3 * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
4 * 2002, 2003, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * A Elbereth Gilthoniel,
13 * silivren penna míriel
14 * o menel aglar elenath!
15 * Na-chaered palan-díriel
16 * o galadhremmin ennorath,
17 * Fanuilos, le linnathon
18 * nef aear, si nef aearon!
20 * [p.238 of _The Lord of the Rings_, II/i: "Many Meetings"]
23 /* utility functions for handling locale-specific stuff like what
24 * character represents the decimal point.
26 * All C programs have an underlying locale. Perl generally doesn't pay any
27 * attention to it except within the scope of a 'use locale'. For most
28 * categories, it accomplishes this by just using different operations if it is
29 * in such scope than if not. However, various libc functions called by Perl
30 * are affected by the LC_NUMERIC category, so there are macros in perl.h that
31 * are used to toggle between the current locale and the C locale depending on
32 * the desired behavior of those functions at the moment. And, LC_MESSAGES is
33 * switched to the C locale for outputting the message unless within the scope
38 #define PERL_IN_LOCALE_C
42 # include <langinfo.h>
47 /* If the environment says to, we can output debugging information during
48 * initialization. This is done before option parsing, and before any thread
49 * creation, so can be a file-level static */
51 # ifdef PERL_GLOBAL_STRUCT
52 /* no global syms allowed */
53 # define debug_initialization 0
54 # define DEBUG_INITIALIZATION_set(v)
56 static bool debug_initialization = FALSE;
57 # define DEBUG_INITIALIZATION_set(v) (debug_initialization = v)
64 * Standardize the locale name from a string returned by 'setlocale', possibly
65 * modifying that string.
67 * The typical return value of setlocale() is either
68 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
69 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
70 * (the space-separated values represent the various sublocales,
71 * in some unspecified order). This is not handled by this function.
73 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
74 * which is harmful for further use of the string in setlocale(). This
75 * function removes the trailing new line and everything up through the '='
79 S_stdize_locale(pTHX_ char *locs)
81 const char * const s = strchr(locs, '=');
84 PERL_ARGS_ASSERT_STDIZE_LOCALE;
87 const char * const t = strchr(s, '.');
90 const char * const u = strchr(t, '\n');
91 if (u && (u[1] == 0)) {
92 const STRLEN len = u - s;
93 Move(s + 1, locs, len, char);
101 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
109 Perl_set_numeric_radix(pTHX)
111 #ifdef USE_LOCALE_NUMERIC
112 # ifdef HAS_LOCALECONV
113 const struct lconv* const lc = localeconv();
115 if (lc && lc->decimal_point) {
116 if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) {
117 SvREFCNT_dec(PL_numeric_radix_sv);
118 PL_numeric_radix_sv = NULL;
121 if (PL_numeric_radix_sv)
122 sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
124 PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
125 if (! is_invariant_string((U8 *) lc->decimal_point, 0)
126 && is_utf8_string((U8 *) lc->decimal_point, 0)
127 && _is_cur_LC_category_utf8(LC_NUMERIC))
129 SvUTF8_on(PL_numeric_radix_sv);
134 PL_numeric_radix_sv = NULL;
137 if (DEBUG_L_TEST || debug_initialization) {
138 PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
139 (PL_numeric_radix_sv)
140 ? SvPVX(PL_numeric_radix_sv)
142 (PL_numeric_radix_sv)
143 ? cBOOL(SvUTF8(PL_numeric_radix_sv))
148 # endif /* HAS_LOCALECONV */
149 #endif /* USE_LOCALE_NUMERIC */
152 /* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
153 * return of setlocale(), then this is extremely likely to be the C or POSIX
154 * locale. However, the output of setlocale() is documented to be opaque, but
155 * the odds are extremely small that it would return these two strings for some
156 * other locale. Note that VMS in these two locales includes many non-ASCII
157 * characters as controls and punctuation (below are hex bytes):
158 * cntrl: 00-1F 7F 84-97 9B-9F
159 * punct: 21-2F 3A-40 5B-60 7B-7E A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
160 * Oddly, none there are listed as alphas, though some represent alphabetics
161 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
162 #define isNAME_C_OR_POSIX(name) ((name) != NULL \
163 && ((*(name) == 'C' && (*(name + 1)) == '\0') \
164 || strEQ((name), "POSIX")))
167 Perl_new_numeric(pTHX_ const char *newnum)
169 #ifdef USE_LOCALE_NUMERIC
171 /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell
172 * core Perl this and that 'newnum' is the name of the new locale.
173 * It installs this locale as the current underlying default.
175 * The default locale and the C locale can be toggled between by use of the
176 * set_numeric_local() and set_numeric_standard() functions, which should
177 * probably not be called directly, but only via macros like
178 * SET_NUMERIC_STANDARD() in perl.h.
180 * The toggling is necessary mainly so that a non-dot radix decimal point
181 * character can be output, while allowing internal calculations to use a
184 * This sets several interpreter-level variables:
185 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
186 * PL_numeric_local A boolean indicating if the toggled state is such
187 * that the current locale is the program's underlying
189 * PL_numeric_standard An int indicating if the toggled state is such
190 * that the current locale is the C locale. If non-zero,
191 * it is in C; if > 1, it means it may not be toggled away
193 * Note that both of the last two variables can be true at the same time,
194 * if the underlying locale is C. (Toggling is a no-op under these
197 * Any code changing the locale (outside this file) should use
198 * POSIX::setlocale, which calls this function. Therefore this function
199 * should be called directly only from this file and from
200 * POSIX::setlocale() */
205 Safefree(PL_numeric_name);
206 PL_numeric_name = NULL;
207 PL_numeric_standard = TRUE;
208 PL_numeric_local = TRUE;
212 save_newnum = stdize_locale(savepv(newnum));
214 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
215 PL_numeric_local = TRUE;
217 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
218 Safefree(PL_numeric_name);
219 PL_numeric_name = save_newnum;
222 Safefree(save_newnum);
225 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
226 * have to worry about the radix being a non-dot. (Core operations that
227 * need the underlying locale change to it temporarily). */
228 set_numeric_standard();
233 PERL_UNUSED_ARG(newnum);
234 #endif /* USE_LOCALE_NUMERIC */
238 Perl_set_numeric_standard(pTHX)
240 #ifdef USE_LOCALE_NUMERIC
241 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
242 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
243 * macro avoids calling this routine if toggling isn't necessary according
244 * to our records (which could be wrong if some XS code has changed the
245 * locale behind our back) */
247 setlocale(LC_NUMERIC, "C");
248 PL_numeric_standard = TRUE;
249 PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
252 if (DEBUG_L_TEST || debug_initialization) {
253 PerlIO_printf(Perl_debug_log,
254 "Underlying LC_NUMERIC locale now is C\n");
258 #endif /* USE_LOCALE_NUMERIC */
262 Perl_set_numeric_local(pTHX)
264 #ifdef USE_LOCALE_NUMERIC
265 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
266 * code should use the macros like SET_NUMERIC_LOCAL() in perl.h instead of
267 * calling this directly. The macro avoids calling this routine if
268 * toggling isn't necessary according to our records (which could be wrong
269 * if some XS code has changed the locale behind our back) */
271 setlocale(LC_NUMERIC, PL_numeric_name);
272 PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
273 PL_numeric_local = TRUE;
276 if (DEBUG_L_TEST || debug_initialization) {
277 PerlIO_printf(Perl_debug_log,
278 "Underlying LC_NUMERIC locale now is %s\n",
283 #endif /* USE_LOCALE_NUMERIC */
287 * Set up for a new ctype locale.
290 Perl_new_ctype(pTHX_ const char *newctype)
292 #ifdef USE_LOCALE_CTYPE
294 /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell
295 * core Perl this and that 'newctype' is the name of the new locale.
297 * This function sets up the folding arrays for all 256 bytes, assuming
298 * that tofold() is tolc() since fold case is not a concept in POSIX,
300 * Any code changing the locale (outside this file) should use
301 * POSIX::setlocale, which calls this function. Therefore this function
302 * should be called directly only from this file and from
303 * POSIX::setlocale() */
308 PERL_ARGS_ASSERT_NEW_CTYPE;
310 /* We will replace any bad locale warning with 1) nothing if the new one is
311 * ok; or 2) a new warning for the bad new locale */
312 if (PL_warn_locale) {
313 SvREFCNT_dec_NN(PL_warn_locale);
314 PL_warn_locale = NULL;
317 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
319 /* A UTF-8 locale gets standard rules. But note that code still has to
320 * handle this specially because of the three problematic code points */
321 if (PL_in_utf8_CTYPE_locale) {
322 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
325 /* Assume enough space for every character being bad. 4 spaces each
326 * for the 94 printable characters that are output like "'x' "; and 5
327 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
329 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
331 bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
333 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
335 unsigned int bad_count = 0; /* Count of bad characters */
337 for (i = 0; i < 256; i++) {
338 if (isUPPER_LC((U8) i))
339 PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
340 else if (isLOWER_LC((U8) i))
341 PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
343 PL_fold_locale[i] = (U8) i;
345 /* If checking for locale problems, see if the native ASCII-range
346 * printables plus \n and \t are in their expected categories in
347 * the new locale. If not, this could mean big trouble, upending
348 * Perl's and most programs' assumptions, like having a
349 * metacharacter with special meaning become a \w. Fortunately,
350 * it's very rare to find locales that aren't supersets of ASCII
351 * nowadays. It isn't a problem for most controls to be changed
352 * into something else; we check only \n and \t, though perhaps \r
353 * could be an issue as well. */
354 if (check_for_problems
355 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
357 if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
358 || (isPUNCT_A(i) && ! isPUNCT_LC(i))
359 || (isBLANK_A(i) && ! isBLANK_LC(i))
360 || (i == '\n' && ! isCNTRL_LC(i)))
362 if (bad_count) { /* Separate multiple entries with a
364 bad_chars_list[bad_count++] = ' ';
366 bad_chars_list[bad_count++] = '\'';
368 bad_chars_list[bad_count++] = (char) i;
371 bad_chars_list[bad_count++] = '\\';
373 bad_chars_list[bad_count++] = 'n';
377 bad_chars_list[bad_count++] = 't';
380 bad_chars_list[bad_count++] = '\'';
381 bad_chars_list[bad_count] = '\0';
387 /* We only handle single-byte locales (outside of UTF-8 ones; so if
388 * this locale requires more than one byte, there are going to be
390 if (check_for_problems && MB_CUR_MAX > 1
392 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
393 * locale. Just assume that the implementation for them (plus
394 * for POSIX) is correct and the > 1 value is spurious. (Since
395 * these are specially handled to never be considered UTF-8
396 * locales, as long as this is the only problem, everything
397 * should work fine */
398 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
400 multi_byte_locale = TRUE;
404 if (bad_count || multi_byte_locale) {
405 PL_warn_locale = Perl_newSVpvf(aTHX_
406 "Locale '%s' may not work well.%s%s%s\n",
409 ? " Some characters in it are not recognized by"
413 ? "\nThe following characters (and maybe others)"
414 " may not have the same meaning as the Perl"
415 " program expects:\n"
421 /* If we are actually in the scope of the locale, output the
422 * message now. Otherwise we save it to be output at the first
423 * operation using this locale, if that actually happens. Most
424 * programs don't use locales, so they are immune to bad ones */
425 if (IN_LC(LC_CTYPE)) {
427 /* We have to save 'newctype' because the setlocale() just
428 * below may destroy it. The next setlocale() further down
429 * should restore it properly so that the intermediate change
430 * here is transparent to this function's caller */
431 const char * const badlocale = savepv(newctype);
433 setlocale(LC_CTYPE, "C");
435 /* The '0' below suppresses a bogus gcc compiler warning */
436 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
437 setlocale(LC_CTYPE, badlocale);
439 SvREFCNT_dec_NN(PL_warn_locale);
440 PL_warn_locale = NULL;
445 #endif /* USE_LOCALE_CTYPE */
446 PERL_ARGS_ASSERT_NEW_CTYPE;
447 PERL_UNUSED_ARG(newctype);
452 Perl__warn_problematic_locale()
455 #ifdef USE_LOCALE_CTYPE
459 /* Internal-to-core function that outputs the message in PL_warn_locale,
460 * and then NULLS it. Should be called only through the macro
461 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
463 if (PL_warn_locale) {
464 /*GCC_DIAG_IGNORE(-Wformat-security); Didn't work */
465 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
466 SvPVX(PL_warn_locale),
467 0 /* dummy to avoid compiler warning */ );
468 /* GCC_DIAG_RESTORE; */
469 SvREFCNT_dec_NN(PL_warn_locale);
470 PL_warn_locale = NULL;
478 Perl_new_collate(pTHX_ const char *newcoll)
480 #ifdef USE_LOCALE_COLLATE
482 /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell
483 * core Perl this and that 'newcoll' is the name of the new locale.
485 * Any code changing the locale (outside this file) should use
486 * POSIX::setlocale, which calls this function. Therefore this function
487 * should be called directly only from this file and from
488 * POSIX::setlocale().
490 * The design of locale collation is that every locale change is given an
491 * index 'PL_collation_ix'. The first time a string particpates in an
492 * operation that requires collation while locale collation is active, it
493 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
494 * magic includes the collation index, and the transformation of the string
495 * by strxfrm(), q.v. That transformation is used when doing comparisons,
496 * instead of the string itself. If a string changes, the magic is
497 * cleared. The next time the locale changes, the index is incremented,
498 * and so we know during a comparison that the transformation is not
499 * necessarily still valid, and so is recomputed. Note that if the locale
500 * changes enough times, the index could wrap (a U32), and it is possible
501 * that a transformation would improperly be considered valid, leading to
505 if (PL_collation_name) {
507 Safefree(PL_collation_name);
508 PL_collation_name = NULL;
510 PL_collation_standard = TRUE;
511 is_standard_collation:
512 PL_collxfrm_base = 0;
513 PL_collxfrm_mult = 2;
514 PL_in_utf8_COLLATE_locale = FALSE;
515 *PL_strxfrm_min_char = '\0';
516 PL_strxfrm_max_cp = 0;
520 /* If this is not the same locale as currently, set the new one up */
521 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
523 Safefree(PL_collation_name);
524 PL_collation_name = stdize_locale(savepv(newcoll));
525 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
526 if (PL_collation_standard) {
527 goto is_standard_collation;
530 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
531 *PL_strxfrm_min_char = '\0';
532 PL_strxfrm_max_cp = 0;
534 /* A locale collation definition includes primary, secondary, tertiary,
535 * etc. weights for each character. To sort, the primary weights are
536 * used, and only if they compare equal, then the secondary weights are
537 * used, and only if they compare equal, then the tertiary, etc.
539 * strxfrm() works by taking the input string, say ABC, and creating an
540 * output transformed string consisting of first the primary weights,
541 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
542 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
543 * may not have weights at every level. In our example, let's say B
544 * doesn't have a tertiary weight, and A doesn't have a secondary
545 * weight. The constructed string is then going to be
546 * A¹B¹C¹ B²C² A³C³ ....
547 * This has the desired effect that strcmp() will look at the secondary
548 * or tertiary weights only if the strings compare equal at all higher
549 * priority weights. The spaces shown here, like in
551 * are not just for readability. In the general case, these must
552 * actually be bytes, which we will call here 'separator weights'; and
553 * they must be smaller than any other weight value, but since these
554 * are C strings, only the terminating one can be a NUL (some
555 * implementations may include a non-NUL separator weight just before
556 * the NUL). Implementations tend to reserve 01 for the separator
557 * weights. They are needed so that a shorter string's secondary
558 * weights won't be misconstrued as primary weights of a longer string,
559 * etc. By making them smaller than any other weight, the shorter
560 * string will sort first. (Actually, if all secondary weights are
561 * smaller than all primary ones, there is no need for a separator
562 * weight between those two levels, etc.)
564 * The length of the transformed string is roughly a linear function of
565 * the input string. It's not exactly linear because some characters
566 * don't have weights at all levels. When we call strxfrm() we have to
567 * allocate some memory to hold the transformed string. The
568 * calculations below try to find coefficients 'm' and 'b' for this
569 * locale so that m*x + b equals how much space we need, given the size
570 * of the input string in 'x'. If we calculate too small, we increase
571 * the size as needed, and call strxfrm() again, but it is better to
572 * get it right the first time to avoid wasted expensive string
573 * transformations. */
576 /* We use the string below to find how long the tranformation of it
577 * is. Almost all locales are supersets of ASCII, or at least the
578 * ASCII letters. We use all of them, half upper half lower,
579 * because if we used fewer, we might hit just the ones that are
580 * outliers in a particular locale. Most of the strings being
581 * collated will contain a preponderance of letters, and even if
582 * they are above-ASCII, they are likely to have the same number of
583 * weight levels as the ASCII ones. It turns out that digits tend
584 * to have fewer levels, and some punctuation has more, but those
585 * are relatively sparse in text, and khw believes this gives a
586 * reasonable result, but it could be changed if experience so
588 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
589 char * x_longer; /* Transformed 'longer' */
590 Size_t x_len_longer; /* Length of 'x_longer' */
592 char * x_shorter; /* We also transform a substring of 'longer' */
593 Size_t x_len_shorter;
595 /* _mem_collxfrm() is used get the transformation (though here we
596 * are interested only in its length). It is used because it has
597 * the intelligence to handle all cases, but to work, it needs some
598 * values of 'm' and 'b' to get it started. For the purposes of
599 * this calculation we use a very conservative estimate of 'm' and
600 * 'b'. This assumes a weight can be multiple bytes, enough to
601 * hold any UV on the platform, and there are 5 levels, 4 weight
602 * bytes, and a trailing NUL. */
603 PL_collxfrm_base = 5;
604 PL_collxfrm_mult = 5 * sizeof(UV);
606 /* Find out how long the transformation really is */
607 x_longer = _mem_collxfrm(longer,
611 /* We avoid converting to UTF-8 in the
612 * called function by telling it the
613 * string is in UTF-8 if the locale is a
614 * UTF-8 one. Since the string passed
615 * here is invariant under UTF-8, we can
616 * claim it's UTF-8 even though it isn't.
618 PL_in_utf8_COLLATE_locale);
621 /* Find out how long the transformation of a substring of 'longer'
622 * is. Together the lengths of these transformations are
623 * sufficient to calculate 'm' and 'b'. The substring is all of
624 * 'longer' except the first character. This minimizes the chances
625 * of being swayed by outliers */
626 x_shorter = _mem_collxfrm(longer + 1,
629 PL_in_utf8_COLLATE_locale);
632 /* If the results are nonsensical for this simple test, the whole
633 * locale definition is suspect. Mark it so that locale collation
634 * is not active at all for it. XXX Should we warn? */
635 if ( x_len_shorter == 0
637 || x_len_shorter >= x_len_longer)
639 PL_collxfrm_mult = 0;
640 PL_collxfrm_base = 0;
643 SSize_t base; /* Temporary */
645 /* We have both: m * strlen(longer) + b = x_len_longer
646 * m * strlen(shorter) + b = x_len_shorter;
647 * subtracting yields:
648 * m * (strlen(longer) - strlen(shorter))
649 * = x_len_longer - x_len_shorter
650 * But we have set things up so that 'shorter' is 1 byte smaller
651 * than 'longer'. Hence:
652 * m = x_len_longer - x_len_shorter
654 * But if something went wrong, make sure the multiplier is at
657 if (x_len_longer > x_len_shorter) {
658 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
661 PL_collxfrm_mult = 1;
666 * but in case something has gone wrong, make sure it is
668 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
673 /* Add 1 for the trailing NUL */
674 PL_collxfrm_base = base + 1;
678 if (DEBUG_L_TEST || debug_initialization) {
679 PerlIO_printf(Perl_debug_log,
680 "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
682 " collate multipler=%zu, collate base=%zu\n",
684 PL_in_utf8_COLLATE_locale,
685 x_len_shorter, x_len_longer,
686 PL_collxfrm_mult, PL_collxfrm_base);
693 PERL_UNUSED_ARG(newcoll);
694 #endif /* USE_LOCALE_COLLATE */
700 Perl_my_setlocale(pTHX_ int category, const char* locale)
702 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
703 * difference unless the input locale is "", which means on Windows to get
704 * the machine default, which is set via the computer's "Regional and
705 * Language Options" (or its current equivalent). In POSIX, it instead
706 * means to find the locale from the user's environment. This routine
707 * looks in the environment, and, if anything is found, uses that instead
708 * of going to the machine default. If there is no environment override,
709 * the machine default is used, as normal, by calling the real setlocale()
710 * with "". The POSIX behavior is to use the LC_ALL variable if set;
711 * otherwise to use the particular category's variable if set; otherwise to
712 * use the LANG variable. */
714 bool override_LC_ALL = FALSE;
717 if (locale && strEQ(locale, "")) {
719 locale = PerlEnv_getenv("LC_ALL");
725 override_LC_ALL = TRUE;
726 break; /* We already know its variable isn't set */
728 # ifdef USE_LOCALE_TIME
730 locale = PerlEnv_getenv("LC_TIME");
733 # ifdef USE_LOCALE_CTYPE
735 locale = PerlEnv_getenv("LC_CTYPE");
738 # ifdef USE_LOCALE_COLLATE
740 locale = PerlEnv_getenv("LC_COLLATE");
743 # ifdef USE_LOCALE_MONETARY
745 locale = PerlEnv_getenv("LC_MONETARY");
748 # ifdef USE_LOCALE_NUMERIC
750 locale = PerlEnv_getenv("LC_NUMERIC");
753 # ifdef USE_LOCALE_MESSAGES
755 locale = PerlEnv_getenv("LC_MESSAGES");
759 /* This is a category, like PAPER_SIZE that we don't
760 * know about; and so can't provide a wrapper. */
764 locale = PerlEnv_getenv("LANG");
774 result = setlocale(category, locale);
775 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
776 _setlocale_debug_string(category, locale, result)));
778 if (! override_LC_ALL) {
782 /* Here the input category was LC_ALL, and we have set it to what is in the
783 * LANG variable or the system default if there is no LANG. But these have
784 * lower priority than the other LC_foo variables, so override it for each
785 * one that is set. (If they are set to "", it means to use the same thing
786 * we just set LC_ALL to, so can skip) */
787 # ifdef USE_LOCALE_TIME
788 result = PerlEnv_getenv("LC_TIME");
789 if (result && strNE(result, "")) {
790 setlocale(LC_TIME, result);
791 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
793 _setlocale_debug_string(LC_TIME, result, "not captured")));
796 # ifdef USE_LOCALE_CTYPE
797 result = PerlEnv_getenv("LC_CTYPE");
798 if (result && strNE(result, "")) {
799 setlocale(LC_CTYPE, result);
800 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
802 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
805 # ifdef USE_LOCALE_COLLATE
806 result = PerlEnv_getenv("LC_COLLATE");
807 if (result && strNE(result, "")) {
808 setlocale(LC_COLLATE, result);
809 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
811 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
814 # ifdef USE_LOCALE_MONETARY
815 result = PerlEnv_getenv("LC_MONETARY");
816 if (result && strNE(result, "")) {
817 setlocale(LC_MONETARY, result);
818 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
820 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
823 # ifdef USE_LOCALE_NUMERIC
824 result = PerlEnv_getenv("LC_NUMERIC");
825 if (result && strNE(result, "")) {
826 setlocale(LC_NUMERIC, result);
827 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
829 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
832 # ifdef USE_LOCALE_MESSAGES
833 result = PerlEnv_getenv("LC_MESSAGES");
834 if (result && strNE(result, "")) {
835 setlocale(LC_MESSAGES, result);
836 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
838 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
842 result = setlocale(LC_ALL, NULL);
843 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
845 _setlocale_debug_string(LC_ALL, NULL, result)));
854 * Initialize locale awareness.
857 Perl_init_i18nl10n(pTHX_ int printwarn)
861 * 0 if not to output warning when setup locale is bad
862 * 1 if to output warning based on value of PERL_BADLANG
863 * >1 if to output regardless of PERL_BADLANG
866 * 1 = set ok or not applicable,
867 * 0 = fallback to a locale of lower priority
868 * -1 = fallback to all locales failed, not even to the C locale
870 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
871 * set, debugging information is output.
873 * This looks more complicated than it is, mainly due to the #ifdefs.
875 * We try to set LC_ALL to the value determined by the environment. If
876 * there is no LC_ALL on this platform, we try the individual categories we
877 * know about. If this works, we are done.
879 * But if it doesn't work, we have to do something else. We search the
880 * environment variables ourselves instead of relying on the system to do
881 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
882 * think there is one), and the ultimate fallback "C". This is all done in
883 * the same loop as above to avoid duplicating code, but it makes things
884 * more complex. After the original failure, we add the fallback
885 * possibilities to the list of locales to try, and iterate the loop
886 * through them all until one succeeds.
888 * On Ultrix, the locale MUST come from the environment, so there is
889 * preliminary code to set it. I (khw) am not sure that it is necessary,
890 * and that this couldn't be folded into the loop, but barring any real
891 * platforms to test on, it's staying as-is
893 * A slight complication is that in embedded Perls, the locale may already
894 * be set-up, and we don't want to get it from the normal environment
895 * variables. This is handled by having a special environment variable
896 * indicate we're in this situation. We simply set setlocale's 2nd
897 * parameter to be a NULL instead of "". That indicates to setlocale that
898 * it is not to change anything, but to return the current value,
899 * effectively initializing perl's db to what the locale already is.
901 * We play the same trick with NULL if a LC_ALL succeeds. We call
902 * setlocale() on the individual categores with NULL to get their existing
903 * values for our db, instead of trying to change them.
908 #if defined(USE_LOCALE)
909 #ifdef USE_LOCALE_CTYPE
910 char *curctype = NULL;
911 #endif /* USE_LOCALE_CTYPE */
912 #ifdef USE_LOCALE_COLLATE
913 char *curcoll = NULL;
914 #endif /* USE_LOCALE_COLLATE */
915 #ifdef USE_LOCALE_NUMERIC
917 #endif /* USE_LOCALE_NUMERIC */
919 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
922 /* NULL uses the existing already set up locale */
923 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
926 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
927 unsigned int trial_locales_count;
928 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
929 const char * const lang = savepv(PerlEnv_getenv("LANG"));
930 bool setlocale_failure = FALSE;
934 /* A later getenv() could zap this, so only use here */
935 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
937 const bool locwarn = (printwarn > 1
939 && (! bad_lang_use_once
941 /* disallow with "" or "0" */
943 && strNE("0", bad_lang_use_once)))));
945 char * sl_result; /* return from setlocale() */
948 /* In some systems you can find out the system default locale
949 * and use that as the fallback locale. */
950 # define SYSTEM_DEFAULT_LOCALE
952 #ifdef SYSTEM_DEFAULT_LOCALE
953 const char *system_default_locale = NULL;
957 DEBUG_INITIALIZATION_set((PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
960 # define DEBUG_LOCALE_INIT(category, locale, result) \
962 if (debug_initialization) { \
963 PerlIO_printf(Perl_debug_log, \
965 __FILE__, __LINE__, \
966 _setlocale_debug_string(category, \
972 # define DEBUG_LOCALE_INIT(a,b,c)
975 #ifndef LOCALE_ENVIRON_REQUIRED
976 PERL_UNUSED_VAR(done);
977 PERL_UNUSED_VAR(locale_param);
981 * Ultrix setlocale(..., "") fails if there are no environment
982 * variables from which to get a locale name.
987 sl_result = my_setlocale(LC_ALL, setlocale_init);
988 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
992 setlocale_failure = TRUE;
994 if (! setlocale_failure) {
995 # ifdef USE_LOCALE_CTYPE
996 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
999 curctype = my_setlocale(LC_CTYPE, locale_param);
1000 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
1002 setlocale_failure = TRUE;
1004 curctype = savepv(curctype);
1005 # endif /* USE_LOCALE_CTYPE */
1006 # ifdef USE_LOCALE_COLLATE
1007 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
1010 curcoll = my_setlocale(LC_COLLATE, locale_param);
1011 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
1013 setlocale_failure = TRUE;
1015 curcoll = savepv(curcoll);
1016 # endif /* USE_LOCALE_COLLATE */
1017 # ifdef USE_LOCALE_NUMERIC
1018 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
1021 curnum = my_setlocale(LC_NUMERIC, locale_param);
1022 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
1024 setlocale_failure = TRUE;
1026 curnum = savepv(curnum);
1027 # endif /* USE_LOCALE_NUMERIC */
1028 # ifdef USE_LOCALE_MESSAGES
1029 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
1032 sl_result = my_setlocale(LC_MESSAGES, locale_param);
1033 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
1035 setlocale_failure = TRUE;
1037 # endif /* USE_LOCALE_MESSAGES */
1038 # ifdef USE_LOCALE_MONETARY
1039 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
1042 sl_result = my_setlocale(LC_MONETARY, locale_param);
1043 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
1045 setlocale_failure = TRUE;
1047 # endif /* USE_LOCALE_MONETARY */
1050 # endif /* LC_ALL */
1052 #endif /* !LOCALE_ENVIRON_REQUIRED */
1054 /* We try each locale in the list until we get one that works, or exhaust
1055 * the list. Normally the loop is executed just once. But if setting the
1056 * locale fails, inside the loop we add fallback trials to the array and so
1057 * will execute the loop multiple times */
1058 trial_locales[0] = setlocale_init;
1059 trial_locales_count = 1;
1060 for (i= 0; i < trial_locales_count; i++) {
1061 const char * trial_locale = trial_locales[i];
1065 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
1066 * when i==0, but I (khw) don't think that behavior makes much
1068 setlocale_failure = FALSE;
1070 #ifdef SYSTEM_DEFAULT_LOCALE
1072 /* On Windows machines, an entry of "" after the 0th means to use
1073 * the system default locale, which we now proceed to get. */
1074 if (strEQ(trial_locale, "")) {
1077 /* Note that this may change the locale, but we are going to do
1078 * that anyway just below */
1079 system_default_locale = setlocale(LC_ALL, "");
1080 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
1082 /* Skip if invalid or it's already on the list of locales to
1084 if (! system_default_locale) {
1085 goto next_iteration;
1087 for (j = 0; j < trial_locales_count; j++) {
1088 if (strEQ(system_default_locale, trial_locales[j])) {
1089 goto next_iteration;
1093 trial_locale = system_default_locale;
1096 #endif /* SYSTEM_DEFAULT_LOCALE */
1100 sl_result = my_setlocale(LC_ALL, trial_locale);
1101 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
1103 setlocale_failure = TRUE;
1106 /* Since LC_ALL succeeded, it should have changed all the other
1107 * categories it can to its value; so we massage things so that the
1108 * setlocales below just return their category's current values.
1109 * This adequately handles the case in NetBSD where LC_COLLATE may
1110 * not be defined for a locale, and setting it individually will
1111 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
1112 * the POSIX locale. */
1113 trial_locale = NULL;
1117 if (!setlocale_failure) {
1118 #ifdef USE_LOCALE_CTYPE
1120 curctype = my_setlocale(LC_CTYPE, trial_locale);
1121 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
1123 setlocale_failure = TRUE;
1125 curctype = savepv(curctype);
1126 #endif /* USE_LOCALE_CTYPE */
1127 #ifdef USE_LOCALE_COLLATE
1129 curcoll = my_setlocale(LC_COLLATE, trial_locale);
1130 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
1132 setlocale_failure = TRUE;
1134 curcoll = savepv(curcoll);
1135 #endif /* USE_LOCALE_COLLATE */
1136 #ifdef USE_LOCALE_NUMERIC
1138 curnum = my_setlocale(LC_NUMERIC, trial_locale);
1139 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
1141 setlocale_failure = TRUE;
1143 curnum = savepv(curnum);
1144 #endif /* USE_LOCALE_NUMERIC */
1145 #ifdef USE_LOCALE_MESSAGES
1146 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
1147 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
1149 setlocale_failure = TRUE;
1150 #endif /* USE_LOCALE_MESSAGES */
1151 #ifdef USE_LOCALE_MONETARY
1152 sl_result = my_setlocale(LC_MONETARY, trial_locale);
1153 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
1155 setlocale_failure = TRUE;
1156 #endif /* USE_LOCALE_MONETARY */
1158 if (! setlocale_failure) { /* Success */
1163 /* Here, something failed; will need to try a fallback. */
1169 if (locwarn) { /* Output failure info only on the first one */
1172 PerlIO_printf(Perl_error_log,
1173 "perl: warning: Setting locale failed.\n");
1177 PerlIO_printf(Perl_error_log,
1178 "perl: warning: Setting locale failed for the categories:\n\t");
1179 # ifdef USE_LOCALE_CTYPE
1181 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
1182 # endif /* USE_LOCALE_CTYPE */
1183 # ifdef USE_LOCALE_COLLATE
1185 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
1186 # endif /* USE_LOCALE_COLLATE */
1187 # ifdef USE_LOCALE_NUMERIC
1189 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
1190 # endif /* USE_LOCALE_NUMERIC */
1191 PerlIO_printf(Perl_error_log, "and possibly others\n");
1195 PerlIO_printf(Perl_error_log,
1196 "perl: warning: Please check that your locale settings:\n");
1199 PerlIO_printf(Perl_error_log,
1200 "\tLANGUAGE = %c%s%c,\n",
1201 language ? '"' : '(',
1202 language ? language : "unset",
1203 language ? '"' : ')');
1206 PerlIO_printf(Perl_error_log,
1207 "\tLC_ALL = %c%s%c,\n",
1209 lc_all ? lc_all : "unset",
1210 lc_all ? '"' : ')');
1212 #if defined(USE_ENVIRON_ARRAY)
1215 for (e = environ; *e; e++) {
1216 if (strnEQ(*e, "LC_", 3)
1217 && strnNE(*e, "LC_ALL=", 7)
1218 && (p = strchr(*e, '=')))
1219 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1220 (int)(p - *e), *e, p + 1);
1224 PerlIO_printf(Perl_error_log,
1225 "\t(possibly more locale environment variables)\n");
1228 PerlIO_printf(Perl_error_log,
1229 "\tLANG = %c%s%c\n",
1231 lang ? lang : "unset",
1234 PerlIO_printf(Perl_error_log,
1235 " are supported and installed on your system.\n");
1238 /* Calculate what fallback locales to try. We have avoided this
1239 * until we have to, because failure is quite unlikely. This will
1240 * usually change the upper bound of the loop we are in.
1242 * Since the system's default way of setting the locale has not
1243 * found one that works, We use Perl's defined ordering: LC_ALL,
1244 * LANG, and the C locale. We don't try the same locale twice, so
1245 * don't add to the list if already there. (On POSIX systems, the
1246 * LC_ALL element will likely be a repeat of the 0th element "",
1247 * but there's no harm done by doing it explicitly.
1249 * Note that this tries the LC_ALL environment variable even on
1250 * systems which have no LC_ALL locale setting. This may or may
1251 * not have been originally intentional, but there's no real need
1252 * to change the behavior. */
1254 for (j = 0; j < trial_locales_count; j++) {
1255 if (strEQ(lc_all, trial_locales[j])) {
1259 trial_locales[trial_locales_count++] = lc_all;
1264 for (j = 0; j < trial_locales_count; j++) {
1265 if (strEQ(lang, trial_locales[j])) {
1269 trial_locales[trial_locales_count++] = lang;
1273 #if defined(WIN32) && defined(LC_ALL)
1274 /* For Windows, we also try the system default locale before "C".
1275 * (If there exists a Windows without LC_ALL we skip this because
1276 * it gets too complicated. For those, the "C" is the next
1277 * fallback possibility). The "" is the same as the 0th element of
1278 * the array, but the code at the loop above knows to treat it
1279 * differently when not the 0th */
1280 trial_locales[trial_locales_count++] = "";
1283 for (j = 0; j < trial_locales_count; j++) {
1284 if (strEQ("C", trial_locales[j])) {
1288 trial_locales[trial_locales_count++] = "C";
1291 } /* end of first time through the loop */
1297 } /* end of looping through the trial locales */
1299 if (ok < 1) { /* If we tried to fallback */
1301 if (! setlocale_failure) { /* fallback succeeded */
1302 msg = "Falling back to";
1304 else { /* fallback failed */
1306 /* We dropped off the end of the loop, so have to decrement i to
1307 * get back to the value the last time through */
1311 msg = "Failed to fall back to";
1313 /* To continue, we should use whatever values we've got */
1314 #ifdef USE_LOCALE_CTYPE
1316 curctype = savepv(setlocale(LC_CTYPE, NULL));
1317 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
1318 #endif /* USE_LOCALE_CTYPE */
1319 #ifdef USE_LOCALE_COLLATE
1321 curcoll = savepv(setlocale(LC_COLLATE, NULL));
1322 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
1323 #endif /* USE_LOCALE_COLLATE */
1324 #ifdef USE_LOCALE_NUMERIC
1326 curnum = savepv(setlocale(LC_NUMERIC, NULL));
1327 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
1328 #endif /* USE_LOCALE_NUMERIC */
1332 const char * description;
1333 const char * name = "";
1334 if (strEQ(trial_locales[i], "C")) {
1335 description = "the standard locale";
1338 #ifdef SYSTEM_DEFAULT_LOCALE
1339 else if (strEQ(trial_locales[i], "")) {
1340 description = "the system default locale";
1341 if (system_default_locale) {
1342 name = system_default_locale;
1345 #endif /* SYSTEM_DEFAULT_LOCALE */
1347 description = "a fallback locale";
1348 name = trial_locales[i];
1350 if (name && strNE(name, "")) {
1351 PerlIO_printf(Perl_error_log,
1352 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1355 PerlIO_printf(Perl_error_log,
1356 "perl: warning: %s %s.\n", msg, description);
1359 } /* End of tried to fallback */
1361 #ifdef USE_LOCALE_CTYPE
1362 new_ctype(curctype);
1363 #endif /* USE_LOCALE_CTYPE */
1365 #ifdef USE_LOCALE_COLLATE
1366 new_collate(curcoll);
1367 #endif /* USE_LOCALE_COLLATE */
1369 #ifdef USE_LOCALE_NUMERIC
1370 new_numeric(curnum);
1371 #endif /* USE_LOCALE_NUMERIC */
1373 #if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
1374 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1375 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1376 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1377 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1379 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
1381 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
1382 This is an alternative to using the -C command line switch
1383 (the -C if present will override this). */
1385 const char *p = PerlEnv_getenv("PERL_UNICODE");
1386 PL_unicode = p ? parse_unicode_opts(&p) : 0;
1387 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1392 #ifdef USE_LOCALE_CTYPE
1394 #endif /* USE_LOCALE_CTYPE */
1395 #ifdef USE_LOCALE_COLLATE
1397 #endif /* USE_LOCALE_COLLATE */
1398 #ifdef USE_LOCALE_NUMERIC
1400 #endif /* USE_LOCALE_NUMERIC */
1409 #else /* !USE_LOCALE */
1410 PERL_UNUSED_ARG(printwarn);
1411 #endif /* USE_LOCALE */
1414 /* So won't continue to output stuff */
1415 DEBUG_INITIALIZATION_set(FALSE);
1421 #ifdef USE_LOCALE_COLLATE
1424 Perl__mem_collxfrm(pTHX_ const char *input_string,
1425 STRLEN len, /* Length of 'input_string' */
1426 STRLEN *xlen, /* Set to length of returned string
1427 (not including the collation index
1429 bool utf8 /* Is the input in UTF-8? */
1433 /* _mem_collxfrm() is a bit like strxfrm() but with two important
1434 * differences. First, it handles embedded NULs. Second, it allocates a bit
1435 * more memory than needed for the transformed data itself. The real
1436 * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
1437 * the length of that, and doesn't include the collation index size.
1438 * Please see sv_collxfrm() to see how this is used. */
1440 #define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
1442 char * s = (char *) input_string;
1443 STRLEN s_strlen = strlen(input_string);
1445 STRLEN xAlloc; /* xalloc is a reserved word in VC */
1446 STRLEN length_in_chars;
1447 bool first_time = TRUE; /* Cleared after first loop iteration */
1449 PERL_ARGS_ASSERT__MEM_COLLXFRM;
1451 /* Must be NUL-terminated */
1452 assert(*(input_string + len) == '\0');
1454 /* If this locale has defective collation, skip */
1455 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
1456 DEBUG_L(PerlIO_printf(Perl_debug_log,
1457 "_mem_collxfrm: locale's collation is defective\n"));
1461 /* Replace any embedded NULs with the control that sorts before any others.
1462 * This will give as good as possible results on strings that don't
1463 * otherwise contain that character, but otherwise there may be
1464 * less-than-perfect results with that character and NUL. This is
1465 * unavoidable unless we replace strxfrm with our own implementation.
1467 * This is one of the few places in the perl core, where we can use
1468 * standard functions like strlen() and strcat(). It's because we're
1469 * looking for NULs. */
1470 if (s_strlen < len) {
1473 STRLEN cur_min_char_len;
1474 int try_non_controls;
1476 /* If we don't know what control character sorts lowest for this
1477 * locale, find it */
1478 if (*PL_strxfrm_min_char == '\0') {
1481 U8 cur_min_cp = 1; /* The code point that sorts lowest, so far */
1483 char * cur_min_x = NULL; /* And its xfrm, (except it also
1484 includes the collation index
1487 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
1489 /* Unlikely, but it may be that no control will work to replace
1490 * NUL, in which case we instead look for any character */
1491 for (try_non_controls = 0;
1492 try_non_controls < 2;
1495 /* Look through all legal code points (NUL isn't) */
1496 for (j = 1; j < 256; j++) {
1497 char * x; /* j's xfrm plus collation index */
1498 STRLEN x_len; /* length of 'x' */
1499 STRLEN trial_len = 1;
1501 /* Create a 1 byte string of the current code point, but
1502 * with room to be 2 bytes */
1503 char cur_source[] = { (char) j, '\0' , '\0' };
1505 if (PL_in_utf8_COLLATE_locale) {
1506 if (! try_non_controls && ! isCNTRL_L1(j)) {
1510 /* If needs to be 2 bytes, find them */
1511 if (! UVCHR_IS_INVARIANT(j)) {
1512 char * d = cur_source;
1513 append_utf8_from_native_byte((U8) j, (U8 **) &d);
1517 else if (! try_non_controls && ! isCNTRL_LC(j)) {
1521 /* Then transform it */
1522 x = _mem_collxfrm(cur_source, trial_len, &x_len,
1523 PL_in_utf8_COLLATE_locale);
1525 /* Ignore any character that didn't successfully transform
1531 /* If this character's transformation is lower than
1532 * the current lowest, this one becomes the lowest */
1533 if ( cur_min_x == NULL
1534 || strLT(x + COLLXFRM_HDR_LEN,
1535 cur_min_x + COLLXFRM_HDR_LEN))
1537 PL_strxfrm_min_char[0] = cur_source[0];
1538 PL_strxfrm_min_char[1] = cur_source[1];
1539 PL_strxfrm_min_char[2] = cur_source[2];
1548 } /* end of loop through all bytes */
1554 /* Unlikely, but possible, if there aren't any controls that
1555 * work in the locale, repeat the loop, looking for any
1556 * character that works */
1557 DEBUG_L(PerlIO_printf(Perl_debug_log,
1558 "_mem_collxfrm: No control worked. Trying non-controls\n"));
1562 DEBUG_L(PerlIO_printf(Perl_debug_log,
1563 "_mem_collxfrm: Couldn't find any character to replace"
1564 " embedded NULs in locale %s with", PL_collation_name));
1568 DEBUG_L(PerlIO_printf(Perl_debug_log,
1569 "_mem_collxfrm: Replacing embedded NULs in locale %s with "
1570 "0x%02X\n", PL_collation_name, cur_min_cp));
1572 Safefree(cur_min_x);
1575 /* The worst case length for the replaced string would be if every
1576 * character in it is NUL. Multiply that by the length of each
1577 * replacement, and allow for a trailing NUL */
1578 cur_min_char_len = strlen(PL_strxfrm_min_char);
1579 Newx(sans_nuls, (len * cur_min_char_len) + 1, char);
1582 /* Replace each NUL with the lowest collating control. Loop until have
1583 * exhausted all the NULs */
1584 while (s + s_strlen < e) {
1585 strcat(sans_nuls, s);
1587 /* Do the actual replacement */
1588 strcat(sans_nuls, PL_strxfrm_min_char);
1590 /* Move past the input NUL */
1592 s_strlen = strlen(s);
1595 /* And add anything that trails the final NUL */
1596 strcat(sans_nuls, s);
1598 /* Switch so below we transform this modified string */
1603 /* Make sure the UTF8ness of the string and locale match */
1604 if (utf8 != PL_in_utf8_COLLATE_locale) {
1605 const char * const t = s; /* Temporary so we can later find where the
1608 /* Here they don't match. Change the string's to be what the locale is
1611 if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
1612 s = (char *) bytes_to_utf8((const U8 *) s, &len);
1615 else { /* locale is not UTF-8; but input is; downgrade the input */
1617 s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
1619 /* If the downgrade was successful we are done, but if the input
1620 * contains things that require UTF-8 to represent, have to do
1621 * damage control ... */
1622 if (UNLIKELY(utf8)) {
1624 /* What we do is construct a non-UTF-8 string with
1625 * 1) the characters representable by a single byte converted
1626 * to be so (if necessary);
1627 * 2) and the rest converted to collate the same as the
1628 * highest collating representable character. That makes
1629 * them collate at the end. This is similar to how we
1630 * handle embedded NULs, but we use the highest collating
1631 * code point instead of the smallest. Like the NUL case,
1632 * this isn't perfect, but is the best we can reasonably
1633 * do. Every above-255 code point will sort the same as
1634 * the highest-sorting 0-255 code point. If that code
1635 * point can combine in a sequence with some other code
1636 * points for weight calculations, us changing something to
1637 * be it can adversely affect the results. But in most
1638 * cases, it should work reasonably. And note that this is
1639 * really an illegal situation: using code points above 255
1640 * on a locale where only 0-255 are valid. If two strings
1641 * sort entirely equal, then the sort order for the
1642 * above-255 code points will be in code point order. */
1646 /* If we haven't calculated the code point with the maximum
1647 * collating order for this locale, do so now */
1648 if (! PL_strxfrm_max_cp) {
1651 /* The current transformed string that collates the
1652 * highest (except it also includes the prefixed collation
1654 char * cur_max_x = NULL;
1656 /* Look through all legal code points (NUL isn't) */
1657 for (j = 1; j < 256; j++) {
1661 /* Create a 1-char string of the current code point. */
1662 char cur_source[] = { (char) j, '\0' };
1664 /* Then transform it */
1665 x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
1667 /* If something went wrong (which it shouldn't), just
1668 * ignore this code point */
1673 /* If this character's transformation is higher than
1674 * the current highest, this one becomes the highest */
1675 if ( cur_max_x == NULL
1676 || strGT(x + COLLXFRM_HDR_LEN,
1677 cur_max_x + COLLXFRM_HDR_LEN))
1679 PL_strxfrm_max_cp = j;
1688 DEBUG_L(PerlIO_printf(Perl_debug_log,
1689 "_mem_collxfrm: Couldn't find any character to"
1690 " replace above-Latin1 chars in locale %s with",
1691 PL_collation_name));
1695 DEBUG_L(PerlIO_printf(Perl_debug_log,
1696 "_mem_collxfrm: highest 1-byte collating character"
1697 " in locale %s is 0x%02X\n",
1699 PL_strxfrm_max_cp));
1701 Safefree(cur_max_x);
1704 /* Here we know which legal code point collates the highest.
1705 * We are ready to construct the non-UTF-8 string. The length
1706 * will be at least 1 byte smaller than the input string
1707 * (because we changed at least one 2-byte character into a
1708 * single byte), but that is eaten up by the trailing NUL */
1715 for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
1717 if (UTF8_IS_INVARIANT(cur_char)) {
1720 else if (UTF8_IS_DOWNGRADEABLE_START(cur_char)) {
1721 s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
1723 else { /* Replace illegal cp with highest collating
1725 s[d++] = PL_strxfrm_max_cp;
1729 Renew(s, d, char); /* Free up unused space */
1734 /* Here, we have constructed a modified version of the input. It could
1735 * be that we already had a modified copy before we did this version.
1736 * If so, that copy is no longer needed */
1737 if (t != input_string) {
1742 length_in_chars = (utf8)
1743 ? utf8_length((U8 *) s, (U8 *) s + len)
1746 /* The first element in the output is the collation id, used by
1747 * sv_collxfrm(); then comes the space for the transformed string. The
1748 * equation should give us a good estimate as to how much is needed */
1749 xAlloc = COLLXFRM_HDR_LEN
1751 + (PL_collxfrm_mult * length_in_chars);
1752 Newx(xbuf, xAlloc, char);
1753 if (UNLIKELY(! xbuf)) {
1754 DEBUG_L(PerlIO_printf(Perl_debug_log,
1755 "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
1759 /* Store the collation id */
1760 *(U32*)xbuf = PL_collation_ix;
1762 /* Then the transformation of the input. We loop until successful, or we
1766 *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
1768 /* If the transformed string occupies less space than we told strxfrm()
1769 * was available, it means it successfully transformed the whole
1771 if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
1773 /* Some systems include a trailing NUL in the returned length.
1774 * Ignore it, using a loop in case multiple trailing NULs are
1777 && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
1782 /* If the first try didn't get it, it means our prediction was low.
1783 * Modify the coefficients so that we predict a larger value in any
1784 * future transformations */
1786 STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
1787 STRLEN computed_guess = PL_collxfrm_base
1788 + (PL_collxfrm_mult * length_in_chars);
1790 /* On zero-length input, just keep current slope instead of
1792 const STRLEN new_m = (length_in_chars != 0)
1793 ? needed / length_in_chars
1796 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
1797 "%s: %d: initial size of %zu bytes for a length "
1798 "%zu string was insufficient, %zu needed\n",
1800 computed_guess, length_in_chars, needed));
1802 /* If slope increased, use it, but discard this result for
1803 * length 1 strings, as we can't be sure that it's a real slope
1805 if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
1807 STRLEN old_m = PL_collxfrm_mult;
1808 STRLEN old_b = PL_collxfrm_base;
1810 PL_collxfrm_mult = new_m;
1811 PL_collxfrm_base = 1; /* +1 For trailing NUL */
1812 computed_guess = PL_collxfrm_base
1813 + (PL_collxfrm_mult * length_in_chars);
1814 if (computed_guess < needed) {
1815 PL_collxfrm_base += needed - computed_guess;
1818 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
1819 "%s: %d: slope is now %zu; was %zu, base "
1820 "is now %zu; was %zu\n",
1822 PL_collxfrm_mult, old_m,
1823 PL_collxfrm_base, old_b));
1825 else { /* Slope didn't change, but 'b' did */
1826 const STRLEN new_b = needed
1829 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
1830 "%s: %d: base is now %zu; was %zu\n",
1832 new_b, PL_collxfrm_base));
1833 PL_collxfrm_base = new_b;
1840 if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
1841 DEBUG_L(PerlIO_printf(Perl_debug_log,
1842 "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
1843 *xlen, PERL_INT_MAX));
1847 /* A well-behaved strxfrm() returns exactly how much space it needs
1848 * (usually not including the trailing NUL) when it fails due to not
1849 * enough space being provided. Assume that this is the case unless
1850 * it's been proven otherwise */
1851 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
1852 xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
1854 else { /* Here, either:
1855 * 1) The strxfrm() has previously shown bad behavior; or
1856 * 2) It isn't the first time through the loop, which means
1857 * that the strxfrm() is now showing bad behavior, because
1858 * we gave it what it said was needed in the previous
1859 * iteration, and it came back saying it needed still more.
1860 * (Many versions of cygwin fit this. When the buffer size
1861 * isn't sufficient, they return the input size instead of
1862 * how much is needed.)
1863 * Increase the buffer size by a fixed percentage and try again.
1865 xAlloc += (xAlloc / 4) + 1;
1866 PL_strxfrm_is_behaved = FALSE;
1869 if (DEBUG_Lv_TEST || debug_initialization) {
1870 PerlIO_printf(Perl_debug_log,
1871 "_mem_collxfrm required more space than previously calculated"
1872 " for locale %s, trying again with new guess=%d+%zu\n",
1873 PL_collation_name, (int) COLLXFRM_HDR_LEN,
1874 xAlloc - COLLXFRM_HDR_LEN);
1879 Renew(xbuf, xAlloc, char);
1880 if (UNLIKELY(! xbuf)) {
1881 DEBUG_L(PerlIO_printf(Perl_debug_log,
1882 "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
1891 if (DEBUG_Lv_TEST || debug_initialization) {
1894 print_collxfrm_input_and_return(s, s + len, xlen, utf8);
1895 PerlIO_printf(Perl_debug_log, "Its xfrm is:");
1896 for (i = COLLXFRM_HDR_LEN; i < *xlen + COLLXFRM_HDR_LEN; i++) {
1897 PerlIO_printf(Perl_debug_log, " %02x", (U8) xbuf[i]);
1899 PerlIO_printf(Perl_debug_log, "\n");
1903 /* Free up unneeded space; retain ehough for trailing NUL */
1904 Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
1906 if (s != input_string) {
1914 if (s != input_string) {
1919 if (DEBUG_Lv_TEST || debug_initialization) {
1920 print_collxfrm_input_and_return(s, s + len, NULL, utf8);
1929 S_print_collxfrm_input_and_return(pTHX_
1930 const char * const s,
1931 const char * const e,
1932 const STRLEN * const xlen,
1936 bool prev_was_printable = TRUE;
1937 bool first_time = TRUE;
1939 PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
1941 PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%d]: returning ",
1944 PerlIO_printf(Perl_debug_log, "%"UVuf"", (UV) *xlen);
1947 PerlIO_printf(Perl_debug_log, "NULL");
1949 PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
1954 ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
1957 if (! prev_was_printable) {
1958 PerlIO_printf(Perl_debug_log, " ");
1960 PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
1961 prev_was_printable = TRUE;
1965 PerlIO_printf(Perl_debug_log, " ");
1967 PerlIO_printf(Perl_debug_log, "%02"UVXf"", cp);
1968 prev_was_printable = FALSE;
1970 t += (is_utf8) ? UTF8SKIP(t) : 1;
1974 PerlIO_printf(Perl_debug_log, "'\n");
1977 #endif /* #ifdef DEBUGGING */
1979 #endif /* USE_LOCALE_COLLATE */
1984 Perl__is_cur_LC_category_utf8(pTHX_ int category)
1986 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1987 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
1988 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
1989 * could give the wrong result. The result will very likely be correct for
1990 * languages that have commonly used non-ASCII characters, but for notably
1991 * English, it comes down to if the locale's name ends in something like
1992 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
1994 char *save_input_locale = NULL;
1998 assert(category != LC_ALL);
2001 /* First dispose of the trivial cases */
2002 save_input_locale = setlocale(category, NULL);
2003 if (! save_input_locale) {
2004 DEBUG_L(PerlIO_printf(Perl_debug_log,
2005 "Could not find current locale for category %d\n",
2007 return FALSE; /* XXX maybe should croak */
2009 save_input_locale = stdize_locale(savepv(save_input_locale));
2010 if (isNAME_C_OR_POSIX(save_input_locale)) {
2011 DEBUG_L(PerlIO_printf(Perl_debug_log,
2012 "Current locale for category %d is %s\n",
2013 category, save_input_locale));
2014 Safefree(save_input_locale);
2018 #if defined(USE_LOCALE_CTYPE) \
2019 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
2021 { /* Next try nl_langinfo or MB_CUR_MAX if available */
2023 char *save_ctype_locale = NULL;
2026 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
2028 /* Get the current LC_CTYPE locale */
2029 save_ctype_locale = setlocale(LC_CTYPE, NULL);
2030 if (! save_ctype_locale) {
2031 DEBUG_L(PerlIO_printf(Perl_debug_log,
2032 "Could not find current locale for LC_CTYPE\n"));
2033 goto cant_use_nllanginfo;
2035 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
2037 /* If LC_CTYPE and the desired category use the same locale, this
2038 * means that finding the value for LC_CTYPE is the same as finding
2039 * the value for the desired category. Otherwise, switch LC_CTYPE
2040 * to the desired category's locale */
2041 if (strEQ(save_ctype_locale, save_input_locale)) {
2042 Safefree(save_ctype_locale);
2043 save_ctype_locale = NULL;
2045 else if (! setlocale(LC_CTYPE, save_input_locale)) {
2046 DEBUG_L(PerlIO_printf(Perl_debug_log,
2047 "Could not change LC_CTYPE locale to %s\n",
2048 save_input_locale));
2049 Safefree(save_ctype_locale);
2050 goto cant_use_nllanginfo;
2054 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
2055 save_input_locale));
2057 /* Here the current LC_CTYPE is set to the locale of the category whose
2058 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
2059 * should give the correct results */
2061 # if defined(HAS_NL_LANGINFO) && defined(CODESET)
2063 char *codeset = nl_langinfo(CODESET);
2064 if (codeset && strNE(codeset, "")) {
2065 codeset = savepv(codeset);
2067 /* If we switched LC_CTYPE, switch back */
2068 if (save_ctype_locale) {
2069 setlocale(LC_CTYPE, save_ctype_locale);
2070 Safefree(save_ctype_locale);
2073 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
2074 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
2076 DEBUG_L(PerlIO_printf(Perl_debug_log,
2077 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
2080 Safefree(save_input_locale);
2088 /* Here, either we don't have nl_langinfo, or it didn't return a
2089 * codeset. Try MB_CUR_MAX */
2091 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
2092 * Unicode code point. Since UTF-8 is the only non-single byte
2093 * encoding we handle, we just say any such encoding is UTF-8, and if
2094 * turns out to be wrong, other things will fail */
2095 is_utf8 = MB_CUR_MAX >= 4;
2097 DEBUG_L(PerlIO_printf(Perl_debug_log,
2098 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
2099 (int) MB_CUR_MAX, is_utf8));
2101 Safefree(save_input_locale);
2105 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
2106 * since they are both in the C99 standard. We can feed a known byte
2107 * string to the latter function, and check that it gives the expected
2111 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
2113 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
2114 != strlen(HYPHEN_UTF8)
2115 || wc != (wchar_t) 0x2010)
2118 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
2119 DEBUG_L(PerlIO_printf(Perl_debug_log,
2120 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
2121 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
2126 /* If we switched LC_CTYPE, switch back */
2127 if (save_ctype_locale) {
2128 setlocale(LC_CTYPE, save_ctype_locale);
2129 Safefree(save_ctype_locale);
2136 cant_use_nllanginfo:
2138 #else /* nl_langinfo should work if available, so don't bother compiling this
2139 fallback code. The final fallback of looking at the name is
2140 compiled, and will be executed if nl_langinfo fails */
2142 /* nl_langinfo not available or failed somehow. Next try looking at the
2143 * currency symbol to see if it disambiguates things. Often that will be
2144 * in the native script, and if the symbol isn't in UTF-8, we know that the
2145 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
2146 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
2149 #ifdef HAS_LOCALECONV
2150 # ifdef USE_LOCALE_MONETARY
2152 char *save_monetary_locale = NULL;
2153 bool only_ascii = FALSE;
2154 bool is_utf8 = FALSE;
2157 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
2158 * the desired category, if it isn't that locale already */
2160 if (category != LC_MONETARY) {
2162 save_monetary_locale = setlocale(LC_MONETARY, NULL);
2163 if (! save_monetary_locale) {
2164 DEBUG_L(PerlIO_printf(Perl_debug_log,
2165 "Could not find current locale for LC_MONETARY\n"));
2166 goto cant_use_monetary;
2168 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
2170 if (strEQ(save_monetary_locale, save_input_locale)) {
2171 Safefree(save_monetary_locale);
2172 save_monetary_locale = NULL;
2174 else if (! setlocale(LC_MONETARY, save_input_locale)) {
2175 DEBUG_L(PerlIO_printf(Perl_debug_log,
2176 "Could not change LC_MONETARY locale to %s\n",
2177 save_input_locale));
2178 Safefree(save_monetary_locale);
2179 goto cant_use_monetary;
2183 /* Here the current LC_MONETARY is set to the locale of the category
2184 * whose information is desired. */
2188 || ! lc->currency_symbol
2189 || is_invariant_string((U8 *) lc->currency_symbol, 0))
2191 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));
2195 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
2198 /* If we changed it, restore LC_MONETARY to its original locale */
2199 if (save_monetary_locale) {
2200 setlocale(LC_MONETARY, save_monetary_locale);
2201 Safefree(save_monetary_locale);
2206 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
2207 * otherwise assume the locale is UTF-8 if and only if the symbol
2208 * is non-ascii UTF-8. */
2209 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
2210 save_input_locale, is_utf8));
2211 Safefree(save_input_locale);
2217 # endif /* USE_LOCALE_MONETARY */
2218 #endif /* HAS_LOCALECONV */
2220 #if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
2222 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
2223 * the names of the months and weekdays, timezone, and am/pm indicator */
2225 char *save_time_locale = NULL;
2227 bool is_dst = FALSE;
2231 char * formatted_time;
2234 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
2235 * desired category, if it isn't that locale already */
2237 if (category != LC_TIME) {
2239 save_time_locale = setlocale(LC_TIME, NULL);
2240 if (! save_time_locale) {
2241 DEBUG_L(PerlIO_printf(Perl_debug_log,
2242 "Could not find current locale for LC_TIME\n"));
2245 save_time_locale = stdize_locale(savepv(save_time_locale));
2247 if (strEQ(save_time_locale, save_input_locale)) {
2248 Safefree(save_time_locale);
2249 save_time_locale = NULL;
2251 else if (! setlocale(LC_TIME, save_input_locale)) {
2252 DEBUG_L(PerlIO_printf(Perl_debug_log,
2253 "Could not change LC_TIME locale to %s\n",
2254 save_input_locale));
2255 Safefree(save_time_locale);
2260 /* Here the current LC_TIME is set to the locale of the category
2261 * whose information is desired. Look at all the days of the week and
2262 * month names, and the timezone and am/pm indicator for UTF-8 variant
2263 * characters. The first such a one found will tell us if the locale
2264 * is UTF-8 or not */
2266 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
2267 formatted_time = my_strftime("%A %B %Z %p",
2268 0, 0, hour, dom, month, 112, 0, 0, is_dst);
2269 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
2271 /* Here, we didn't find a non-ASCII. Try the next time through
2272 * with the complemented dst and am/pm, and try with the next
2273 * weekday. After we have gotten all weekdays, try the next
2276 hour = (hour + 12) % 24;
2284 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
2285 * false otherwise. But first, restore LC_TIME to its original
2286 * locale if we changed it */
2287 if (save_time_locale) {
2288 setlocale(LC_TIME, save_time_locale);
2289 Safefree(save_time_locale);
2292 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
2294 is_utf8_string((U8 *) formatted_time, 0)));
2295 Safefree(save_input_locale);
2296 return is_utf8_string((U8 *) formatted_time, 0);
2299 /* Falling off the end of the loop indicates all the names were just
2300 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
2301 * to its original locale */
2302 if (save_time_locale) {
2303 setlocale(LC_TIME, save_time_locale);
2304 Safefree(save_time_locale);
2306 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));
2312 #if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
2314 /* This code is ifdefd out because it was found to not be necessary in testing
2315 * on our dromedary test machine, which has over 700 locales. There, this
2316 * added no value to looking at the currency symbol and the time strings. I
2317 * left it in so as to avoid rewriting it if real-world experience indicates
2318 * that dromedary is an outlier. Essentially, instead of returning abpve if we
2319 * haven't found illegal utf8, we continue on and examine all the strerror()
2320 * messages on the platform for utf8ness. If all are ASCII, we still don't
2321 * know the answer; but otherwise we have a pretty good indication of the
2322 * utf8ness. The reason this doesn't help much is that the messages may not
2323 * have been translated into the locale. The currency symbol and time strings
2324 * are much more likely to have been translated. */
2327 bool is_utf8 = FALSE;
2328 bool non_ascii = FALSE;
2329 char *save_messages_locale = NULL;
2330 const char * errmsg = NULL;
2332 /* Like above, we set LC_MESSAGES to the locale of the desired
2333 * category, if it isn't that locale already */
2335 if (category != LC_MESSAGES) {
2337 save_messages_locale = setlocale(LC_MESSAGES, NULL);
2338 if (! save_messages_locale) {
2339 DEBUG_L(PerlIO_printf(Perl_debug_log,
2340 "Could not find current locale for LC_MESSAGES\n"));
2341 goto cant_use_messages;
2343 save_messages_locale = stdize_locale(savepv(save_messages_locale));
2345 if (strEQ(save_messages_locale, save_input_locale)) {
2346 Safefree(save_messages_locale);
2347 save_messages_locale = NULL;
2349 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
2350 DEBUG_L(PerlIO_printf(Perl_debug_log,
2351 "Could not change LC_MESSAGES locale to %s\n",
2352 save_input_locale));
2353 Safefree(save_messages_locale);
2354 goto cant_use_messages;
2358 /* Here the current LC_MESSAGES is set to the locale of the category
2359 * whose information is desired. Look through all the messages. We
2360 * can't use Strerror() here because it may expand to code that
2361 * segfaults in miniperl */
2363 for (e = 0; e <= sys_nerr; e++) {
2365 errmsg = sys_errlist[e];
2366 if (errno || !errmsg) {
2369 errmsg = savepv(errmsg);
2370 if (! is_invariant_string((U8 *) errmsg, 0)) {
2372 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
2378 /* And, if we changed it, restore LC_MESSAGES to its original locale */
2379 if (save_messages_locale) {
2380 setlocale(LC_MESSAGES, save_messages_locale);
2381 Safefree(save_messages_locale);
2386 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
2387 * any non-ascii means it is one; otherwise we assume it isn't */
2388 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
2391 Safefree(save_input_locale);
2395 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));
2401 #endif /* the code that is compiled when no nl_langinfo */
2403 #ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
2405 /* As a last resort, look at the locale name to see if it matches
2406 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
2407 * return of setlocale(), is actually defined to be opaque, so we can't
2408 * really rely on the absence of various substrings in the name to indicate
2409 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
2410 * be a UTF-8 locale. Similarly for the other common names */
2412 final_pos = strlen(save_input_locale) - 1;
2413 if (final_pos >= 3) {
2414 char *name = save_input_locale;
2416 /* Find next 'U' or 'u' and look from there */
2417 while ((name += strcspn(name, "Uu") + 1)
2418 <= save_input_locale + final_pos - 2)
2420 if (!isALPHA_FOLD_NE(*name, 't')
2421 || isALPHA_FOLD_NE(*(name + 1), 'f'))
2426 if (*(name) == '-') {
2427 if ((name > save_input_locale + final_pos - 1)) {
2432 if (*(name) == '8') {
2433 DEBUG_L(PerlIO_printf(Perl_debug_log,
2434 "Locale %s ends with UTF-8 in name\n",
2435 save_input_locale));
2436 Safefree(save_input_locale);
2440 DEBUG_L(PerlIO_printf(Perl_debug_log,
2441 "Locale %s doesn't end with UTF-8 in name\n",
2442 save_input_locale));
2447 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
2449 && *(save_input_locale + final_pos - 0) == '1'
2450 && *(save_input_locale + final_pos - 1) == '0'
2451 && *(save_input_locale + final_pos - 2) == '0'
2452 && *(save_input_locale + final_pos - 3) == '5'
2453 && *(save_input_locale + final_pos - 4) == '6')
2455 DEBUG_L(PerlIO_printf(Perl_debug_log,
2456 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
2457 save_input_locale));
2458 Safefree(save_input_locale);
2463 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
2464 * since we are about to return FALSE anyway, there is no point in doing
2465 * this extra work */
2467 if (instr(save_input_locale, "8859")) {
2468 DEBUG_L(PerlIO_printf(Perl_debug_log,
2469 "Locale %s has 8859 in name, not UTF-8 locale\n",
2470 save_input_locale));
2471 Safefree(save_input_locale);
2476 DEBUG_L(PerlIO_printf(Perl_debug_log,
2477 "Assuming locale %s is not a UTF-8 locale\n",
2478 save_input_locale));
2479 Safefree(save_input_locale);
2487 Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
2490 /* Internal function which returns if we are in the scope of a pragma that
2491 * enables the locale category 'category'. 'compiling' should indicate if
2492 * this is during the compilation phase (TRUE) or not (FALSE). */
2494 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
2496 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
2497 if (! categories || categories == &PL_sv_placeholder) {
2501 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
2502 * a valid unsigned */
2503 assert(category >= -1);
2504 return cBOOL(SvUV(categories) & (1U << (category + 1)));
2508 Perl_my_strerror(pTHX_ const int errnum)
2510 /* Returns a mortalized copy of the text of the error message associated
2511 * with 'errnum'. It uses the current locale's text unless the platform
2512 * doesn't have the LC_MESSAGES category or we are not being called from
2513 * within the scope of 'use locale'. In the former case, it uses whatever
2514 * strerror returns; in the latter case it uses the text from the C locale.
2516 * The function just calls strerror(), but temporarily switches, if needed,
2517 * to the C locale */
2521 #ifdef USE_LOCALE_MESSAGES /* If platform doesn't have messages category, we
2522 don't do any switching to the C locale; we just
2523 use whatever strerror() returns */
2524 const bool within_locale_scope = IN_LC(LC_MESSAGES);
2528 # ifdef USE_THREAD_SAFE_LOCALE
2529 locale_t save_locale;
2532 bool locale_is_C = FALSE;
2534 /* We have a critical section to prevent another thread from changing the
2535 * locale out from under us (or zapping the buffer returned from
2541 if (! within_locale_scope) {
2544 # ifdef USE_THREAD_SAFE_LOCALE /* Use the thread-safe locale functions */
2546 save_locale = uselocale(PL_C_locale_obj);
2547 if (! save_locale) {
2548 DEBUG_L(PerlIO_printf(Perl_debug_log,
2549 "uselocale failed, errno=%d\n", errno));
2552 # else /* Not thread-safe build */
2554 save_locale = setlocale(LC_MESSAGES, NULL);
2555 if (! save_locale) {
2556 DEBUG_L(PerlIO_printf(Perl_debug_log,
2557 "setlocale failed, errno=%d\n", errno));
2560 locale_is_C = isNAME_C_OR_POSIX(save_locale);
2562 /* Switch to the C locale if not already in it */
2563 if (! locale_is_C) {
2565 /* The setlocale() just below likely will zap 'save_locale', so
2567 save_locale = savepv(save_locale);
2568 setlocale(LC_MESSAGES, "C");
2574 } /* end of ! within_locale_scope */
2578 errstr = Strerror(errnum);
2580 errstr = savepv(errstr);
2584 #ifdef USE_LOCALE_MESSAGES
2586 if (! within_locale_scope) {
2589 # ifdef USE_THREAD_SAFE_LOCALE
2591 if (save_locale && ! uselocale(save_locale)) {
2592 DEBUG_L(PerlIO_printf(Perl_debug_log,
2593 "uselocale restore failed, errno=%d\n", errno));
2599 if (save_locale && ! locale_is_C) {
2600 if (! setlocale(LC_MESSAGES, save_locale)) {
2601 DEBUG_L(PerlIO_printf(Perl_debug_log,
2602 "setlocale restore failed, errno=%d\n", errno));
2604 Safefree(save_locale);
2618 =head1 Locale-related functions and macros
2620 =for apidoc sync_locale
2622 Changing the program's locale should be avoided by XS code. Nevertheless,
2623 certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
2624 happens, Perl needs to be told that the locale has changed. Use this function
2625 to do so, before returning to Perl.
2631 Perl_sync_locale(pTHX)
2634 #ifdef USE_LOCALE_CTYPE
2635 new_ctype(setlocale(LC_CTYPE, NULL));
2636 #endif /* USE_LOCALE_CTYPE */
2638 #ifdef USE_LOCALE_COLLATE
2639 new_collate(setlocale(LC_COLLATE, NULL));
2642 #ifdef USE_LOCALE_NUMERIC
2643 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
2644 new_numeric(setlocale(LC_NUMERIC, NULL));
2645 #endif /* USE_LOCALE_NUMERIC */
2649 #if defined(DEBUGGING) && defined(USE_LOCALE)
2652 Perl__setlocale_debug_string(const int category, /* category number,
2654 const char* const locale, /* locale name */
2656 /* return value from setlocale() when attempting to
2657 * set 'category' to 'locale' */
2658 const char* const retval)
2660 /* Returns a pointer to a NUL-terminated string in static storage with
2661 * added text about the info passed in. This is not thread safe and will
2662 * be overwritten by the next call, so this should be used just to
2663 * formulate a string to immediately print or savepv() on. */
2665 /* initialise to a non-null value to keep it out of BSS and so keep
2666 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
2667 static char ret[128] = "x";
2669 my_strlcpy(ret, "setlocale(", sizeof(ret));
2673 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
2677 my_strlcat(ret, "LC_ALL", sizeof(ret));
2682 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
2687 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
2692 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
2697 my_strlcat(ret, "LC_TIME", sizeof(ret));
2702 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
2707 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
2712 my_strlcat(ret, ", ", sizeof(ret));
2715 my_strlcat(ret, "\"", sizeof(ret));
2716 my_strlcat(ret, locale, sizeof(ret));
2717 my_strlcat(ret, "\"", sizeof(ret));
2720 my_strlcat(ret, "NULL", sizeof(ret));
2723 my_strlcat(ret, ") returned ", sizeof(ret));
2726 my_strlcat(ret, "\"", sizeof(ret));
2727 my_strlcat(ret, retval, sizeof(ret));
2728 my_strlcat(ret, "\"", sizeof(ret));
2731 my_strlcat(ret, "NULL", sizeof(ret));
2734 assert(strlen(ret) < sizeof(ret));
2743 * ex: set ts=8 sts=4 sw=4 et: