ifdef USE_LOCALE_NUMERIC
557 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
558 * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h
559 * instead of calling this directly. The macro avoids calling this routine
560 * if toggling isn't necessary according to our records (which could be
561 * wrong if some XS code has changed the locale behind our back) */
563 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
564 PL_numeric_standard = PL_numeric_underlying_is_standard;
565 PL_numeric_underlying = TRUE;
566 set_numeric_radix(! PL_numeric_standard);
570 if (DEBUG_L_TEST || debug_initialization) {
571 PerlIO_printf(Perl_debug_log,
572 "LC_NUMERIC locale now is %s\n",
577 #endif /* USE_LOCALE_NUMERIC */
582 * Set up for a new ctype locale.
585 S_new_ctype(pTHX_ const char *newctype)
588 #ifndef USE_LOCALE_CTYPE
590 PERL_ARGS_ASSERT_NEW_CTYPE;
591 PERL_UNUSED_ARG(newctype);
596 /* Called after each libc setlocale() call affecting LC_CTYPE, to tell
597 * core Perl this and that 'newctype' is the name of the new locale.
599 * This function sets up the folding arrays for all 256 bytes, assuming
600 * that tofold() is tolc() since fold case is not a concept in POSIX,
602 * Any code changing the locale (outside this file) should use
603 * POSIX::setlocale, which calls this function. Therefore this function
604 * should be called directly only from this file and from
605 * POSIX::setlocale() */
610 PERL_ARGS_ASSERT_NEW_CTYPE;
612 /* We will replace any bad locale warning with 1) nothing if the new one is
613 * ok; or 2) a new warning for the bad new locale */
614 if (PL_warn_locale) {
615 SvREFCNT_dec_NN(PL_warn_locale);
616 PL_warn_locale = NULL;
619 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
621 /* A UTF-8 locale gets standard rules. But note that code still has to
622 * handle this specially because of the three problematic code points */
623 if (PL_in_utf8_CTYPE_locale) {
624 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
627 /* Assume enough space for every character being bad. 4 spaces each
628 * for the 94 printable characters that are output like "'x' "; and 5
629 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
631 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
633 /* Don't check for problems if we are suppressing the warnings */
634 bool check_for_problems = ckWARN_d(WARN_LOCALE)
635 || UNLIKELY(DEBUG_L_TEST);
636 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
638 unsigned int bad_count = 0; /* Count of bad characters */
640 for (i = 0; i < 256; i++) {
642 PL_fold_locale[i] = (U8) tolower(i);
644 PL_fold_locale[i] = (U8) toupper(i);
646 PL_fold_locale[i] = (U8) i;
648 /* If checking for locale problems, see if the native ASCII-range
649 * printables plus \n and \t are in their expected categories in
650 * the new locale. If not, this could mean big trouble, upending
651 * Perl's and most programs' assumptions, like having a
652 * metacharacter with special meaning become a \w. Fortunately,
653 * it's very rare to find locales that aren't supersets of ASCII
654 * nowadays. It isn't a problem for most controls to be changed
655 * into something else; we check only \n and \t, though perhaps \r
656 * could be an issue as well. */
657 if ( check_for_problems
658 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
660 if ( cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC(i))
661 || cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i))
662 || cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i))
663 || cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i))
664 || cBOOL(islower(i)) != cBOOL(isLOWER_A(i))
665 || cBOOL(isprint(i)) != cBOOL(isPRINT_A(i))
666 || cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i))
667 || cBOOL(isspace(i)) != cBOOL(isSPACE_A(i))
668 || cBOOL(isupper(i)) != cBOOL(isUPPER_A(i))
669 || cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i))
670 || tolower(i) != (int) toLOWER_A(i)
671 || toupper(i) != (int) toUPPER_A(i)
672 || (i == '\n' && ! isCNTRL_LC(i)))
674 if (bad_count) { /* Separate multiple entries with a
676 bad_chars_list[bad_count++] = ' ';
678 bad_chars_list[bad_count++] = '\'';
680 bad_chars_list[bad_count++] = (char) i;
683 bad_chars_list[bad_count++] = '\\';
685 bad_chars_list[bad_count++] = 'n';
689 bad_chars_list[bad_count++] = 't';
692 bad_chars_list[bad_count++] = '\'';
693 bad_chars_list[bad_count] = '\0';
700 /* We only handle single-byte locales (outside of UTF-8 ones; so if
701 * this locale requires more than one byte, there are going to be
703 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
704 "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n",
705 __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX));
707 if (check_for_problems && MB_CUR_MAX > 1
709 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
710 * locale. Just assume that the implementation for them (plus
711 * for POSIX) is correct and the > 1 value is spurious. (Since
712 * these are specially handled to never be considered UTF-8
713 * locales, as long as this is the only problem, everything
714 * should work fine */
715 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
717 multi_byte_locale = TRUE;
722 if (bad_count || multi_byte_locale) {
723 PL_warn_locale = Perl_newSVpvf(aTHX_
724 "Locale '%s' may not work well.%s%s%s\n",
727 ? " Some characters in it are not recognized by"
731 ? "\nThe following characters (and maybe others)"
732 " may not have the same meaning as the Perl"
733 " program expects:\n"
739 /* If we are actually in the scope of the locale or are debugging,
740 * output the message now. If not in that scope, we save the
741 * message to be output at the first operation using this locale,
742 * if that actually happens. Most programs don't use locales, so
743 * they are immune to bad ones. */
744 if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) {
746 /* We have to save 'newctype' because the setlocale() just
747 * below may destroy it. The next setlocale() further down
748 * should restore it properly so that the intermediate change
749 * here is transparent to this function's caller */
750 const char * const badlocale = savepv(newctype);
752 do_setlocale_c(LC_CTYPE, "C");
754 /* The '0' below suppresses a bogus gcc compiler warning */
755 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
757 do_setlocale_c(LC_CTYPE, badlocale);
760 if (IN_LC(LC_CTYPE)) {
761 SvREFCNT_dec_NN(PL_warn_locale);
762 PL_warn_locale = NULL;
768 #endif /* USE_LOCALE_CTYPE */
773 Perl__warn_problematic_locale()
776 #ifdef USE_LOCALE_CTYPE
780 /* Internal-to-core function that outputs the message in PL_warn_locale,
781 * and then NULLS it. Should be called only through the macro
782 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
784 if (PL_warn_locale) {
785 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
786 SvPVX(PL_warn_locale),
787 0 /* dummy to avoid compiler warning */ );
788 SvREFCNT_dec_NN(PL_warn_locale);
789 PL_warn_locale = NULL;
797 S_new_collate(pTHX_ const char *newcoll)
800 #ifndef USE_LOCALE_COLLATE
802 PERL_UNUSED_ARG(newcoll);
807 /* Called after each libc setlocale() call affecting LC_COLLATE, to tell
808 * core Perl this and that 'newcoll' is the name of the new locale.
810 * The design of locale collation is that every locale change is given an
811 * index 'PL_collation_ix'. The first time a string particpates in an
812 * operation that requires collation while locale collation is active, it
813 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
814 * magic includes the collation index, and the transformation of the string
815 * by strxfrm(), q.v. That transformation is used when doing comparisons,
816 * instead of the string itself. If a string changes, the magic is
817 * cleared. The next time the locale changes, the index is incremented,
818 * and so we know during a comparison that the transformation is not
819 * necessarily still valid, and so is recomputed. Note that if the locale
820 * changes enough times, the index could wrap (a U32), and it is possible
821 * that a transformation would improperly be considered valid, leading to
825 if (PL_collation_name) {
827 Safefree(PL_collation_name);
828 PL_collation_name = NULL;
830 PL_collation_standard = TRUE;
831 is_standard_collation:
832 PL_collxfrm_base = 0;
833 PL_collxfrm_mult = 2;
834 PL_in_utf8_COLLATE_locale = FALSE;
835 PL_strxfrm_NUL_replacement = '\0';
836 PL_strxfrm_max_cp = 0;
840 /* If this is not the same locale as currently, set the new one up */
841 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
843 Safefree(PL_collation_name);
844 PL_collation_name = stdize_locale(savepv(newcoll));
845 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
846 if (PL_collation_standard) {
847 goto is_standard_collation;
850 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
851 PL_strxfrm_NUL_replacement = '\0';
852 PL_strxfrm_max_cp = 0;
854 /* A locale collation definition includes primary, secondary, tertiary,
855 * etc. weights for each character. To sort, the primary weights are
856 * used, and only if they compare equal, then the secondary weights are
857 * used, and only if they compare equal, then the tertiary, etc.
859 * strxfrm() works by taking the input string, say ABC, and creating an
860 * output transformed string consisting of first the primary weights,
861 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
862 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
863 * may not have weights at every level. In our example, let's say B
864 * doesn't have a tertiary weight, and A doesn't have a secondary
865 * weight. The constructed string is then going to be
866 * A¹B¹C¹ B²C² A³C³ ....
867 * This has the desired effect that strcmp() will look at the secondary
868 * or tertiary weights only if the strings compare equal at all higher
869 * priority weights. The spaces shown here, like in
871 * are not just for readability. In the general case, these must
872 * actually be bytes, which we will call here 'separator weights'; and
873 * they must be smaller than any other weight value, but since these
874 * are C strings, only the terminating one can be a NUL (some
875 * implementations may include a non-NUL separator weight just before
876 * the NUL). Implementations tend to reserve 01 for the separator
877 * weights. They are needed so that a shorter string's secondary
878 * weights won't be misconstrued as primary weights of a longer string,
879 * etc. By making them smaller than any other weight, the shorter
880 * string will sort first. (Actually, if all secondary weights are
881 * smaller than all primary ones, there is no need for a separator
882 * weight between those two levels, etc.)
884 * The length of the transformed string is roughly a linear function of
885 * the input string. It's not exactly linear because some characters
886 * don't have weights at all levels. When we call strxfrm() we have to
887 * allocate some memory to hold the transformed string. The
888 * calculations below try to find coefficients 'm' and 'b' for this
889 * locale so that m*x + b equals how much space we need, given the size
890 * of the input string in 'x'. If we calculate too small, we increase
891 * the size as needed, and call strxfrm() again, but it is better to
892 * get it right the first time to avoid wasted expensive string
893 * transformations. */
896 /* We use the string below to find how long the tranformation of it
897 * is. Almost all locales are supersets of ASCII, or at least the
898 * ASCII letters. We use all of them, half upper half lower,
899 * because if we used fewer, we might hit just the ones that are
900 * outliers in a particular locale. Most of the strings being
901 * collated will contain a preponderance of letters, and even if
902 * they are above-ASCII, they are likely to have the same number of
903 * weight levels as the ASCII ones. It turns out that digits tend
904 * to have fewer levels, and some punctuation has more, but those
905 * are relatively sparse in text, and khw believes this gives a
906 * reasonable result, but it could be changed if experience so
908 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
909 char * x_longer; /* Transformed 'longer' */
910 Size_t x_len_longer; /* Length of 'x_longer' */
912 char * x_shorter; /* We also transform a substring of 'longer' */
913 Size_t x_len_shorter;
915 /* _mem_collxfrm() is used get the transformation (though here we
916 * are interested only in its length). It is used because it has
917 * the intelligence to handle all cases, but to work, it needs some
918 * values of 'm' and 'b' to get it started. For the purposes of
919 * this calculation we use a very conservative estimate of 'm' and
920 * 'b'. This assumes a weight can be multiple bytes, enough to
921 * hold any UV on the platform, and there are 5 levels, 4 weight
922 * bytes, and a trailing NUL. */
923 PL_collxfrm_base = 5;
924 PL_collxfrm_mult = 5 * sizeof(UV);
926 /* Find out how long the transformation really is */
927 x_longer = _mem_collxfrm(longer,
931 /* We avoid converting to UTF-8 in the
932 * called function by telling it the
933 * string is in UTF-8 if the locale is a
934 * UTF-8 one. Since the string passed
935 * here is invariant under UTF-8, we can
936 * claim it's UTF-8 even though it isn't.
938 PL_in_utf8_COLLATE_locale);
941 /* Find out how long the transformation of a substring of 'longer'
942 * is. Together the lengths of these transformations are
943 * sufficient to calculate 'm' and 'b'. The substring is all of
944 * 'longer' except the first character. This minimizes the chances
945 * of being swayed by outliers */
946 x_shorter = _mem_collxfrm(longer + 1,
949 PL_in_utf8_COLLATE_locale);
952 /* If the results are nonsensical for this simple test, the whole
953 * locale definition is suspect. Mark it so that locale collation
954 * is not active at all for it. XXX Should we warn? */
955 if ( x_len_shorter == 0
957 || x_len_shorter >= x_len_longer)
959 PL_collxfrm_mult = 0;
960 PL_collxfrm_base = 0;
963 SSize_t base; /* Temporary */
965 /* We have both: m * strlen(longer) + b = x_len_longer
966 * m * strlen(shorter) + b = x_len_shorter;
967 * subtracting yields:
968 * m * (strlen(longer) - strlen(shorter))
969 * = x_len_longer - x_len_shorter
970 * But we have set things up so that 'shorter' is 1 byte smaller
971 * than 'longer'. Hence:
972 * m = x_len_longer - x_len_shorter
974 * But if something went wrong, make sure the multiplier is at
977 if (x_len_longer > x_len_shorter) {
978 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
981 PL_collxfrm_mult = 1;
986 * but in case something has gone wrong, make sure it is
988 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
993 /* Add 1 for the trailing NUL */
994 PL_collxfrm_base = base + 1;
999 if (DEBUG_L_TEST || debug_initialization) {
1000 PerlIO_printf(Perl_debug_log,
1001 "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
1003 " collate multipler=%zu, collate base=%zu\n",
1005 PL_in_utf8_COLLATE_locale,
1006 x_len_shorter, x_len_longer,
1007 PL_collxfrm_mult, PL_collxfrm_base);
1014 #endif /* USE_LOCALE_COLLATE */
1021 S_win32_setlocale(pTHX_ int category, const char* locale)
1023 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
1024 * difference between the two unless the input locale is "", which normally
1025 * means on Windows to get the machine default, which is set via the
1026 * computer's "Regional and Language Options" (or its current equivalent).
1027 * In POSIX, it instead means to find the locale from the user's
1028 * environment. This routine changes the Windows behavior to first look in
1029 * the environment, and, if anything is found, use that instead of going to
1030 * the machine default. If there is no environment override, the machine
1031 * default is used, by calling the real setlocale() with "".
1033 * The POSIX behavior is to use the LC_ALL variable if set; otherwise to
1034 * use the particular category's variable if set; otherwise to use the LANG
1037 bool override_LC_ALL = FALSE;
1041 if (locale && strEQ(locale, "")) {
1045 locale = PerlEnv_getenv("LC_ALL");
1047 if (category == LC_ALL) {
1048 override_LC_ALL = TRUE;
1054 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
1055 if (category == categories[i]) {
1056 locale = PerlEnv_getenv(category_names[i]);
1061 locale = PerlEnv_getenv("LANG");
1077 result = setlocale(category, locale);
1078 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
1079 setlocale_debug_string(category, locale, result)));
1081 if (! override_LC_ALL) {
1085 /* Here the input category was LC_ALL, and we have set it to what is in the
1086 * LANG variable or the system default if there is no LANG. But these have
1087 * lower priority than the other LC_foo variables, so override it for each
1088 * one that is set. (If they are set to "", it means to use the same thing
1089 * we just set LC_ALL to, so can skip) */
1091 for (i = 0; i < LC_ALL_INDEX; i++) {
1092 result = PerlEnv_getenv(category_names[i]);
1093 if (result && strNE(result, "")) {
1094 setlocale(categories[i], result);
1095 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1097 setlocale_debug_string(categories[i], result, "not captured")));
1101 result = setlocale(LC_ALL, NULL);
1102 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1104 setlocale_debug_string(LC_ALL, NULL, result)));
1112 Perl_setlocale(int category, const char * locale)
1114 /* This wraps POSIX::setlocale() */
1120 #ifdef USE_LOCALE_NUMERIC
1122 /* A NULL locale means only query what the current one is. We have the
1123 * LC_NUMERIC name saved, because we are normally switched into the C
1124 * locale for it. For an LC_ALL query, switch back to get the correct
1125 * results. All other categories don't require special handling */
1126 if (locale == NULL) {
1127 if (category == LC_NUMERIC) {
1128 return savepv(PL_numeric_name);
1133 else if (category == LC_ALL && ! PL_numeric_underlying) {
1135 SET_NUMERIC_UNDERLYING();
1144 /* Save retval since subsequent setlocale() calls may overwrite it. */
1145 retval = savepv(do_setlocale_r(category, locale));
1147 DEBUG_L(PerlIO_printf(Perl_debug_log,
1148 "%s:%d: %s\n", __FILE__, __LINE__,
1149 setlocale_debug_string(category, locale, retval)));
1151 /* Should never happen that a query would return an error, but be
1152 * sure and reset to C locale */
1154 SET_NUMERIC_STANDARD();
1160 /* If locale == NULL, we are just querying the state, but may have switched
1161 * to NUMERIC_UNDERLYING. Switch back before returning. */
1162 if (locale == NULL) {
1163 SET_NUMERIC_STANDARD();
1167 /* Now that have switched locales, we have to update our records to
1172 #ifdef USE_LOCALE_CTYPE
1179 #ifdef USE_LOCALE_COLLATE
1182 new_collate(retval);
1186 #ifdef USE_LOCALE_NUMERIC
1189 new_numeric(retval);
1197 /* LC_ALL updates all the things we care about. The values may not
1198 * be the same as 'retval', as the locale "" may have set things
1201 # ifdef USE_LOCALE_CTYPE
1203 newlocale = do_setlocale_c(LC_CTYPE, NULL);
1204 new_ctype(newlocale);
1206 # endif /* USE_LOCALE_CTYPE */
1207 # ifdef USE_LOCALE_COLLATE
1209 newlocale = do_setlocale_c(LC_COLLATE, NULL);
1210 new_collate(newlocale);
1213 # ifdef USE_LOCALE_NUMERIC
1215 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
1216 new_numeric(newlocale);
1218 # endif /* USE_LOCALE_NUMERIC */
1230 PERL_STATIC_INLINE const char *
1231 S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset)
1233 /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size',
1234 * growing it if necessary */
1236 const Size_t string_size = strlen(string) + offset + 1;
1238 PERL_ARGS_ASSERT_SAVE_TO_BUFFER;
1240 if (*buf_size == 0) {
1241 Newx(*buf, string_size, char);
1242 *buf_size = string_size;
1244 else if (string_size > *buf_size) {
1245 Renew(*buf, string_size, char);
1246 *buf_size = string_size;
1249 Copy(string, *buf + offset, string_size - offset, char);
1255 =head1 Locale-related functions and macros
1257 =for apidoc Perl_langinfo
1259 This is an (almost ª) drop-in replacement for the system C<L<nl_langinfo(3)>>,
1260 taking the same C<item> parameter values, and returning the same information.
1261 But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks
1262 of Perl's locale handling from your code, and can be used on systems that lack
1263 a native C<nl_langinfo>.
1271 It delivers the correct results for the C<RADIXCHAR> and C<THOUSESEP> items,
1272 without you having to write extra code. The reason for the extra code would be
1273 because these are from the C<LC_NUMERIC> locale category, which is normally
1274 kept set to the C locale by Perl, no matter what the underlying locale is
1275 supposed to be, and so to get the expected results, you have to temporarily
1276 toggle into the underlying locale, and later toggle back. (You could use
1277 plain C<nl_langinfo> and C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this
1278 but then you wouldn't get the other advantages of C<Perl_langinfo()>; not
1279 keeping C<LC_NUMERIC> in the C locale would break a lot of CPAN, which is
1280 expecting the radix (decimal point) character to be a dot.)
1284 Depending on C<item>, it works on systems that don't have C<nl_langinfo>, hence
1285 makes your code more portable. Of the fifty-some possible items specified by
1286 the POSIX 2008 standard,
1287 L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>,
1288 only two are completely unimplemented. It uses various techniques to recover
1289 the other items, including calling C<L<localeconv(3)>>, and C<L<strftime(3)>>,
1290 both of which are specified in C89, so should be always be available. Later
1291 C<strftime()> versions have additional capabilities; C<""> is returned for
1292 those not available on your system.
1294 The details for those items which may differ from what this emulation returns
1295 and what a native C<nl_langinfo()> would return are:
1303 Unimplemented, so returns C<"">.
1313 Only the values for English are returned. C<YESSTR> and C<NOSTR> have been
1314 removed from POSIX 2008, and are retained for backwards compatibility. Your
1315 platform's C<nl_langinfo> may not support them.
1319 Always evaluates to C<%x>, the locale's appropriate date representation.
1323 Always evaluates to C<%X>, the locale's appropriate time representation.
1327 Always evaluates to C<%c>, the locale's appropriate date and time
1332 The return may be incorrect for those rare locales where the currency symbol
1333 replaces the radix character.
1334 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1335 to work differently.
1339 Currently this gives the same results as Linux does.
1340 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1341 to work differently.
1347 =item C<ERA_D_T_FMT>
1351 These are derived by using C<strftime()>, and not all versions of that function
1352 know about them. C<""> is returned for these on such systems.
1356 When using C<Perl_langinfo> on systems that don't have a native
1357 C<nl_langinfo()>, you must
1359 #include "perl_langinfo.h"
1361 before the C<perl.h> C<#include>. You can replace your C<langinfo.h>
1362 C<#include> with this one. (Doing it this way keeps out the symbols that plain
1363 C<langinfo.h> imports into the namespace for code that doesn't need it.)
1365 You also should not use the bare C<langinfo.h> item names, but should preface
1366 them with C<PERL_>, so use C<PERL_RADIXCHAR> instead of plain C<RADIXCHAR>.
1367 The C<PERL_I<foo>> versions will also work for this function on systems that do
1368 have a native C<nl_langinfo>.
1372 It is thread-friendly, returning its result in a buffer that won't be
1373 overwritten by another thread, so you don't have to code for that possibility.
1374 The buffer can be overwritten by the next call to C<nl_langinfo> or
1375 C<Perl_langinfo> in the same thread.
1379 ª It returns S<C<const char *>>, whereas plain C<nl_langinfo()> returns S<C<char
1380 *>>, but you are (only by documentation) forbidden to write into the buffer.
1381 By declaring this C<const>, the compiler enforces this restriction. The extra
1382 C<const> is why this isn't an unequivocal drop-in replacement for
1387 The original impetus for C<Perl_langinfo()> was so that code that needs to
1388 find out the current currency symbol, floating point radix character, or digit
1389 grouping separator can use, on all systems, the simpler and more
1390 thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a
1391 pain to make thread-friendly. For other fields returned by C<localeconv>, it
1392 is better to use the methods given in L<perlcall> to call
1393 L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly.
1400 #ifdef HAS_NL_LANGINFO
1401 Perl_langinfo(const nl_item item)
1403 Perl_langinfo(const int item)
1406 return my_nl_langinfo(item, TRUE);
1410 #ifdef HAS_NL_LANGINFO
1411 S_my_nl_langinfo(const nl_item item, bool toggle)
1413 S_my_nl_langinfo(const int item, bool toggle)
1418 /* We only need to toggle into the underlying LC_NUMERIC locale for these
1419 * two items, and only if not already there */
1420 if (toggle && (( item != PERL_RADIXCHAR && item != PERL_THOUSEP)
1421 || PL_numeric_underlying))
1426 #if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */
1427 #if ! defined(HAS_POSIX_2008_LOCALE)
1429 /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC
1430 * for those items dependent on it. This must be copied to a buffer before
1431 * switching back, as some systems destroy the buffer when setlocale() is
1435 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
1438 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
1441 LOCALE_LOCK; /* Prevent interference from another thread executing
1442 this code section (the only call to nl_langinfo in
1445 save_to_buffer(nl_langinfo(item), &PL_langinfo_buf,
1446 &PL_langinfo_bufsize, 0);
1451 RESTORE_LC_NUMERIC();
1455 # else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */
1458 bool do_free = FALSE;
1459 locale_t cur = uselocale((locale_t) 0);
1461 if (cur == LC_GLOBAL_LOCALE) {
1462 cur = duplocale(LC_GLOBAL_LOCALE);
1467 if (PL_underlying_numeric_obj) {
1468 cur = PL_underlying_numeric_obj;
1471 cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur);
1476 save_to_buffer(nl_langinfo_l(item, cur),
1477 &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
1485 if (strEQ(PL_langinfo_buf, "")) {
1486 if (item == PERL_YESSTR) {
1489 if (item == PERL_NOSTR) {
1494 return PL_langinfo_buf;
1496 #else /* Below, emulate nl_langinfo as best we can */
1500 # ifdef HAS_LOCALECONV
1502 const struct lconv* lc;
1503 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
1506 # ifdef HAS_STRFTIME
1509 bool return_format = FALSE; /* Return the %format, not the value */
1510 const char * format;
1514 /* We copy the results to a per-thread buffer, even if not
1515 * multi-threaded. This is in part to simplify this code, and partly
1516 * because we need a buffer anyway for strftime(), and partly because a
1517 * call of localeconv() could otherwise wipe out the buffer, and the
1518 * programmer would not be expecting this, as this is a nl_langinfo()
1519 * substitute after all, so s/he might be thinking their localeconv()
1520 * is safe until another localeconv() call. */
1524 const char * retval;
1526 /* These 2 are unimplemented */
1528 case PERL_ERA: /* For use with strftime() %E modifier */
1533 /* We use only an English set, since we don't know any more */
1534 case PERL_YESEXPR: return "^[+1yY]";
1535 case PERL_YESSTR: return "yes";
1536 case PERL_NOEXPR: return "^[-0nN]";
1537 case PERL_NOSTR: return "no";
1539 # ifdef HAS_LOCALECONV
1543 /* We don't bother with localeconv_l() because any system that
1544 * has it is likely to also have nl_langinfo() */
1546 LOCALE_LOCK; /* Prevent interference with other threads
1547 using localeconv() */
1551 || ! lc->currency_symbol
1552 || strEQ("", lc->currency_symbol))
1558 /* Leave the first spot empty to be filled in below */
1559 save_to_buffer(lc->currency_symbol, &PL_langinfo_buf,
1560 &PL_langinfo_bufsize, 1);
1561 if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, ""))
1562 { /* khw couldn't figure out how the localedef specifications
1563 would show that the $ should replace the radix; this is
1564 just a guess as to how it might work.*/
1565 *PL_langinfo_buf = '.';
1567 else if (lc->p_cs_precedes) {
1568 *PL_langinfo_buf = '-';
1571 *PL_langinfo_buf = '+';
1577 case PERL_RADIXCHAR:
1581 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
1584 LOCALE_LOCK; /* Prevent interference with other threads
1585 using localeconv() */
1592 retval = (item == PERL_RADIXCHAR)
1594 : lc->thousands_sep;
1600 save_to_buffer(retval, &PL_langinfo_buf,
1601 &PL_langinfo_bufsize, 0);
1606 RESTORE_LC_NUMERIC();
1612 # ifdef HAS_STRFTIME
1614 /* These are defined by C89, so we assume that strftime supports
1615 * them, and so are returned unconditionally; they may not be what
1616 * the locale actually says, but should give good enough results
1617 * for someone using them as formats (as opposed to trying to parse
1618 * them to figure out what the locale says). The other format
1619 * items are actually tested to verify they work on the platform */
1620 case PERL_D_FMT: return "%x";
1621 case PERL_T_FMT: return "%X";
1622 case PERL_D_T_FMT: return "%c";
1624 /* These formats are only available in later strfmtime's */
1625 case PERL_ERA_D_FMT: case PERL_ERA_T_FMT: case PERL_ERA_D_T_FMT:
1626 case PERL_T_FMT_AMPM:
1628 /* The rest can be gotten from most versions of strftime(). */
1629 case PERL_ABDAY_1: case PERL_ABDAY_2: case PERL_ABDAY_3:
1630 case PERL_ABDAY_4: case PERL_ABDAY_5: case PERL_ABDAY_6:
1632 case PERL_ALT_DIGITS:
1633 case PERL_AM_STR: case PERL_PM_STR:
1634 case PERL_ABMON_1: case PERL_ABMON_2: case PERL_ABMON_3:
1635 case PERL_ABMON_4: case PERL_ABMON_5: case PERL_ABMON_6:
1636 case PERL_ABMON_7: case PERL_ABMON_8: case PERL_ABMON_9:
1637 case PERL_ABMON_10: case PERL_ABMON_11: case PERL_ABMON_12:
1638 case PERL_DAY_1: case PERL_DAY_2: case PERL_DAY_3: case PERL_DAY_4:
1639 case PERL_DAY_5: case PERL_DAY_6: case PERL_DAY_7:
1640 case PERL_MON_1: case PERL_MON_2: case PERL_MON_3: case PERL_MON_4:
1641 case PERL_MON_5: case PERL_MON_6: case PERL_MON_7: case PERL_MON_8:
1642 case PERL_MON_9: case PERL_MON_10: case PERL_MON_11:
1647 init_tm(&tm); /* Precaution against core dumps */
1651 tm.tm_year = 2017 - 1900;
1658 "panic: %s: %d: switch case: %d problem",
1659 __FILE__, __LINE__, item);
1660 NOT_REACHED; /* NOTREACHED */
1662 case PERL_PM_STR: tm.tm_hour = 18;
1667 case PERL_ABDAY_7: tm.tm_wday++;
1668 case PERL_ABDAY_6: tm.tm_wday++;
1669 case PERL_ABDAY_5: tm.tm_wday++;
1670 case PERL_ABDAY_4: tm.tm_wday++;
1671 case PERL_ABDAY_3: tm.tm_wday++;
1672 case PERL_ABDAY_2: tm.tm_wday++;
1677 case PERL_DAY_7: tm.tm_wday++;
1678 case PERL_DAY_6: tm.tm_wday++;
1679 case PERL_DAY_5: tm.tm_wday++;
1680 case PERL_DAY_4: tm.tm_wday++;
1681 case PERL_DAY_3: tm.tm_wday++;
1682 case PERL_DAY_2: tm.tm_wday++;
1687 case PERL_ABMON_12: tm.tm_mon++;
1688 case PERL_ABMON_11: tm.tm_mon++;
1689 case PERL_ABMON_10: tm.tm_mon++;
1690 case PERL_ABMON_9: tm.tm_mon++;
1691 case PERL_ABMON_8: tm.tm_mon++;
1692 case PERL_ABMON_7: tm.tm_mon++;
1693 case PERL_ABMON_6: tm.tm_mon++;
1694 case PERL_ABMON_5: tm.tm_mon++;
1695 case PERL_ABMON_4: tm.tm_mon++;
1696 case PERL_ABMON_3: tm.tm_mon++;
1697 case PERL_ABMON_2: tm.tm_mon++;
1702 case PERL_MON_12: tm.tm_mon++;
1703 case PERL_MON_11: tm.tm_mon++;
1704 case PERL_MON_10: tm.tm_mon++;
1705 case PERL_MON_9: tm.tm_mon++;
1706 case PERL_MON_8: tm.tm_mon++;
1707 case PERL_MON_7: tm.tm_mon++;
1708 case PERL_MON_6: tm.tm_mon++;
1709 case PERL_MON_5: tm.tm_mon++;
1710 case PERL_MON_4: tm.tm_mon++;
1711 case PERL_MON_3: tm.tm_mon++;
1712 case PERL_MON_2: tm.tm_mon++;
1717 case PERL_T_FMT_AMPM:
1719 return_format = TRUE;
1722 case PERL_ERA_D_FMT:
1724 return_format = TRUE;
1727 case PERL_ERA_T_FMT:
1729 return_format = TRUE;
1732 case PERL_ERA_D_T_FMT:
1734 return_format = TRUE;
1737 case PERL_ALT_DIGITS:
1739 format = "%Ow"; /* Find the alternate digit for 0 */
1743 /* We can't use my_strftime() because it doesn't look at
1745 while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize,
1748 /* A zero return means one of:
1749 * a) there wasn't enough space in PL_langinfo_buf
1750 * b) the format, like a plain %p, returns empty
1751 * c) it was an illegal format, though some
1752 * implementations of strftime will just return the
1753 * illegal format as a plain character sequence.
1755 * To quickly test for case 'b)', try again but precede
1756 * the format with a plain character. If that result is
1757 * still empty, the problem is either 'a)' or 'c)' */
1759 Size_t format_size = strlen(format) + 1;
1760 Size_t mod_size = format_size + 1;
1764 Newx(mod_format, mod_size, char);
1765 Newx(temp_result, PL_langinfo_bufsize, char);
1767 my_strlcpy(mod_format + 1, format, mod_size);
1768 len = strftime(temp_result,
1769 PL_langinfo_bufsize,
1771 Safefree(mod_format);
1772 Safefree(temp_result);
1774 /* If 'len' is non-zero, it means that we had a case like
1775 * %p which means the current locale doesn't use a.m. or
1776 * p.m., and that is valid */
1779 /* Here, still didn't work. If we get well beyond a
1780 * reasonable size, bail out to prevent an infinite
1783 if (PL_langinfo_bufsize > 100 * format_size) {
1784 *PL_langinfo_buf = '\0';
1787 /* Double the buffer size to retry; Add 1 in case
1788 * original was 0, so we aren't stuck at 0. */
1789 PL_langinfo_bufsize *= 2;
1790 PL_langinfo_bufsize++;
1791 Renew(PL_langinfo_buf, PL_langinfo_bufsize, char);
1799 /* Here, we got a result.
1801 * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the
1802 * alternate format for wday 0. If the value is the same as
1803 * the normal 0, there isn't an alternate, so clear the buffer.
1805 if ( item == PERL_ALT_DIGITS
1806 && strEQ(PL_langinfo_buf, "0"))
1808 *PL_langinfo_buf = '\0';
1811 /* ALT_DIGITS is problematic. Experiments on it showed that
1812 * strftime() did not always work properly when going from
1813 * alt-9 to alt-10. Only a few locales have this item defined,
1814 * and in all of them on Linux that khw was able to find,
1815 * nl_langinfo() merely returned the alt-0 character, possibly
1816 * doubled. Most Unicode digits are in blocks of 10
1817 * consecutive code points, so that is sufficient information
1818 * for those scripts, as we can infer alt-1, alt-2, .... But
1819 * for a Japanese locale, a CJK ideographic 0 is returned, and
1820 * the CJK digits are not in code point order, so you can't
1821 * really infer anything. The localedef for this locale did
1822 * specify the succeeding digits, so that strftime() works
1823 * properly on them, without needing to infer anything. But
1824 * the nl_langinfo() return did not give sufficient information
1825 * for the caller to understand what's going on. So until
1826 * there is evidence that it should work differently, this
1827 * returns the alt-0 string for ALT_DIGITS.
1829 * wday was chosen because its range is all a single digit.
1830 * Things like tm_sec have two digits as the minimum: '00' */
1834 /* If to return the format, not the value, overwrite the buffer
1835 * with it. But some strftime()s will keep the original format
1836 * if illegal, so change those to "" */
1837 if (return_format) {
1838 if (strEQ(PL_langinfo_buf, format)) {
1839 *PL_langinfo_buf = '\0';
1842 save_to_buffer(format, &PL_langinfo_buf,
1843 &PL_langinfo_bufsize, 0);
1854 return PL_langinfo_buf;
1861 * Initialize locale awareness.
1864 Perl_init_i18nl10n(pTHX_ int printwarn)
1868 * 0 if not to output warning when setup locale is bad
1869 * 1 if to output warning based on value of PERL_BADLANG
1870 * >1 if to output regardless of PERL_BADLANG
1873 * 1 = set ok or not applicable,
1874 * 0 = fallback to a locale of lower priority
1875 * -1 = fallback to all locales failed, not even to the C locale
1877 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
1878 * set, debugging information is output.
1880 * This looks more complicated than it is, mainly due to the #ifdefs.
1882 * We try to set LC_ALL to the value determined by the environment. If
1883 * there is no LC_ALL on this platform, we try the individual categories we
1884 * know about. If this works, we are done.
1886 * But if it doesn't work, we have to do something else. We search the
1887 * environment variables ourselves instead of relying on the system to do
1888 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
1889 * think there is one), and the ultimate fallback "C". This is all done in
1890 * the same loop as above to avoid duplicating code, but it makes things
1891 * more complex. The 'trial_locales' array is initialized with just one
1892 * element; it causes the behavior described in the paragraph above this to
1893 * happen. If that fails, we add elements to 'trial_locales', and do extra
1894 * loop iterations to cause the behavior described in this paragraph.
1896 * On Ultrix, the locale MUST come from the environment, so there is
1897 * preliminary code to set it. I (khw) am not sure that it is necessary,
1898 * and that this couldn't be folded into the loop, but barring any real
1899 * platforms to test on, it's staying as-is
1901 * A slight complication is that in embedded Perls, the locale may already
1902 * be set-up, and we don't want to get it from the normal environment
1903 * variables. This is handled by having a special environment variable
1904 * indicate we're in this situation. We simply set setlocale's 2nd
1905 * parameter to be a NULL instead of "". That indicates to setlocale that
1906 * it is not to change anything, but to return the current value,
1907 * effectively initializing perl's db to what the locale already is.
1909 * We play the same trick with NULL if a LC_ALL succeeds. We call
1910 * setlocale() on the individual categores with NULL to get their existing
1911 * values for our db, instead of trying to change them.
1918 PERL_UNUSED_ARG(printwarn);
1920 #else /* USE_LOCALE */
1923 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
1927 /* NULL uses the existing already set up locale */
1928 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
1931 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
1932 unsigned int trial_locales_count;
1933 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
1934 const char * const lang = savepv(PerlEnv_getenv("LANG"));
1935 bool setlocale_failure = FALSE;
1938 /* A later getenv() could zap this, so only use here */
1939 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
1941 const bool locwarn = (printwarn > 1
1943 && ( ! bad_lang_use_once
1945 /* disallow with "" or "0" */
1947 && strNE("0", bad_lang_use_once)))));
1949 /* setlocale() return vals; not copied so must be looked at immediately */
1950 const char * sl_result[NOMINAL_LC_ALL_INDEX + 1];
1952 /* current locale for given category; should have been copied so aren't
1954 const char * curlocales[NOMINAL_LC_ALL_INDEX + 1];
1958 /* In some systems you can find out the system default locale
1959 * and use that as the fallback locale. */
1960 # define SYSTEM_DEFAULT_LOCALE
1962 # ifdef SYSTEM_DEFAULT_LOCALE
1964 const char *system_default_locale = NULL;
1969 # define DEBUG_LOCALE_INIT(a,b,c)
1972 DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT")));
1974 # define DEBUG_LOCALE_INIT(category, locale, result) \
1976 if (debug_initialization) { \
1977 PerlIO_printf(Perl_debug_log, \
1979 __FILE__, __LINE__, \
1980 setlocale_debug_string(category, \
1986 /* Make sure the parallel arrays are properly set up */
1987 # ifdef USE_LOCALE_NUMERIC
1988 assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC);
1989 assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC"));
1991 # ifdef USE_LOCALE_CTYPE
1992 assert(categories[LC_CTYPE_INDEX] == LC_CTYPE);
1993 assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE"));
1995 # ifdef USE_LOCALE_COLLATE
1996 assert(categories[LC_COLLATE_INDEX] == LC_COLLATE);
1997 assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE"));
1999 # ifdef USE_LOCALE_TIME
2000 assert(categories[LC_TIME_INDEX] == LC_TIME);
2001 assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME"));
2003 # ifdef USE_LOCALE_MESSAGES
2004 assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES);
2005 assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES"));
2007 # ifdef USE_LOCALE_MONETARY
2008 assert(categories[LC_MONETARY_INDEX] == LC_MONETARY);
2009 assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY"));
2011 # ifdef USE_LOCALE_ADDRESS
2012 assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS);
2013 assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS"));
2015 # ifdef USE_LOCALE_IDENTIFICATION
2016 assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION);
2017 assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION"));
2019 # ifdef USE_LOCALE_MEASUREMENT
2020 assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT);
2021 assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT"));
2023 # ifdef USE_LOCALE_PAPER
2024 assert(categories[LC_PAPER_INDEX] == LC_PAPER);
2025 assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER"));
2027 # ifdef USE_LOCALE_TELEPHONE
2028 assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE);
2029 assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE"));
2032 assert(categories[LC_ALL_INDEX] == LC_ALL);
2033 assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL"));
2034 assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX);
2036 # endif /* DEBUGGING */
2038 /* Initialize the cache of the program's UTF-8ness for the always known
2039 * locales C and POSIX */
2040 my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness,
2041 sizeof(PL_locale_utf8ness));
2043 PL_numeric_radix_sv = newSVpvs(".");
2045 # ifdef LOCALE_ENVIRON_REQUIRED
2048 * Ultrix setlocale(..., "") fails if there are no environment
2049 * variables from which to get a locale name.
2053 # error Ultrix without LC_ALL not implemented
2059 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init);
2060 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]);
2061 if (sl_result[LC_ALL_INDEX])
2064 setlocale_failure = TRUE;
2066 if (! setlocale_failure) {
2067 const char * locale_param;
2068 for (i = 0; i < LC_ALL_INDEX; i++) {
2069 locale_param = (! done && (lang || PerlEnv_getenv(category_names[i])))
2072 sl_result[i] = do_setlocale_r(categories[i], locale_param);
2073 if (! sl_result[i]) {
2074 setlocale_failure = TRUE;
2076 DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]);
2081 # endif /* LC_ALL */
2082 # endif /* LOCALE_ENVIRON_REQUIRED */
2084 /* We try each locale in the list until we get one that works, or exhaust
2085 * the list. Normally the loop is executed just once. But if setting the
2086 * locale fails, inside the loop we add fallback trials to the array and so
2087 * will execute the loop multiple times */
2088 trial_locales[0] = setlocale_init;
2089 trial_locales_count = 1;
2091 for (i= 0; i < trial_locales_count; i++) {
2092 const char * trial_locale = trial_locales[i];
2096 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
2097 * when i==0, but I (khw) don't think that behavior makes much
2099 setlocale_failure = FALSE;
2101 # ifdef SYSTEM_DEFAULT_LOCALE
2102 # ifdef WIN32 /* Note that assumes Win32 has LC_ALL */
2104 /* On Windows machines, an entry of "" after the 0th means to use
2105 * the system default locale, which we now proceed to get. */
2106 if (strEQ(trial_locale, "")) {
2109 /* Note that this may change the locale, but we are going to do
2110 * that anyway just below */
2111 system_default_locale = do_setlocale_c(LC_ALL, "");
2112 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
2114 /* Skip if invalid or if it's already on the list of locales to
2116 if (! system_default_locale) {
2117 goto next_iteration;
2119 for (j = 0; j < trial_locales_count; j++) {
2120 if (strEQ(system_default_locale, trial_locales[j])) {
2121 goto next_iteration;
2125 trial_locale = system_default_locale;
2128 # error SYSTEM_DEFAULT_LOCALE only implemented for Win32
2130 # endif /* SYSTEM_DEFAULT_LOCALE */
2136 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale);
2137 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]);
2138 if (! sl_result[LC_ALL_INDEX]) {
2139 setlocale_failure = TRUE;
2142 /* Since LC_ALL succeeded, it should have changed all the other
2143 * categories it can to its value; so we massage things so that the
2144 * setlocales below just return their category's current values.
2145 * This adequately handles the case in NetBSD where LC_COLLATE may
2146 * not be defined for a locale, and setting it individually will
2147 * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to
2148 * the POSIX locale. */
2149 trial_locale = NULL;
2152 # endif /* LC_ALL */
2154 if (! setlocale_failure) {
2156 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2158 = savepv(do_setlocale_r(categories[j], trial_locale));
2159 if (! curlocales[j]) {
2160 setlocale_failure = TRUE;
2162 DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]);
2165 if (! setlocale_failure) { /* All succeeded */
2166 break; /* Exit trial_locales loop */
2170 /* Here, something failed; will need to try a fallback. */
2176 if (locwarn) { /* Output failure info only on the first one */
2180 PerlIO_printf(Perl_error_log,
2181 "perl: warning: Setting locale failed.\n");
2183 # else /* !LC_ALL */
2185 PerlIO_printf(Perl_error_log,
2186 "perl: warning: Setting locale failed for the categories:\n\t");
2188 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2189 if (! curlocales[j]) {
2190 PerlIO_printf(Perl_error_log, category_names[j]);
2193 Safefree(curlocales[j]);
2197 # endif /* LC_ALL */
2199 PerlIO_printf(Perl_error_log,
2200 "perl: warning: Please check that your locale settings:\n");
2204 PerlIO_printf(Perl_error_log,
2205 "\tLANGUAGE = %c%s%c,\n",
2206 language ? '"' : '(',
2207 language ? language : "unset",
2208 language ? '"' : ')');
2211 PerlIO_printf(Perl_error_log,
2212 "\tLC_ALL = %c%s%c,\n",
2214 lc_all ? lc_all : "unset",
2215 lc_all ? '"' : ')');
2217 # if defined(USE_ENVIRON_ARRAY)
2222 /* Look through the environment for any variables of the
2223 * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was
2224 * already handled above. These are assumed to be locale
2225 * settings. Output them and their values. */
2226 for (e = environ; *e; e++) {
2227 const STRLEN prefix_len = sizeof("LC_") - 1;
2230 if ( strBEGINs(*e, "LC_")
2231 && ! strBEGINs(*e, "LC_ALL=")
2232 && (uppers_len = strspn(*e + prefix_len,
2233 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"))
2234 && ((*e)[prefix_len + uppers_len] == '='))
2236 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
2237 (int) (prefix_len + uppers_len), *e,
2238 *e + prefix_len + uppers_len + 1);
2245 PerlIO_printf(Perl_error_log,
2246 "\t(possibly more locale environment variables)\n");
2250 PerlIO_printf(Perl_error_log,
2251 "\tLANG = %c%s%c\n",
2253 lang ? lang : "unset",
2256 PerlIO_printf(Perl_error_log,
2257 " are supported and installed on your system.\n");
2260 /* Calculate what fallback locales to try. We have avoided this
2261 * until we have to, because failure is quite unlikely. This will
2262 * usually change the upper bound of the loop we are in.
2264 * Since the system's default way of setting the locale has not
2265 * found one that works, We use Perl's defined ordering: LC_ALL,
2266 * LANG, and the C locale. We don't try the same locale twice, so
2267 * don't add to the list if already there. (On POSIX systems, the
2268 * LC_ALL element will likely be a repeat of the 0th element "",
2269 * but there's no harm done by doing it explicitly.
2271 * Note that this tries the LC_ALL environment variable even on
2272 * systems which have no LC_ALL locale setting. This may or may
2273 * not have been originally intentional, but there's no real need
2274 * to change the behavior. */
2276 for (j = 0; j < trial_locales_count; j++) {
2277 if (strEQ(lc_all, trial_locales[j])) {
2281 trial_locales[trial_locales_count++] = lc_all;
2286 for (j = 0; j < trial_locales_count; j++) {
2287 if (strEQ(lang, trial_locales[j])) {
2291 trial_locales[trial_locales_count++] = lang;
2295 # if defined(WIN32) && defined(LC_ALL)
2297 /* For Windows, we also try the system default locale before "C".
2298 * (If there exists a Windows without LC_ALL we skip this because
2299 * it gets too complicated. For those, the "C" is the next
2300 * fallback possibility). The "" is the same as the 0th element of
2301 * the array, but the code at the loop above knows to treat it
2302 * differently when not the 0th */
2303 trial_locales[trial_locales_count++] = "";
2307 for (j = 0; j < trial_locales_count; j++) {
2308 if (strEQ("C", trial_locales[j])) {
2312 trial_locales[trial_locales_count++] = "C";
2315 } /* end of first time through the loop */
2323 } /* end of looping through the trial locales */
2325 if (ok < 1) { /* If we tried to fallback */
2327 if (! setlocale_failure) { /* fallback succeeded */
2328 msg = "Falling back to";
2330 else { /* fallback failed */
2333 /* We dropped off the end of the loop, so have to decrement i to
2334 * get back to the value the last time through */
2338 msg = "Failed to fall back to";
2340 /* To continue, we should use whatever values we've got */
2342 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2343 Safefree(curlocales[j]);
2344 curlocales[j] = savepv(do_setlocale_r(categories[j], NULL));
2345 DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]);
2350 const char * description;
2351 const char * name = "";
2352 if (strEQ(trial_locales[i], "C")) {
2353 description = "the standard locale";
2357 # ifdef SYSTEM_DEFAULT_LOCALE
2359 else if (strEQ(trial_locales[i], "")) {
2360 description = "the system default locale";
2361 if (system_default_locale) {
2362 name = system_default_locale;
2366 # endif /* SYSTEM_DEFAULT_LOCALE */
2369 description = "a fallback locale";
2370 name = trial_locales[i];
2372 if (name && strNE(name, "")) {
2373 PerlIO_printf(Perl_error_log,
2374 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
2377 PerlIO_printf(Perl_error_log,
2378 "perl: warning: %s %s.\n", msg, description);
2381 } /* End of tried to fallback */
2383 /* Done with finding the locales; update our records */
2385 # ifdef USE_LOCALE_CTYPE
2387 new_ctype(curlocales[LC_CTYPE_INDEX]);
2390 # ifdef USE_LOCALE_COLLATE
2392 new_collate(curlocales[LC_COLLATE_INDEX]);
2395 # ifdef USE_LOCALE_NUMERIC
2397 new_numeric(curlocales[LC_NUMERIC_INDEX]);
2401 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
2403 # if defined(USE_ITHREADS)
2405 /* This caches whether each category's locale is UTF-8 or not. This
2406 * may involve changing the locale. It is ok to do this at
2407 * initialization time before any threads have started, but not later.
2408 * Caching means that if the program heeds our dictate not to change
2409 * locales in threaded applications, this data will remain valid, and
2410 * it may get queried without changing locales. If the environment is
2411 * such that all categories have the same locale, this isn't needed, as
2412 * the code will not change the locale; but this handles the uncommon
2413 * case where the environment has disparate locales for the categories
2415 (void) _is_cur_LC_category_utf8(categories[i]);
2419 Safefree(curlocales[i]);
2422 # if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
2424 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
2425 * locale is UTF-8. The call to new_ctype() just above has already
2426 * calculated the latter value and saved it in PL_in_utf8_CTYPE_locale. If
2427 * both PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE})
2428 * are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer on
2429 * STDIN, STDOUT, STDERR, _and_ the default open discipline. */
2430 PL_utf8locale = PL_in_utf8_CTYPE_locale;
2432 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
2433 This is an alternative to using the -C command line switch
2434 (the -C if present will override this). */
2436 const char *p = PerlEnv_getenv("PERL_UNICODE");
2437 PL_unicode = p ? parse_unicode_opts(&p) : 0;
2438 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
2452 #endif /* USE_LOCALE */
2455 /* So won't continue to output stuff */
2456 DEBUG_INITIALIZATION_set(FALSE);
2463 #ifdef USE_LOCALE_COLLATE
2466 Perl__mem_collxfrm(pTHX_ const char *input_string,
2467 STRLEN len, /* Length of 'input_string' */
2468 STRLEN *xlen, /* Set to length of returned string
2469 (not including the collation index
2471 bool utf8 /* Is the input in UTF-8? */
2475 /* _mem_collxfrm() is a bit like strxfrm() but with two important
2476 * differences. First, it handles embedded NULs. Second, it allocates a bit
2477 * more memory than needed for the transformed data itself. The real
2478 * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
2479 * the length of that, and doesn't include the collation index size.
2480 * Please see sv_collxfrm() to see how this is used. */
2482 #define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
2484 char * s = (char *) input_string;
2485 STRLEN s_strlen = strlen(input_string);
2487 STRLEN xAlloc; /* xalloc is a reserved word in VC */
2488 STRLEN length_in_chars;
2489 bool first_time = TRUE; /* Cleared after first loop iteration */
2491 PERL_ARGS_ASSERT__MEM_COLLXFRM;
2493 /* Must be NUL-terminated */
2494 assert(*(input_string + len) == '\0');
2496 /* If this locale has defective collation, skip */
2497 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
2498 DEBUG_L(PerlIO_printf(Perl_debug_log,
2499 "_mem_collxfrm: locale's collation is defective\n"));
2503 /* Replace any embedded NULs with the control that sorts before any others.
2504 * This will give as good as possible results on strings that don't
2505 * otherwise contain that character, but otherwise there may be
2506 * less-than-perfect results with that character and NUL. This is
2507 * unavoidable unless we replace strxfrm with our own implementation. */
2508 if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded
2512 STRLEN sans_nuls_len;
2513 int try_non_controls;
2514 char this_replacement_char[] = "?\0"; /* Room for a two-byte string,
2515 making sure 2nd byte is NUL.
2517 STRLEN this_replacement_len;
2519 /* If we don't know what non-NUL control character sorts lowest for
2520 * this locale, find it */
2521 if (PL_strxfrm_NUL_replacement == '\0') {
2523 char * cur_min_x = NULL; /* The min_char's xfrm, (except it also
2524 includes the collation index
2527 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
2529 /* Unlikely, but it may be that no control will work to replace
2530 * NUL, in which case we instead look for any character. Controls
2531 * are preferred because collation order is, in general, context
2532 * sensitive, with adjoining characters affecting the order, and
2533 * controls are less likely to have such interactions, allowing the
2534 * NUL-replacement to stand on its own. (Another way to look at it
2535 * is to imagine what would happen if the NUL were replaced by a
2536 * combining character; it wouldn't work out all that well.) */
2537 for (try_non_controls = 0;
2538 try_non_controls < 2;
2541 /* Look through all legal code points (NUL isn't) */
2542 for (j = 1; j < 256; j++) {
2543 char * x; /* j's xfrm plus collation index */
2544 STRLEN x_len; /* length of 'x' */
2545 STRLEN trial_len = 1;
2546 char cur_source[] = { '\0', '\0' };
2548 /* Skip non-controls the first time through the loop. The
2549 * controls in a UTF-8 locale are the L1 ones */
2550 if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
2557 /* Create a 1-char string of the current code point */
2558 cur_source[0] = (char) j;
2560 /* Then transform it */
2561 x = _mem_collxfrm(cur_source, trial_len, &x_len,
2562 0 /* The string is not in UTF-8 */);
2564 /* Ignore any character that didn't successfully transform.
2570 /* If this character's transformation is lower than
2571 * the current lowest, this one becomes the lowest */
2572 if ( cur_min_x == NULL
2573 || strLT(x + COLLXFRM_HDR_LEN,
2574 cur_min_x + COLLXFRM_HDR_LEN))
2576 PL_strxfrm_NUL_replacement = j;
2582 } /* end of loop through all 255 characters */
2584 /* Stop looking if found */
2589 /* Unlikely, but possible, if there aren't any controls that
2590 * work in the locale, repeat the loop, looking for any
2591 * character that works */
2592 DEBUG_L(PerlIO_printf(Perl_debug_log,
2593 "_mem_collxfrm: No control worked. Trying non-controls\n"));
2594 } /* End of loop to try first the controls, then any char */
2597 DEBUG_L(PerlIO_printf(Perl_debug_log,
2598 "_mem_collxfrm: Couldn't find any character to replace"
2599 " embedded NULs in locale %s with", PL_collation_name));
2603 DEBUG_L(PerlIO_printf(Perl_debug_log,
2604 "_mem_collxfrm: Replacing embedded NULs in locale %s with "
2605 "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement));
2607 Safefree(cur_min_x);
2608 } /* End of determining the character that is to replace NULs */
2610 /* If the replacement is variant under UTF-8, it must match the
2611 * UTF8-ness of the original */
2612 if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) {
2613 this_replacement_char[0] =
2614 UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement);
2615 this_replacement_char[1] =
2616 UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement);
2617 this_replacement_len = 2;
2620 this_replacement_char[0] = PL_strxfrm_NUL_replacement;
2621 /* this_replacement_char[1] = '\0' was done at initialization */
2622 this_replacement_len = 1;
2625 /* The worst case length for the replaced string would be if every
2626 * character in it is NUL. Multiply that by the length of each
2627 * replacement, and allow for a trailing NUL */
2628 sans_nuls_len = (len * this_replacement_len) + 1;
2629 Newx(sans_nuls, sans_nuls_len, char);
2632 /* Replace each NUL with the lowest collating control. Loop until have
2633 * exhausted all the NULs */
2634 while (s + s_strlen < e) {
2635 my_strlcat(sans_nuls, s, sans_nuls_len);
2637 /* Do the actual replacement */
2638 my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len);
2640 /* Move past the input NUL */
2642 s_strlen = strlen(s);
2645 /* And add anything that trails the final NUL */
2646 my_strlcat(sans_nuls, s, sans_nuls_len);
2648 /* Switch so below we transform this modified string */
2651 } /* End of replacing NULs */
2653 /* Make sure the UTF8ness of the string and locale match */
2654 if (utf8 != PL_in_utf8_COLLATE_locale) {
2655 const char * const t = s; /* Temporary so we can later find where the
2658 /* Here they don't match. Change the string's to be what the locale is
2661 if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
2662 s = (char *) bytes_to_utf8((const U8 *) s, &len);
2665 else { /* locale is not UTF-8; but input is; downgrade the input */
2667 s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
2669 /* If the downgrade was successful we are done, but if the input
2670 * contains things that require UTF-8 to represent, have to do
2671 * damage control ... */
2672 if (UNLIKELY(utf8)) {
2674 /* What we do is construct a non-UTF-8 string with
2675 * 1) the characters representable by a single byte converted
2676 * to be so (if necessary);
2677 * 2) and the rest converted to collate the same as the
2678 * highest collating representable character. That makes
2679 * them collate at the end. This is similar to how we
2680 * handle embedded NULs, but we use the highest collating
2681 * code point instead of the smallest. Like the NUL case,
2682 * this isn't perfect, but is the best we can reasonably
2683 * do. Every above-255 code point will sort the same as
2684 * the highest-sorting 0-255 code point. If that code
2685 * point can combine in a sequence with some other code
2686 * points for weight calculations, us changing something to
2687 * be it can adversely affect the results. But in most
2688 * cases, it should work reasonably. And note that this is
2689 * really an illegal situation: using code points above 255
2690 * on a locale where only 0-255 are valid. If two strings
2691 * sort entirely equal, then the sort order for the
2692 * above-255 code points will be in code point order. */
2696 /* If we haven't calculated the code point with the maximum
2697 * collating order for this locale, do so now */
2698 if (! PL_strxfrm_max_cp) {
2701 /* The current transformed string that collates the
2702 * highest (except it also includes the prefixed collation
2704 char * cur_max_x = NULL;
2706 /* Look through all legal code points (NUL isn't) */
2707 for (j = 1; j < 256; j++) {
2710 char cur_source[] = { '\0', '\0' };
2712 /* Create a 1-char string of the current code point */
2713 cur_source[0] = (char) j;
2715 /* Then transform it */
2716 x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
2718 /* If something went wrong (which it shouldn't), just
2719 * ignore this code point */
2724 /* If this character's transformation is higher than
2725 * the current highest, this one becomes the highest */
2726 if ( cur_max_x == NULL
2727 || strGT(x + COLLXFRM_HDR_LEN,
2728 cur_max_x + COLLXFRM_HDR_LEN))
2730 PL_strxfrm_max_cp = j;
2739 DEBUG_L(PerlIO_printf(Perl_debug_log,
2740 "_mem_collxfrm: Couldn't find any character to"
2741 " replace above-Latin1 chars in locale %s with",
2742 PL_collation_name));
2746 DEBUG_L(PerlIO_printf(Perl_debug_log,
2747 "_mem_collxfrm: highest 1-byte collating character"
2748 " in locale %s is 0x%02X\n",
2750 PL_strxfrm_max_cp));
2752 Safefree(cur_max_x);
2755 /* Here we know which legal code point collates the highest.
2756 * We are ready to construct the non-UTF-8 string. The length
2757 * will be at least 1 byte smaller than the input string
2758 * (because we changed at least one 2-byte character into a
2759 * single byte), but that is eaten up by the trailing NUL */
2765 char * e = (char *) t + len;
2767 for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
2769 if (UTF8_IS_INVARIANT(cur_char)) {
2772 else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) {
2773 s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
2775 else { /* Replace illegal cp with highest collating
2777 s[d++] = PL_strxfrm_max_cp;
2781 Renew(s, d, char); /* Free up unused space */
2786 /* Here, we have constructed a modified version of the input. It could
2787 * be that we already had a modified copy before we did this version.
2788 * If so, that copy is no longer needed */
2789 if (t != input_string) {
2794 length_in_chars = (utf8)
2795 ? utf8_length((U8 *) s, (U8 *) s + len)
2798 /* The first element in the output is the collation id, used by
2799 * sv_collxfrm(); then comes the space for the transformed string. The
2800 * equation should give us a good estimate as to how much is needed */
2801 xAlloc = COLLXFRM_HDR_LEN
2803 + (PL_collxfrm_mult * length_in_chars);
2804 Newx(xbuf, xAlloc, char);
2805 if (UNLIKELY(! xbuf)) {
2806 DEBUG_L(PerlIO_printf(Perl_debug_log,
2807 "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
2811 /* Store the collation id */
2812 *(U32*)xbuf = PL_collation_ix;
2814 /* Then the transformation of the input. We loop until successful, or we
2818 *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
2820 /* If the transformed string occupies less space than we told strxfrm()
2821 * was available, it means it successfully transformed the whole
2823 if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
2825 /* Some systems include a trailing NUL in the returned length.
2826 * Ignore it, using a loop in case multiple trailing NULs are
2829 && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
2834 /* If the first try didn't get it, it means our prediction was low.
2835 * Modify the coefficients so that we predict a larger value in any
2836 * future transformations */
2838 STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
2839 STRLEN computed_guess = PL_collxfrm_base
2840 + (PL_collxfrm_mult * length_in_chars);
2842 /* On zero-length input, just keep current slope instead of
2844 const STRLEN new_m = (length_in_chars != 0)
2845 ? needed / length_in_chars
2848 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2849 "%s: %d: initial size of %zu bytes for a length "
2850 "%zu string was insufficient, %zu needed\n",
2852 computed_guess, length_in_chars, needed));
2854 /* If slope increased, use it, but discard this result for
2855 * length 1 strings, as we can't be sure that it's a real slope
2857 if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
2861 STRLEN old_m = PL_collxfrm_mult;
2862 STRLEN old_b = PL_collxfrm_base;
2866 PL_collxfrm_mult = new_m;
2867 PL_collxfrm_base = 1; /* +1 For trailing NUL */
2868 computed_guess = PL_collxfrm_base
2869 + (PL_collxfrm_mult * length_in_chars);
2870 if (computed_guess < needed) {
2871 PL_collxfrm_base += needed - computed_guess;
2874 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2875 "%s: %d: slope is now %zu; was %zu, base "
2876 "is now %zu; was %zu\n",
2878 PL_collxfrm_mult, old_m,
2879 PL_collxfrm_base, old_b));
2881 else { /* Slope didn't change, but 'b' did */
2882 const STRLEN new_b = needed
2885 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2886 "%s: %d: base is now %zu; was %zu\n",
2888 new_b, PL_collxfrm_base));
2889 PL_collxfrm_base = new_b;
2896 if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
2897 DEBUG_L(PerlIO_printf(Perl_debug_log,
2898 "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
2899 *xlen, PERL_INT_MAX));
2903 /* A well-behaved strxfrm() returns exactly how much space it needs
2904 * (usually not including the trailing NUL) when it fails due to not
2905 * enough space being provided. Assume that this is the case unless
2906 * it's been proven otherwise */
2907 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
2908 xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
2910 else { /* Here, either:
2911 * 1) The strxfrm() has previously shown bad behavior; or
2912 * 2) It isn't the first time through the loop, which means
2913 * that the strxfrm() is now showing bad behavior, because
2914 * we gave it what it said was needed in the previous
2915 * iteration, and it came back saying it needed still more.
2916 * (Many versions of cygwin fit this. When the buffer size
2917 * isn't sufficient, they return the input size instead of
2918 * how much is needed.)
2919 * Increase the buffer size by a fixed percentage and try again.
2921 xAlloc += (xAlloc / 4) + 1;
2922 PL_strxfrm_is_behaved = FALSE;
2926 if (DEBUG_Lv_TEST || debug_initialization) {
2927 PerlIO_printf(Perl_debug_log,
2928 "_mem_collxfrm required more space than previously calculated"
2929 " for locale %s, trying again with new guess=%d+%zu\n",
2930 PL_collation_name, (int) COLLXFRM_HDR_LEN,
2931 xAlloc - COLLXFRM_HDR_LEN);
2938 Renew(xbuf, xAlloc, char);
2939 if (UNLIKELY(! xbuf)) {
2940 DEBUG_L(PerlIO_printf(Perl_debug_log,
2941 "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
2951 if (DEBUG_Lv_TEST || debug_initialization) {
2953 print_collxfrm_input_and_return(s, s + len, xlen, utf8);
2954 PerlIO_printf(Perl_debug_log, "Its xfrm is:");
2955 PerlIO_printf(Perl_debug_log, "%s\n",
2956 _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN,
2962 /* Free up unneeded space; retain ehough for trailing NUL */
2963 Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
2965 if (s != input_string) {
2973 if (s != input_string) {
2980 if (DEBUG_Lv_TEST || debug_initialization) {
2981 print_collxfrm_input_and_return(s, s + len, NULL, utf8);
2992 S_print_collxfrm_input_and_return(pTHX_
2993 const char * const s,
2994 const char * const e,
2995 const STRLEN * const xlen,
2999 PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
3001 PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ",
3002 (UV)PL_collation_ix);
3004 PerlIO_printf(Perl_debug_log, "%zu", *xlen);
3007 PerlIO_printf(Perl_debug_log, "NULL");
3009 PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
3011 print_bytes_for_locale(s, e, is_utf8);
3013 PerlIO_printf(Perl_debug_log, "'\n");
3017 S_print_bytes_for_locale(pTHX_
3018 const char * const s,
3019 const char * const e,
3023 bool prev_was_printable = TRUE;
3024 bool first_time = TRUE;
3026 PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE;
3030 ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
3033 if (! prev_was_printable) {
3034 PerlIO_printf(Perl_debug_log, " ");
3036 PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
3037 prev_was_printable = TRUE;
3041 PerlIO_printf(Perl_debug_log, " ");
3043 PerlIO_printf(Perl_debug_log, "%02" UVXf, cp);
3044 prev_was_printable = FALSE;
3046 t += (is_utf8) ? UTF8SKIP(t) : 1;
3051 # endif /* #ifdef DEBUGGING */
3052 #endif /* USE_LOCALE_COLLATE */
3057 Perl__is_cur_LC_category_utf8(pTHX_ int category)
3059 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
3060 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
3061 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
3062 * could give the wrong result. The result will very likely be correct for
3063 * languages that have commonly used non-ASCII characters, but for notably
3064 * English, it comes down to if the locale's name ends in something like
3065 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
3067 /* Name of current locale corresponding to the input category */
3068 const char *save_input_locale = NULL;
3070 bool is_utf8 = FALSE; /* The return value */
3073 /* The variables below are for the cache of previous lookups using this
3074 * function. The cache is a C string, described at the definition for
3075 * 'C_and_POSIX_utf8ness'.
3077 * The first part of the cache is fixed, for the C and POSIX locales. The
3078 * varying part starts just after them. */
3079 char * utf8ness_cache = PL_locale_utf8ness + STRLENs(C_and_POSIX_utf8ness);
3081 Size_t utf8ness_cache_size; /* Size of the varying portion */
3082 Size_t input_name_len; /* Length in bytes of save_input_locale */
3083 Size_t input_name_len_with_overhead; /* plus extra chars used to store
3084 the name in the cache */
3085 char * delimited; /* The name plus the delimiters used to store
3087 char * name_pos; /* position of 'delimited' in the cache, or 0
3093 assert(category != LC_ALL);
3097 /* Get the desired category's locale */
3098 save_input_locale = do_setlocale_r(category, NULL);
3099 if (! save_input_locale) {
3101 "panic: %s: %d: Could not find current %s locale, errno=%d\n",
3102 __FILE__, __LINE__, category_name(category), errno);
3105 save_input_locale = stdize_locale(savepv(save_input_locale));
3106 DEBUG_L(PerlIO_printf(Perl_debug_log,
3107 "Current locale for %s is %s\n",
3108 category_name(category), save_input_locale));
3110 input_name_len = strlen(save_input_locale);
3112 /* In our cache, each name is accompanied by two delimiters and a single
3114 input_name_len_with_overhead = input_name_len + 3;
3116 /* Allocate and populate space for a copy of the name surrounded by the
3118 Newx(delimited, input_name_len_with_overhead, char);
3119 delimited[0] = UTF8NESS_SEP[0];
3120 Copy(save_input_locale, delimited + 1, input_name_len, char);
3121 delimited[input_name_len+1] = UTF8NESS_PREFIX[0];
3122 delimited[input_name_len+2] = '\0';
3124 /* And see if that is in the cache */
3125 name_pos = instr(PL_locale_utf8ness, delimited);
3127 is_utf8 = *(name_pos + input_name_len_with_overhead - 1) - '0';
3131 if (DEBUG_Lv_TEST || debug_initialization) {
3132 PerlIO_printf(Perl_debug_log, "UTF8ness for locale %s=%d, \n",
3133 save_input_locale, is_utf8);
3138 /* And, if not already in that position, move it to the beginning of
3139 * the non-constant portion of the list, since it is the most recently
3140 * used. (We don't have to worry about overflow, since just moving
3141 * existing names around) */
3142 if (name_pos > utf8ness_cache) {
3143 Move(utf8ness_cache,
3144 utf8ness_cache + input_name_len_with_overhead,
3145 name_pos - utf8ness_cache, char);
3148 input_name_len_with_overhead - 1, char);
3149 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3152 Safefree(delimited);
3153 Safefree(save_input_locale);
3157 /* Here we don't have stored the utf8ness for the input locale. We have to
3160 # if defined(USE_LOCALE_CTYPE) \
3161 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
3163 { /* Next try nl_langinfo or MB_CUR_MAX if available */
3165 char *save_ctype_locale = NULL;
3167 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
3169 /* Get the current LC_CTYPE locale */
3170 save_ctype_locale = do_setlocale_c(LC_CTYPE, NULL);
3171 if (! save_ctype_locale) {
3173 "panic: %s: %d: Could not find current LC_CTYPE locale,"
3174 " errno=%d\n", __FILE__, __LINE__, errno);
3176 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
3178 /* If LC_CTYPE and the desired category use the same locale, this
3179 * means that finding the value for LC_CTYPE is the same as finding
3180 * the value for the desired category. Otherwise, switch LC_CTYPE
3181 * to the desired category's locale */
3182 if (strEQ(save_ctype_locale, save_input_locale)) {
3183 Safefree(save_ctype_locale);
3184 save_ctype_locale = NULL;
3186 else if (! do_setlocale_c(LC_CTYPE, save_input_locale)) {
3187 Safefree(save_ctype_locale);
3189 "panic: %s: %d: Could not change LC_CTYPE locale to %s,"
3190 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3194 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
3195 save_input_locale));
3197 /* Here the current LC_CTYPE is set to the locale of the category whose
3198 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
3199 * should give the correct results */
3201 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
3203 { /* The task is easiest if the platform has this POSIX 2001 function */
3204 const char *codeset = my_nl_langinfo(PERL_CODESET, FALSE);
3205 /* FALSE => already in dest locale */
3207 DEBUG_L(PerlIO_printf(Perl_debug_log,
3208 "\tnllanginfo returned CODESET '%s'\n", codeset));
3210 if (codeset && strNE(codeset, "")) {
3211 /* If we switched LC_CTYPE, switch back */
3212 if (save_ctype_locale) {
3213 do_setlocale_c(LC_CTYPE, save_ctype_locale);
3214 Safefree(save_ctype_locale);
3217 /* If the implementation of foldEQ() somehow were
3218 * to change to not go byte-by-byte, this could
3219 * read past end of string, as only one length is
3220 * checked. But currently, a premature NUL will
3221 * compare false, and it will stop there */
3222 is_utf8 = cBOOL( foldEQ(codeset, STR_WITH_LEN("UTF-8"))
3223 || foldEQ(codeset, STR_WITH_LEN("UTF8")));
3225 DEBUG_L(PerlIO_printf(Perl_debug_log,
3226 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
3228 goto finish_and_return;
3235 /* Here, either we don't have nl_langinfo, or it didn't return a
3236 * codeset. Try MB_CUR_MAX */
3238 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
3239 * Unicode code point. Since UTF-8 is the only non-single byte
3240 * encoding we handle, we just say any such encoding is UTF-8, and if
3241 * turns out to be wrong, other things will fail */
3242 is_utf8 = (unsigned) MB_CUR_MAX >= STRLENs(MAX_UNICODE_UTF8);
3244 DEBUG_L(PerlIO_printf(Perl_debug_log,
3245 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
3246 (int) MB_CUR_MAX, is_utf8));
3248 Safefree(save_input_locale);
3252 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
3253 * since they are both in the C99 standard. We can feed a known byte
3254 * string to the latter function, and check that it gives the expected
3260 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
3262 len = mbtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8));
3265 if ( len != STRLENs(REPLACEMENT_CHARACTER_UTF8)
3266 || wc != (wchar_t) UNICODE_REPLACEMENT)
3269 DEBUG_L(PerlIO_printf(Perl_debug_log, "\replacement=U+%x\n",
3271 DEBUG_L(PerlIO_printf(Perl_debug_log,
3272 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
3279 /* If we switched LC_CTYPE, switch back */
3280 if (save_ctype_locale) {
3281 do_setlocale_c(LC_CTYPE, save_ctype_locale);
3282 Safefree(save_ctype_locale);
3285 goto finish_and_return;
3291 # else /* nl_langinfo should work if available, so don't bother compiling this
3292 fallback code. The final fallback of looking at the name is
3293 compiled, and will be executed if nl_langinfo fails */
3295 /* nl_langinfo not available or failed somehow. Next try looking at the
3296 * currency symbol to see if it disambiguates things. Often that will be
3297 * in the native script, and if the symbol isn't in UTF-8, we know that the
3298 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
3299 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
3302 # ifdef HAS_LOCALECONV
3303 # ifdef USE_LOCALE_MONETARY
3306 char *save_monetary_locale = NULL;
3307 bool only_ascii = FALSE;
3310 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
3311 * the desired category, if it isn't that locale already */
3313 if (category != LC_MONETARY) {
3315 save_monetary_locale = do_setlocale_c(LC_MONETARY, NULL);
3316 if (! save_monetary_locale) {
3318 "panic: %s: %d: Could not find current LC_MONETARY locale,"
3319 " errno=%d\n", __FILE__, __LINE__, errno);
3321 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
3323 if (strEQ(save_monetary_locale, save_input_locale)) {
3324 Safefree(save_monetary_locale);
3325 save_monetary_locale = NULL;
3327 else if (! do_setlocale_c(LC_MONETARY, save_input_locale)) {
3328 Safefree(save_monetary_locale);
3330 "panic: %s: %d: Could not change LC_MONETARY locale to %s,"
3331 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3335 /* Here the current LC_MONETARY is set to the locale of the category
3336 * whose information is desired. */
3340 || ! lc->currency_symbol
3341 || is_utf8_invariant_string((U8 *) lc->currency_symbol, 0))
3343 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));
3347 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
3350 /* If we changed it, restore LC_MONETARY to its original locale */
3351 if (save_monetary_locale) {
3352 do_setlocale_c(LC_MONETARY, save_monetary_locale);
3353 Safefree(save_monetary_locale);
3358 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
3359 * otherwise assume the locale is UTF-8 if and only if the symbol
3360 * is non-ascii UTF-8. */
3361 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
3362 save_input_locale, is_utf8));
3363 goto finish_and_return;
3367 # endif /* USE_LOCALE_MONETARY */
3368 # endif /* HAS_LOCALECONV */
3370 # if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
3372 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
3373 * the names of the months and weekdays, timezone, and am/pm indicator */
3375 char *save_time_locale = NULL;
3377 bool is_dst = FALSE;
3381 char * formatted_time;
3384 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
3385 * desired category, if it isn't that locale already */
3387 if (category != LC_TIME) {
3389 save_time_locale = do_setlocale_c(LC_TIME, NULL);
3390 if (! save_time_locale) {
3392 "panic: %s: %d: Could not find current LC_TIME locale,"
3393 " errno=%d\n", __FILE__, __LINE__, errno);
3395 save_time_locale = stdize_locale(savepv(save_time_locale));
3397 if (strEQ(save_time_locale, save_input_locale)) {
3398 Safefree(save_time_locale);
3399 save_time_locale = NULL;
3401 else if (! do_setlocale_c(LC_TIME, save_input_locale)) {
3402 Safefree(save_time_locale);
3404 "panic: %s: %d: Could not change LC_TIME locale to %s,"
3405 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3409 /* Here the current LC_TIME is set to the locale of the category
3410 * whose information is desired. Look at all the days of the week and
3411 * month names, and the timezone and am/pm indicator for UTF-8 variant
3412 * characters. The first such a one found will tell us if the locale
3413 * is UTF-8 or not */
3415 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
3416 formatted_time = my_strftime("%A %B %Z %p",
3417 0, 0, hour, dom, month, 2012 - 1900, 0, 0, is_dst);
3418 if ( ! formatted_time
3419 || is_utf8_invariant_string((U8 *) formatted_time, 0))
3422 /* Here, we didn't find a non-ASCII. Try the next time through
3423 * with the complemented dst and am/pm, and try with the next
3424 * weekday. After we have gotten all weekdays, try the next
3427 hour = (hour + 12) % 24;
3435 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
3436 * false otherwise. But first, restore LC_TIME to its original
3437 * locale if we changed it */
3438 if (save_time_locale) {
3439 do_setlocale_c(LC_TIME, save_time_locale);
3440 Safefree(save_time_locale);
3443 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
3445 is_utf8_string((U8 *) formatted_time, 0)));
3446 is_utf8 = is_utf8_string((U8 *) formatted_time, 0);
3447 goto finish_and_return;
3450 /* Falling off the end of the loop indicates all the names were just
3451 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
3452 * to its original locale */
3453 if (save_time_locale) {
3454 do_setlocale_c(LC_TIME, save_time_locale);
3455 Safefree(save_time_locale);
3457 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));
3462 # if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
3464 /* This code is ifdefd out because it was found to not be necessary in testing
3465 * on our dromedary test machine, which has over 700 locales. There, this
3466 * added no value to looking at the currency symbol and the time strings. I
3467 * left it in so as to avoid rewriting it if real-world experience indicates
3468 * that dromedary is an outlier. Essentially, instead of returning abpve if we
3469 * haven't found illegal utf8, we continue on and examine all the strerror()
3470 * messages on the platform for utf8ness. If all are ASCII, we still don't
3471 * know the answer; but otherwise we have a pretty good indication of the
3472 * utf8ness. The reason this doesn't help much is that the messages may not
3473 * have been translated into the locale. The currency symbol and time strings
3474 * are much more likely to have been translated. */
3477 bool non_ascii = FALSE;
3478 char *save_messages_locale = NULL;
3479 const char * errmsg = NULL;
3481 /* Like above, we set LC_MESSAGES to the locale of the desired
3482 * category, if it isn't that locale already */
3484 if (category != LC_MESSAGES) {
3486 save_messages_locale = do_setlocale_c(LC_MESSAGES, NULL);
3487 if (! save_messages_locale) {
3489 "panic: %s: %d: Could not find current LC_MESSAGES locale,"
3490 " errno=%d\n", __FILE__, __LINE__, errno);
3492 save_messages_locale = stdize_locale(savepv(save_messages_locale));
3494 if (strEQ(save_messages_locale, save_input_locale)) {
3495 Safefree(save_messages_locale);
3496 save_messages_locale = NULL;
3498 else if (! do_setlocale_c(LC_MESSAGES, save_input_locale)) {
3499 Safefree(save_messages_locale);
3501 "panic: %s: %d: Could not change LC_MESSAGES locale to %s,"
3502 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3506 /* Here the current LC_MESSAGES is set to the locale of the category
3507 * whose information is desired. Look through all the messages. We
3508 * can't use Strerror() here because it may expand to code that
3509 * segfaults in miniperl */
3511 for (e = 0; e <= sys_nerr; e++) {
3513 errmsg = sys_errlist[e];
3514 if (errno || !errmsg) {
3517 errmsg = savepv(errmsg);
3518 if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
3520 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
3526 /* And, if we changed it, restore LC_MESSAGES to its original locale */
3527 if (save_messages_locale) {
3528 do_setlocale_c(LC_MESSAGES, save_messages_locale);
3529 Safefree(save_messages_locale);
3534 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
3535 * any non-ascii means it is one; otherwise we assume it isn't */
3536 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
3539 goto finish_and_return;
3542 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));
3546 # endif /* the code that is compiled when no nl_langinfo */
3548 # ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
3551 /* As a last resort, look at the locale name to see if it matches
3552 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
3553 * return of setlocale(), is actually defined to be opaque, so we can't
3554 * really rely on the absence of various substrings in the name to indicate
3555 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
3556 * be a UTF-8 locale. Similarly for the other common names */
3558 final_pos = strlen(save_input_locale) - 1;
3559 if (final_pos >= 3) {
3560 const char *name = save_input_locale;
3562 /* Find next 'U' or 'u' and look from there */
3563 while ((name += strcspn(name, "Uu") + 1)
3564 <= save_input_locale + final_pos - 2)
3566 if ( isALPHA_FOLD_NE(*name, 't')
3567 || isALPHA_FOLD_NE(*(name + 1), 'f'))
3572 if (*(name) == '-') {
3573 if ((name > save_input_locale + final_pos - 1)) {
3578 if (*(name) == '8') {
3579 DEBUG_L(PerlIO_printf(Perl_debug_log,
3580 "Locale %s ends with UTF-8 in name\n",
3581 save_input_locale));
3583 goto finish_and_return;
3586 DEBUG_L(PerlIO_printf(Perl_debug_log,
3587 "Locale %s doesn't end with UTF-8 in name\n",
3588 save_input_locale));
3593 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
3594 if (memENDs(save_input_locale, final_pos, "65001")) {
3595 DEBUG_L(PerlIO_printf(Perl_debug_log,
3596 "Locale %s ends with 65001 in name, is UTF-8 locale\n",
3597 save_input_locale));
3599 goto finish_and_return;
3605 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
3606 * since we are about to return FALSE anyway, there is no point in doing
3607 * this extra work */
3610 if (instr(save_input_locale, "8859")) {
3611 DEBUG_L(PerlIO_printf(Perl_debug_log,
3612 "Locale %s has 8859 in name, not UTF-8 locale\n",
3613 save_input_locale));
3615 goto finish_and_return;
3619 DEBUG_L(PerlIO_printf(Perl_debug_log,
3620 "Assuming locale %s is not a UTF-8 locale\n",
3621 save_input_locale));
3626 /* Cache this result so we don't have to go through all this next time. */
3627 utf8ness_cache_size = sizeof(PL_locale_utf8ness)
3628 - (utf8ness_cache - PL_locale_utf8ness);
3630 /* But we can't save it if it is too large for the total space available */
3631 if (LIKELY(input_name_len_with_overhead < utf8ness_cache_size)) {
3632 Size_t utf8ness_cache_len = strlen(utf8ness_cache);
3634 /* Here it can fit, but we may need to clear out the oldest cached
3635 * result(s) to do so. Check */
3636 if (utf8ness_cache_len + input_name_len_with_overhead
3637 >= utf8ness_cache_size)
3639 /* Here we have to clear something out to make room for this.
3640 * Start looking at the rightmost place where it could fit and find
3641 * the beginning of the entry that extends past that. */
3642 char * cutoff = (char *) my_memrchr(utf8ness_cache,
3645 - input_name_len_with_overhead);
3648 assert(cutoff >= utf8ness_cache);
3650 /* This and all subsequent entries must be removed */
3652 utf8ness_cache_len = strlen(utf8ness_cache);
3655 /* Make space for the new entry */
3656 Move(utf8ness_cache,
3657 utf8ness_cache + input_name_len_with_overhead,
3658 utf8ness_cache_len + 1 /* Incl. trailing NUL */, char);
3661 Copy(delimited, utf8ness_cache, input_name_len_with_overhead - 1, char);
3662 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3664 if ((PL_locale_utf8ness[strlen(PL_locale_utf8ness)-1]
3665 & (PERL_UINTMAX_T) ~1) != '0')
3668 "panic: %s: %d: Corrupt utf8ness_cache=%s\nlen=%u,"
3669 " inserted_name=%s, its_len=%u\n",
3671 PL_locale_utf8ness, strlen(PL_locale_utf8ness),
3672 delimited, input_name_len_with_overhead);
3678 if (DEBUG_Lv_TEST || debug_initialization) {
3679 PerlIO_printf(Perl_debug_log,
3680 "PL_locale_utf8ness is now %s; returning %d\n",
3681 PL_locale_utf8ness, is_utf8);
3686 Safefree(delimited);
3687 Safefree(save_input_locale);
3695 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
3698 /* Internal function which returns if we are in the scope of a pragma that
3699 * enables the locale category 'category'. 'compiling' should indicate if
3700 * this is during the compilation phase (TRUE) or not (FALSE). */
3702 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
3704 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
3705 if (! categories || categories == &PL_sv_placeholder) {
3709 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
3710 * a valid unsigned */
3711 assert(category >= -1);
3712 return cBOOL(SvUV(categories) & (1U << (category + 1)));
3716 Perl_my_strerror(pTHX_ const int errnum)
3718 /* Returns a mortalized copy of the text of the error message associated
3719 * with 'errnum'. It uses the current locale's text unless the platform
3720 * doesn't have the LC_MESSAGES category or we are not being called from
3721 * within the scope of 'use locale'. In the former case, it uses whatever
3722 * strerror returns; in the latter case it uses the text from the C locale.
3724 * The function just calls strerror(), but temporarily switches, if needed,
3725 * to the C locale */
3730 #ifndef USE_LOCALE_MESSAGES
3732 /* If platform doesn't have messages category, we don't do any switching to
3733 * the C locale; we just use whatever strerror() returns */
3735 errstr = savepv(Strerror(errnum));
3737 #else /* Has locale messages */
3739 const bool within_locale_scope = IN_LC(LC_MESSAGES);
3741 # if defined(HAS_POSIX_2008_LOCALE) && defined(HAS_STRERROR_L)
3743 /* This function is trivial if we don't have to worry about thread safety
3744 * and have strerror_l(), as it handles the switch of locales so we don't
3745 * have to deal with that. We don't have to worry about thread safety if
3746 * this is an unthreaded build, or if strerror_r() is also available. Both
3747 * it and strerror_l() are thread-safe. Plain strerror() isn't thread
3748 * safe. But on threaded builds when strerror_r() is available, the
3749 * apparent call to strerror() below is actually a macro that
3750 * behind-the-scenes calls strerror_r().
3753 # if ! defined(USE_ITHREADS) || defined(HAS_STRERROR_R)
3755 if (within_locale_scope) {
3756 errstr = savepv(strerror(errnum));
3759 errstr = savepv(strerror_l(errnum, PL_C_locale_obj));
3764 /* Here we have strerror_l(), but not strerror_r() and we are on a
3765 * threaded-build. We use strerror_l() for everything, constructing a
3766 * locale to pass to it if necessary */
3768 bool do_free = FALSE;
3769 locale_t locale_to_use;
3771 if (within_locale_scope) {
3772 locale_to_use = uselocale((locale_t) 0);
3773 if (locale_to_use == LC_GLOBAL_LOCALE) {
3774 locale_to_use = duplocale(LC_GLOBAL_LOCALE);
3778 else { /* Use C locale if not within 'use locale' scope */
3779 locale_to_use = PL_C_locale_obj;
3782 errstr = savepv(strerror_l(errnum, locale_to_use));
3785 freelocale(locale_to_use);
3789 # else /* Doesn't have strerror_l() */
3791 # ifdef USE_POSIX_2008_LOCALE
3793 locale_t save_locale = NULL;
3797 const char * save_locale = NULL;
3798 bool locale_is_C = FALSE;
3800 /* We have a critical section to prevent another thread from changing the
3801 * locale out from under us (or zapping the buffer returned from
3807 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3808 "my_strerror called with errnum %d\n", errnum));
3809 if (! within_locale_scope) {
3812 # ifdef USE_POSIX_2008_LOCALE /* Use the thread-safe locale functions */
3814 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3815 "Not within locale scope, about to call"
3816 " uselocale(0x%p)\n", PL_C_locale_obj));
3817 save_locale = uselocale(PL_C_locale_obj);
3818 if (! save_locale) {
3819 DEBUG_L(PerlIO_printf(Perl_debug_log,
3820 "uselocale failed, errno=%d\n", errno));
3823 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3824 "uselocale returned 0x%p\n", save_locale));
3827 # else /* Not thread-safe build */
3829 save_locale = do_setlocale_c(LC_MESSAGES, NULL);
3830 if (! save_locale) {
3832 "panic: %s: %d: Could not find current LC_MESSAGES locale,"
3833 " errno=%d\n", __FILE__, __LINE__, errno);
3836 locale_is_C = isNAME_C_OR_POSIX(save_locale);
3838 /* Switch to the C locale if not already in it */
3839 if (! locale_is_C) {
3841 /* The setlocale() just below likely will zap 'save_locale', so
3843 save_locale = savepv(save_locale);
3844 do_setlocale_c(LC_MESSAGES, "C");
3850 } /* end of ! within_locale_scope */
3852 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s: %d: WITHIN locale scope\n",
3853 __FILE__, __LINE__));
3856 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3857 "Any locale change has been done; about to call Strerror\n"));
3858 errstr = savepv(Strerror(errnum));
3860 if (! within_locale_scope) {
3863 # ifdef USE_POSIX_2008_LOCALE
3865 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3866 "%s: %d: not within locale scope, restoring the locale\n",
3867 __FILE__, __LINE__));
3868 if (save_locale && ! uselocale(save_locale)) {
3869 DEBUG_L(PerlIO_printf(Perl_debug_log,
3870 "uselocale restore failed, errno=%d\n", errno));
3876 if (save_locale && ! locale_is_C) {
3877 if (! do_setlocale_c(LC_MESSAGES, save_locale)) {
3879 "panic: %s: %d: setlocale restore failed, errno=%d\n",
3880 __FILE__, __LINE__, errno);
3882 Safefree(save_locale);
3889 # endif /* End of doesn't have strerror_l */
3890 #endif /* End of does have locale messages */
3894 if (DEBUG_Lv_TEST) {
3895 PerlIO_printf(Perl_debug_log, "Strerror returned; saving a copy: '");
3896 print_bytes_for_locale(errstr, errstr + strlen(errstr), 0);
3897 PerlIO_printf(Perl_debug_log, "'\n");
3908 =for apidoc sync_locale
3910 Changing the program's locale should be avoided by XS code. Nevertheless,
3911 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
3912 happens, Perl needs to be told that the locale has changed. Use this function
3913 to do so, before returning to Perl.
3919 Perl_sync_locale(pTHX)
3923 #ifdef USE_LOCALE_CTYPE
3925 newlocale = do_setlocale_c(LC_CTYPE, NULL);
3926 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3927 "%s:%d: %s\n", __FILE__, __LINE__,
3928 setlocale_debug_string(LC_CTYPE, NULL, newlocale)));
3929 new_ctype(newlocale);
3931 #endif /* USE_LOCALE_CTYPE */
3932 #ifdef USE_LOCALE_COLLATE
3934 newlocale = do_setlocale_c(LC_COLLATE, NULL);
3935 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3936 "%s:%d: %s\n", __FILE__, __LINE__,
3937 setlocale_debug_string(LC_COLLATE, NULL, newlocale)));
3938 new_collate(newlocale);
3941 #ifdef USE_LOCALE_NUMERIC
3943 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
3944 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3945 "%s:%d: %s\n", __FILE__, __LINE__,
3946 setlocale_debug_string(LC_NUMERIC, NULL, newlocale)));
3947 new_numeric(newlocale);
3949 #endif /* USE_LOCALE_NUMERIC */
3953 #if defined(DEBUGGING) && defined(USE_LOCALE)
3956 S_setlocale_debug_string(const int category, /* category number,
3958 const char* const locale, /* locale name */
3960 /* return value from setlocale() when attempting to
3961 * set 'category' to 'locale' */
3962 const char* const retval)
3964 /* Returns a pointer to a NUL-terminated string in static storage with
3965 * added text about the info passed in. This is not thread safe and will
3966 * be overwritten by the next call, so this should be used just to
3967 * formulate a string to immediately print or savepv() on. */
3969 /* initialise to a non-null value to keep it out of BSS and so keep
3970 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
3971 static char ret[128] = "If you can read this, thank your buggy C"
3972 " library strlcpy(), and change your hints file"
3975 my_strlcpy(ret, "setlocale(", sizeof(ret));
3976 my_strlcat(ret, category_name(category), sizeof(ret));
3977 my_strlcat(ret, ", ", sizeof(ret));
3980 my_strlcat(ret, "\"", sizeof(ret));
3981 my_strlcat(ret, locale, sizeof(ret));
3982 my_strlcat(ret, "\"", sizeof(ret));
3985 my_strlcat(ret, "NULL", sizeof(ret));
3988 my_strlcat(ret, ") returned ", sizeof(ret));
3991 my_strlcat(ret, "\"", sizeof(ret));
3992 my_strlcat(ret, retval, sizeof(ret));
3993 my_strlcat(ret, "\"", sizeof(ret));
3996 my_strlcat(ret, "NULL", sizeof(ret));
3999 assert(strlen(ret) < sizeof(ret));
4008 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob