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"
48 /* If the environment says to, we can output debugging information during
49 * initialization. This is done before option parsing, and before any thread
50 * creation, so can be a file-level static */
51 #if ! defined(DEBUGGING) || defined(PERL_GLOBAL_STRUCT)
52 # define debug_initialization 0
53 # define DEBUG_INITIALIZATION_set(v)
55 static bool debug_initialization = FALSE;
56 # define DEBUG_INITIALIZATION_set(v) (debug_initialization = v)
59 /* strlen() of a literal string constant. We might want this more general,
60 * but using it in just this file for now. A problem with more generality is
61 * the compiler warnings about comparing unlike signs */
62 #define STRLENs(s) (sizeof("" s "") - 1)
64 /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
65 * return of setlocale(), then this is extremely likely to be the C or POSIX
66 * locale. However, the output of setlocale() is documented to be opaque, but
67 * the odds are extremely small that it would return these two strings for some
68 * other locale. Note that VMS in these two locales includes many non-ASCII
69 * characters as controls and punctuation (below are hex bytes):
71 * punct: A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
72 * Oddly, none there are listed as alphas, though some represent alphabetics
73 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
74 #define isNAME_C_OR_POSIX(name) \
76 && (( *(name) == 'C' && (*(name + 1)) == '\0') \
77 || strEQ((name), "POSIX")))
81 /* This code keeps a LRU cache of the UTF-8ness of the locales it has so-far
82 * looked up. This is in the form of a C string: */
84 #define UTF8NESS_SEP "\v"
85 #define UTF8NESS_PREFIX "\f"
87 /* So, the string looks like:
89 * \vC\a0\vPOSIX\a0\vam_ET\a0\vaf_ZA.utf8\a1\ven_US.UTF-8\a1\0
91 * where the digit 0 after the \a indicates that the locale starting just
92 * after the preceding \v is not UTF-8, and the digit 1 mean it is. */
94 STATIC_ASSERT_DECL(STRLENs(UTF8NESS_SEP) == 1);
95 STATIC_ASSERT_DECL(STRLENs(UTF8NESS_PREFIX) == 1);
97 #define C_and_POSIX_utf8ness UTF8NESS_SEP "C" UTF8NESS_PREFIX "0" \
98 UTF8NESS_SEP "POSIX" UTF8NESS_PREFIX "0"
100 /* The cache is initialized to C_and_POSIX_utf8ness at start up. These are
101 * kept there always. The remining portion of the cache is LRU, with the
102 * oldest looked-up locale at the tail end */
105 S_stdize_locale(pTHX_ char *locs)
107 /* Standardize the locale name from a string returned by 'setlocale',
108 * possibly modifying that string.
110 * The typical return value of setlocale() is either
111 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
112 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
113 * (the space-separated values represent the various sublocales,
114 * in some unspecified order). This is not handled by this function.
116 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
117 * which is harmful for further use of the string in setlocale(). This
118 * function removes the trailing new line and everything up through the '='
121 const char * const s = strchr(locs, '=');
124 PERL_ARGS_ASSERT_STDIZE_LOCALE;
127 const char * const t = strchr(s, '.');
130 const char * const u = strchr(t, '\n');
131 if (u && (u[1] == 0)) {
132 const STRLEN len = u - s;
133 Move(s + 1, locs, len, char);
141 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
146 /* Two parallel arrays; first the locale categories Perl uses on this system;
147 * the second array is their names. These arrays are in mostly arbitrary
150 const int categories[] = {
152 # ifdef USE_LOCALE_NUMERIC
155 # ifdef USE_LOCALE_CTYPE
158 # ifdef USE_LOCALE_COLLATE
161 # ifdef USE_LOCALE_TIME
164 # ifdef USE_LOCALE_MESSAGES
167 # ifdef USE_LOCALE_MONETARY
170 # ifdef USE_LOCALE_ADDRESS
173 # ifdef USE_LOCALE_IDENTIFICATION
176 # ifdef USE_LOCALE_MEASUREMENT
179 # ifdef USE_LOCALE_PAPER
182 # ifdef USE_LOCALE_TELEPHONE
188 -1 /* Placeholder because C doesn't allow a
189 trailing comma, and it would get complicated
190 with all the #ifdef's */
193 /* The top-most real element is LC_ALL */
195 const char * category_names[] = {
197 # ifdef USE_LOCALE_NUMERIC
200 # ifdef USE_LOCALE_CTYPE
203 # ifdef USE_LOCALE_COLLATE
206 # ifdef USE_LOCALE_TIME
209 # ifdef USE_LOCALE_MESSAGES
212 # ifdef USE_LOCALE_MONETARY
215 # ifdef USE_LOCALE_ADDRESS
218 # ifdef USE_LOCALE_IDENTIFICATION
221 # ifdef USE_LOCALE_MEASUREMENT
224 # ifdef USE_LOCALE_PAPER
227 # ifdef USE_LOCALE_TELEPHONE
233 NULL /* Placeholder */
238 /* On systems with LC_ALL, it is kept in the highest index position. (-2
239 * to account for the final unused placeholder element.) */
240 # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 2)
244 /* On systems without LC_ALL, we pretend it is there, one beyond the real
245 * top element, hence in the unused placeholder element. */
246 # define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 1)
250 /* Pretending there is an LC_ALL element just above allows us to avoid most
251 * special cases. Most loops through these arrays in the code below are
252 * written like 'for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++)'. They will work
253 * on either type of system. But the code must be written to not access the
254 * element at 'LC_ALL_INDEX' except on platforms that have it. This can be
255 * checked for at compile time by using the #define LC_ALL_INDEX which is only
256 * defined if we do have LC_ALL. */
259 S_category_name(const int category)
265 if (category == LC_ALL) {
271 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
272 if (category == categories[i]) {
273 return category_names[i];
278 const char suffix[] = " (unknown)";
280 Size_t length = sizeof(suffix) + 1;
289 /* Calculate the number of digits */
295 Newx(unknown, length, char);
296 my_snprintf(unknown, length, "%d%s", category, suffix);
302 /* Now create LC_foo_INDEX #defines for just those categories on this system */
303 # ifdef USE_LOCALE_NUMERIC
304 # define LC_NUMERIC_INDEX 0
305 # define _DUMMY_NUMERIC LC_NUMERIC_INDEX
307 # define _DUMMY_NUMERIC -1
309 # ifdef USE_LOCALE_CTYPE
310 # define LC_CTYPE_INDEX _DUMMY_NUMERIC + 1
311 # define _DUMMY_CTYPE LC_CTYPE_INDEX
313 # define _DUMMY_CTYPE _DUMMY_NUMERIC
315 # ifdef USE_LOCALE_COLLATE
316 # define LC_COLLATE_INDEX _DUMMY_CTYPE + 1
317 # define _DUMMY_COLLATE LC_COLLATE_INDEX
319 # define _DUMMY_COLLATE _DUMMY_COLLATE
321 # ifdef USE_LOCALE_TIME
322 # define LC_TIME_INDEX _DUMMY_COLLATE + 1
323 # define _DUMMY_TIME LC_TIME_INDEX
325 # define _DUMMY_TIME _DUMMY_COLLATE
327 # ifdef USE_LOCALE_MESSAGES
328 # define LC_MESSAGES_INDEX _DUMMY_TIME + 1
329 # define _DUMMY_MESSAGES LC_MESSAGES_INDEX
331 # define _DUMMY_MESSAGES _DUMMY_TIME
333 # ifdef USE_LOCALE_MONETARY
334 # define LC_MONETARY_INDEX _DUMMY_MESSAGES + 1
335 # define _DUMMY_MONETARY LC_MONETARY_INDEX
337 # define _DUMMY_MONETARY _DUMMY_MESSAGES
339 # ifdef USE_LOCALE_ADDRESS
340 # define LC_ADDRESS_INDEX _DUMMY_MONETARY + 1
341 # define _DUMMY_ADDRESS LC_ADDRESS_INDEX
343 # define _DUMMY_ADDRESS _DUMMY_MONETARY
345 # ifdef USE_LOCALE_IDENTIFICATION
346 # define LC_IDENTIFICATION_INDEX _DUMMY_ADDRESS + 1
347 # define _DUMMY_IDENTIFICATION LC_IDENTIFICATION_INDEX
349 # define _DUMMY_IDENTIFICATION _DUMMY_ADDRESS
351 # ifdef USE_LOCALE_MEASUREMENT
352 # define LC_MEASUREMENT_INDEX _DUMMY_IDENTIFICATION + 1
353 # define _DUMMY_MEASUREMENT LC_MEASUREMENT_INDEX
355 # define _DUMMY_MEASUREMENT _DUMMY_IDENTIFICATION
357 # ifdef USE_LOCALE_PAPER
358 # define LC_PAPER_INDEX _DUMMY_MEASUREMENT + 1
359 # define _DUMMY_PAPER LC_PAPER_INDEX
361 # define _DUMMY_PAPER _DUMMY_MEASUREMENT
363 # ifdef USE_LOCALE_TELEPHONE
364 # define LC_TELEPHONE_INDEX _DUMMY_PAPER + 1
365 # define _DUMMY_TELEPHONE LC_TELEPHONE_INDEX
367 # define _DUMMY_TELEPHONE _DUMMY_PAPER
370 # define LC_ALL_INDEX _DUMMY_TELEPHONE + 1
372 #endif /* ifdef USE_LOCALE */
374 /* Windows requres a customized base-level setlocale() */
376 # define my_setlocale(cat, locale) win32_setlocale(cat, locale)
378 # define my_setlocale(cat, locale) setlocale(cat, locale)
381 /* Just placeholders for now. "_c" is intended to be called when the category
382 * is a constant known at compile time; "_r", not known until run time */
383 # define do_setlocale_c(category, locale) my_setlocale(category, locale)
384 # define do_setlocale_r(category, locale) my_setlocale(category, locale)
387 S_set_numeric_radix(pTHX_ const bool use_locale)
389 /* If 'use_locale' is FALSE, set to use a dot for the radix character. If
390 * TRUE, use the radix character derived from the current locale */
392 #if defined(USE_LOCALE_NUMERIC) && ( defined(HAS_LOCALECONV) \
393 || defined(HAS_NL_LANGINFO))
395 const char * radix = (use_locale)
396 ? my_nl_langinfo(PERL_RADIXCHAR, FALSE)
397 /* FALSE => already in dest locale */
400 sv_setpv(PL_numeric_radix_sv, radix);
402 /* If this is valid UTF-8 that isn't totally ASCII, and we are in
403 * a UTF-8 locale, then mark the radix as being in UTF-8 */
404 if (is_utf8_non_invariant_string((U8 *) SvPVX(PL_numeric_radix_sv),
405 SvCUR(PL_numeric_radix_sv))
406 && _is_cur_LC_category_utf8(LC_NUMERIC))
408 SvUTF8_on(PL_numeric_radix_sv);
413 if (DEBUG_L_TEST || debug_initialization) {
414 PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
415 SvPVX(PL_numeric_radix_sv),
416 cBOOL(SvUTF8(PL_numeric_radix_sv)));
420 #endif /* USE_LOCALE_NUMERIC and can find the radix char */
426 Perl_new_numeric(pTHX_ const char *newnum)
429 #ifndef USE_LOCALE_NUMERIC
431 PERL_UNUSED_ARG(newnum);
435 /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell
436 * core Perl this and that 'newnum' is the name of the new locale.
437 * It installs this locale as the current underlying default.
439 * The default locale and the C locale can be toggled between by use of the
440 * set_numeric_underlying() and set_numeric_standard() functions, which
441 * should probably not be called directly, but only via macros like
442 * SET_NUMERIC_STANDARD() in perl.h.
444 * The toggling is necessary mainly so that a non-dot radix decimal point
445 * character can be output, while allowing internal calculations to use a
448 * This sets several interpreter-level variables:
449 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
450 * PL_numeric_underlying A boolean indicating if the toggled state is such
451 * that the current locale is the program's underlying
453 * PL_numeric_standard An int indicating if the toggled state is such
454 * that the current locale is the C locale or
455 * indistinguishable from the C locale. If non-zero, it
456 * is in C; if > 1, it means it may not be toggled away
458 * PL_numeric_underlying_is_standard A bool kept by this function
459 * indicating that the underlying locale and the standard
460 * C locale are indistinguishable for the purposes of
461 * LC_NUMERIC. This happens when both of the above two
462 * variables are true at the same time. (Toggling is a
463 * no-op under these circumstances.) This variable is
464 * used to avoid having to recalculate.
465 * Any code changing the locale (outside this file) should use
466 * POSIX::setlocale, which calls this function. Therefore this function
467 * should be called directly only from this file and from
468 * POSIX::setlocale() */
473 Safefree(PL_numeric_name);
474 PL_numeric_name = NULL;
475 PL_numeric_standard = TRUE;
476 PL_numeric_underlying = TRUE;
477 PL_numeric_underlying_is_standard = TRUE;
481 save_newnum = stdize_locale(savepv(newnum));
482 PL_numeric_underlying = TRUE;
483 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
485 /* If its name isn't C nor POSIX, it could still be indistinguishable from
487 if (! PL_numeric_standard) {
488 PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(PERL_RADIXCHAR,
489 FALSE /* Don't toggle locale */ ))
490 && strEQ("", my_nl_langinfo(PERL_THOUSEP,
494 /* Save the new name if it isn't the same as the previous one, if any */
495 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
496 Safefree(PL_numeric_name);
497 PL_numeric_name = save_newnum;
500 Safefree(save_newnum);
503 PL_numeric_underlying_is_standard = PL_numeric_standard;
505 # if defined(HAS_NEWLOCALE) && ! defined(NO_POSIX_2008_LOCALE)
507 PL_underlying_numeric_obj = newlocale(LC_NUMERIC_MASK,
509 PL_underlying_numeric_obj);
513 if (DEBUG_L_TEST || debug_initialization) {
514 PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name);
517 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
518 * have to worry about the radix being a non-dot. (Core operations that
519 * need the underlying locale change to it temporarily). */
520 set_numeric_standard();
522 #endif /* USE_LOCALE_NUMERIC */
527 Perl_set_numeric_standard(pTHX)
530 #ifdef USE_LOCALE_NUMERIC
532 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
533 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
534 * macro avoids calling this routine if toggling isn't necessary according
535 * to our records (which could be wrong if some XS code has changed the
536 * locale behind our back) */
538 do_setlocale_c(LC_NUMERIC, "C");
539 PL_numeric_standard = TRUE;
540 PL_numeric_underlying = PL_numeric_underlying_is_standard;
541 set_numeric_radix(0);
545 if (DEBUG_L_TEST || debug_initialization) {
546 PerlIO_printf(Perl_debug_log,
547 "LC_NUMERIC locale now is standard C\n");
551 #endif /* USE_LOCALE_NUMERIC */
556 Perl_set_numeric_underlying(pTHX)
559 #ifdef USE_LOCALE_NUMERIC
561 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
562 * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h
563 * instead of calling this directly. The macro avoids calling this routine
564 * if toggling isn't necessary according to our records (which could be
565 * wrong if some XS code has changed the locale behind our back) */
567 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
568 PL_numeric_standard = PL_numeric_underlying_is_standard;
569 PL_numeric_underlying = TRUE;
570 set_numeric_radix(! PL_numeric_standard);
574 if (DEBUG_L_TEST || debug_initialization) {
575 PerlIO_printf(Perl_debug_log,
576 "LC_NUMERIC locale now is %s\n",
581 #endif /* USE_LOCALE_NUMERIC */
586 * Set up for a new ctype locale.
589 S_new_ctype(pTHX_ const char *newctype)
592 #ifndef USE_LOCALE_CTYPE
594 PERL_ARGS_ASSERT_NEW_CTYPE;
595 PERL_UNUSED_ARG(newctype);
600 /* Called after each libc setlocale() call affecting LC_CTYPE, to tell
601 * core Perl this and that 'newctype' is the name of the new locale.
603 * This function sets up the folding arrays for all 256 bytes, assuming
604 * that tofold() is tolc() since fold case is not a concept in POSIX,
606 * Any code changing the locale (outside this file) should use
607 * POSIX::setlocale, which calls this function. Therefore this function
608 * should be called directly only from this file and from
609 * POSIX::setlocale() */
614 PERL_ARGS_ASSERT_NEW_CTYPE;
616 /* We will replace any bad locale warning with 1) nothing if the new one is
617 * ok; or 2) a new warning for the bad new locale */
618 if (PL_warn_locale) {
619 SvREFCNT_dec_NN(PL_warn_locale);
620 PL_warn_locale = NULL;
623 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
625 /* A UTF-8 locale gets standard rules. But note that code still has to
626 * handle this specially because of the three problematic code points */
627 if (PL_in_utf8_CTYPE_locale) {
628 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
631 /* Assume enough space for every character being bad. 4 spaces each
632 * for the 94 printable characters that are output like "'x' "; and 5
633 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
635 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
637 /* Don't check for problems if we are suppressing the warnings */
638 bool check_for_problems = ckWARN_d(WARN_LOCALE)
639 || UNLIKELY(DEBUG_L_TEST);
640 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
642 unsigned int bad_count = 0; /* Count of bad characters */
644 for (i = 0; i < 256; i++) {
646 PL_fold_locale[i] = (U8) tolower(i);
648 PL_fold_locale[i] = (U8) toupper(i);
650 PL_fold_locale[i] = (U8) i;
652 /* If checking for locale problems, see if the native ASCII-range
653 * printables plus \n and \t are in their expected categories in
654 * the new locale. If not, this could mean big trouble, upending
655 * Perl's and most programs' assumptions, like having a
656 * metacharacter with special meaning become a \w. Fortunately,
657 * it's very rare to find locales that aren't supersets of ASCII
658 * nowadays. It isn't a problem for most controls to be changed
659 * into something else; we check only \n and \t, though perhaps \r
660 * could be an issue as well. */
661 if ( check_for_problems
662 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
664 if ( cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC(i))
665 || cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i))
666 || cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i))
667 || cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i))
668 || cBOOL(islower(i)) != cBOOL(isLOWER_A(i))
669 || cBOOL(isprint(i)) != cBOOL(isPRINT_A(i))
670 || cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i))
671 || cBOOL(isspace(i)) != cBOOL(isSPACE_A(i))
672 || cBOOL(isupper(i)) != cBOOL(isUPPER_A(i))
673 || cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i))
674 || tolower(i) != (int) toLOWER_A(i)
675 || toupper(i) != (int) toUPPER_A(i)
676 || (i == '\n' && ! isCNTRL_LC(i)))
678 if (bad_count) { /* Separate multiple entries with a
680 bad_chars_list[bad_count++] = ' ';
682 bad_chars_list[bad_count++] = '\'';
684 bad_chars_list[bad_count++] = (char) i;
687 bad_chars_list[bad_count++] = '\\';
689 bad_chars_list[bad_count++] = 'n';
693 bad_chars_list[bad_count++] = 't';
696 bad_chars_list[bad_count++] = '\'';
697 bad_chars_list[bad_count] = '\0';
704 /* We only handle single-byte locales (outside of UTF-8 ones; so if
705 * this locale requires more than one byte, there are going to be
707 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
708 "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n",
709 __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX));
711 if (check_for_problems && MB_CUR_MAX > 1
713 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
714 * locale. Just assume that the implementation for them (plus
715 * for POSIX) is correct and the > 1 value is spurious. (Since
716 * these are specially handled to never be considered UTF-8
717 * locales, as long as this is the only problem, everything
718 * should work fine */
719 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
721 multi_byte_locale = TRUE;
726 if (bad_count || multi_byte_locale) {
727 PL_warn_locale = Perl_newSVpvf(aTHX_
728 "Locale '%s' may not work well.%s%s%s\n",
731 ? " Some characters in it are not recognized by"
735 ? "\nThe following characters (and maybe others)"
736 " may not have the same meaning as the Perl"
737 " program expects:\n"
743 /* If we are actually in the scope of the locale or are debugging,
744 * output the message now. If not in that scope, we save the
745 * message to be output at the first operation using this locale,
746 * if that actually happens. Most programs don't use locales, so
747 * they are immune to bad ones. */
748 if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) {
750 /* We have to save 'newctype' because the setlocale() just
751 * below may destroy it. The next setlocale() further down
752 * should restore it properly so that the intermediate change
753 * here is transparent to this function's caller */
754 const char * const badlocale = savepv(newctype);
756 do_setlocale_c(LC_CTYPE, "C");
758 /* The '0' below suppresses a bogus gcc compiler warning */
759 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
761 do_setlocale_c(LC_CTYPE, badlocale);
764 if (IN_LC(LC_CTYPE)) {
765 SvREFCNT_dec_NN(PL_warn_locale);
766 PL_warn_locale = NULL;
772 #endif /* USE_LOCALE_CTYPE */
777 Perl__warn_problematic_locale()
780 #ifdef USE_LOCALE_CTYPE
784 /* Internal-to-core function that outputs the message in PL_warn_locale,
785 * and then NULLS it. Should be called only through the macro
786 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
788 if (PL_warn_locale) {
789 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
790 SvPVX(PL_warn_locale),
791 0 /* dummy to avoid compiler warning */ );
792 SvREFCNT_dec_NN(PL_warn_locale);
793 PL_warn_locale = NULL;
801 S_new_collate(pTHX_ const char *newcoll)
804 #ifndef USE_LOCALE_COLLATE
806 PERL_UNUSED_ARG(newcoll);
811 /* Called after each libc setlocale() call affecting LC_COLLATE, to tell
812 * core Perl this and that 'newcoll' is the name of the new locale.
814 * The design of locale collation is that every locale change is given an
815 * index 'PL_collation_ix'. The first time a string particpates in an
816 * operation that requires collation while locale collation is active, it
817 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
818 * magic includes the collation index, and the transformation of the string
819 * by strxfrm(), q.v. That transformation is used when doing comparisons,
820 * instead of the string itself. If a string changes, the magic is
821 * cleared. The next time the locale changes, the index is incremented,
822 * and so we know during a comparison that the transformation is not
823 * necessarily still valid, and so is recomputed. Note that if the locale
824 * changes enough times, the index could wrap (a U32), and it is possible
825 * that a transformation would improperly be considered valid, leading to
829 if (PL_collation_name) {
831 Safefree(PL_collation_name);
832 PL_collation_name = NULL;
834 PL_collation_standard = TRUE;
835 is_standard_collation:
836 PL_collxfrm_base = 0;
837 PL_collxfrm_mult = 2;
838 PL_in_utf8_COLLATE_locale = FALSE;
839 PL_strxfrm_NUL_replacement = '\0';
840 PL_strxfrm_max_cp = 0;
844 /* If this is not the same locale as currently, set the new one up */
845 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
847 Safefree(PL_collation_name);
848 PL_collation_name = stdize_locale(savepv(newcoll));
849 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
850 if (PL_collation_standard) {
851 goto is_standard_collation;
854 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
855 PL_strxfrm_NUL_replacement = '\0';
856 PL_strxfrm_max_cp = 0;
858 /* A locale collation definition includes primary, secondary, tertiary,
859 * etc. weights for each character. To sort, the primary weights are
860 * used, and only if they compare equal, then the secondary weights are
861 * used, and only if they compare equal, then the tertiary, etc.
863 * strxfrm() works by taking the input string, say ABC, and creating an
864 * output transformed string consisting of first the primary weights,
865 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
866 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
867 * may not have weights at every level. In our example, let's say B
868 * doesn't have a tertiary weight, and A doesn't have a secondary
869 * weight. The constructed string is then going to be
870 * A¹B¹C¹ B²C² A³C³ ....
871 * This has the desired effect that strcmp() will look at the secondary
872 * or tertiary weights only if the strings compare equal at all higher
873 * priority weights. The spaces shown here, like in
875 * are not just for readability. In the general case, these must
876 * actually be bytes, which we will call here 'separator weights'; and
877 * they must be smaller than any other weight value, but since these
878 * are C strings, only the terminating one can be a NUL (some
879 * implementations may include a non-NUL separator weight just before
880 * the NUL). Implementations tend to reserve 01 for the separator
881 * weights. They are needed so that a shorter string's secondary
882 * weights won't be misconstrued as primary weights of a longer string,
883 * etc. By making them smaller than any other weight, the shorter
884 * string will sort first. (Actually, if all secondary weights are
885 * smaller than all primary ones, there is no need for a separator
886 * weight between those two levels, etc.)
888 * The length of the transformed string is roughly a linear function of
889 * the input string. It's not exactly linear because some characters
890 * don't have weights at all levels. When we call strxfrm() we have to
891 * allocate some memory to hold the transformed string. The
892 * calculations below try to find coefficients 'm' and 'b' for this
893 * locale so that m*x + b equals how much space we need, given the size
894 * of the input string in 'x'. If we calculate too small, we increase
895 * the size as needed, and call strxfrm() again, but it is better to
896 * get it right the first time to avoid wasted expensive string
897 * transformations. */
900 /* We use the string below to find how long the tranformation of it
901 * is. Almost all locales are supersets of ASCII, or at least the
902 * ASCII letters. We use all of them, half upper half lower,
903 * because if we used fewer, we might hit just the ones that are
904 * outliers in a particular locale. Most of the strings being
905 * collated will contain a preponderance of letters, and even if
906 * they are above-ASCII, they are likely to have the same number of
907 * weight levels as the ASCII ones. It turns out that digits tend
908 * to have fewer levels, and some punctuation has more, but those
909 * are relatively sparse in text, and khw believes this gives a
910 * reasonable result, but it could be changed if experience so
912 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
913 char * x_longer; /* Transformed 'longer' */
914 Size_t x_len_longer; /* Length of 'x_longer' */
916 char * x_shorter; /* We also transform a substring of 'longer' */
917 Size_t x_len_shorter;
919 /* _mem_collxfrm() is used get the transformation (though here we
920 * are interested only in its length). It is used because it has
921 * the intelligence to handle all cases, but to work, it needs some
922 * values of 'm' and 'b' to get it started. For the purposes of
923 * this calculation we use a very conservative estimate of 'm' and
924 * 'b'. This assumes a weight can be multiple bytes, enough to
925 * hold any UV on the platform, and there are 5 levels, 4 weight
926 * bytes, and a trailing NUL. */
927 PL_collxfrm_base = 5;
928 PL_collxfrm_mult = 5 * sizeof(UV);
930 /* Find out how long the transformation really is */
931 x_longer = _mem_collxfrm(longer,
935 /* We avoid converting to UTF-8 in the
936 * called function by telling it the
937 * string is in UTF-8 if the locale is a
938 * UTF-8 one. Since the string passed
939 * here is invariant under UTF-8, we can
940 * claim it's UTF-8 even though it isn't.
942 PL_in_utf8_COLLATE_locale);
945 /* Find out how long the transformation of a substring of 'longer'
946 * is. Together the lengths of these transformations are
947 * sufficient to calculate 'm' and 'b'. The substring is all of
948 * 'longer' except the first character. This minimizes the chances
949 * of being swayed by outliers */
950 x_shorter = _mem_collxfrm(longer + 1,
953 PL_in_utf8_COLLATE_locale);
956 /* If the results are nonsensical for this simple test, the whole
957 * locale definition is suspect. Mark it so that locale collation
958 * is not active at all for it. XXX Should we warn? */
959 if ( x_len_shorter == 0
961 || x_len_shorter >= x_len_longer)
963 PL_collxfrm_mult = 0;
964 PL_collxfrm_base = 0;
967 SSize_t base; /* Temporary */
969 /* We have both: m * strlen(longer) + b = x_len_longer
970 * m * strlen(shorter) + b = x_len_shorter;
971 * subtracting yields:
972 * m * (strlen(longer) - strlen(shorter))
973 * = x_len_longer - x_len_shorter
974 * But we have set things up so that 'shorter' is 1 byte smaller
975 * than 'longer'. Hence:
976 * m = x_len_longer - x_len_shorter
978 * But if something went wrong, make sure the multiplier is at
981 if (x_len_longer > x_len_shorter) {
982 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
985 PL_collxfrm_mult = 1;
990 * but in case something has gone wrong, make sure it is
992 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
997 /* Add 1 for the trailing NUL */
998 PL_collxfrm_base = base + 1;
1003 if (DEBUG_L_TEST || debug_initialization) {
1004 PerlIO_printf(Perl_debug_log,
1005 "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
1007 " collate multipler=%zu, collate base=%zu\n",
1009 PL_in_utf8_COLLATE_locale,
1010 x_len_shorter, x_len_longer,
1011 PL_collxfrm_mult, PL_collxfrm_base);
1018 #endif /* USE_LOCALE_COLLATE */
1025 S_win32_setlocale(pTHX_ int category, const char* locale)
1027 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
1028 * difference between the two unless the input locale is "", which normally
1029 * means on Windows to get the machine default, which is set via the
1030 * computer's "Regional and Language Options" (or its current equivalent).
1031 * In POSIX, it instead means to find the locale from the user's
1032 * environment. This routine changes the Windows behavior to first look in
1033 * the environment, and, if anything is found, use that instead of going to
1034 * the machine default. If there is no environment override, the machine
1035 * default is used, by calling the real setlocale() with "".
1037 * The POSIX behavior is to use the LC_ALL variable if set; otherwise to
1038 * use the particular category's variable if set; otherwise to use the LANG
1041 bool override_LC_ALL = FALSE;
1045 if (locale && strEQ(locale, "")) {
1049 locale = PerlEnv_getenv("LC_ALL");
1051 if (category == LC_ALL) {
1052 override_LC_ALL = TRUE;
1058 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
1059 if (category == categories[i]) {
1060 locale = PerlEnv_getenv(category_names[i]);
1065 locale = PerlEnv_getenv("LANG");
1081 result = setlocale(category, locale);
1082 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
1083 setlocale_debug_string(category, locale, result)));
1085 if (! override_LC_ALL) {
1089 /* Here the input category was LC_ALL, and we have set it to what is in the
1090 * LANG variable or the system default if there is no LANG. But these have
1091 * lower priority than the other LC_foo variables, so override it for each
1092 * one that is set. (If they are set to "", it means to use the same thing
1093 * we just set LC_ALL to, so can skip) */
1095 for (i = 0; i < LC_ALL_INDEX; i++) {
1096 result = PerlEnv_getenv(category_names[i]);
1097 if (result && strNE(result, "")) {
1098 setlocale(categories[i], result);
1099 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1101 setlocale_debug_string(categories[i], result, "not captured")));
1105 result = setlocale(LC_ALL, NULL);
1106 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1108 setlocale_debug_string(LC_ALL, NULL, result)));
1116 Perl_setlocale(int category, const char * locale)
1118 /* This wraps POSIX::setlocale() */
1124 #ifdef USE_LOCALE_NUMERIC
1126 /* A NULL locale means only query what the current one is. We have the
1127 * LC_NUMERIC name saved, because we are normally switched into the C
1128 * locale for it. For an LC_ALL query, switch back to get the correct
1129 * results. All other categories don't require special handling */
1130 if (locale == NULL) {
1131 if (category == LC_NUMERIC) {
1132 return savepv(PL_numeric_name);
1137 else if (category == LC_ALL && ! PL_numeric_underlying) {
1139 SET_NUMERIC_UNDERLYING();
1148 /* Save retval since subsequent setlocale() calls may overwrite it. */
1149 retval = savepv(do_setlocale_r(category, locale));
1151 DEBUG_L(PerlIO_printf(Perl_debug_log,
1152 "%s:%d: %s\n", __FILE__, __LINE__,
1153 setlocale_debug_string(category, locale, retval)));
1155 /* Should never happen that a query would return an error, but be
1156 * sure and reset to C locale */
1158 SET_NUMERIC_STANDARD();
1164 /* If locale == NULL, we are just querying the state, but may have switched
1165 * to NUMERIC_UNDERLYING. Switch back before returning. */
1166 if (locale == NULL) {
1167 SET_NUMERIC_STANDARD();
1171 /* Now that have switched locales, we have to update our records to
1176 #ifdef USE_LOCALE_CTYPE
1183 #ifdef USE_LOCALE_COLLATE
1186 new_collate(retval);
1190 #ifdef USE_LOCALE_NUMERIC
1193 new_numeric(retval);
1201 /* LC_ALL updates all the things we care about. The values may not
1202 * be the same as 'retval', as the locale "" may have set things
1205 # ifdef USE_LOCALE_CTYPE
1207 newlocale = do_setlocale_c(LC_CTYPE, NULL);
1208 new_ctype(newlocale);
1210 # endif /* USE_LOCALE_CTYPE */
1211 # ifdef USE_LOCALE_COLLATE
1213 newlocale = do_setlocale_c(LC_COLLATE, NULL);
1214 new_collate(newlocale);
1217 # ifdef USE_LOCALE_NUMERIC
1219 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
1220 new_numeric(newlocale);
1222 # endif /* USE_LOCALE_NUMERIC */
1234 PERL_STATIC_INLINE const char *
1235 S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset)
1237 /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size',
1238 * growing it if necessary */
1240 const Size_t string_size = strlen(string) + offset + 1;
1242 PERL_ARGS_ASSERT_SAVE_TO_BUFFER;
1244 if (*buf_size == 0) {
1245 Newx(*buf, string_size, char);
1246 *buf_size = string_size;
1248 else if (string_size > *buf_size) {
1249 Renew(*buf, string_size, char);
1250 *buf_size = string_size;
1253 Copy(string, *buf + offset, string_size - offset, char);
1259 =head1 Locale-related functions and macros
1261 =for apidoc Perl_langinfo
1263 This is an (almost ª) drop-in replacement for the system C<L<nl_langinfo(3)>>,
1264 taking the same C<item> parameter values, and returning the same information.
1265 But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks
1266 of Perl's locale handling from your code, and can be used on systems that lack
1267 a native C<nl_langinfo>.
1275 It delivers the correct results for the C<RADIXCHAR> and C<THOUSESEP> items,
1276 without you having to write extra code. The reason for the extra code would be
1277 because these are from the C<LC_NUMERIC> locale category, which is normally
1278 kept set to the C locale by Perl, no matter what the underlying locale is
1279 supposed to be, and so to get the expected results, you have to temporarily
1280 toggle into the underlying locale, and later toggle back. (You could use
1281 plain C<nl_langinfo> and C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this
1282 but then you wouldn't get the other advantages of C<Perl_langinfo()>; not
1283 keeping C<LC_NUMERIC> in the C locale would break a lot of CPAN, which is
1284 expecting the radix (decimal point) character to be a dot.)
1288 Depending on C<item>, it works on systems that don't have C<nl_langinfo>, hence
1289 makes your code more portable. Of the fifty-some possible items specified by
1290 the POSIX 2008 standard,
1291 L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>,
1292 only two are completely unimplemented. It uses various techniques to recover
1293 the other items, including calling C<L<localeconv(3)>>, and C<L<strftime(3)>>,
1294 both of which are specified in C89, so should be always be available. Later
1295 C<strftime()> versions have additional capabilities; C<""> is returned for
1296 those not available on your system.
1298 The details for those items which may differ from what this emulation returns
1299 and what a native C<nl_langinfo()> would return are:
1307 Unimplemented, so returns C<"">.
1317 Only the values for English are returned. C<YESSTR> and C<NOSTR> have been
1318 removed from POSIX 2008, and are retained for backwards compatibility. Your
1319 platform's C<nl_langinfo> may not support them.
1323 Always evaluates to C<%x>, the locale's appropriate date representation.
1327 Always evaluates to C<%X>, the locale's appropriate time representation.
1331 Always evaluates to C<%c>, the locale's appropriate date and time
1336 The return may be incorrect for those rare locales where the currency symbol
1337 replaces the radix character.
1338 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1339 to work differently.
1343 Currently this gives the same results as Linux does.
1344 Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1345 to work differently.
1351 =item C<ERA_D_T_FMT>
1355 These are derived by using C<strftime()>, and not all versions of that function
1356 know about them. C<""> is returned for these on such systems.
1360 When using C<Perl_langinfo> on systems that don't have a native
1361 C<nl_langinfo()>, you must
1363 #include "perl_langinfo.h"
1365 before the C<perl.h> C<#include>. You can replace your C<langinfo.h>
1366 C<#include> with this one. (Doing it this way keeps out the symbols that plain
1367 C<langinfo.h> imports into the namespace for code that doesn't need it.)
1369 You also should not use the bare C<langinfo.h> item names, but should preface
1370 them with C<PERL_>, so use C<PERL_RADIXCHAR> instead of plain C<RADIXCHAR>.
1371 The C<PERL_I<foo>> versions will also work for this function on systems that do
1372 have a native C<nl_langinfo>.
1376 It is thread-friendly, returning its result in a buffer that won't be
1377 overwritten by another thread, so you don't have to code for that possibility.
1378 The buffer can be overwritten by the next call to C<nl_langinfo> or
1379 C<Perl_langinfo> in the same thread.
1383 ª It returns S<C<const char *>>, whereas plain C<nl_langinfo()> returns S<C<char
1384 *>>, but you are (only by documentation) forbidden to write into the buffer.
1385 By declaring this C<const>, the compiler enforces this restriction. The extra
1386 C<const> is why this isn't an unequivocal drop-in replacement for
1391 The original impetus for C<Perl_langinfo()> was so that code that needs to
1392 find out the current currency symbol, floating point radix character, or digit
1393 grouping separator can use, on all systems, the simpler and more
1394 thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a
1395 pain to make thread-friendly. For other fields returned by C<localeconv>, it
1396 is better to use the methods given in L<perlcall> to call
1397 L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly.
1404 #ifdef HAS_NL_LANGINFO
1405 Perl_langinfo(const nl_item item)
1407 Perl_langinfo(const int item)
1410 return my_nl_langinfo(item, TRUE);
1414 #ifdef HAS_NL_LANGINFO
1415 S_my_nl_langinfo(const nl_item item, bool toggle)
1417 S_my_nl_langinfo(const int item, bool toggle)
1422 /* We only need to toggle into the underlying LC_NUMERIC locale for these
1423 * two items, and only if not already there */
1424 if (toggle && (( item != PERL_RADIXCHAR && item != PERL_THOUSEP)
1425 || PL_numeric_underlying))
1430 #if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */
1431 #if ! defined(HAS_POSIX_2008_LOCALE)
1433 /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC
1434 * for those items dependent on it. This must be copied to a buffer before
1435 * switching back, as some systems destroy the buffer when setlocale() is
1439 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
1442 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
1445 LOCALE_LOCK; /* Prevent interference from another thread executing
1446 this code section (the only call to nl_langinfo in
1449 save_to_buffer(nl_langinfo(item), &PL_langinfo_buf,
1450 &PL_langinfo_bufsize, 0);
1455 RESTORE_LC_NUMERIC();
1459 # else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */
1462 bool do_free = FALSE;
1463 locale_t cur = uselocale((locale_t) 0);
1465 if (cur == LC_GLOBAL_LOCALE) {
1466 cur = duplocale(LC_GLOBAL_LOCALE);
1471 if (PL_underlying_numeric_obj) {
1472 cur = PL_underlying_numeric_obj;
1475 cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur);
1480 save_to_buffer(nl_langinfo_l(item, cur),
1481 &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
1489 if (strEQ(PL_langinfo_buf, "")) {
1490 if (item == PERL_YESSTR) {
1493 if (item == PERL_NOSTR) {
1498 return PL_langinfo_buf;
1500 #else /* Below, emulate nl_langinfo as best we can */
1504 # ifdef HAS_LOCALECONV
1506 const struct lconv* lc;
1507 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
1510 # ifdef HAS_STRFTIME
1513 bool return_format = FALSE; /* Return the %format, not the value */
1514 const char * format;
1518 /* We copy the results to a per-thread buffer, even if not
1519 * multi-threaded. This is in part to simplify this code, and partly
1520 * because we need a buffer anyway for strftime(), and partly because a
1521 * call of localeconv() could otherwise wipe out the buffer, and the
1522 * programmer would not be expecting this, as this is a nl_langinfo()
1523 * substitute after all, so s/he might be thinking their localeconv()
1524 * is safe until another localeconv() call. */
1528 const char * retval;
1530 /* These 2 are unimplemented */
1532 case PERL_ERA: /* For use with strftime() %E modifier */
1537 /* We use only an English set, since we don't know any more */
1538 case PERL_YESEXPR: return "^[+1yY]";
1539 case PERL_YESSTR: return "yes";
1540 case PERL_NOEXPR: return "^[-0nN]";
1541 case PERL_NOSTR: return "no";
1543 # ifdef HAS_LOCALECONV
1547 /* We don't bother with localeconv_l() because any system that
1548 * has it is likely to also have nl_langinfo() */
1550 LOCALE_LOCK; /* Prevent interference with other threads
1551 using localeconv() */
1555 || ! lc->currency_symbol
1556 || strEQ("", lc->currency_symbol))
1562 /* Leave the first spot empty to be filled in below */
1563 save_to_buffer(lc->currency_symbol, &PL_langinfo_buf,
1564 &PL_langinfo_bufsize, 1);
1565 if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, ""))
1566 { /* khw couldn't figure out how the localedef specifications
1567 would show that the $ should replace the radix; this is
1568 just a guess as to how it might work.*/
1569 *PL_langinfo_buf = '.';
1571 else if (lc->p_cs_precedes) {
1572 *PL_langinfo_buf = '-';
1575 *PL_langinfo_buf = '+';
1581 case PERL_RADIXCHAR:
1585 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
1588 LOCALE_LOCK; /* Prevent interference with other threads
1589 using localeconv() */
1596 retval = (item == PERL_RADIXCHAR)
1598 : lc->thousands_sep;
1604 save_to_buffer(retval, &PL_langinfo_buf,
1605 &PL_langinfo_bufsize, 0);
1610 RESTORE_LC_NUMERIC();
1616 # ifdef HAS_STRFTIME
1618 /* These are defined by C89, so we assume that strftime supports
1619 * them, and so are returned unconditionally; they may not be what
1620 * the locale actually says, but should give good enough results
1621 * for someone using them as formats (as opposed to trying to parse
1622 * them to figure out what the locale says). The other format
1623 * items are actually tested to verify they work on the platform */
1624 case PERL_D_FMT: return "%x";
1625 case PERL_T_FMT: return "%X";
1626 case PERL_D_T_FMT: return "%c";
1628 /* These formats are only available in later strfmtime's */
1629 case PERL_ERA_D_FMT: case PERL_ERA_T_FMT: case PERL_ERA_D_T_FMT:
1630 case PERL_T_FMT_AMPM:
1632 /* The rest can be gotten from most versions of strftime(). */
1633 case PERL_ABDAY_1: case PERL_ABDAY_2: case PERL_ABDAY_3:
1634 case PERL_ABDAY_4: case PERL_ABDAY_5: case PERL_ABDAY_6:
1636 case PERL_ALT_DIGITS:
1637 case PERL_AM_STR: case PERL_PM_STR:
1638 case PERL_ABMON_1: case PERL_ABMON_2: case PERL_ABMON_3:
1639 case PERL_ABMON_4: case PERL_ABMON_5: case PERL_ABMON_6:
1640 case PERL_ABMON_7: case PERL_ABMON_8: case PERL_ABMON_9:
1641 case PERL_ABMON_10: case PERL_ABMON_11: case PERL_ABMON_12:
1642 case PERL_DAY_1: case PERL_DAY_2: case PERL_DAY_3: case PERL_DAY_4:
1643 case PERL_DAY_5: case PERL_DAY_6: case PERL_DAY_7:
1644 case PERL_MON_1: case PERL_MON_2: case PERL_MON_3: case PERL_MON_4:
1645 case PERL_MON_5: case PERL_MON_6: case PERL_MON_7: case PERL_MON_8:
1646 case PERL_MON_9: case PERL_MON_10: case PERL_MON_11:
1651 init_tm(&tm); /* Precaution against core dumps */
1655 tm.tm_year = 2017 - 1900;
1662 "panic: %s: %d: switch case: %d problem",
1663 __FILE__, __LINE__, item);
1664 NOT_REACHED; /* NOTREACHED */
1666 case PERL_PM_STR: tm.tm_hour = 18;
1671 case PERL_ABDAY_7: tm.tm_wday++;
1672 case PERL_ABDAY_6: tm.tm_wday++;
1673 case PERL_ABDAY_5: tm.tm_wday++;
1674 case PERL_ABDAY_4: tm.tm_wday++;
1675 case PERL_ABDAY_3: tm.tm_wday++;
1676 case PERL_ABDAY_2: tm.tm_wday++;
1681 case PERL_DAY_7: tm.tm_wday++;
1682 case PERL_DAY_6: tm.tm_wday++;
1683 case PERL_DAY_5: tm.tm_wday++;
1684 case PERL_DAY_4: tm.tm_wday++;
1685 case PERL_DAY_3: tm.tm_wday++;
1686 case PERL_DAY_2: tm.tm_wday++;
1691 case PERL_ABMON_12: tm.tm_mon++;
1692 case PERL_ABMON_11: tm.tm_mon++;
1693 case PERL_ABMON_10: tm.tm_mon++;
1694 case PERL_ABMON_9: tm.tm_mon++;
1695 case PERL_ABMON_8: tm.tm_mon++;
1696 case PERL_ABMON_7: tm.tm_mon++;
1697 case PERL_ABMON_6: tm.tm_mon++;
1698 case PERL_ABMON_5: tm.tm_mon++;
1699 case PERL_ABMON_4: tm.tm_mon++;
1700 case PERL_ABMON_3: tm.tm_mon++;
1701 case PERL_ABMON_2: tm.tm_mon++;
1706 case PERL_MON_12: tm.tm_mon++;
1707 case PERL_MON_11: tm.tm_mon++;
1708 case PERL_MON_10: tm.tm_mon++;
1709 case PERL_MON_9: tm.tm_mon++;
1710 case PERL_MON_8: tm.tm_mon++;
1711 case PERL_MON_7: tm.tm_mon++;
1712 case PERL_MON_6: tm.tm_mon++;
1713 case PERL_MON_5: tm.tm_mon++;
1714 case PERL_MON_4: tm.tm_mon++;
1715 case PERL_MON_3: tm.tm_mon++;
1716 case PERL_MON_2: tm.tm_mon++;
1721 case PERL_T_FMT_AMPM:
1723 return_format = TRUE;
1726 case PERL_ERA_D_FMT:
1728 return_format = TRUE;
1731 case PERL_ERA_T_FMT:
1733 return_format = TRUE;
1736 case PERL_ERA_D_T_FMT:
1738 return_format = TRUE;
1741 case PERL_ALT_DIGITS:
1743 format = "%Ow"; /* Find the alternate digit for 0 */
1747 /* We can't use my_strftime() because it doesn't look at
1749 while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize,
1752 /* A zero return means one of:
1753 * a) there wasn't enough space in PL_langinfo_buf
1754 * b) the format, like a plain %p, returns empty
1755 * c) it was an illegal format, though some
1756 * implementations of strftime will just return the
1757 * illegal format as a plain character sequence.
1759 * To quickly test for case 'b)', try again but precede
1760 * the format with a plain character. If that result is
1761 * still empty, the problem is either 'a)' or 'c)' */
1763 Size_t format_size = strlen(format) + 1;
1764 Size_t mod_size = format_size + 1;
1768 Newx(mod_format, mod_size, char);
1769 Newx(temp_result, PL_langinfo_bufsize, char);
1771 my_strlcpy(mod_format + 1, format, mod_size);
1772 len = strftime(temp_result,
1773 PL_langinfo_bufsize,
1775 Safefree(mod_format);
1776 Safefree(temp_result);
1778 /* If 'len' is non-zero, it means that we had a case like
1779 * %p which means the current locale doesn't use a.m. or
1780 * p.m., and that is valid */
1783 /* Here, still didn't work. If we get well beyond a
1784 * reasonable size, bail out to prevent an infinite
1787 if (PL_langinfo_bufsize > 100 * format_size) {
1788 *PL_langinfo_buf = '\0';
1791 /* Double the buffer size to retry; Add 1 in case
1792 * original was 0, so we aren't stuck at 0. */
1793 PL_langinfo_bufsize *= 2;
1794 PL_langinfo_bufsize++;
1795 Renew(PL_langinfo_buf, PL_langinfo_bufsize, char);
1803 /* Here, we got a result.
1805 * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the
1806 * alternate format for wday 0. If the value is the same as
1807 * the normal 0, there isn't an alternate, so clear the buffer.
1809 if ( item == PERL_ALT_DIGITS
1810 && strEQ(PL_langinfo_buf, "0"))
1812 *PL_langinfo_buf = '\0';
1815 /* ALT_DIGITS is problematic. Experiments on it showed that
1816 * strftime() did not always work properly when going from
1817 * alt-9 to alt-10. Only a few locales have this item defined,
1818 * and in all of them on Linux that khw was able to find,
1819 * nl_langinfo() merely returned the alt-0 character, possibly
1820 * doubled. Most Unicode digits are in blocks of 10
1821 * consecutive code points, so that is sufficient information
1822 * for those scripts, as we can infer alt-1, alt-2, .... But
1823 * for a Japanese locale, a CJK ideographic 0 is returned, and
1824 * the CJK digits are not in code point order, so you can't
1825 * really infer anything. The localedef for this locale did
1826 * specify the succeeding digits, so that strftime() works
1827 * properly on them, without needing to infer anything. But
1828 * the nl_langinfo() return did not give sufficient information
1829 * for the caller to understand what's going on. So until
1830 * there is evidence that it should work differently, this
1831 * returns the alt-0 string for ALT_DIGITS.
1833 * wday was chosen because its range is all a single digit.
1834 * Things like tm_sec have two digits as the minimum: '00' */
1838 /* If to return the format, not the value, overwrite the buffer
1839 * with it. But some strftime()s will keep the original format
1840 * if illegal, so change those to "" */
1841 if (return_format) {
1842 if (strEQ(PL_langinfo_buf, format)) {
1843 *PL_langinfo_buf = '\0';
1846 save_to_buffer(format, &PL_langinfo_buf,
1847 &PL_langinfo_bufsize, 0);
1858 return PL_langinfo_buf;
1865 * Initialize locale awareness.
1868 Perl_init_i18nl10n(pTHX_ int printwarn)
1872 * 0 if not to output warning when setup locale is bad
1873 * 1 if to output warning based on value of PERL_BADLANG
1874 * >1 if to output regardless of PERL_BADLANG
1877 * 1 = set ok or not applicable,
1878 * 0 = fallback to a locale of lower priority
1879 * -1 = fallback to all locales failed, not even to the C locale
1881 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
1882 * set, debugging information is output.
1884 * This looks more complicated than it is, mainly due to the #ifdefs.
1886 * We try to set LC_ALL to the value determined by the environment. If
1887 * there is no LC_ALL on this platform, we try the individual categories we
1888 * know about. If this works, we are done.
1890 * But if it doesn't work, we have to do something else. We search the
1891 * environment variables ourselves instead of relying on the system to do
1892 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
1893 * think there is one), and the ultimate fallback "C". This is all done in
1894 * the same loop as above to avoid duplicating code, but it makes things
1895 * more complex. The 'trial_locales' array is initialized with just one
1896 * element; it causes the behavior described in the paragraph above this to
1897 * happen. If that fails, we add elements to 'trial_locales', and do extra
1898 * loop iterations to cause the behavior described in this paragraph.
1900 * On Ultrix, the locale MUST come from the environment, so there is
1901 * preliminary code to set it. I (khw) am not sure that it is necessary,
1902 * and that this couldn't be folded into the loop, but barring any real
1903 * platforms to test on, it's staying as-is
1905 * A slight complication is that in embedded Perls, the locale may already
1906 * be set-up, and we don't want to get it from the normal environment
1907 * variables. This is handled by having a special environment variable
1908 * indicate we're in this situation. We simply set setlocale's 2nd
1909 * parameter to be a NULL instead of "". That indicates to setlocale that
1910 * it is not to change anything, but to return the current value,
1911 * effectively initializing perl's db to what the locale already is.
1913 * We play the same trick with NULL if a LC_ALL succeeds. We call
1914 * setlocale() on the individual categores with NULL to get their existing
1915 * values for our db, instead of trying to change them.
1922 PERL_UNUSED_ARG(printwarn);
1924 #else /* USE_LOCALE */
1927 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
1931 /* NULL uses the existing already set up locale */
1932 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
1935 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
1936 unsigned int trial_locales_count;
1937 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
1938 const char * const lang = savepv(PerlEnv_getenv("LANG"));
1939 bool setlocale_failure = FALSE;
1942 /* A later getenv() could zap this, so only use here */
1943 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
1945 const bool locwarn = (printwarn > 1
1947 && ( ! bad_lang_use_once
1949 /* disallow with "" or "0" */
1951 && strNE("0", bad_lang_use_once)))));
1953 /* setlocale() return vals; not copied so must be looked at immediately */
1954 const char * sl_result[NOMINAL_LC_ALL_INDEX + 1];
1956 /* current locale for given category; should have been copied so aren't
1958 const char * curlocales[NOMINAL_LC_ALL_INDEX + 1];
1962 /* In some systems you can find out the system default locale
1963 * and use that as the fallback locale. */
1964 # define SYSTEM_DEFAULT_LOCALE
1966 # ifdef SYSTEM_DEFAULT_LOCALE
1968 const char *system_default_locale = NULL;
1973 # define DEBUG_LOCALE_INIT(a,b,c)
1976 DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT")));
1978 # define DEBUG_LOCALE_INIT(category, locale, result) \
1980 if (debug_initialization) { \
1981 PerlIO_printf(Perl_debug_log, \
1983 __FILE__, __LINE__, \
1984 setlocale_debug_string(category, \
1990 /* Make sure the parallel arrays are properly set up */
1991 # ifdef USE_LOCALE_NUMERIC
1992 assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC);
1993 assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC"));
1995 # ifdef USE_LOCALE_CTYPE
1996 assert(categories[LC_CTYPE_INDEX] == LC_CTYPE);
1997 assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE"));
1999 # ifdef USE_LOCALE_COLLATE
2000 assert(categories[LC_COLLATE_INDEX] == LC_COLLATE);
2001 assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE"));
2003 # ifdef USE_LOCALE_TIME
2004 assert(categories[LC_TIME_INDEX] == LC_TIME);
2005 assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME"));
2007 # ifdef USE_LOCALE_MESSAGES
2008 assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES);
2009 assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES"));
2011 # ifdef USE_LOCALE_MONETARY
2012 assert(categories[LC_MONETARY_INDEX] == LC_MONETARY);
2013 assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY"));
2015 # ifdef USE_LOCALE_ADDRESS
2016 assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS);
2017 assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS"));
2019 # ifdef USE_LOCALE_IDENTIFICATION
2020 assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION);
2021 assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION"));
2023 # ifdef USE_LOCALE_MEASUREMENT
2024 assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT);
2025 assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT"));
2027 # ifdef USE_LOCALE_PAPER
2028 assert(categories[LC_PAPER_INDEX] == LC_PAPER);
2029 assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER"));
2031 # ifdef USE_LOCALE_TELEPHONE
2032 assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE);
2033 assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE"));
2036 assert(categories[LC_ALL_INDEX] == LC_ALL);
2037 assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL"));
2038 assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX);
2040 # endif /* DEBUGGING */
2042 /* Initialize the cache of the program's UTF-8ness for the always known
2043 * locales C and POSIX */
2044 my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness,
2045 sizeof(PL_locale_utf8ness));
2047 PL_numeric_radix_sv = newSVpvs(".");
2049 # ifdef LOCALE_ENVIRON_REQUIRED
2052 * Ultrix setlocale(..., "") fails if there are no environment
2053 * variables from which to get a locale name.
2057 # error Ultrix without LC_ALL not implemented
2063 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init);
2064 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]);
2065 if (sl_result[LC_ALL_INDEX])
2068 setlocale_failure = TRUE;
2070 if (! setlocale_failure) {
2071 const char * locale_param;
2072 for (i = 0; i < LC_ALL_INDEX; i++) {
2073 locale_param = (! done && (lang || PerlEnv_getenv(category_names[i])))
2076 sl_result[i] = do_setlocale_r(categories[i], locale_param);
2077 if (! sl_result[i]) {
2078 setlocale_failure = TRUE;
2080 DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]);
2085 # endif /* LC_ALL */
2086 # endif /* LOCALE_ENVIRON_REQUIRED */
2088 /* We try each locale in the list until we get one that works, or exhaust
2089 * the list. Normally the loop is executed just once. But if setting the
2090 * locale fails, inside the loop we add fallback trials to the array and so
2091 * will execute the loop multiple times */
2092 trial_locales[0] = setlocale_init;
2093 trial_locales_count = 1;
2095 for (i= 0; i < trial_locales_count; i++) {
2096 const char * trial_locale = trial_locales[i];
2100 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
2101 * when i==0, but I (khw) don't think that behavior makes much
2103 setlocale_failure = FALSE;
2105 # ifdef SYSTEM_DEFAULT_LOCALE
2106 # ifdef WIN32 /* Note that assumes Win32 has LC_ALL */
2108 /* On Windows machines, an entry of "" after the 0th means to use
2109 * the system default locale, which we now proceed to get. */
2110 if (strEQ(trial_locale, "")) {
2113 /* Note that this may change the locale, but we are going to do
2114 * that anyway just below */
2115 system_default_locale = do_setlocale_c(LC_ALL, "");
2116 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
2118 /* Skip if invalid or if it's already on the list of locales to
2120 if (! system_default_locale) {
2121 goto next_iteration;
2123 for (j = 0; j < trial_locales_count; j++) {
2124 if (strEQ(system_default_locale, trial_locales[j])) {
2125 goto next_iteration;
2129 trial_locale = system_default_locale;
2132 # error SYSTEM_DEFAULT_LOCALE only implemented for Win32
2134 # endif /* SYSTEM_DEFAULT_LOCALE */
2140 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale);
2141 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]);
2142 if (! sl_result[LC_ALL_INDEX]) {
2143 setlocale_failure = TRUE;
2146 /* Since LC_ALL succeeded, it should have changed all the other
2147 * categories it can to its value; so we massage things so that the
2148 * setlocales below just return their category's current values.
2149 * This adequately handles the case in NetBSD where LC_COLLATE may
2150 * not be defined for a locale, and setting it individually will
2151 * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to
2152 * the POSIX locale. */
2153 trial_locale = NULL;
2156 # endif /* LC_ALL */
2158 if (! setlocale_failure) {
2160 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2162 = savepv(do_setlocale_r(categories[j], trial_locale));
2163 if (! curlocales[j]) {
2164 setlocale_failure = TRUE;
2166 DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]);
2169 if (! setlocale_failure) { /* All succeeded */
2170 break; /* Exit trial_locales loop */
2174 /* Here, something failed; will need to try a fallback. */
2180 if (locwarn) { /* Output failure info only on the first one */
2184 PerlIO_printf(Perl_error_log,
2185 "perl: warning: Setting locale failed.\n");
2187 # else /* !LC_ALL */
2189 PerlIO_printf(Perl_error_log,
2190 "perl: warning: Setting locale failed for the categories:\n\t");
2192 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2193 if (! curlocales[j]) {
2194 PerlIO_printf(Perl_error_log, category_names[j]);
2197 Safefree(curlocales[j]);
2201 # endif /* LC_ALL */
2203 PerlIO_printf(Perl_error_log,
2204 "perl: warning: Please check that your locale settings:\n");
2208 PerlIO_printf(Perl_error_log,
2209 "\tLANGUAGE = %c%s%c,\n",
2210 language ? '"' : '(',
2211 language ? language : "unset",
2212 language ? '"' : ')');
2215 PerlIO_printf(Perl_error_log,
2216 "\tLC_ALL = %c%s%c,\n",
2218 lc_all ? lc_all : "unset",
2219 lc_all ? '"' : ')');
2221 # if defined(USE_ENVIRON_ARRAY)
2226 /* Look through the environment for any variables of the
2227 * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was
2228 * already handled above. These are assumed to be locale
2229 * settings. Output them and their values. */
2230 for (e = environ; *e; e++) {
2231 const STRLEN prefix_len = sizeof("LC_") - 1;
2234 if ( strBEGINs(*e, "LC_")
2235 && ! strBEGINs(*e, "LC_ALL=")
2236 && (uppers_len = strspn(*e + prefix_len,
2237 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"))
2238 && ((*e)[prefix_len + uppers_len] == '='))
2240 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
2241 (int) (prefix_len + uppers_len), *e,
2242 *e + prefix_len + uppers_len + 1);
2249 PerlIO_printf(Perl_error_log,
2250 "\t(possibly more locale environment variables)\n");
2254 PerlIO_printf(Perl_error_log,
2255 "\tLANG = %c%s%c\n",
2257 lang ? lang : "unset",
2260 PerlIO_printf(Perl_error_log,
2261 " are supported and installed on your system.\n");
2264 /* Calculate what fallback locales to try. We have avoided this
2265 * until we have to, because failure is quite unlikely. This will
2266 * usually change the upper bound of the loop we are in.
2268 * Since the system's default way of setting the locale has not
2269 * found one that works, We use Perl's defined ordering: LC_ALL,
2270 * LANG, and the C locale. We don't try the same locale twice, so
2271 * don't add to the list if already there. (On POSIX systems, the
2272 * LC_ALL element will likely be a repeat of the 0th element "",
2273 * but there's no harm done by doing it explicitly.
2275 * Note that this tries the LC_ALL environment variable even on
2276 * systems which have no LC_ALL locale setting. This may or may
2277 * not have been originally intentional, but there's no real need
2278 * to change the behavior. */
2280 for (j = 0; j < trial_locales_count; j++) {
2281 if (strEQ(lc_all, trial_locales[j])) {
2285 trial_locales[trial_locales_count++] = lc_all;
2290 for (j = 0; j < trial_locales_count; j++) {
2291 if (strEQ(lang, trial_locales[j])) {
2295 trial_locales[trial_locales_count++] = lang;
2299 # if defined(WIN32) && defined(LC_ALL)
2301 /* For Windows, we also try the system default locale before "C".
2302 * (If there exists a Windows without LC_ALL we skip this because
2303 * it gets too complicated. For those, the "C" is the next
2304 * fallback possibility). The "" is the same as the 0th element of
2305 * the array, but the code at the loop above knows to treat it
2306 * differently when not the 0th */
2307 trial_locales[trial_locales_count++] = "";
2311 for (j = 0; j < trial_locales_count; j++) {
2312 if (strEQ("C", trial_locales[j])) {
2316 trial_locales[trial_locales_count++] = "C";
2319 } /* end of first time through the loop */
2327 } /* end of looping through the trial locales */
2329 if (ok < 1) { /* If we tried to fallback */
2331 if (! setlocale_failure) { /* fallback succeeded */
2332 msg = "Falling back to";
2334 else { /* fallback failed */
2337 /* We dropped off the end of the loop, so have to decrement i to
2338 * get back to the value the last time through */
2342 msg = "Failed to fall back to";
2344 /* To continue, we should use whatever values we've got */
2346 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2347 Safefree(curlocales[j]);
2348 curlocales[j] = savepv(do_setlocale_r(categories[j], NULL));
2349 DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]);
2354 const char * description;
2355 const char * name = "";
2356 if (strEQ(trial_locales[i], "C")) {
2357 description = "the standard locale";
2361 # ifdef SYSTEM_DEFAULT_LOCALE
2363 else if (strEQ(trial_locales[i], "")) {
2364 description = "the system default locale";
2365 if (system_default_locale) {
2366 name = system_default_locale;
2370 # endif /* SYSTEM_DEFAULT_LOCALE */
2373 description = "a fallback locale";
2374 name = trial_locales[i];
2376 if (name && strNE(name, "")) {
2377 PerlIO_printf(Perl_error_log,
2378 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
2381 PerlIO_printf(Perl_error_log,
2382 "perl: warning: %s %s.\n", msg, description);
2385 } /* End of tried to fallback */
2387 /* Done with finding the locales; update our records */
2389 # ifdef USE_LOCALE_CTYPE
2391 new_ctype(curlocales[LC_CTYPE_INDEX]);
2394 # ifdef USE_LOCALE_COLLATE
2396 new_collate(curlocales[LC_COLLATE_INDEX]);
2399 # ifdef USE_LOCALE_NUMERIC
2401 new_numeric(curlocales[LC_NUMERIC_INDEX]);
2405 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
2407 # if defined(USE_ITHREADS)
2409 /* This caches whether each category's locale is UTF-8 or not. This
2410 * may involve changing the locale. It is ok to do this at
2411 * initialization time before any threads have started, but not later.
2412 * Caching means that if the program heeds our dictate not to change
2413 * locales in threaded applications, this data will remain valid, and
2414 * it may get queried without changing locales. If the environment is
2415 * such that all categories have the same locale, this isn't needed, as
2416 * the code will not change the locale; but this handles the uncommon
2417 * case where the environment has disparate locales for the categories
2419 (void) _is_cur_LC_category_utf8(categories[i]);
2423 Safefree(curlocales[i]);
2426 # if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
2428 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
2429 * locale is UTF-8. The call to new_ctype() just above has already
2430 * calculated the latter value and saved it in PL_in_utf8_CTYPE_locale. If
2431 * both PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE})
2432 * are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer on
2433 * STDIN, STDOUT, STDERR, _and_ the default open discipline. */
2434 PL_utf8locale = PL_in_utf8_CTYPE_locale;
2436 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
2437 This is an alternative to using the -C command line switch
2438 (the -C if present will override this). */
2440 const char *p = PerlEnv_getenv("PERL_UNICODE");
2441 PL_unicode = p ? parse_unicode_opts(&p) : 0;
2442 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
2456 #endif /* USE_LOCALE */
2459 /* So won't continue to output stuff */
2460 DEBUG_INITIALIZATION_set(FALSE);
2467 #ifdef USE_LOCALE_COLLATE
2470 Perl__mem_collxfrm(pTHX_ const char *input_string,
2471 STRLEN len, /* Length of 'input_string' */
2472 STRLEN *xlen, /* Set to length of returned string
2473 (not including the collation index
2475 bool utf8 /* Is the input in UTF-8? */
2479 /* _mem_collxfrm() is a bit like strxfrm() but with two important
2480 * differences. First, it handles embedded NULs. Second, it allocates a bit
2481 * more memory than needed for the transformed data itself. The real
2482 * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
2483 * the length of that, and doesn't include the collation index size.
2484 * Please see sv_collxfrm() to see how this is used. */
2486 #define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
2488 char * s = (char *) input_string;
2489 STRLEN s_strlen = strlen(input_string);
2491 STRLEN xAlloc; /* xalloc is a reserved word in VC */
2492 STRLEN length_in_chars;
2493 bool first_time = TRUE; /* Cleared after first loop iteration */
2495 PERL_ARGS_ASSERT__MEM_COLLXFRM;
2497 /* Must be NUL-terminated */
2498 assert(*(input_string + len) == '\0');
2500 /* If this locale has defective collation, skip */
2501 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
2502 DEBUG_L(PerlIO_printf(Perl_debug_log,
2503 "_mem_collxfrm: locale's collation is defective\n"));
2507 /* Replace any embedded NULs with the control that sorts before any others.
2508 * This will give as good as possible results on strings that don't
2509 * otherwise contain that character, but otherwise there may be
2510 * less-than-perfect results with that character and NUL. This is
2511 * unavoidable unless we replace strxfrm with our own implementation. */
2512 if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded
2516 STRLEN sans_nuls_len;
2517 int try_non_controls;
2518 char this_replacement_char[] = "?\0"; /* Room for a two-byte string,
2519 making sure 2nd byte is NUL.
2521 STRLEN this_replacement_len;
2523 /* If we don't know what non-NUL control character sorts lowest for
2524 * this locale, find it */
2525 if (PL_strxfrm_NUL_replacement == '\0') {
2527 char * cur_min_x = NULL; /* The min_char's xfrm, (except it also
2528 includes the collation index
2531 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
2533 /* Unlikely, but it may be that no control will work to replace
2534 * NUL, in which case we instead look for any character. Controls
2535 * are preferred because collation order is, in general, context
2536 * sensitive, with adjoining characters affecting the order, and
2537 * controls are less likely to have such interactions, allowing the
2538 * NUL-replacement to stand on its own. (Another way to look at it
2539 * is to imagine what would happen if the NUL were replaced by a
2540 * combining character; it wouldn't work out all that well.) */
2541 for (try_non_controls = 0;
2542 try_non_controls < 2;
2545 /* Look through all legal code points (NUL isn't) */
2546 for (j = 1; j < 256; j++) {
2547 char * x; /* j's xfrm plus collation index */
2548 STRLEN x_len; /* length of 'x' */
2549 STRLEN trial_len = 1;
2550 char cur_source[] = { '\0', '\0' };
2552 /* Skip non-controls the first time through the loop. The
2553 * controls in a UTF-8 locale are the L1 ones */
2554 if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
2561 /* Create a 1-char string of the current code point */
2562 cur_source[0] = (char) j;
2564 /* Then transform it */
2565 x = _mem_collxfrm(cur_source, trial_len, &x_len,
2566 0 /* The string is not in UTF-8 */);
2568 /* Ignore any character that didn't successfully transform.
2574 /* If this character's transformation is lower than
2575 * the current lowest, this one becomes the lowest */
2576 if ( cur_min_x == NULL
2577 || strLT(x + COLLXFRM_HDR_LEN,
2578 cur_min_x + COLLXFRM_HDR_LEN))
2580 PL_strxfrm_NUL_replacement = j;
2586 } /* end of loop through all 255 characters */
2588 /* Stop looking if found */
2593 /* Unlikely, but possible, if there aren't any controls that
2594 * work in the locale, repeat the loop, looking for any
2595 * character that works */
2596 DEBUG_L(PerlIO_printf(Perl_debug_log,
2597 "_mem_collxfrm: No control worked. Trying non-controls\n"));
2598 } /* End of loop to try first the controls, then any char */
2601 DEBUG_L(PerlIO_printf(Perl_debug_log,
2602 "_mem_collxfrm: Couldn't find any character to replace"
2603 " embedded NULs in locale %s with", PL_collation_name));
2607 DEBUG_L(PerlIO_printf(Perl_debug_log,
2608 "_mem_collxfrm: Replacing embedded NULs in locale %s with "
2609 "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement));
2611 Safefree(cur_min_x);
2612 } /* End of determining the character that is to replace NULs */
2614 /* If the replacement is variant under UTF-8, it must match the
2615 * UTF8-ness of the original */
2616 if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) {
2617 this_replacement_char[0] =
2618 UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement);
2619 this_replacement_char[1] =
2620 UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement);
2621 this_replacement_len = 2;
2624 this_replacement_char[0] = PL_strxfrm_NUL_replacement;
2625 /* this_replacement_char[1] = '\0' was done at initialization */
2626 this_replacement_len = 1;
2629 /* The worst case length for the replaced string would be if every
2630 * character in it is NUL. Multiply that by the length of each
2631 * replacement, and allow for a trailing NUL */
2632 sans_nuls_len = (len * this_replacement_len) + 1;
2633 Newx(sans_nuls, sans_nuls_len, char);
2636 /* Replace each NUL with the lowest collating control. Loop until have
2637 * exhausted all the NULs */
2638 while (s + s_strlen < e) {
2639 my_strlcat(sans_nuls, s, sans_nuls_len);
2641 /* Do the actual replacement */
2642 my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len);
2644 /* Move past the input NUL */
2646 s_strlen = strlen(s);
2649 /* And add anything that trails the final NUL */
2650 my_strlcat(sans_nuls, s, sans_nuls_len);
2652 /* Switch so below we transform this modified string */
2655 } /* End of replacing NULs */
2657 /* Make sure the UTF8ness of the string and locale match */
2658 if (utf8 != PL_in_utf8_COLLATE_locale) {
2659 const char * const t = s; /* Temporary so we can later find where the
2662 /* Here they don't match. Change the string's to be what the locale is
2665 if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
2666 s = (char *) bytes_to_utf8((const U8 *) s, &len);
2669 else { /* locale is not UTF-8; but input is; downgrade the input */
2671 s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
2673 /* If the downgrade was successful we are done, but if the input
2674 * contains things that require UTF-8 to represent, have to do
2675 * damage control ... */
2676 if (UNLIKELY(utf8)) {
2678 /* What we do is construct a non-UTF-8 string with
2679 * 1) the characters representable by a single byte converted
2680 * to be so (if necessary);
2681 * 2) and the rest converted to collate the same as the
2682 * highest collating representable character. That makes
2683 * them collate at the end. This is similar to how we
2684 * handle embedded NULs, but we use the highest collating
2685 * code point instead of the smallest. Like the NUL case,
2686 * this isn't perfect, but is the best we can reasonably
2687 * do. Every above-255 code point will sort the same as
2688 * the highest-sorting 0-255 code point. If that code
2689 * point can combine in a sequence with some other code
2690 * points for weight calculations, us changing something to
2691 * be it can adversely affect the results. But in most
2692 * cases, it should work reasonably. And note that this is
2693 * really an illegal situation: using code points above 255
2694 * on a locale where only 0-255 are valid. If two strings
2695 * sort entirely equal, then the sort order for the
2696 * above-255 code points will be in code point order. */
2700 /* If we haven't calculated the code point with the maximum
2701 * collating order for this locale, do so now */
2702 if (! PL_strxfrm_max_cp) {
2705 /* The current transformed string that collates the
2706 * highest (except it also includes the prefixed collation
2708 char * cur_max_x = NULL;
2710 /* Look through all legal code points (NUL isn't) */
2711 for (j = 1; j < 256; j++) {
2714 char cur_source[] = { '\0', '\0' };
2716 /* Create a 1-char string of the current code point */
2717 cur_source[0] = (char) j;
2719 /* Then transform it */
2720 x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
2722 /* If something went wrong (which it shouldn't), just
2723 * ignore this code point */
2728 /* If this character's transformation is higher than
2729 * the current highest, this one becomes the highest */
2730 if ( cur_max_x == NULL
2731 || strGT(x + COLLXFRM_HDR_LEN,
2732 cur_max_x + COLLXFRM_HDR_LEN))
2734 PL_strxfrm_max_cp = j;
2743 DEBUG_L(PerlIO_printf(Perl_debug_log,
2744 "_mem_collxfrm: Couldn't find any character to"
2745 " replace above-Latin1 chars in locale %s with",
2746 PL_collation_name));
2750 DEBUG_L(PerlIO_printf(Perl_debug_log,
2751 "_mem_collxfrm: highest 1-byte collating character"
2752 " in locale %s is 0x%02X\n",
2754 PL_strxfrm_max_cp));
2756 Safefree(cur_max_x);
2759 /* Here we know which legal code point collates the highest.
2760 * We are ready to construct the non-UTF-8 string. The length
2761 * will be at least 1 byte smaller than the input string
2762 * (because we changed at least one 2-byte character into a
2763 * single byte), but that is eaten up by the trailing NUL */
2769 char * e = (char *) t + len;
2771 for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
2773 if (UTF8_IS_INVARIANT(cur_char)) {
2776 else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) {
2777 s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
2779 else { /* Replace illegal cp with highest collating
2781 s[d++] = PL_strxfrm_max_cp;
2785 Renew(s, d, char); /* Free up unused space */
2790 /* Here, we have constructed a modified version of the input. It could
2791 * be that we already had a modified copy before we did this version.
2792 * If so, that copy is no longer needed */
2793 if (t != input_string) {
2798 length_in_chars = (utf8)
2799 ? utf8_length((U8 *) s, (U8 *) s + len)
2802 /* The first element in the output is the collation id, used by
2803 * sv_collxfrm(); then comes the space for the transformed string. The
2804 * equation should give us a good estimate as to how much is needed */
2805 xAlloc = COLLXFRM_HDR_LEN
2807 + (PL_collxfrm_mult * length_in_chars);
2808 Newx(xbuf, xAlloc, char);
2809 if (UNLIKELY(! xbuf)) {
2810 DEBUG_L(PerlIO_printf(Perl_debug_log,
2811 "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
2815 /* Store the collation id */
2816 *(U32*)xbuf = PL_collation_ix;
2818 /* Then the transformation of the input. We loop until successful, or we
2822 *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
2824 /* If the transformed string occupies less space than we told strxfrm()
2825 * was available, it means it successfully transformed the whole
2827 if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
2829 /* Some systems include a trailing NUL in the returned length.
2830 * Ignore it, using a loop in case multiple trailing NULs are
2833 && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
2838 /* If the first try didn't get it, it means our prediction was low.
2839 * Modify the coefficients so that we predict a larger value in any
2840 * future transformations */
2842 STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
2843 STRLEN computed_guess = PL_collxfrm_base
2844 + (PL_collxfrm_mult * length_in_chars);
2846 /* On zero-length input, just keep current slope instead of
2848 const STRLEN new_m = (length_in_chars != 0)
2849 ? needed / length_in_chars
2852 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2853 "%s: %d: initial size of %zu bytes for a length "
2854 "%zu string was insufficient, %zu needed\n",
2856 computed_guess, length_in_chars, needed));
2858 /* If slope increased, use it, but discard this result for
2859 * length 1 strings, as we can't be sure that it's a real slope
2861 if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
2865 STRLEN old_m = PL_collxfrm_mult;
2866 STRLEN old_b = PL_collxfrm_base;
2870 PL_collxfrm_mult = new_m;
2871 PL_collxfrm_base = 1; /* +1 For trailing NUL */
2872 computed_guess = PL_collxfrm_base
2873 + (PL_collxfrm_mult * length_in_chars);
2874 if (computed_guess < needed) {
2875 PL_collxfrm_base += needed - computed_guess;
2878 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2879 "%s: %d: slope is now %zu; was %zu, base "
2880 "is now %zu; was %zu\n",
2882 PL_collxfrm_mult, old_m,
2883 PL_collxfrm_base, old_b));
2885 else { /* Slope didn't change, but 'b' did */
2886 const STRLEN new_b = needed
2889 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
2890 "%s: %d: base is now %zu; was %zu\n",
2892 new_b, PL_collxfrm_base));
2893 PL_collxfrm_base = new_b;
2900 if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
2901 DEBUG_L(PerlIO_printf(Perl_debug_log,
2902 "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
2903 *xlen, PERL_INT_MAX));
2907 /* A well-behaved strxfrm() returns exactly how much space it needs
2908 * (usually not including the trailing NUL) when it fails due to not
2909 * enough space being provided. Assume that this is the case unless
2910 * it's been proven otherwise */
2911 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
2912 xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
2914 else { /* Here, either:
2915 * 1) The strxfrm() has previously shown bad behavior; or
2916 * 2) It isn't the first time through the loop, which means
2917 * that the strxfrm() is now showing bad behavior, because
2918 * we gave it what it said was needed in the previous
2919 * iteration, and it came back saying it needed still more.
2920 * (Many versions of cygwin fit this. When the buffer size
2921 * isn't sufficient, they return the input size instead of
2922 * how much is needed.)
2923 * Increase the buffer size by a fixed percentage and try again.
2925 xAlloc += (xAlloc / 4) + 1;
2926 PL_strxfrm_is_behaved = FALSE;
2930 if (DEBUG_Lv_TEST || debug_initialization) {
2931 PerlIO_printf(Perl_debug_log,
2932 "_mem_collxfrm required more space than previously calculated"
2933 " for locale %s, trying again with new guess=%d+%zu\n",
2934 PL_collation_name, (int) COLLXFRM_HDR_LEN,
2935 xAlloc - COLLXFRM_HDR_LEN);
2942 Renew(xbuf, xAlloc, char);
2943 if (UNLIKELY(! xbuf)) {
2944 DEBUG_L(PerlIO_printf(Perl_debug_log,
2945 "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
2955 if (DEBUG_Lv_TEST || debug_initialization) {
2957 print_collxfrm_input_and_return(s, s + len, xlen, utf8);
2958 PerlIO_printf(Perl_debug_log, "Its xfrm is:");
2959 PerlIO_printf(Perl_debug_log, "%s\n",
2960 _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN,
2966 /* Free up unneeded space; retain ehough for trailing NUL */
2967 Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
2969 if (s != input_string) {
2977 if (s != input_string) {
2984 if (DEBUG_Lv_TEST || debug_initialization) {
2985 print_collxfrm_input_and_return(s, s + len, NULL, utf8);
2996 S_print_collxfrm_input_and_return(pTHX_
2997 const char * const s,
2998 const char * const e,
2999 const STRLEN * const xlen,
3003 PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
3005 PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ",
3006 (UV)PL_collation_ix);
3008 PerlIO_printf(Perl_debug_log, "%zu", *xlen);
3011 PerlIO_printf(Perl_debug_log, "NULL");
3013 PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
3015 print_bytes_for_locale(s, e, is_utf8);
3017 PerlIO_printf(Perl_debug_log, "'\n");
3021 S_print_bytes_for_locale(pTHX_
3022 const char * const s,
3023 const char * const e,
3027 bool prev_was_printable = TRUE;
3028 bool first_time = TRUE;
3030 PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE;
3034 ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
3037 if (! prev_was_printable) {
3038 PerlIO_printf(Perl_debug_log, " ");
3040 PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
3041 prev_was_printable = TRUE;
3045 PerlIO_printf(Perl_debug_log, " ");
3047 PerlIO_printf(Perl_debug_log, "%02" UVXf, cp);
3048 prev_was_printable = FALSE;
3050 t += (is_utf8) ? UTF8SKIP(t) : 1;
3055 # endif /* #ifdef DEBUGGING */
3056 #endif /* USE_LOCALE_COLLATE */
3061 Perl__is_cur_LC_category_utf8(pTHX_ int category)
3063 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
3064 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
3065 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
3066 * could give the wrong result. The result will very likely be correct for
3067 * languages that have commonly used non-ASCII characters, but for notably
3068 * English, it comes down to if the locale's name ends in something like
3069 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
3071 /* Name of current locale corresponding to the input category */
3072 const char *save_input_locale = NULL;
3074 bool is_utf8 = FALSE; /* The return value */
3076 /* The variables below are for the cache of previous lookups using this
3077 * function. The cache is a C string, described at the definition for
3078 * 'C_and_POSIX_utf8ness'.
3080 * The first part of the cache is fixed, for the C and POSIX locales. The
3081 * varying part starts just after them. */
3082 char * utf8ness_cache = PL_locale_utf8ness + STRLENs(C_and_POSIX_utf8ness);
3084 Size_t utf8ness_cache_size; /* Size of the varying portion */
3085 Size_t input_name_len; /* Length in bytes of save_input_locale */
3086 Size_t input_name_len_with_overhead; /* plus extra chars used to store
3087 the name in the cache */
3088 char * delimited; /* The name plus the delimiters used to store
3090 char * name_pos; /* position of 'delimited' in the cache, or 0
3096 assert(category != LC_ALL);
3100 /* Get the desired category's locale */
3101 save_input_locale = do_setlocale_r(category, NULL);
3102 if (! save_input_locale) {
3104 "panic: %s: %d: Could not find current %s locale, errno=%d\n",
3105 __FILE__, __LINE__, category_name(category), errno);
3108 save_input_locale = stdize_locale(savepv(save_input_locale));
3109 DEBUG_L(PerlIO_printf(Perl_debug_log,
3110 "Current locale for %s is %s\n",
3111 category_name(category), save_input_locale));
3113 input_name_len = strlen(save_input_locale);
3115 /* In our cache, each name is accompanied by two delimiters and a single
3117 input_name_len_with_overhead = input_name_len + 3;
3119 /* Allocate and populate space for a copy of the name surrounded by the
3121 Newx(delimited, input_name_len_with_overhead, char);
3122 delimited[0] = UTF8NESS_SEP[0];
3123 Copy(save_input_locale, delimited + 1, input_name_len, char);
3124 delimited[input_name_len+1] = UTF8NESS_PREFIX[0];
3125 delimited[input_name_len+2] = '\0';
3127 /* And see if that is in the cache */
3128 name_pos = instr(PL_locale_utf8ness, delimited);
3130 is_utf8 = *(name_pos + input_name_len_with_overhead - 1) - '0';
3134 if (DEBUG_Lv_TEST || debug_initialization) {
3135 PerlIO_printf(Perl_debug_log, "UTF8ness for locale %s=%d, \n",
3136 save_input_locale, is_utf8);
3141 /* And, if not already in that position, move it to the beginning of
3142 * the non-constant portion of the list, since it is the most recently
3143 * used. (We don't have to worry about overflow, since just moving
3144 * existing names around) */
3145 if (name_pos > utf8ness_cache) {
3146 Move(utf8ness_cache,
3147 utf8ness_cache + input_name_len_with_overhead,
3148 name_pos - utf8ness_cache, char);
3151 input_name_len_with_overhead - 1, char);
3152 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3155 Safefree(delimited);
3156 Safefree(save_input_locale);
3160 /* Here we don't have stored the utf8ness for the input locale. We have to
3163 # if defined(USE_LOCALE_CTYPE) \
3164 && ( (defined(HAS_NL_LANGINFO) && defined(CODESET)) \
3165 || defined(HAS_MBTOWC))
3168 const char *save_ctype_locale = NULL;
3170 if (category != LC_CTYPE) {
3172 /* Get the current LC_CTYPE locale */
3173 save_ctype_locale = do_setlocale_c(LC_CTYPE, NULL);
3174 if (! save_ctype_locale) {
3176 "panic: %s: %d: Could not find current LC_CTYPE locale,"
3177 " errno=%d\n", __FILE__, __LINE__, errno);
3179 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
3181 /* If LC_CTYPE and the desired category use the same locale, this
3182 * means that finding the value for LC_CTYPE is the same as finding
3183 * the value for the desired category. Otherwise, switch LC_CTYPE
3184 * to the desired category's locale */
3185 if (strEQ(save_ctype_locale, save_input_locale)) {
3186 Safefree(save_ctype_locale);
3187 save_ctype_locale = NULL;
3189 else if (! do_setlocale_c(LC_CTYPE, save_input_locale)) {
3190 Safefree(save_ctype_locale);
3192 "panic: %s: %d: Could not change LC_CTYPE locale to %s,"
3193 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3197 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
3198 save_input_locale));
3200 /* Here the current LC_CTYPE is set to the locale of the category whose
3201 * information is desired. This means that nl_langinfo() and mbtowc()
3202 * should give the correct results */
3204 # ifdef MB_CUR_MAX /* But we can potentially rule out UTF-8ness, avoiding
3205 calling the functions if we have this */
3207 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
3208 * Unicode code point. */
3210 DEBUG_L(PerlIO_printf(Perl_debug_log, "\tMB_CUR_MAX=%d\n",
3212 if ((unsigned) MB_CUR_MAX < STRLENs(MAX_UNICODE_UTF8)) {
3218 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
3220 { /* The task is easiest if the platform has this POSIX 2001 function.
3221 Except on some platforms it can wrongly return "", so have to have
3222 a fallback. And it can return that it's UTF-8, even if there are
3223 variances from that. For example, Turkish locales may use the
3224 alternate dotted I rules, and sometimes it appears to be a
3225 defective locale definition. XXX We should probably check for
3226 these in the Latin1 range and warn */
3227 const char *codeset = my_nl_langinfo(PERL_CODESET, FALSE);
3228 /* FALSE => already in dest locale */
3230 DEBUG_L(PerlIO_printf(Perl_debug_log,
3231 "\tnllanginfo returned CODESET '%s'\n", codeset));
3233 if (codeset && strNE(codeset, "")) {
3235 /* If the implementation of foldEQ() somehow were
3236 * to change to not go byte-by-byte, this could
3237 * read past end of string, as only one length is
3238 * checked. But currently, a premature NUL will
3239 * compare false, and it will stop there */
3240 is_utf8 = cBOOL( foldEQ(codeset, STR_WITH_LEN("UTF-8"))
3241 || foldEQ(codeset, STR_WITH_LEN("UTF8")));
3243 DEBUG_L(PerlIO_printf(Perl_debug_log,
3244 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
3251 # if defined(HAS_MBTOWC)
3252 /* We can see if this is a UTF-8-like locale if have mbtowc(). It was a
3253 * late adder to C89, so very likely to have it. However, testing has
3254 * shown that, like nl_langinfo() above, there are locales that are not
3255 * strictly UTF-8 that this will return that they are */
3261 /* mbtowc() converts a byte string to a wide character. Feed a byte
3262 * string to it and check that the result is the expected Unicode
3265 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
3267 len = mbtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8));
3269 DEBUG_L(PerlIO_printf(Perl_debug_log,
3270 "\treturn from mbtowc; len=%d; code_point=%x; errno=%d\n",
3271 len, (unsigned int) wc, errno));
3273 is_utf8 = cBOOL( len == STRLENs(REPLACEMENT_CHARACTER_UTF8)
3274 && wc == (wchar_t) UNICODE_REPLACEMENT);
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 /* Here, we must have a C89 compiler that doesn't have mbtowc(). Next
3292 * try looking at the currency symbol to see if it disambiguates
3293 * things. Often that will be in the native script, and if the symbol
3294 * isn't in UTF-8, we know that the locale isn't. If it is non-ASCII
3295 * UTF-8, we infer that the locale is too, as the odds of a non-UTF8
3296 * string being valid UTF-8 are quite small */
3298 # ifdef HAS_LOCALECONV
3299 # ifdef USE_LOCALE_MONETARY
3302 char *save_monetary_locale = NULL;
3303 bool only_ascii = FALSE;
3306 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
3307 * the desired category, if it isn't that locale already */
3309 if (category != LC_MONETARY) {
3311 save_monetary_locale = do_setlocale_c(LC_MONETARY, NULL);
3312 if (! save_monetary_locale) {
3314 "panic: %s: %d: Could not find current LC_MONETARY locale,"
3315 " errno=%d\n", __FILE__, __LINE__, errno);
3317 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
3319 if (strEQ(save_monetary_locale, save_input_locale)) {
3320 Safefree(save_monetary_locale);
3321 save_monetary_locale = NULL;
3323 else if (! do_setlocale_c(LC_MONETARY, save_input_locale)) {
3324 Safefree(save_monetary_locale);
3326 "panic: %s: %d: Could not change LC_MONETARY locale to %s,"
3327 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3331 /* Here the current LC_MONETARY is set to the locale of the category
3332 * whose information is desired. */
3336 || ! lc->currency_symbol
3337 || is_utf8_invariant_string((U8 *) lc->currency_symbol, 0))
3339 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));
3343 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
3346 /* If we changed it, restore LC_MONETARY to its original locale */
3347 if (save_monetary_locale) {
3348 do_setlocale_c(LC_MONETARY, save_monetary_locale);
3349 Safefree(save_monetary_locale);
3354 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
3355 * otherwise assume the locale is UTF-8 if and only if the symbol
3356 * is non-ascii UTF-8. */
3357 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
3358 save_input_locale, is_utf8));
3359 goto finish_and_return;
3363 # endif /* USE_LOCALE_MONETARY */
3364 # endif /* HAS_LOCALECONV */
3366 # if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
3368 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
3369 * the names of the months and weekdays, timezone, and am/pm indicator */
3371 char *save_time_locale = NULL;
3373 bool is_dst = FALSE;
3377 char * formatted_time;
3380 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
3381 * desired category, if it isn't that locale already */
3383 if (category != LC_TIME) {
3385 save_time_locale = do_setlocale_c(LC_TIME, NULL);
3386 if (! save_time_locale) {
3388 "panic: %s: %d: Could not find current LC_TIME locale,"
3389 " errno=%d\n", __FILE__, __LINE__, errno);
3391 save_time_locale = stdize_locale(savepv(save_time_locale));
3393 if (strEQ(save_time_locale, save_input_locale)) {
3394 Safefree(save_time_locale);
3395 save_time_locale = NULL;
3397 else if (! do_setlocale_c(LC_TIME, save_input_locale)) {
3398 Safefree(save_time_locale);
3400 "panic: %s: %d: Could not change LC_TIME locale to %s,"
3401 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3405 /* Here the current LC_TIME is set to the locale of the category
3406 * whose information is desired. Look at all the days of the week and
3407 * month names, and the timezone and am/pm indicator for UTF-8 variant
3408 * characters. The first such a one found will tell us if the locale
3409 * is UTF-8 or not */
3411 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
3412 formatted_time = my_strftime("%A %B %Z %p",
3413 0, 0, hour, dom, month, 2012 - 1900, 0, 0, is_dst);
3414 if ( ! formatted_time
3415 || is_utf8_invariant_string((U8 *) formatted_time, 0))
3418 /* Here, we didn't find a non-ASCII. Try the next time through
3419 * with the complemented dst and am/pm, and try with the next
3420 * weekday. After we have gotten all weekdays, try the next
3423 hour = (hour + 12) % 24;
3431 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
3432 * false otherwise. But first, restore LC_TIME to its original
3433 * locale if we changed it */
3434 if (save_time_locale) {
3435 do_setlocale_c(LC_TIME, save_time_locale);
3436 Safefree(save_time_locale);
3439 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
3441 is_utf8_string((U8 *) formatted_time, 0)));
3442 is_utf8 = is_utf8_string((U8 *) formatted_time, 0);
3443 goto finish_and_return;
3446 /* Falling off the end of the loop indicates all the names were just
3447 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
3448 * to its original locale */
3449 if (save_time_locale) {
3450 do_setlocale_c(LC_TIME, save_time_locale);
3451 Safefree(save_time_locale);
3453 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));
3458 # if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
3460 /* This code is ifdefd out because it was found to not be necessary in testing
3461 * on our dromedary test machine, which has over 700 locales. There, this
3462 * added no value to looking at the currency symbol and the time strings. I
3463 * left it in so as to avoid rewriting it if real-world experience indicates
3464 * that dromedary is an outlier. Essentially, instead of returning abpve if we
3465 * haven't found illegal utf8, we continue on and examine all the strerror()
3466 * messages on the platform for utf8ness. If all are ASCII, we still don't
3467 * know the answer; but otherwise we have a pretty good indication of the
3468 * utf8ness. The reason this doesn't help much is that the messages may not
3469 * have been translated into the locale. The currency symbol and time strings
3470 * are much more likely to have been translated. */
3473 bool non_ascii = FALSE;
3474 char *save_messages_locale = NULL;
3475 const char * errmsg = NULL;
3477 /* Like above, we set LC_MESSAGES to the locale of the desired
3478 * category, if it isn't that locale already */
3480 if (category != LC_MESSAGES) {
3482 save_messages_locale = do_setlocale_c(LC_MESSAGES, NULL);
3483 if (! save_messages_locale) {
3485 "panic: %s: %d: Could not find current LC_MESSAGES locale,"
3486 " errno=%d\n", __FILE__, __LINE__, errno);
3488 save_messages_locale = stdize_locale(savepv(save_messages_locale));
3490 if (strEQ(save_messages_locale, save_input_locale)) {
3491 Safefree(save_messages_locale);
3492 save_messages_locale = NULL;
3494 else if (! do_setlocale_c(LC_MESSAGES, save_input_locale)) {
3495 Safefree(save_messages_locale);
3497 "panic: %s: %d: Could not change LC_MESSAGES locale to %s,"
3498 " errno=%d\n", __FILE__, __LINE__, save_input_locale, errno);
3502 /* Here the current LC_MESSAGES is set to the locale of the category
3503 * whose information is desired. Look through all the messages. We
3504 * can't use Strerror() here because it may expand to code that
3505 * segfaults in miniperl */
3507 for (e = 0; e <= sys_nerr; e++) {
3509 errmsg = sys_errlist[e];
3510 if (errno || !errmsg) {
3513 errmsg = savepv(errmsg);
3514 if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
3516 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
3522 /* And, if we changed it, restore LC_MESSAGES to its original locale */
3523 if (save_messages_locale) {
3524 do_setlocale_c(LC_MESSAGES, save_messages_locale);
3525 Safefree(save_messages_locale);
3530 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
3531 * any non-ascii means it is one; otherwise we assume it isn't */
3532 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
3535 goto finish_and_return;
3538 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));
3542 # ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
3545 /* As a last resort, look at the locale name to see if it matches
3546 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
3547 * return of setlocale(), is actually defined to be opaque, so we can't
3548 * really rely on the absence of various substrings in the name to indicate
3549 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
3550 * be a UTF-8 locale. Similarly for the other common names */
3553 const Size_t final_pos = strlen(save_input_locale) - 1;
3555 if (final_pos >= 3) {
3556 const char *name = save_input_locale;
3558 /* Find next 'U' or 'u' and look from there */
3559 while ((name += strcspn(name, "Uu") + 1)
3560 <= save_input_locale + final_pos - 2)
3562 if ( isALPHA_FOLD_NE(*name, 't')
3563 || isALPHA_FOLD_NE(*(name + 1), 'f'))
3568 if (*(name) == '-') {
3569 if ((name > save_input_locale + final_pos - 1)) {
3574 if (*(name) == '8') {
3575 DEBUG_L(PerlIO_printf(Perl_debug_log,
3576 "Locale %s ends with UTF-8 in name\n",
3577 save_input_locale));
3579 goto finish_and_return;
3582 DEBUG_L(PerlIO_printf(Perl_debug_log,
3583 "Locale %s doesn't end with UTF-8 in name\n",
3584 save_input_locale));
3589 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
3590 if (memENDs(save_input_locale, final_pos, "65001")) {
3591 DEBUG_L(PerlIO_printf(Perl_debug_log,
3592 "Locale %s ends with 65001 in name, is UTF-8 locale\n",
3593 save_input_locale));
3595 goto finish_and_return;
3602 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
3603 * since we are about to return FALSE anyway, there is no point in doing
3604 * this extra work */
3607 if (instr(save_input_locale, "8859")) {
3608 DEBUG_L(PerlIO_printf(Perl_debug_log,
3609 "Locale %s has 8859 in name, not UTF-8 locale\n",
3610 save_input_locale));
3612 goto finish_and_return;
3616 DEBUG_L(PerlIO_printf(Perl_debug_log,
3617 "Assuming locale %s is not a UTF-8 locale\n",
3618 save_input_locale));
3621 # endif /* the code that is compiled when no modern LC_CTYPE */
3625 /* Cache this result so we don't have to go through all this next time. */
3626 utf8ness_cache_size = sizeof(PL_locale_utf8ness)
3627 - (utf8ness_cache - PL_locale_utf8ness);
3629 /* But we can't save it if it is too large for the total space available */
3630 if (LIKELY(input_name_len_with_overhead < utf8ness_cache_size)) {
3631 Size_t utf8ness_cache_len = strlen(utf8ness_cache);
3633 /* Here it can fit, but we may need to clear out the oldest cached
3634 * result(s) to do so. Check */
3635 if (utf8ness_cache_len + input_name_len_with_overhead
3636 >= utf8ness_cache_size)
3638 /* Here we have to clear something out to make room for this.
3639 * Start looking at the rightmost place where it could fit and find
3640 * the beginning of the entry that extends past that. */
3641 char * cutoff = (char *) my_memrchr(utf8ness_cache,
3644 - input_name_len_with_overhead);
3647 assert(cutoff >= utf8ness_cache);
3649 /* This and all subsequent entries must be removed */
3651 utf8ness_cache_len = strlen(utf8ness_cache);
3654 /* Make space for the new entry */
3655 Move(utf8ness_cache,
3656 utf8ness_cache + input_name_len_with_overhead,
3657 utf8ness_cache_len + 1 /* Incl. trailing NUL */, char);
3660 Copy(delimited, utf8ness_cache, input_name_len_with_overhead - 1, char);
3661 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3663 if ((PL_locale_utf8ness[strlen(PL_locale_utf8ness)-1]
3664 & (PERL_UINTMAX_T) ~1) != '0')
3667 "panic: %s: %d: Corrupt utf8ness_cache=%s\nlen=%u,"
3668 " inserted_name=%s, its_len=%u\n",
3670 PL_locale_utf8ness, strlen(PL_locale_utf8ness),
3671 delimited, input_name_len_with_overhead);
3677 if (DEBUG_Lv_TEST || debug_initialization) {
3678 PerlIO_printf(Perl_debug_log,
3679 "PL_locale_utf8ness is now %s; returning %d\n",
3680 PL_locale_utf8ness, is_utf8);
3685 Safefree(delimited);
3686 Safefree(save_input_locale);
3693 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
3696 /* Internal function which returns if we are in the scope of a pragma that
3697 * enables the locale category 'category'. 'compiling' should indicate if
3698 * this is during the compilation phase (TRUE) or not (FALSE). */
3700 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
3702 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
3703 if (! categories || categories == &PL_sv_placeholder) {
3707 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
3708 * a valid unsigned */
3709 assert(category >= -1);
3710 return cBOOL(SvUV(categories) & (1U << (category + 1)));
3714 Perl_my_strerror(pTHX_ const int errnum)
3716 /* Returns a mortalized copy of the text of the error message associated
3717 * with 'errnum'. It uses the current locale's text unless the platform
3718 * doesn't have the LC_MESSAGES category or we are not being called from
3719 * within the scope of 'use locale'. In the former case, it uses whatever
3720 * strerror returns; in the latter case it uses the text from the C locale.
3722 * The function just calls strerror(), but temporarily switches, if needed,
3723 * to the C locale */
3728 #ifndef USE_LOCALE_MESSAGES
3730 /* If platform doesn't have messages category, we don't do any switching to
3731 * the C locale; we just use whatever strerror() returns */
3733 errstr = savepv(Strerror(errnum));
3735 #else /* Has locale messages */
3737 const bool within_locale_scope = IN_LC(LC_MESSAGES);
3739 # if defined(HAS_POSIX_2008_LOCALE) && defined(HAS_STRERROR_L)
3741 /* This function is trivial if we don't have to worry about thread safety
3742 * and have strerror_l(), as it handles the switch of locales so we don't
3743 * have to deal with that. We don't have to worry about thread safety if
3744 * this is an unthreaded build, or if strerror_r() is also available. Both
3745 * it and strerror_l() are thread-safe. Plain strerror() isn't thread
3746 * safe. But on threaded builds when strerror_r() is available, the
3747 * apparent call to strerror() below is actually a macro that
3748 * behind-the-scenes calls strerror_r().
3751 # if ! defined(USE_ITHREADS) || defined(HAS_STRERROR_R)
3753 if (within_locale_scope) {
3754 errstr = savepv(strerror(errnum));
3757 errstr = savepv(strerror_l(errnum, PL_C_locale_obj));
3762 /* Here we have strerror_l(), but not strerror_r() and we are on a
3763 * threaded-build. We use strerror_l() for everything, constructing a
3764 * locale to pass to it if necessary */
3766 bool do_free = FALSE;
3767 locale_t locale_to_use;
3769 if (within_locale_scope) {
3770 locale_to_use = uselocale((locale_t) 0);
3771 if (locale_to_use == LC_GLOBAL_LOCALE) {
3772 locale_to_use = duplocale(LC_GLOBAL_LOCALE);
3776 else { /* Use C locale if not within 'use locale' scope */
3777 locale_to_use = PL_C_locale_obj;
3780 errstr = savepv(strerror_l(errnum, locale_to_use));
3783 freelocale(locale_to_use);
3787 # else /* Doesn't have strerror_l() */
3789 # ifdef USE_POSIX_2008_LOCALE
3791 locale_t save_locale = NULL;
3795 const char * save_locale = NULL;
3796 bool locale_is_C = FALSE;
3798 /* We have a critical section to prevent another thread from changing the
3799 * locale out from under us (or zapping the buffer returned from
3805 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3806 "my_strerror called with errnum %d\n", errnum));
3807 if (! within_locale_scope) {
3810 # ifdef USE_POSIX_2008_LOCALE /* Use the thread-safe locale functions */
3812 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3813 "Not within locale scope, about to call"
3814 " uselocale(0x%p)\n", PL_C_locale_obj));
3815 save_locale = uselocale(PL_C_locale_obj);
3816 if (! save_locale) {
3817 DEBUG_L(PerlIO_printf(Perl_debug_log,
3818 "uselocale failed, errno=%d\n", errno));
3821 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3822 "uselocale returned 0x%p\n", save_locale));
3825 # else /* Not thread-safe build */
3827 save_locale = do_setlocale_c(LC_MESSAGES, NULL);
3828 if (! save_locale) {
3830 "panic: %s: %d: Could not find current LC_MESSAGES locale,"
3831 " errno=%d\n", __FILE__, __LINE__, errno);
3834 locale_is_C = isNAME_C_OR_POSIX(save_locale);
3836 /* Switch to the C locale if not already in it */
3837 if (! locale_is_C) {
3839 /* The setlocale() just below likely will zap 'save_locale', so
3841 save_locale = savepv(save_locale);
3842 do_setlocale_c(LC_MESSAGES, "C");
3848 } /* end of ! within_locale_scope */
3850 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s: %d: WITHIN locale scope\n",
3851 __FILE__, __LINE__));
3854 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3855 "Any locale change has been done; about to call Strerror\n"));
3856 errstr = savepv(Strerror(errnum));
3858 if (! within_locale_scope) {
3861 # ifdef USE_POSIX_2008_LOCALE
3863 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3864 "%s: %d: not within locale scope, restoring the locale\n",
3865 __FILE__, __LINE__));
3866 if (save_locale && ! uselocale(save_locale)) {
3867 DEBUG_L(PerlIO_printf(Perl_debug_log,
3868 "uselocale restore failed, errno=%d\n", errno));
3874 if (save_locale && ! locale_is_C) {
3875 if (! do_setlocale_c(LC_MESSAGES, save_locale)) {
3877 "panic: %s: %d: setlocale restore failed, errno=%d\n",
3878 __FILE__, __LINE__, errno);
3880 Safefree(save_locale);
3887 # endif /* End of doesn't have strerror_l */
3888 #endif /* End of does have locale messages */
3892 if (DEBUG_Lv_TEST) {
3893 PerlIO_printf(Perl_debug_log, "Strerror returned; saving a copy: '");
3894 print_bytes_for_locale(errstr, errstr + strlen(errstr), 0);
3895 PerlIO_printf(Perl_debug_log, "'\n");
3906 =for apidoc sync_locale
3908 Changing the program's locale should be avoided by XS code. Nevertheless,
3909 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
3910 happens, Perl needs to be told that the locale has changed. Use this function
3911 to do so, before returning to Perl.
3917 Perl_sync_locale(pTHX)
3921 #ifdef USE_LOCALE_CTYPE
3923 newlocale = do_setlocale_c(LC_CTYPE, NULL);
3924 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3925 "%s:%d: %s\n", __FILE__, __LINE__,
3926 setlocale_debug_string(LC_CTYPE, NULL, newlocale)));
3927 new_ctype(newlocale);
3929 #endif /* USE_LOCALE_CTYPE */
3930 #ifdef USE_LOCALE_COLLATE
3932 newlocale = do_setlocale_c(LC_COLLATE, NULL);
3933 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3934 "%s:%d: %s\n", __FILE__, __LINE__,
3935 setlocale_debug_string(LC_COLLATE, NULL, newlocale)));
3936 new_collate(newlocale);
3939 #ifdef USE_LOCALE_NUMERIC
3941 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
3942 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
3943 "%s:%d: %s\n", __FILE__, __LINE__,
3944 setlocale_debug_string(LC_NUMERIC, NULL, newlocale)));
3945 new_numeric(newlocale);
3947 #endif /* USE_LOCALE_NUMERIC */
3951 #if defined(DEBUGGING) && defined(USE_LOCALE)
3954 S_setlocale_debug_string(const int category, /* category number,
3956 const char* const locale, /* locale name */
3958 /* return value from setlocale() when attempting to
3959 * set 'category' to 'locale' */
3960 const char* const retval)
3962 /* Returns a pointer to a NUL-terminated string in static storage with
3963 * added text about the info passed in. This is not thread safe and will
3964 * be overwritten by the next call, so this should be used just to
3965 * formulate a string to immediately print or savepv() on. */
3967 /* initialise to a non-null value to keep it out of BSS and so keep
3968 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
3969 static char ret[128] = "If you can read this, thank your buggy C"
3970 " library strlcpy(), and change your hints file"
3973 my_strlcpy(ret, "setlocale(", sizeof(ret));
3974 my_strlcat(ret, category_name(category), sizeof(ret));
3975 my_strlcat(ret, ", ", sizeof(ret));
3978 my_strlcat(ret, "\"", sizeof(ret));
3979 my_strlcat(ret, locale, sizeof(ret));
3980 my_strlcat(ret, "\"", sizeof(ret));
3983 my_strlcat(ret, "NULL", sizeof(ret));
3986 my_strlcat(ret, ") returned ", sizeof(ret));
3989 my_strlcat(ret, "\"", sizeof(ret));
3990 my_strlcat(ret, retval, sizeof(ret));
3991 my_strlcat(ret, "\"", sizeof(ret));
3994 my_strlcat(ret, "NULL", sizeof(ret));
3997 assert(strlen(ret) < sizeof(ret));
4006 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob