This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perllocale: Change headings so two aren't identical
[perl5.git] / locale.c
CommitLineData
98994639
HS
1/* locale.c
2 *
1129b882
NC
3 * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
4 * 2002, 2003, 2005, 2006, 2007, 2008 by Larry Wall and others
98994639
HS
5 *
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.
8 *
9 */
10
11/*
4ac71550 12 * A Elbereth Gilthoniel,
cdad3b53 13 * silivren penna míriel
4ac71550 14 * o menel aglar elenath!
cdad3b53 15 * Na-chaered palan-díriel
4ac71550
TC
16 * o galadhremmin ennorath,
17 * Fanuilos, le linnathon
18 * nef aear, si nef aearon!
19 *
20 * [p.238 of _The Lord of the Rings_, II/i: "Many Meetings"]
98994639
HS
21 */
22
166f8a29
DM
23/* utility functions for handling locale-specific stuff like what
24 * character represents the decimal point.
0d071d52
KW
25 *
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
a9ad02a8
KW
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
34 * of 'use locale'.
166f8a29
DM
35 */
36
98994639
HS
37#include "EXTERN.h"
38#define PERL_IN_LOCALE_C
39#include "perl.h"
40
b310b053
JH
41#ifdef I_LANGINFO
42# include <langinfo.h>
43#endif
44
a4af207c
JH
45#include "reentr.h"
46
8ef6e574
KW
47#ifdef USE_LOCALE
48
98994639 49/*
0d071d52
KW
50 * Standardize the locale name from a string returned by 'setlocale', possibly
51 * modifying that string.
98994639 52 *
0ef2a2b2 53 * The typical return value of setlocale() is either
98994639
HS
54 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
55 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
56 * (the space-separated values represent the various sublocales,
0ef2a2b2 57 * in some unspecified order). This is not handled by this function.
98994639
HS
58 *
59 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
0ef2a2b2
KW
60 * which is harmful for further use of the string in setlocale(). This
61 * function removes the trailing new line and everything up through the '='
98994639
HS
62 *
63 */
64STATIC char *
65S_stdize_locale(pTHX_ char *locs)
66{
7452cf6a 67 const char * const s = strchr(locs, '=');
98994639
HS
68 bool okay = TRUE;
69
7918f24d
NC
70 PERL_ARGS_ASSERT_STDIZE_LOCALE;
71
8772537c
AL
72 if (s) {
73 const char * const t = strchr(s, '.');
98994639 74 okay = FALSE;
8772537c
AL
75 if (t) {
76 const char * const u = strchr(t, '\n');
77 if (u && (u[1] == 0)) {
78 const STRLEN len = u - s;
79 Move(s + 1, locs, len, char);
80 locs[len] = 0;
81 okay = TRUE;
98994639
HS
82 }
83 }
84 }
85
86 if (!okay)
87 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
88
89 return locs;
90}
91
8ef6e574
KW
92#endif
93
98994639
HS
94void
95Perl_set_numeric_radix(pTHX)
96{
97#ifdef USE_LOCALE_NUMERIC
98# ifdef HAS_LOCALECONV
7452cf6a 99 const struct lconv* const lc = localeconv();
98994639 100
98994639
HS
101 if (lc && lc->decimal_point) {
102 if (lc->decimal_point[0] == '.' && lc->decimal_point[1] == 0) {
103 SvREFCNT_dec(PL_numeric_radix_sv);
a0714e2c 104 PL_numeric_radix_sv = NULL;
98994639
HS
105 }
106 else {
107 if (PL_numeric_radix_sv)
108 sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
109 else
110 PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
9f10db87 111 if (! is_invariant_string((U8 *) lc->decimal_point, 0)
28acfe03 112 && is_utf8_string((U8 *) lc->decimal_point, 0)
c1284011 113 && _is_cur_LC_category_utf8(LC_NUMERIC))
28acfe03
KW
114 {
115 SvUTF8_on(PL_numeric_radix_sv);
116 }
98994639
HS
117 }
118 }
119 else
a0714e2c 120 PL_numeric_radix_sv = NULL;
69014004 121
ba99daed 122 DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
69014004 123 (PL_numeric_radix_sv)
37b7e435
KW
124 ? SvPVX(PL_numeric_radix_sv)
125 : "NULL",
126 (PL_numeric_radix_sv)
39eb7305 127 ? cBOOL(SvUTF8(PL_numeric_radix_sv))
37b7e435 128 : 0));
69014004 129
98994639
HS
130# endif /* HAS_LOCALECONV */
131#endif /* USE_LOCALE_NUMERIC */
132}
133
a39edc4c
KW
134/* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
135 * return of setlocale(), then this is extremely likely to be the C or POSIX
136 * locale. However, the output of setlocale() is documented to be opaque, but
137 * the odds are extremely small that it would return these two strings for some
138 * other locale. Note that VMS in these two locales includes many non-ASCII
139 * characters as controls and punctuation (below are hex bytes):
140 * cntrl: 00-1F 7F 84-97 9B-9F
141 * punct: 21-2F 3A-40 5B-60 7B-7E A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
142 * Oddly, none there are listed as alphas, though some represent alphabetics
143 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
98b630b3
KW
144#define isNAME_C_OR_POSIX(name) ((name) != NULL \
145 && ((*(name) == 'C' && (*(name + 1)) == '\0') \
a39edc4c
KW
146 || strEQ((name), "POSIX")))
147
98994639 148void
8772537c 149Perl_new_numeric(pTHX_ const char *newnum)
98994639
HS
150{
151#ifdef USE_LOCALE_NUMERIC
0d071d52
KW
152
153 /* Called after all libc setlocale() calls affecting LC_NUMERIC, to tell
154 * core Perl this and that 'newnum' is the name of the new locale.
155 * It installs this locale as the current underlying default.
156 *
157 * The default locale and the C locale can be toggled between by use of the
158 * set_numeric_local() and set_numeric_standard() functions, which should
159 * probably not be called directly, but only via macros like
160 * SET_NUMERIC_STANDARD() in perl.h.
161 *
162 * The toggling is necessary mainly so that a non-dot radix decimal point
163 * character can be output, while allowing internal calculations to use a
164 * dot.
165 *
166 * This sets several interpreter-level variables:
bb304765 167 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
0d071d52 168 * PL_numeric_local A boolean indicating if the toggled state is such
7738054c
KW
169 * that the current locale is the program's underlying
170 * locale
171 * PL_numeric_standard An int indicating if the toggled state is such
172 * that the current locale is the C locale. If non-zero,
173 * it is in C; if > 1, it means it may not be toggled away
174 * from C.
0d071d52
KW
175 * Note that both of the last two variables can be true at the same time,
176 * if the underlying locale is C. (Toggling is a no-op under these
177 * circumstances.)
178 *
179 * Any code changing the locale (outside this file) should use
180 * POSIX::setlocale, which calls this function. Therefore this function
181 * should be called directly only from this file and from
182 * POSIX::setlocale() */
183
b03f34cf 184 char *save_newnum;
98994639
HS
185
186 if (! newnum) {
43c5f42d
NC
187 Safefree(PL_numeric_name);
188 PL_numeric_name = NULL;
98994639
HS
189 PL_numeric_standard = TRUE;
190 PL_numeric_local = TRUE;
191 return;
192 }
193
b03f34cf 194 save_newnum = stdize_locale(savepv(newnum));
abe1abcf
KW
195
196 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
197 PL_numeric_local = TRUE;
198
b03f34cf 199 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
98994639 200 Safefree(PL_numeric_name);
b03f34cf 201 PL_numeric_name = save_newnum;
b03f34cf 202 }
abe1abcf
KW
203 else {
204 Safefree(save_newnum);
205 }
4c28b29c
KW
206
207 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
208 * have to worry about the radix being a non-dot. (Core operations that
209 * need the underlying locale change to it temporarily). */
210 set_numeric_standard();
211
e19f01cb 212 set_numeric_radix();
6959d69d 213
f2ce9e1c
JH
214#else
215 PERL_UNUSED_ARG(newnum);
98994639
HS
216#endif /* USE_LOCALE_NUMERIC */
217}
218
219void
220Perl_set_numeric_standard(pTHX)
221{
222#ifdef USE_LOCALE_NUMERIC
28c1bf33
KW
223 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
224 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
225 * macro avoids calling this routine if toggling isn't necessary according
226 * to our records (which could be wrong if some XS code has changed the
227 * locale behind our back) */
0d071d52 228
a9b8c0d8
KW
229 setlocale(LC_NUMERIC, "C");
230 PL_numeric_standard = TRUE;
231 PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
232 set_numeric_radix();
69014004
KW
233 DEBUG_L(PerlIO_printf(Perl_debug_log,
234 "Underlying LC_NUMERIC locale now is C\n"));
98994639
HS
235
236#endif /* USE_LOCALE_NUMERIC */
237}
238
239void
240Perl_set_numeric_local(pTHX)
241{
242#ifdef USE_LOCALE_NUMERIC
28c1bf33
KW
243 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
244 * code should use the macros like SET_NUMERIC_LOCAL() in perl.h instead of
245 * calling this directly. The macro avoids calling this routine if
246 * toggling isn't necessary according to our records (which could be wrong
247 * if some XS code has changed the locale behind our back) */
a9b8c0d8
KW
248
249 setlocale(LC_NUMERIC, PL_numeric_name);
250 PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
251 PL_numeric_local = TRUE;
252 set_numeric_radix();
69014004
KW
253 DEBUG_L(PerlIO_printf(Perl_debug_log,
254 "Underlying LC_NUMERIC locale now is %s\n",
255 PL_numeric_name));
98994639
HS
256
257#endif /* USE_LOCALE_NUMERIC */
258}
259
260/*
261 * Set up for a new ctype locale.
262 */
263void
8772537c 264Perl_new_ctype(pTHX_ const char *newctype)
98994639
HS
265{
266#ifdef USE_LOCALE_CTYPE
0d071d52
KW
267
268 /* Called after all libc setlocale() calls affecting LC_CTYPE, to tell
269 * core Perl this and that 'newctype' is the name of the new locale.
270 *
271 * This function sets up the folding arrays for all 256 bytes, assuming
272 * that tofold() is tolc() since fold case is not a concept in POSIX,
273 *
274 * Any code changing the locale (outside this file) should use
275 * POSIX::setlocale, which calls this function. Therefore this function
276 * should be called directly only from this file and from
277 * POSIX::setlocale() */
278
27da23d5 279 dVAR;
68067e4e 280 UV i;
98994639 281
7918f24d
NC
282 PERL_ARGS_ASSERT_NEW_CTYPE;
283
215c5139
KW
284 /* We will replace any bad locale warning with 1) nothing if the new one is
285 * ok; or 2) a new warning for the bad new locale */
286 if (PL_warn_locale) {
287 SvREFCNT_dec_NN(PL_warn_locale);
288 PL_warn_locale = NULL;
289 }
290
c1284011 291 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
31f05a37
KW
292
293 /* A UTF-8 locale gets standard rules. But note that code still has to
294 * handle this specially because of the three problematic code points */
295 if (PL_in_utf8_CTYPE_locale) {
296 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
297 }
298 else {
8c6180a9
KW
299 /* Assume enough space for every character being bad. 4 spaces each
300 * for the 94 printable characters that are output like "'x' "; and 5
301 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
302 * NUL */
303 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
304
305 bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
306 no check */
307 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
308 to start */
309 unsigned int bad_count = 0; /* Count of bad characters */
310
baa60164
KW
311 for (i = 0; i < 256; i++) {
312 if (isUPPER_LC((U8) i))
313 PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
314 else if (isLOWER_LC((U8) i))
315 PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
316 else
317 PL_fold_locale[i] = (U8) i;
8c6180a9
KW
318
319 /* If checking for locale problems, see if the native ASCII-range
320 * printables plus \n and \t are in their expected categories in
321 * the new locale. If not, this could mean big trouble, upending
322 * Perl's and most programs' assumptions, like having a
323 * metacharacter with special meaning become a \w. Fortunately,
324 * it's very rare to find locales that aren't supersets of ASCII
325 * nowadays. It isn't a problem for most controls to be changed
326 * into something else; we check only \n and \t, though perhaps \r
327 * could be an issue as well. */
328 if (check_for_problems
329 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
330 {
331 if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
332 || (isPUNCT_A(i) && ! isPUNCT_LC(i))
333 || (isBLANK_A(i) && ! isBLANK_LC(i))
334 || (i == '\n' && ! isCNTRL_LC(i)))
335 {
336 if (bad_count) { /* Separate multiple entries with a
337 blank */
338 bad_chars_list[bad_count++] = ' ';
339 }
340 bad_chars_list[bad_count++] = '\'';
341 if (isPRINT_A(i)) {
342 bad_chars_list[bad_count++] = (char) i;
343 }
344 else {
345 bad_chars_list[bad_count++] = '\\';
346 if (i == '\n') {
347 bad_chars_list[bad_count++] = 'n';
348 }
349 else {
350 assert(i == '\t');
351 bad_chars_list[bad_count++] = 't';
352 }
353 }
354 bad_chars_list[bad_count++] = '\'';
355 bad_chars_list[bad_count] = '\0';
356 }
357 }
358 }
359
360#ifdef MB_CUR_MAX
361 /* We only handle single-byte locales (outside of UTF-8 ones; so if
d35fca5f 362 * this locale requires more than one byte, there are going to be
8c6180a9 363 * problems. */
ba1a4362
KW
364 if (check_for_problems && MB_CUR_MAX > 1
365
366 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
367 * locale. Just assume that the implementation for them (plus
368 * for POSIX) is correct and the > 1 value is spurious. (Since
369 * these are specially handled to never be considered UTF-8
370 * locales, as long as this is the only problem, everything
371 * should work fine */
372 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
373 {
8c6180a9
KW
374 multi_byte_locale = TRUE;
375 }
376#endif
377
378 if (bad_count || multi_byte_locale) {
780fcc9f 379 PL_warn_locale = Perl_newSVpvf(aTHX_
8c6180a9 380 "Locale '%s' may not work well.%s%s%s\n",
780fcc9f 381 newctype,
8c6180a9
KW
382 (multi_byte_locale)
383 ? " Some characters in it are not recognized by"
384 " Perl."
385 : "",
386 (bad_count)
387 ? "\nThe following characters (and maybe others)"
388 " may not have the same meaning as the Perl"
389 " program expects:\n"
390 : "",
391 (bad_count)
392 ? bad_chars_list
393 : ""
394 );
780fcc9f
KW
395 /* If we are actually in the scope of the locale, output the
396 * message now. Otherwise we save it to be output at the first
397 * operation using this locale, if that actually happens. Most
398 * programs don't use locales, so they are immune to bad ones */
399 if (IN_LC(LC_CTYPE)) {
400
401 /* We have to save 'newctype' because the setlocale() just
402 * below may destroy it. The next setlocale() further down
403 * should restore it properly so that the intermediate change
404 * here is transparent to this function's caller */
405 const char * const badlocale = savepv(newctype);
406
407 setlocale(LC_CTYPE, "C");
408
409 /* The '0' below suppresses a bogus gcc compiler warning */
410 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
411 setlocale(LC_CTYPE, badlocale);
c0f3a893 412 Safefree(badlocale);
780fcc9f
KW
413 SvREFCNT_dec_NN(PL_warn_locale);
414 PL_warn_locale = NULL;
415 }
baa60164 416 }
31f05a37 417 }
98994639
HS
418
419#endif /* USE_LOCALE_CTYPE */
7918f24d 420 PERL_ARGS_ASSERT_NEW_CTYPE;
8772537c 421 PERL_UNUSED_ARG(newctype);
96a5add6 422 PERL_UNUSED_CONTEXT;
98994639
HS
423}
424
98994639 425void
2726666d
KW
426Perl__warn_problematic_locale()
427{
2726666d
KW
428
429#ifdef USE_LOCALE_CTYPE
430
5f04a188
KW
431 dTHX;
432
433 /* Internal-to-core function that outputs the message in PL_warn_locale,
434 * and then NULLS it. Should be called only through the macro
435 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
436
2726666d
KW
437 if (PL_warn_locale) {
438 /*GCC_DIAG_IGNORE(-Wformat-security); Didn't work */
439 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
440 SvPVX(PL_warn_locale),
441 0 /* dummy to avoid compiler warning */ );
442 /* GCC_DIAG_RESTORE; */
443 SvREFCNT_dec_NN(PL_warn_locale);
444 PL_warn_locale = NULL;
445 }
446
447#endif
448
449}
450
451void
8772537c 452Perl_new_collate(pTHX_ const char *newcoll)
98994639
HS
453{
454#ifdef USE_LOCALE_COLLATE
0d071d52
KW
455
456 /* Called after all libc setlocale() calls affecting LC_COLLATE, to tell
457 * core Perl this and that 'newcoll' is the name of the new locale.
458 *
459 * Any code changing the locale (outside this file) should use
460 * POSIX::setlocale, which calls this function. Therefore this function
461 * should be called directly only from this file and from
d35fca5f
KW
462 * POSIX::setlocale().
463 *
464 * The design of locale collation is that every locale change is given an
465 * index 'PL_collation_ix'. The first time a string particpates in an
466 * operation that requires collation while locale collation is active, it
467 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
468 * magic includes the collation index, and the transformation of the string
469 * by strxfrm(), q.v. That transformation is used when doing comparisons,
470 * instead of the string itself. If a string changes, the magic is
471 * cleared. The next time the locale changes, the index is incremented,
472 * and so we know during a comparison that the transformation is not
473 * necessarily still valid, and so is recomputed. Note that if the locale
474 * changes enough times, the index could wrap (a U32), and it is possible
475 * that a transformation would improperly be considered valid, leading to
476 * an unlikely bug */
0d071d52 477
98994639
HS
478 if (! newcoll) {
479 if (PL_collation_name) {
480 ++PL_collation_ix;
481 Safefree(PL_collation_name);
482 PL_collation_name = NULL;
483 }
484 PL_collation_standard = TRUE;
00bf60ca 485 is_standard_collation:
98994639
HS
486 PL_collxfrm_base = 0;
487 PL_collxfrm_mult = 2;
165a1c52 488 PL_in_utf8_COLLATE_locale = FALSE;
6696cfa7 489 *PL_strxfrm_min_char = '\0';
98994639
HS
490 return;
491 }
492
d35fca5f 493 /* If this is not the same locale as currently, set the new one up */
98994639
HS
494 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
495 ++PL_collation_ix;
496 Safefree(PL_collation_name);
497 PL_collation_name = stdize_locale(savepv(newcoll));
a39edc4c 498 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
00bf60ca
KW
499 if (PL_collation_standard) {
500 goto is_standard_collation;
501 }
98994639 502
165a1c52 503 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
6696cfa7 504 *PL_strxfrm_min_char = '\0';
165a1c52 505
59c018b9
KW
506 /* A locale collation definition includes primary, secondary, tertiary,
507 * etc. weights for each character. To sort, the primary weights are
508 * used, and only if they compare equal, then the secondary weights are
509 * used, and only if they compare equal, then the tertiary, etc.
510 *
511 * strxfrm() works by taking the input string, say ABC, and creating an
512 * output transformed string consisting of first the primary weights,
513 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
514 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
515 * may not have weights at every level. In our example, let's say B
516 * doesn't have a tertiary weight, and A doesn't have a secondary
517 * weight. The constructed string is then going to be
518 * A¹B¹C¹ B²C² A³C³ ....
519 * This has the desired effect that strcmp() will look at the secondary
520 * or tertiary weights only if the strings compare equal at all higher
521 * priority weights. The spaces shown here, like in
522 * "A¹B¹C¹ * A²B²C² "
523 * are not just for readability. In the general case, these must
524 * actually be bytes, which we will call here 'separator weights'; and
525 * they must be smaller than any other weight value, but since these
526 * are C strings, only the terminating one can be a NUL (some
527 * implementations may include a non-NUL separator weight just before
528 * the NUL). Implementations tend to reserve 01 for the separator
529 * weights. They are needed so that a shorter string's secondary
530 * weights won't be misconstrued as primary weights of a longer string,
531 * etc. By making them smaller than any other weight, the shorter
532 * string will sort first. (Actually, if all secondary weights are
533 * smaller than all primary ones, there is no need for a separator
534 * weight between those two levels, etc.)
535 *
536 * The length of the transformed string is roughly a linear function of
537 * the input string. It's not exactly linear because some characters
538 * don't have weights at all levels. When we call strxfrm() we have to
539 * allocate some memory to hold the transformed string. The
540 * calculations below try to find coefficients 'm' and 'b' for this
541 * locale so that m*x + b equals how much space we need, given the size
542 * of the input string in 'x'. If we calculate too small, we increase
543 * the size as needed, and call strxfrm() again, but it is better to
544 * get it right the first time to avoid wasted expensive string
545 * transformations. */
546
98994639 547 {
79f120c8
KW
548 /* We use the string below to find how long the tranformation of it
549 * is. Almost all locales are supersets of ASCII, or at least the
550 * ASCII letters. We use all of them, half upper half lower,
551 * because if we used fewer, we might hit just the ones that are
552 * outliers in a particular locale. Most of the strings being
553 * collated will contain a preponderance of letters, and even if
554 * they are above-ASCII, they are likely to have the same number of
555 * weight levels as the ASCII ones. It turns out that digits tend
556 * to have fewer levels, and some punctuation has more, but those
557 * are relatively sparse in text, and khw believes this gives a
558 * reasonable result, but it could be changed if experience so
559 * dictates. */
560 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
561 char * x_longer; /* Transformed 'longer' */
562 Size_t x_len_longer; /* Length of 'x_longer' */
563
564 char * x_shorter; /* We also transform a substring of 'longer' */
565 Size_t x_len_shorter;
566
567 /* mem_collxfrm() is used get the transformation (though here we
568 * are interested only in its length). It is used because it has
569 * the intelligence to handle all cases, but to work, it needs some
570 * values of 'm' and 'b' to get it started. For the purposes of
571 * this calculation we use a very conservative estimate of 'm' and
572 * 'b'. This assumes a weight can be multiple bytes, enough to
573 * hold any UV on the platform, and there are 5 levels, 4 weight
574 * bytes, and a trailing NUL. */
575 PL_collxfrm_base = 5;
576 PL_collxfrm_mult = 5 * sizeof(UV);
577
578 /* Find out how long the transformation really is */
579 x_longer = mem_collxfrm(longer,
580 sizeof(longer) - 1,
581 &x_len_longer);
582 Safefree(x_longer);
583
584 /* Find out how long the transformation of a substring of 'longer'
585 * is. Together the lengths of these transformations are
586 * sufficient to calculate 'm' and 'b'. The substring is all of
587 * 'longer' except the first character. This minimizes the chances
588 * of being swayed by outliers */
589 x_shorter = mem_collxfrm(longer + 1,
590 sizeof(longer) - 2,
591 &x_len_shorter);
592 Safefree(x_shorter);
593
594 /* If the results are nonsensical for this simple test, the whole
595 * locale definition is suspect. Mark it so that locale collation
596 * is not active at all for it. XXX Should we warn? */
597 if ( x_len_shorter == 0
598 || x_len_longer == 0
599 || x_len_shorter >= x_len_longer)
600 {
601 PL_collxfrm_mult = 0;
602 PL_collxfrm_base = 0;
603 }
604 else {
605 SSize_t base; /* Temporary */
606
607 /* We have both: m * strlen(longer) + b = x_len_longer
608 * m * strlen(shorter) + b = x_len_shorter;
609 * subtracting yields:
610 * m * (strlen(longer) - strlen(shorter))
611 * = x_len_longer - x_len_shorter
612 * But we have set things up so that 'shorter' is 1 byte smaller
613 * than 'longer'. Hence:
614 * m = x_len_longer - x_len_shorter
615 *
616 * But if something went wrong, make sure the multiplier is at
617 * least 1.
618 */
619 if (x_len_longer > x_len_shorter) {
620 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
621 }
622 else {
623 PL_collxfrm_mult = 1;
624 }
625
626 /* mx + b = len
627 * so: b = len - mx
628 * but in case something has gone wrong, make sure it is
629 * non-negative */
630 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
631 if (base < 0) {
632 base = 0;
633 }
634
635 /* Add 1 for the trailing NUL */
636 PL_collxfrm_base = base + 1;
637 }
98994639
HS
638 }
639 }
640
f2ce9e1c
JH
641#else
642 PERL_UNUSED_ARG(newcoll);
98994639
HS
643#endif /* USE_LOCALE_COLLATE */
644}
645
b385bb4d
KW
646#ifdef WIN32
647
648char *
649Perl_my_setlocale(pTHX_ int category, const char* locale)
650{
651 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
652 * difference unless the input locale is "", which means on Windows to get
653 * the machine default, which is set via the computer's "Regional and
654 * Language Options" (or its current equivalent). In POSIX, it instead
655 * means to find the locale from the user's environment. This routine
656 * looks in the environment, and, if anything is found, uses that instead
657 * of going to the machine default. If there is no environment override,
658 * the machine default is used, as normal, by calling the real setlocale()
659 * with "". The POSIX behavior is to use the LC_ALL variable if set;
660 * otherwise to use the particular category's variable if set; otherwise to
661 * use the LANG variable. */
662
175c4cf9 663 bool override_LC_ALL = FALSE;
89f7b9aa
KW
664 char * result;
665
b385bb4d
KW
666 if (locale && strEQ(locale, "")) {
667# ifdef LC_ALL
668 locale = PerlEnv_getenv("LC_ALL");
669 if (! locale) {
670#endif
671 switch (category) {
672# ifdef LC_ALL
673 case LC_ALL:
481465ea 674 override_LC_ALL = TRUE;
b385bb4d
KW
675 break; /* We already know its variable isn't set */
676# endif
677# ifdef USE_LOCALE_TIME
678 case LC_TIME:
679 locale = PerlEnv_getenv("LC_TIME");
680 break;
681# endif
682# ifdef USE_LOCALE_CTYPE
683 case LC_CTYPE:
684 locale = PerlEnv_getenv("LC_CTYPE");
685 break;
686# endif
687# ifdef USE_LOCALE_COLLATE
688 case LC_COLLATE:
689 locale = PerlEnv_getenv("LC_COLLATE");
690 break;
691# endif
692# ifdef USE_LOCALE_MONETARY
693 case LC_MONETARY:
694 locale = PerlEnv_getenv("LC_MONETARY");
695 break;
696# endif
697# ifdef USE_LOCALE_NUMERIC
698 case LC_NUMERIC:
699 locale = PerlEnv_getenv("LC_NUMERIC");
700 break;
701# endif
702# ifdef USE_LOCALE_MESSAGES
703 case LC_MESSAGES:
704 locale = PerlEnv_getenv("LC_MESSAGES");
705 break;
706# endif
707 default:
708 /* This is a category, like PAPER_SIZE that we don't
709 * know about; and so can't provide a wrapper. */
710 break;
711 }
712 if (! locale) {
713 locale = PerlEnv_getenv("LANG");
481465ea 714 if (! locale) {
b385bb4d
KW
715 locale = "";
716 }
717 }
718# ifdef LC_ALL
719 }
720# endif
721 }
722
89f7b9aa 723 result = setlocale(category, locale);
bbc98134
KW
724 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
725 _setlocale_debug_string(category, locale, result)));
89f7b9aa 726
481465ea 727 if (! override_LC_ALL) {
89f7b9aa
KW
728 return result;
729 }
730
dfd77d7a 731 /* Here the input category was LC_ALL, and we have set it to what is in the
481465ea
KW
732 * LANG variable or the system default if there is no LANG. But these have
733 * lower priority than the other LC_foo variables, so override it for each
734 * one that is set. (If they are set to "", it means to use the same thing
735 * we just set LC_ALL to, so can skip) */
89f7b9aa
KW
736# ifdef USE_LOCALE_TIME
737 result = PerlEnv_getenv("LC_TIME");
730252b2 738 if (result && strNE(result, "")) {
89f7b9aa 739 setlocale(LC_TIME, result);
bbc98134
KW
740 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
741 __FILE__, __LINE__,
742 _setlocale_debug_string(LC_TIME, result, "not captured")));
89f7b9aa
KW
743 }
744# endif
745# ifdef USE_LOCALE_CTYPE
746 result = PerlEnv_getenv("LC_CTYPE");
730252b2 747 if (result && strNE(result, "")) {
89f7b9aa 748 setlocale(LC_CTYPE, result);
bbc98134
KW
749 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
750 __FILE__, __LINE__,
751 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
89f7b9aa
KW
752 }
753# endif
754# ifdef USE_LOCALE_COLLATE
755 result = PerlEnv_getenv("LC_COLLATE");
730252b2 756 if (result && strNE(result, "")) {
89f7b9aa 757 setlocale(LC_COLLATE, result);
bbc98134
KW
758 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
759 __FILE__, __LINE__,
760 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
89f7b9aa
KW
761 }
762# endif
763# ifdef USE_LOCALE_MONETARY
764 result = PerlEnv_getenv("LC_MONETARY");
730252b2 765 if (result && strNE(result, "")) {
89f7b9aa 766 setlocale(LC_MONETARY, result);
bbc98134
KW
767 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
768 __FILE__, __LINE__,
769 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
89f7b9aa
KW
770 }
771# endif
772# ifdef USE_LOCALE_NUMERIC
773 result = PerlEnv_getenv("LC_NUMERIC");
730252b2 774 if (result && strNE(result, "")) {
89f7b9aa 775 setlocale(LC_NUMERIC, result);
bbc98134
KW
776 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
777 __FILE__, __LINE__,
778 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
89f7b9aa
KW
779 }
780# endif
781# ifdef USE_LOCALE_MESSAGES
782 result = PerlEnv_getenv("LC_MESSAGES");
730252b2 783 if (result && strNE(result, "")) {
89f7b9aa 784 setlocale(LC_MESSAGES, result);
bbc98134
KW
785 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
786 __FILE__, __LINE__,
787 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
89f7b9aa
KW
788 }
789# endif
790
bbc98134
KW
791 result = setlocale(LC_ALL, NULL);
792 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
793 __FILE__, __LINE__,
794 _setlocale_debug_string(LC_ALL, NULL, result)));
89f7b9aa 795
bbc98134 796 return result;
b385bb4d
KW
797}
798
799#endif
800
801
98994639
HS
802/*
803 * Initialize locale awareness.
804 */
805int
806Perl_init_i18nl10n(pTHX_ int printwarn)
807{
0e92a118
KW
808 /* printwarn is
809 *
810 * 0 if not to output warning when setup locale is bad
811 * 1 if to output warning based on value of PERL_BADLANG
812 * >1 if to output regardless of PERL_BADLANG
813 *
814 * returns
98994639 815 * 1 = set ok or not applicable,
0e92a118
KW
816 * 0 = fallback to a locale of lower priority
817 * -1 = fallback to all locales failed, not even to the C locale
6b058d42
KW
818 *
819 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
820 * set, debugging information is output.
821 *
822 * This looks more complicated than it is, mainly due to the #ifdefs.
823 *
824 * We try to set LC_ALL to the value determined by the environment. If
825 * there is no LC_ALL on this platform, we try the individual categories we
826 * know about. If this works, we are done.
827 *
828 * But if it doesn't work, we have to do something else. We search the
829 * environment variables ourselves instead of relying on the system to do
830 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
831 * think there is one), and the ultimate fallback "C". This is all done in
832 * the same loop as above to avoid duplicating code, but it makes things
833 * more complex. After the original failure, we add the fallback
834 * possibilities to the list of locales to try, and iterate the loop
835 * through them all until one succeeds.
836 *
837 * On Ultrix, the locale MUST come from the environment, so there is
838 * preliminary code to set it. I (khw) am not sure that it is necessary,
839 * and that this couldn't be folded into the loop, but barring any real
840 * platforms to test on, it's staying as-is
841 *
842 * A slight complication is that in embedded Perls, the locale may already
843 * be set-up, and we don't want to get it from the normal environment
844 * variables. This is handled by having a special environment variable
845 * indicate we're in this situation. We simply set setlocale's 2nd
846 * parameter to be a NULL instead of "". That indicates to setlocale that
847 * it is not to change anything, but to return the current value,
848 * effectively initializing perl's db to what the locale already is.
849 *
850 * We play the same trick with NULL if a LC_ALL succeeds. We call
851 * setlocale() on the individual categores with NULL to get their existing
852 * values for our db, instead of trying to change them.
853 * */
98994639 854
0e92a118
KW
855 int ok = 1;
856
98994639 857#if defined(USE_LOCALE)
98994639
HS
858#ifdef USE_LOCALE_CTYPE
859 char *curctype = NULL;
860#endif /* USE_LOCALE_CTYPE */
861#ifdef USE_LOCALE_COLLATE
862 char *curcoll = NULL;
863#endif /* USE_LOCALE_COLLATE */
864#ifdef USE_LOCALE_NUMERIC
865 char *curnum = NULL;
866#endif /* USE_LOCALE_NUMERIC */
867#ifdef __GLIBC__
175c4cf9 868 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
98994639 869#endif
65ebb059 870
ccd65d51
KW
871 /* NULL uses the existing already set up locale */
872 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
873 ? NULL
874 : "";
5d1187d1
KW
875#ifdef DEBUGGING
876 const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
877 ? TRUE
878 : FALSE;
879# define DEBUG_LOCALE_INIT(category, locale, result) \
880 STMT_START { \
881 if (debug) { \
882 PerlIO_printf(Perl_debug_log, \
883 "%s:%d: %s\n", \
884 __FILE__, __LINE__, \
885 _setlocale_debug_string(category, \
886 locale, \
887 result)); \
888 } \
889 } STMT_END
890#else
891# define DEBUG_LOCALE_INIT(a,b,c)
892#endif
c3fcd832
KW
893 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
894 unsigned int trial_locales_count;
175c4cf9
KW
895 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
896 const char * const lang = savepv(PerlEnv_getenv("LANG"));
98994639 897 bool setlocale_failure = FALSE;
65ebb059
KW
898 unsigned int i;
899 char *p;
175c4cf9
KW
900
901 /* A later getenv() could zap this, so only use here */
902 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
903
904 const bool locwarn = (printwarn > 1
905 || (printwarn
906 && (! bad_lang_use_once
22ff3130
HS
907 || (
908 /* disallow with "" or "0" */
909 *bad_lang_use_once
910 && strNE("0", bad_lang_use_once)))));
0e92a118 911 bool done = FALSE;
5d1187d1
KW
912 char * sl_result; /* return from setlocale() */
913 char * locale_param;
6bce99ee
JH
914#ifdef WIN32
915 /* In some systems you can find out the system default locale
916 * and use that as the fallback locale. */
917# define SYSTEM_DEFAULT_LOCALE
918#endif
919#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059 920 const char *system_default_locale = NULL;
6bce99ee 921#endif
98994639 922
0e92a118
KW
923#ifndef LOCALE_ENVIRON_REQUIRED
924 PERL_UNUSED_VAR(done);
5d1187d1 925 PERL_UNUSED_VAR(locale_param);
0e92a118 926#else
98994639
HS
927
928 /*
929 * Ultrix setlocale(..., "") fails if there are no environment
930 * variables from which to get a locale name.
931 */
932
b3e384bf 933# ifdef LC_ALL
98994639 934 if (lang) {
5d1187d1
KW
935 sl_result = my_setlocale(LC_ALL, setlocale_init);
936 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
937 if (sl_result)
98994639
HS
938 done = TRUE;
939 else
940 setlocale_failure = TRUE;
941 }
5d1187d1 942 if (! setlocale_failure) {
b3e384bf 943# ifdef USE_LOCALE_CTYPE
5d1187d1
KW
944 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
945 ? setlocale_init
946 : NULL;
947 curctype = my_setlocale(LC_CTYPE, locale_param);
948 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
949 if (! curctype)
98994639
HS
950 setlocale_failure = TRUE;
951 else
952 curctype = savepv(curctype);
b3e384bf
KW
953# endif /* USE_LOCALE_CTYPE */
954# ifdef USE_LOCALE_COLLATE
5d1187d1
KW
955 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
956 ? setlocale_init
957 : NULL;
958 curcoll = my_setlocale(LC_COLLATE, locale_param);
959 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
960 if (! curcoll)
98994639
HS
961 setlocale_failure = TRUE;
962 else
963 curcoll = savepv(curcoll);
b3e384bf
KW
964# endif /* USE_LOCALE_COLLATE */
965# ifdef USE_LOCALE_NUMERIC
5d1187d1
KW
966 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
967 ? setlocale_init
968 : NULL;
969 curnum = my_setlocale(LC_NUMERIC, locale_param);
970 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
971 if (! curnum)
98994639
HS
972 setlocale_failure = TRUE;
973 else
974 curnum = savepv(curnum);
b3e384bf 975# endif /* USE_LOCALE_NUMERIC */
a782673d 976# ifdef USE_LOCALE_MESSAGES
5d1187d1
KW
977 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
978 ? setlocale_init
979 : NULL;
980 sl_result = my_setlocale(LC_MESSAGES, locale_param);
981 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
982 if (! sl_result)
a782673d
KW
983 setlocale_failure = TRUE;
984 }
985# endif /* USE_LOCALE_MESSAGES */
c835d6be 986# ifdef USE_LOCALE_MONETARY
5d1187d1
KW
987 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
988 ? setlocale_init
989 : NULL;
990 sl_result = my_setlocale(LC_MONETARY, locale_param);
991 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
992 if (! sl_result) {
c835d6be
KW
993 setlocale_failure = TRUE;
994 }
995# endif /* USE_LOCALE_MONETARY */
98994639
HS
996 }
997
b3e384bf 998# endif /* LC_ALL */
98994639
HS
999
1000#endif /* !LOCALE_ENVIRON_REQUIRED */
1001
65ebb059 1002 /* We try each locale in the list until we get one that works, or exhaust
20a240df
KW
1003 * the list. Normally the loop is executed just once. But if setting the
1004 * locale fails, inside the loop we add fallback trials to the array and so
1005 * will execute the loop multiple times */
c3fcd832
KW
1006 trial_locales[0] = setlocale_init;
1007 trial_locales_count = 1;
65ebb059
KW
1008 for (i= 0; i < trial_locales_count; i++) {
1009 const char * trial_locale = trial_locales[i];
1010
1011 if (i > 0) {
1012
1013 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
1014 * when i==0, but I (khw) don't think that behavior makes much
1015 * sense */
1016 setlocale_failure = FALSE;
1017
6bce99ee
JH
1018#ifdef SYSTEM_DEFAULT_LOCALE
1019# ifdef WIN32
65ebb059
KW
1020 /* On Windows machines, an entry of "" after the 0th means to use
1021 * the system default locale, which we now proceed to get. */
1022 if (strEQ(trial_locale, "")) {
1023 unsigned int j;
1024
1025 /* Note that this may change the locale, but we are going to do
1026 * that anyway just below */
1027 system_default_locale = setlocale(LC_ALL, "");
5d1187d1 1028 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
65ebb059
KW
1029
1030 /* Skip if invalid or it's already on the list of locales to
1031 * try */
1032 if (! system_default_locale) {
1033 goto next_iteration;
1034 }
1035 for (j = 0; j < trial_locales_count; j++) {
1036 if (strEQ(system_default_locale, trial_locales[j])) {
1037 goto next_iteration;
1038 }
1039 }
1040
1041 trial_locale = system_default_locale;
1042 }
6bce99ee
JH
1043# endif /* WIN32 */
1044#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
1045 }
1046
98994639 1047#ifdef LC_ALL
5d1187d1
KW
1048 sl_result = my_setlocale(LC_ALL, trial_locale);
1049 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
1050 if (! sl_result) {
49c85077 1051 setlocale_failure = TRUE;
7cd8b568
KW
1052 }
1053 else {
1054 /* Since LC_ALL succeeded, it should have changed all the other
1055 * categories it can to its value; so we massage things so that the
1056 * setlocales below just return their category's current values.
1057 * This adequately handles the case in NetBSD where LC_COLLATE may
1058 * not be defined for a locale, and setting it individually will
1059 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
1060 * the POSIX locale. */
1061 trial_locale = NULL;
1062 }
98994639
HS
1063#endif /* LC_ALL */
1064
49c85077 1065 if (!setlocale_failure) {
98994639 1066#ifdef USE_LOCALE_CTYPE
49c85077 1067 Safefree(curctype);
5d1187d1
KW
1068 curctype = my_setlocale(LC_CTYPE, trial_locale);
1069 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
1070 if (! curctype)
49c85077
KW
1071 setlocale_failure = TRUE;
1072 else
1073 curctype = savepv(curctype);
98994639
HS
1074#endif /* USE_LOCALE_CTYPE */
1075#ifdef USE_LOCALE_COLLATE
49c85077 1076 Safefree(curcoll);
5d1187d1
KW
1077 curcoll = my_setlocale(LC_COLLATE, trial_locale);
1078 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
1079 if (! curcoll)
49c85077
KW
1080 setlocale_failure = TRUE;
1081 else
1082 curcoll = savepv(curcoll);
98994639
HS
1083#endif /* USE_LOCALE_COLLATE */
1084#ifdef USE_LOCALE_NUMERIC
49c85077 1085 Safefree(curnum);
5d1187d1
KW
1086 curnum = my_setlocale(LC_NUMERIC, trial_locale);
1087 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
1088 if (! curnum)
49c85077
KW
1089 setlocale_failure = TRUE;
1090 else
1091 curnum = savepv(curnum);
98994639 1092#endif /* USE_LOCALE_NUMERIC */
a782673d 1093#ifdef USE_LOCALE_MESSAGES
5d1187d1
KW
1094 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
1095 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
1096 if (! (sl_result))
a782673d
KW
1097 setlocale_failure = TRUE;
1098#endif /* USE_LOCALE_MESSAGES */
c835d6be 1099#ifdef USE_LOCALE_MONETARY
5d1187d1
KW
1100 sl_result = my_setlocale(LC_MONETARY, trial_locale);
1101 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
1102 if (! (sl_result))
c835d6be
KW
1103 setlocale_failure = TRUE;
1104#endif /* USE_LOCALE_MONETARY */
1105
49c85077
KW
1106 if (! setlocale_failure) { /* Success */
1107 break;
1108 }
65ebb059 1109 }
98994639 1110
49c85077
KW
1111 /* Here, something failed; will need to try a fallback. */
1112 ok = 0;
65ebb059 1113
49c85077
KW
1114 if (i == 0) {
1115 unsigned int j;
98994639 1116
65ebb059 1117 if (locwarn) { /* Output failure info only on the first one */
98994639
HS
1118#ifdef LC_ALL
1119
49c85077
KW
1120 PerlIO_printf(Perl_error_log,
1121 "perl: warning: Setting locale failed.\n");
98994639
HS
1122
1123#else /* !LC_ALL */
1124
49c85077
KW
1125 PerlIO_printf(Perl_error_log,
1126 "perl: warning: Setting locale failed for the categories:\n\t");
20a240df 1127# ifdef USE_LOCALE_CTYPE
49c85077
KW
1128 if (! curctype)
1129 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
20a240df
KW
1130# endif /* USE_LOCALE_CTYPE */
1131# ifdef USE_LOCALE_COLLATE
49c85077
KW
1132 if (! curcoll)
1133 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
20a240df
KW
1134# endif /* USE_LOCALE_COLLATE */
1135# ifdef USE_LOCALE_NUMERIC
49c85077
KW
1136 if (! curnum)
1137 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
20a240df 1138# endif /* USE_LOCALE_NUMERIC */
a782673d 1139 PerlIO_printf(Perl_error_log, "and possibly others\n");
98994639
HS
1140
1141#endif /* LC_ALL */
1142
49c85077
KW
1143 PerlIO_printf(Perl_error_log,
1144 "perl: warning: Please check that your locale settings:\n");
98994639
HS
1145
1146#ifdef __GLIBC__
49c85077
KW
1147 PerlIO_printf(Perl_error_log,
1148 "\tLANGUAGE = %c%s%c,\n",
1149 language ? '"' : '(',
1150 language ? language : "unset",
1151 language ? '"' : ')');
98994639
HS
1152#endif
1153
49c85077
KW
1154 PerlIO_printf(Perl_error_log,
1155 "\tLC_ALL = %c%s%c,\n",
1156 lc_all ? '"' : '(',
1157 lc_all ? lc_all : "unset",
1158 lc_all ? '"' : ')');
98994639
HS
1159
1160#if defined(USE_ENVIRON_ARRAY)
49c85077
KW
1161 {
1162 char **e;
1163 for (e = environ; *e; e++) {
1164 if (strnEQ(*e, "LC_", 3)
1165 && strnNE(*e, "LC_ALL=", 7)
1166 && (p = strchr(*e, '=')))
1167 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1168 (int)(p - *e), *e, p + 1);
1169 }
1170 }
98994639 1171#else
49c85077
KW
1172 PerlIO_printf(Perl_error_log,
1173 "\t(possibly more locale environment variables)\n");
98994639
HS
1174#endif
1175
49c85077
KW
1176 PerlIO_printf(Perl_error_log,
1177 "\tLANG = %c%s%c\n",
1178 lang ? '"' : '(',
1179 lang ? lang : "unset",
1180 lang ? '"' : ')');
98994639 1181
49c85077
KW
1182 PerlIO_printf(Perl_error_log,
1183 " are supported and installed on your system.\n");
1184 }
98994639 1185
65ebb059 1186 /* Calculate what fallback locales to try. We have avoided this
f6bab5f6 1187 * until we have to, because failure is quite unlikely. This will
65ebb059
KW
1188 * usually change the upper bound of the loop we are in.
1189 *
1190 * Since the system's default way of setting the locale has not
1191 * found one that works, We use Perl's defined ordering: LC_ALL,
1192 * LANG, and the C locale. We don't try the same locale twice, so
1193 * don't add to the list if already there. (On POSIX systems, the
1194 * LC_ALL element will likely be a repeat of the 0th element "",
6b058d42
KW
1195 * but there's no harm done by doing it explicitly.
1196 *
1197 * Note that this tries the LC_ALL environment variable even on
1198 * systems which have no LC_ALL locale setting. This may or may
1199 * not have been originally intentional, but there's no real need
1200 * to change the behavior. */
65ebb059
KW
1201 if (lc_all) {
1202 for (j = 0; j < trial_locales_count; j++) {
1203 if (strEQ(lc_all, trial_locales[j])) {
1204 goto done_lc_all;
1205 }
1206 }
1207 trial_locales[trial_locales_count++] = lc_all;
1208 }
1209 done_lc_all:
98994639 1210
65ebb059
KW
1211 if (lang) {
1212 for (j = 0; j < trial_locales_count; j++) {
1213 if (strEQ(lang, trial_locales[j])) {
1214 goto done_lang;
1215 }
1216 }
1217 trial_locales[trial_locales_count++] = lang;
1218 }
1219 done_lang:
1220
1221#if defined(WIN32) && defined(LC_ALL)
1222 /* For Windows, we also try the system default locale before "C".
1223 * (If there exists a Windows without LC_ALL we skip this because
1224 * it gets too complicated. For those, the "C" is the next
1225 * fallback possibility). The "" is the same as the 0th element of
1226 * the array, but the code at the loop above knows to treat it
1227 * differently when not the 0th */
1228 trial_locales[trial_locales_count++] = "";
1229#endif
1230
1231 for (j = 0; j < trial_locales_count; j++) {
1232 if (strEQ("C", trial_locales[j])) {
1233 goto done_C;
1234 }
1235 }
1236 trial_locales[trial_locales_count++] = "C";
98994639 1237
65ebb059
KW
1238 done_C: ;
1239 } /* end of first time through the loop */
98994639 1240
65ebb059
KW
1241#ifdef WIN32
1242 next_iteration: ;
1243#endif
1244
1245 } /* end of looping through the trial locales */
1246
1247 if (ok < 1) { /* If we tried to fallback */
1248 const char* msg;
1249 if (! setlocale_failure) { /* fallback succeeded */
1250 msg = "Falling back to";
1251 }
1252 else { /* fallback failed */
98994639 1253
65ebb059
KW
1254 /* We dropped off the end of the loop, so have to decrement i to
1255 * get back to the value the last time through */
1256 i--;
98994639 1257
65ebb059
KW
1258 ok = -1;
1259 msg = "Failed to fall back to";
1260
1261 /* To continue, we should use whatever values we've got */
98994639 1262#ifdef USE_LOCALE_CTYPE
49c85077
KW
1263 Safefree(curctype);
1264 curctype = savepv(setlocale(LC_CTYPE, NULL));
5d1187d1 1265 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
98994639
HS
1266#endif /* USE_LOCALE_CTYPE */
1267#ifdef USE_LOCALE_COLLATE
49c85077
KW
1268 Safefree(curcoll);
1269 curcoll = savepv(setlocale(LC_COLLATE, NULL));
5d1187d1 1270 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
98994639
HS
1271#endif /* USE_LOCALE_COLLATE */
1272#ifdef USE_LOCALE_NUMERIC
49c85077
KW
1273 Safefree(curnum);
1274 curnum = savepv(setlocale(LC_NUMERIC, NULL));
5d1187d1 1275 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
98994639 1276#endif /* USE_LOCALE_NUMERIC */
65ebb059
KW
1277 }
1278
1279 if (locwarn) {
1280 const char * description;
1281 const char * name = "";
1282 if (strEQ(trial_locales[i], "C")) {
1283 description = "the standard locale";
1284 name = "C";
1285 }
6bce99ee 1286#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059
KW
1287 else if (strEQ(trial_locales[i], "")) {
1288 description = "the system default locale";
1289 if (system_default_locale) {
1290 name = system_default_locale;
1291 }
1292 }
6bce99ee 1293#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
1294 else {
1295 description = "a fallback locale";
1296 name = trial_locales[i];
1297 }
1298 if (name && strNE(name, "")) {
1299 PerlIO_printf(Perl_error_log,
1300 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1301 }
1302 else {
1303 PerlIO_printf(Perl_error_log,
1304 "perl: warning: %s %s.\n", msg, description);
1305 }
1306 }
1307 } /* End of tried to fallback */
98994639
HS
1308
1309#ifdef USE_LOCALE_CTYPE
1310 new_ctype(curctype);
1311#endif /* USE_LOCALE_CTYPE */
1312
1313#ifdef USE_LOCALE_COLLATE
1314 new_collate(curcoll);
1315#endif /* USE_LOCALE_COLLATE */
1316
1317#ifdef USE_LOCALE_NUMERIC
1318 new_numeric(curnum);
1319#endif /* USE_LOCALE_NUMERIC */
b310b053 1320
8ef6e574 1321#if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
49c85077
KW
1322 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1323 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1324 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1325 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1326 * discipline. */
c1284011 1327 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
49c85077 1328
a05d7ebb 1329 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
fde18df1
JH
1330 This is an alternative to using the -C command line switch
1331 (the -C if present will override this). */
1332 {
dd374669 1333 const char *p = PerlEnv_getenv("PERL_UNICODE");
a05d7ebb 1334 PL_unicode = p ? parse_unicode_opts(&p) : 0;
5a22a2bb
NC
1335 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1336 PL_utf8cache = -1;
b310b053 1337 }
ec71e770 1338#endif
b310b053 1339
98994639 1340#ifdef USE_LOCALE_CTYPE
43c5f42d 1341 Safefree(curctype);
98994639
HS
1342#endif /* USE_LOCALE_CTYPE */
1343#ifdef USE_LOCALE_COLLATE
43c5f42d 1344 Safefree(curcoll);
98994639
HS
1345#endif /* USE_LOCALE_COLLATE */
1346#ifdef USE_LOCALE_NUMERIC
43c5f42d 1347 Safefree(curnum);
98994639 1348#endif /* USE_LOCALE_NUMERIC */
8ef6e574 1349
175c4cf9
KW
1350#ifdef __GLIBC__
1351 Safefree(language);
1352#endif
1353
1354 Safefree(lc_all);
1355 Safefree(lang);
1356
e3305790
KW
1357#else /* !USE_LOCALE */
1358 PERL_UNUSED_ARG(printwarn);
1359#endif /* USE_LOCALE */
1360
98994639
HS
1361 return ok;
1362}
1363
49c85077 1364
98994639
HS
1365#ifdef USE_LOCALE_COLLATE
1366
1367/*
1368 * mem_collxfrm() is a bit like strxfrm() but with two important
1369 * differences. First, it handles embedded NULs. Second, it allocates
1370 * a bit more memory than needed for the transformed data itself.
1371 * The real transformed data begins at offset sizeof(collationix).
d35fca5f
KW
1372 * *xlen is set to the length of that, and doesn't include the collation index
1373 * size.
98994639
HS
1374 * Please see sv_collxfrm() to see how this is used.
1375 */
166f8a29 1376
98994639 1377char *
6696cfa7
KW
1378Perl_mem_collxfrm(pTHX_ const char *input_string,
1379 STRLEN len,
1380 STRLEN *xlen
1381 )
98994639 1382{
6696cfa7
KW
1383 char * s = (char *) input_string;
1384 STRLEN s_strlen = strlen(input_string);
79f120c8 1385 char *xbuf = NULL;
6696cfa7 1386 STRLEN xAlloc, xout; /* xalloc is a reserved word in VC */
c664130f 1387 bool first_time = TRUE; /* Cleared after first loop iteration */
98994639 1388
7918f24d
NC
1389 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1390
79f120c8
KW
1391 /* If this locale has defective collation, skip */
1392 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
1393 goto bad;
1394 }
1395
6696cfa7
KW
1396 /* Replace any embedded NULs with the control that sorts before any others.
1397 * This will give as good as possible results on strings that don't
1398 * otherwise contain that character, but otherwise there may be
1399 * less-than-perfect results with that character and NUL. This is
1400 * unavoidable unless we replace strxfrm with our own implementation.
1401 *
1402 * XXX This code may be overkill. khw wrote it before realizing that if
1403 * you change a NUL into some other character, that that may change the
1404 * strxfrm results if that character is part of a sequence with other
1405 * characters for weight calculations. To minimize the chances of this,
1406 * now the replacement is restricted to another control (likely to be
1407 * \001). But the full generality has been retained.
1408 *
1409 * This is one of the few places in the perl core, where we can use
1410 * standard functions like strlen() and strcat(). It's because we're
1411 * looking for NULs. */
1412 if (s_strlen < len) {
1413 char * e = s + len;
1414 char * sans_nuls;
1415 STRLEN cur_min_char_len;
1416
1417 /* If we don't know what control character sorts lowest for this
1418 * locale, find it */
1419 if (*PL_strxfrm_min_char == '\0') {
1420 int j;
1421 char * cur_min_x = NULL; /* Cur cp's xfrm, (except it also
1422 includes the collation index
1423 prefixed. */
1424
1425 /* Look through all legal code points (NUL isn't) */
1426 for (j = 1; j < 256; j++) {
1427 char * x; /* j's xfrm plus collation index */
1428 STRLEN x_len; /* length of 'x' */
1429 STRLEN trial_len = 1;
1430
1431 /* Create a 1 byte string of the current code point, but with
1432 * room to be 2 bytes */
1433 char cur_source[] = { (char) j, '\0' , '\0' };
1434
1435 if (PL_in_utf8_COLLATE_locale) {
1436 if (! isCNTRL_L1(j)) {
1437 continue;
1438 }
1439
1440 /* If needs to be 2 bytes, find them */
1441 if (! UVCHR_IS_INVARIANT(j)) {
1442 continue; /* Can't handle variants yet */
1443 }
1444 }
1445 else if (! isCNTRL_LC(j)) {
1446 continue;
1447 }
1448
1449 /* Then transform it */
1450 x = mem_collxfrm(cur_source, trial_len, &x_len);
1451
1452 /* If something went wrong (which it shouldn't), just
1453 * ignore this code point */
1454 if ( x_len == 0
1455 || strlen(x + sizeof(PL_collation_ix)) < x_len)
1456 {
1457 continue;
1458 }
1459
1460 /* If this character's transformation is lower than
1461 * the current lowest, this one becomes the lowest */
1462 if ( cur_min_x == NULL
1463 || strLT(x + sizeof(PL_collation_ix),
1464 cur_min_x + sizeof(PL_collation_ix)))
1465 {
1466 strcpy(PL_strxfrm_min_char, cur_source);
1467 cur_min_x = x;
1468 }
1469 else {
1470 Safefree(x);
1471 }
1472 } /* end of loop through all bytes */
1473
1474 /* Unlikely, but possible, if there aren't any controls in the
1475 * locale, arbitrarily use \001 */
1476 if (cur_min_x == NULL) {
1477 STRLEN x_len; /* temporary */
1478 cur_min_x = mem_collxfrm("\001", 1, &x_len);
1479 /* cur_min_cp was already initialized to 1 */
1480 }
1481
1482 Safefree(cur_min_x);
1483 }
1484
1485 /* The worst case length for the replaced string would be if every
1486 * character in it is NUL. Multiply that by the length of each
1487 * replacement, and allow for a trailing NUL */
1488 cur_min_char_len = strlen(PL_strxfrm_min_char);
1489 Newx(sans_nuls, (len * cur_min_char_len) + 1, char);
1490 *sans_nuls = '\0';
1491
1492
1493 /* Replace each NUL with the lowest collating control. Loop until have
1494 * exhausted all the NULs */
1495 while (s + s_strlen < e) {
1496 strcat(sans_nuls, s);
1497
1498 /* Do the actual replacement */
1499 strcat(sans_nuls, PL_strxfrm_min_char);
1500
1501 /* Move past the input NUL */
1502 s += s_strlen + 1;
1503 s_strlen = strlen(s);
1504 }
1505
1506 /* And add anything that trails the final NUL */
1507 strcat(sans_nuls, s);
1508
1509 /* Switch so below we transform this modified string */
1510 s = sans_nuls;
1511 len = strlen(s);
1512 }
1513
59c018b9
KW
1514 /* The first element in the output is the collation id, used by
1515 * sv_collxfrm(); then comes the space for the transformed string. The
1516 * equation should give us a good estimate as to how much is needed */
98994639 1517 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
a02a5408 1518 Newx(xbuf, xAlloc, char);
bb0f664e 1519 if (UNLIKELY(! xbuf))
98994639
HS
1520 goto bad;
1521
d35fca5f 1522 /* Store the collation id */
98994639
HS
1523 *(U32*)xbuf = PL_collation_ix;
1524 xout = sizeof(PL_collation_ix);
d35fca5f
KW
1525
1526 /* Then the transformation of the input. We loop until successful, or we
1527 * give up */
4ebeff16
KW
1528 for (;;) {
1529 STRLEN xused = strxfrm(xbuf + xout, s, xAlloc - xout);
1530
1531 /* If the transformed string occupies less space than we told strxfrm()
1532 * was available, it means it successfully transformed the whole
1533 * string. */
1534 if (xused < xAlloc - xout) {
1535 xout += xused;
1536 break;
1537 }
bb0f664e 1538
4ebeff16
KW
1539 if (UNLIKELY(xused >= PERL_INT_MAX))
1540 goto bad;
d35fca5f 1541
c664130f
KW
1542 /* A well-behaved strxfrm() returns exactly how much space it needs
1543 * (not including the trailing NUL) when it fails due to not enough
1544 * space being provided. Assume that this is the case unless it's been
1545 * proven otherwise */
1546 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
1547 xAlloc = xused + sizeof(PL_collation_ix) + 1;
1548 }
1549 else { /* Here, either:
1550 * 1) The strxfrm() has previously shown bad behavior; or
1551 * 2) It isn't the first time through the loop, which means
1552 * that the strxfrm() is now showing bad behavior, because
1553 * we gave it what it said was needed in the previous
1554 * iteration, and it came back saying it needed still more.
1555 * (Many versions of cygwin fit this. When the buffer size
1556 * isn't sufficient, they return the input size instead of
1557 * how much is needed.)
1558 * Increase the buffer size by a fixed percentage and try again. */
1559 xAlloc = (2 * xAlloc) + 1;
1560 PL_strxfrm_is_behaved = FALSE;
1561 }
1562
1563
4ebeff16
KW
1564 Renew(xbuf, xAlloc, char);
1565 if (UNLIKELY(! xbuf))
1566 goto bad;
c664130f
KW
1567
1568 first_time = FALSE;
4ebeff16 1569 }
98994639 1570
6696cfa7
KW
1571 *xlen = xout - sizeof(PL_collation_ix);
1572
3c5f993e
KW
1573 /* Free up unneeded space; retain ehough for trailing NUL */
1574 Renew(xbuf, xout + 1, char);
98994639 1575
6696cfa7
KW
1576 if (s != input_string) {
1577 Safefree(s);
98994639
HS
1578 }
1579
98994639
HS
1580 return xbuf;
1581
1582 bad:
1583 Safefree(xbuf);
6696cfa7
KW
1584 if (s != input_string) {
1585 Safefree(s);
1586 }
98994639
HS
1587 *xlen = 0;
1588 return NULL;
1589}
1590
1591#endif /* USE_LOCALE_COLLATE */
8ef6e574
KW
1592#ifdef USE_LOCALE
1593
c1284011
KW
1594bool
1595Perl__is_cur_LC_category_utf8(pTHX_ int category)
7d74bb61
KW
1596{
1597 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1598 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
119ee68b 1599 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
609548d2
KW
1600 * could give the wrong result. The result will very likely be correct for
1601 * languages that have commonly used non-ASCII characters, but for notably
1602 * English, it comes down to if the locale's name ends in something like
1603 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
7d74bb61
KW
1604
1605 char *save_input_locale = NULL;
7d74bb61
KW
1606 STRLEN final_pos;
1607
8ef6e574 1608#ifdef LC_ALL
7d74bb61 1609 assert(category != LC_ALL);
8ef6e574 1610#endif
7d74bb61
KW
1611
1612 /* First dispose of the trivial cases */
b07fffd1 1613 save_input_locale = setlocale(category, NULL);
7d74bb61 1614 if (! save_input_locale) {
69014004
KW
1615 DEBUG_L(PerlIO_printf(Perl_debug_log,
1616 "Could not find current locale for category %d\n",
1617 category));
7d74bb61
KW
1618 return FALSE; /* XXX maybe should croak */
1619 }
b07fffd1 1620 save_input_locale = stdize_locale(savepv(save_input_locale));
a39edc4c 1621 if (isNAME_C_OR_POSIX(save_input_locale)) {
69014004
KW
1622 DEBUG_L(PerlIO_printf(Perl_debug_log,
1623 "Current locale for category %d is %s\n",
1624 category, save_input_locale));
b07fffd1 1625 Safefree(save_input_locale);
7d74bb61
KW
1626 return FALSE;
1627 }
1628
1d958db2
KW
1629#if defined(USE_LOCALE_CTYPE) \
1630 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
7d74bb61 1631
1d958db2 1632 { /* Next try nl_langinfo or MB_CUR_MAX if available */
7d74bb61
KW
1633
1634 char *save_ctype_locale = NULL;
119ee68b 1635 bool is_utf8;
7d74bb61 1636
119ee68b 1637 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
7d74bb61
KW
1638
1639 /* Get the current LC_CTYPE locale */
4f72bb37 1640 save_ctype_locale = setlocale(LC_CTYPE, NULL);
7d74bb61 1641 if (! save_ctype_locale) {
69014004
KW
1642 DEBUG_L(PerlIO_printf(Perl_debug_log,
1643 "Could not find current locale for LC_CTYPE\n"));
7d74bb61
KW
1644 goto cant_use_nllanginfo;
1645 }
4f72bb37 1646 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
7d74bb61
KW
1647
1648 /* If LC_CTYPE and the desired category use the same locale, this
1649 * means that finding the value for LC_CTYPE is the same as finding
1650 * the value for the desired category. Otherwise, switch LC_CTYPE
1651 * to the desired category's locale */
1652 if (strEQ(save_ctype_locale, save_input_locale)) {
1653 Safefree(save_ctype_locale);
1654 save_ctype_locale = NULL;
1655 }
1656 else if (! setlocale(LC_CTYPE, save_input_locale)) {
69014004
KW
1657 DEBUG_L(PerlIO_printf(Perl_debug_log,
1658 "Could not change LC_CTYPE locale to %s\n",
1659 save_input_locale));
7d74bb61
KW
1660 Safefree(save_ctype_locale);
1661 goto cant_use_nllanginfo;
1662 }
1663 }
1664
69014004
KW
1665 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1666 save_input_locale));
1667
7d74bb61 1668 /* Here the current LC_CTYPE is set to the locale of the category whose
1d958db2
KW
1669 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1670 * should give the correct results */
119ee68b 1671
1d958db2
KW
1672# if defined(HAS_NL_LANGINFO) && defined(CODESET)
1673 {
4f72bb37 1674 char *codeset = nl_langinfo(CODESET);
1d958db2 1675 if (codeset && strNE(codeset, "")) {
4f72bb37 1676 codeset = savepv(codeset);
119ee68b 1677
1d958db2
KW
1678 /* If we switched LC_CTYPE, switch back */
1679 if (save_ctype_locale) {
1680 setlocale(LC_CTYPE, save_ctype_locale);
1681 Safefree(save_ctype_locale);
1682 }
1683
1684 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1685 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1686
69014004
KW
1687 DEBUG_L(PerlIO_printf(Perl_debug_log,
1688 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1689 codeset, is_utf8));
1d958db2
KW
1690 Safefree(codeset);
1691 Safefree(save_input_locale);
1692 return is_utf8;
1693 }
119ee68b
KW
1694 }
1695
1d958db2
KW
1696# endif
1697# ifdef MB_CUR_MAX
1698
1699 /* Here, either we don't have nl_langinfo, or it didn't return a
1700 * codeset. Try MB_CUR_MAX */
1701
119ee68b
KW
1702 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1703 * Unicode code point. Since UTF-8 is the only non-single byte
1704 * encoding we handle, we just say any such encoding is UTF-8, and if
1705 * turns out to be wrong, other things will fail */
1706 is_utf8 = MB_CUR_MAX >= 4;
1707
69014004
KW
1708 DEBUG_L(PerlIO_printf(Perl_debug_log,
1709 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1710 (int) MB_CUR_MAX, is_utf8));
1711
119ee68b
KW
1712 Safefree(save_input_locale);
1713
1714# ifdef HAS_MBTOWC
1715
1716 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1717 * since they are both in the C99 standard. We can feed a known byte
1718 * string to the latter function, and check that it gives the expected
1719 * result */
1720 if (is_utf8) {
1721 wchar_t wc;
856b881c 1722 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
69014004 1723 errno = 0;
f019f68f 1724 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
119ee68b
KW
1725 != strlen(HYPHEN_UTF8)
1726 || wc != (wchar_t) 0x2010)
1727 {
1728 is_utf8 = FALSE;
abdcbdb8 1729 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
69014004
KW
1730 DEBUG_L(PerlIO_printf(Perl_debug_log,
1731 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1732 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
119ee68b
KW
1733 }
1734 }
119ee68b
KW
1735# endif
1736
1d958db2
KW
1737 /* If we switched LC_CTYPE, switch back */
1738 if (save_ctype_locale) {
1739 setlocale(LC_CTYPE, save_ctype_locale);
1740 Safefree(save_ctype_locale);
119ee68b 1741 }
7d74bb61 1742
1d958db2 1743 return is_utf8;
119ee68b 1744# endif
7d74bb61 1745 }
119ee68b 1746
7d74bb61
KW
1747 cant_use_nllanginfo:
1748
0080c90a
KW
1749#else /* nl_langinfo should work if available, so don't bother compiling this
1750 fallback code. The final fallback of looking at the name is
1751 compiled, and will be executed if nl_langinfo fails */
7d74bb61 1752
97f4de96
KW
1753 /* nl_langinfo not available or failed somehow. Next try looking at the
1754 * currency symbol to see if it disambiguates things. Often that will be
1755 * in the native script, and if the symbol isn't in UTF-8, we know that the
1756 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
609548d2
KW
1757 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1758 * */
fa9b773e
KW
1759
1760#ifdef HAS_LOCALECONV
fa9b773e 1761# ifdef USE_LOCALE_MONETARY
fa9b773e
KW
1762 {
1763 char *save_monetary_locale = NULL;
fa9b773e 1764 bool only_ascii = FALSE;
13542a67
KW
1765 bool is_utf8 = FALSE;
1766 struct lconv* lc;
fa9b773e 1767
97f4de96
KW
1768 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1769 * the desired category, if it isn't that locale already */
1770
fa9b773e
KW
1771 if (category != LC_MONETARY) {
1772
4f72bb37 1773 save_monetary_locale = setlocale(LC_MONETARY, NULL);
fa9b773e 1774 if (! save_monetary_locale) {
69014004
KW
1775 DEBUG_L(PerlIO_printf(Perl_debug_log,
1776 "Could not find current locale for LC_MONETARY\n"));
fa9b773e
KW
1777 goto cant_use_monetary;
1778 }
4f72bb37 1779 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
fa9b773e 1780
13542a67
KW
1781 if (strEQ(save_monetary_locale, save_input_locale)) {
1782 Safefree(save_monetary_locale);
1783 save_monetary_locale = NULL;
1784 }
1785 else if (! setlocale(LC_MONETARY, save_input_locale)) {
59c234b4
KW
1786 DEBUG_L(PerlIO_printf(Perl_debug_log,
1787 "Could not change LC_MONETARY locale to %s\n",
1788 save_input_locale));
1789 Safefree(save_monetary_locale);
1790 goto cant_use_monetary;
fa9b773e
KW
1791 }
1792 }
1793
1794 /* Here the current LC_MONETARY is set to the locale of the category
1795 * whose information is desired. */
1796
13542a67
KW
1797 lc = localeconv();
1798 if (! lc
1799 || ! lc->currency_symbol
9f10db87 1800 || is_invariant_string((U8 *) lc->currency_symbol, 0))
13542a67
KW
1801 {
1802 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));
1803 only_ascii = TRUE;
1804 }
1805 else {
1806 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
fa9b773e
KW
1807 }
1808
1809 /* If we changed it, restore LC_MONETARY to its original locale */
1810 if (save_monetary_locale) {
1811 setlocale(LC_MONETARY, save_monetary_locale);
1812 Safefree(save_monetary_locale);
1813 }
1814
13542a67 1815 if (! only_ascii) {
fa9b773e 1816
59c234b4
KW
1817 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1818 * otherwise assume the locale is UTF-8 if and only if the symbol
1819 * is non-ascii UTF-8. */
1820 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1821 save_input_locale, is_utf8));
1822 Safefree(save_input_locale);
1823 return is_utf8;
13542a67 1824 }
fa9b773e
KW
1825 }
1826 cant_use_monetary:
1827
1828# endif /* USE_LOCALE_MONETARY */
1829#endif /* HAS_LOCALECONV */
1830
15f7e74e
KW
1831#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1832
1833/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1834 * the names of the months and weekdays, timezone, and am/pm indicator */
1835 {
1836 char *save_time_locale = NULL;
1837 int hour = 10;
1838 bool is_dst = FALSE;
1839 int dom = 1;
1840 int month = 0;
1841 int i;
1842 char * formatted_time;
1843
1844
1845 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1846 * desired category, if it isn't that locale already */
1847
1848 if (category != LC_TIME) {
1849
1850 save_time_locale = setlocale(LC_TIME, NULL);
1851 if (! save_time_locale) {
1852 DEBUG_L(PerlIO_printf(Perl_debug_log,
1853 "Could not find current locale for LC_TIME\n"));
1854 goto cant_use_time;
1855 }
1856 save_time_locale = stdize_locale(savepv(save_time_locale));
1857
1858 if (strEQ(save_time_locale, save_input_locale)) {
1859 Safefree(save_time_locale);
1860 save_time_locale = NULL;
1861 }
1862 else if (! setlocale(LC_TIME, save_input_locale)) {
1863 DEBUG_L(PerlIO_printf(Perl_debug_log,
1864 "Could not change LC_TIME locale to %s\n",
1865 save_input_locale));
1866 Safefree(save_time_locale);
1867 goto cant_use_time;
1868 }
1869 }
1870
1871 /* Here the current LC_TIME is set to the locale of the category
1872 * whose information is desired. Look at all the days of the week and
9f10db87 1873 * month names, and the timezone and am/pm indicator for UTF-8 variant
15f7e74e
KW
1874 * characters. The first such a one found will tell us if the locale
1875 * is UTF-8 or not */
1876
1877 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1878 formatted_time = my_strftime("%A %B %Z %p",
1879 0, 0, hour, dom, month, 112, 0, 0, is_dst);
9f10db87 1880 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
15f7e74e
KW
1881
1882 /* Here, we didn't find a non-ASCII. Try the next time through
1883 * with the complemented dst and am/pm, and try with the next
1884 * weekday. After we have gotten all weekdays, try the next
1885 * month */
1886 is_dst = ! is_dst;
1887 hour = (hour + 12) % 24;
1888 dom++;
1889 if (i > 6) {
1890 month++;
1891 }
1892 continue;
1893 }
1894
1895 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1896 * false otherwise. But first, restore LC_TIME to its original
1897 * locale if we changed it */
1898 if (save_time_locale) {
1899 setlocale(LC_TIME, save_time_locale);
1900 Safefree(save_time_locale);
1901 }
1902
1903 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1904 save_input_locale,
1905 is_utf8_string((U8 *) formatted_time, 0)));
1906 Safefree(save_input_locale);
1907 return is_utf8_string((U8 *) formatted_time, 0);
1908 }
1909
1910 /* Falling off the end of the loop indicates all the names were just
1911 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1912 * to its original locale */
1913 if (save_time_locale) {
1914 setlocale(LC_TIME, save_time_locale);
1915 Safefree(save_time_locale);
1916 }
1917 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));
1918 }
1919 cant_use_time:
1920
1921#endif
1922
1923#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
855aeb93
JH
1924
1925/* This code is ifdefd out because it was found to not be necessary in testing
5857e934
KW
1926 * on our dromedary test machine, which has over 700 locales. There, this
1927 * added no value to looking at the currency symbol and the time strings. I
1928 * left it in so as to avoid rewriting it if real-world experience indicates
1929 * that dromedary is an outlier. Essentially, instead of returning abpve if we
855aeb93
JH
1930 * haven't found illegal utf8, we continue on and examine all the strerror()
1931 * messages on the platform for utf8ness. If all are ASCII, we still don't
1932 * know the answer; but otherwise we have a pretty good indication of the
5857e934
KW
1933 * utf8ness. The reason this doesn't help much is that the messages may not
1934 * have been translated into the locale. The currency symbol and time strings
1935 * are much more likely to have been translated. */
1936 {
855aeb93 1937 int e;
5857e934
KW
1938 bool is_utf8 = FALSE;
1939 bool non_ascii = FALSE;
855aeb93 1940 char *save_messages_locale = NULL;
5857e934 1941 const char * errmsg = NULL;
855aeb93 1942
5857e934
KW
1943 /* Like above, we set LC_MESSAGES to the locale of the desired
1944 * category, if it isn't that locale already */
855aeb93
JH
1945
1946 if (category != LC_MESSAGES) {
1947
5857e934 1948 save_messages_locale = setlocale(LC_MESSAGES, NULL);
855aeb93 1949 if (! save_messages_locale) {
5857e934
KW
1950 DEBUG_L(PerlIO_printf(Perl_debug_log,
1951 "Could not find current locale for LC_MESSAGES\n"));
855aeb93
JH
1952 goto cant_use_messages;
1953 }
5857e934 1954 save_messages_locale = stdize_locale(savepv(save_messages_locale));
855aeb93
JH
1955
1956 if (strEQ(save_messages_locale, save_input_locale)) {
5857e934
KW
1957 Safefree(save_messages_locale);
1958 save_messages_locale = NULL;
855aeb93
JH
1959 }
1960 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
5857e934
KW
1961 DEBUG_L(PerlIO_printf(Perl_debug_log,
1962 "Could not change LC_MESSAGES locale to %s\n",
1963 save_input_locale));
855aeb93
JH
1964 Safefree(save_messages_locale);
1965 goto cant_use_messages;
1966 }
1967 }
1968
1969 /* Here the current LC_MESSAGES is set to the locale of the category
5857e934
KW
1970 * whose information is desired. Look through all the messages. We
1971 * can't use Strerror() here because it may expand to code that
1972 * segfaults in miniperl */
855aeb93 1973
5857e934
KW
1974 for (e = 0; e <= sys_nerr; e++) {
1975 errno = 0;
1976 errmsg = sys_errlist[e];
1977 if (errno || !errmsg) {
855aeb93
JH
1978 break;
1979 }
5857e934 1980 errmsg = savepv(errmsg);
9f10db87 1981 if (! is_invariant_string((U8 *) errmsg, 0)) {
5857e934
KW
1982 non_ascii = TRUE;
1983 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1984 break;
855aeb93
JH
1985 }
1986 }
5857e934 1987 Safefree(errmsg);
855aeb93
JH
1988
1989 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1990 if (save_messages_locale) {
1991 setlocale(LC_MESSAGES, save_messages_locale);
1992 Safefree(save_messages_locale);
1993 }
1994
5857e934
KW
1995 if (non_ascii) {
1996
1997 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1998 * any non-ascii means it is one; otherwise we assume it isn't */
1999 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
2000 save_input_locale,
2001 is_utf8));
2002 Safefree(save_input_locale);
2003 return is_utf8;
2004 }
855aeb93 2005
5857e934 2006 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));
855aeb93
JH
2007 }
2008 cant_use_messages:
2009
2010#endif
fa9b773e 2011
0080c90a
KW
2012#endif /* the code that is compiled when no nl_langinfo */
2013
92c0a900
KW
2014#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
2015 UTF-8 locale */
97f4de96
KW
2016 /* As a last resort, look at the locale name to see if it matches
2017 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
2018 * return of setlocale(), is actually defined to be opaque, so we can't
2019 * really rely on the absence of various substrings in the name to indicate
2020 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
2021 * be a UTF-8 locale. Similarly for the other common names */
2022
2023 final_pos = strlen(save_input_locale) - 1;
2024 if (final_pos >= 3) {
2025 char *name = save_input_locale;
2026
2027 /* Find next 'U' or 'u' and look from there */
2028 while ((name += strcspn(name, "Uu") + 1)
2029 <= save_input_locale + final_pos - 2)
2030 {
305b8651
KW
2031 if (!isALPHA_FOLD_NE(*name, 't')
2032 || isALPHA_FOLD_NE(*(name + 1), 'f'))
97f4de96
KW
2033 {
2034 continue;
2035 }
2036 name += 2;
2037 if (*(name) == '-') {
2038 if ((name > save_input_locale + final_pos - 1)) {
2039 break;
2040 }
2041 name++;
2042 }
2043 if (*(name) == '8') {
97f4de96
KW
2044 DEBUG_L(PerlIO_printf(Perl_debug_log,
2045 "Locale %s ends with UTF-8 in name\n",
2046 save_input_locale));
00c54b9c 2047 Safefree(save_input_locale);
97f4de96
KW
2048 return TRUE;
2049 }
2050 }
2051 DEBUG_L(PerlIO_printf(Perl_debug_log,
2052 "Locale %s doesn't end with UTF-8 in name\n",
2053 save_input_locale));
2054 }
92c0a900 2055#endif
97f4de96
KW
2056
2057#ifdef WIN32
2058 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
2059 if (final_pos >= 4
2060 && *(save_input_locale + final_pos - 0) == '1'
2061 && *(save_input_locale + final_pos - 1) == '0'
2062 && *(save_input_locale + final_pos - 2) == '0'
2063 && *(save_input_locale + final_pos - 3) == '5'
2064 && *(save_input_locale + final_pos - 4) == '6')
2065 {
2066 DEBUG_L(PerlIO_printf(Perl_debug_log,
2067 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
2068 save_input_locale));
2069 Safefree(save_input_locale);
2070 return TRUE;
2071 }
2072#endif
2073
2074 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
2075 * since we are about to return FALSE anyway, there is no point in doing
2076 * this extra work */
2077#if 0
2078 if (instr(save_input_locale, "8859")) {
2079 DEBUG_L(PerlIO_printf(Perl_debug_log,
2080 "Locale %s has 8859 in name, not UTF-8 locale\n",
2081 save_input_locale));
2082 Safefree(save_input_locale);
2083 return FALSE;
2084 }
2085#endif
2086
69014004
KW
2087 DEBUG_L(PerlIO_printf(Perl_debug_log,
2088 "Assuming locale %s is not a UTF-8 locale\n",
2089 save_input_locale));
fa9b773e 2090 Safefree(save_input_locale);
7d74bb61
KW
2091 return FALSE;
2092}
2093
8ef6e574 2094#endif
7d74bb61 2095
d6ded950
KW
2096
2097bool
2098Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
2099{
1a4f13e1 2100 dVAR;
d6ded950
KW
2101 /* Internal function which returns if we are in the scope of a pragma that
2102 * enables the locale category 'category'. 'compiling' should indicate if
2103 * this is during the compilation phase (TRUE) or not (FALSE). */
2104
2105 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
2106
2107 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
2108 if (! categories || categories == &PL_sv_placeholder) {
2109 return FALSE;
2110 }
2111
2112 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
2113 * a valid unsigned */
2114 assert(category >= -1);
2115 return cBOOL(SvUV(categories) & (1U << (category + 1)));
2116}
2117
2c6ee1a7
KW
2118char *
2119Perl_my_strerror(pTHX_ const int errnum) {
a0b53297 2120 dVAR;
2c6ee1a7
KW
2121
2122 /* Uses C locale for the error text unless within scope of 'use locale' for
2123 * LC_MESSAGES */
2124
2125#ifdef USE_LOCALE_MESSAGES
8ee4b769 2126 if (! IN_LC(LC_MESSAGES)) {
a0b53297
KW
2127 char * save_locale;
2128
2129 /* We have a critical section to prevent another thread from changing
2130 * the locale out from under us (or zapping the buffer returned from
2131 * setlocale() ) */
2132 LOCALE_LOCK;
2133
2134 save_locale = setlocale(LC_MESSAGES, NULL);
a39edc4c 2135 if (! isNAME_C_OR_POSIX(save_locale)) {
2c6ee1a7
KW
2136 char *errstr;
2137
2138 /* The next setlocale likely will zap this, so create a copy */
2139 save_locale = savepv(save_locale);
2140
2141 setlocale(LC_MESSAGES, "C");
2142
2143 /* This points to the static space in Strerror, with all its
2144 * limitations */
2145 errstr = Strerror(errnum);
2146
2147 setlocale(LC_MESSAGES, save_locale);
2148 Safefree(save_locale);
a0b53297
KW
2149
2150 LOCALE_UNLOCK;
2151
2c6ee1a7
KW
2152 return errstr;
2153 }
a0b53297
KW
2154
2155 LOCALE_UNLOCK;
2c6ee1a7
KW
2156 }
2157#endif
2158
2159 return Strerror(errnum);
2160}
2161
66610fdd 2162/*
747c467a
KW
2163
2164=head1 Locale-related functions and macros
2165
2166=for apidoc sync_locale
2167
2168Changing the program's locale should be avoided by XS code. Nevertheless,
2169certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
2170happens, Perl needs to be told that the locale has changed. Use this function
2171to do so, before returning to Perl.
2172
2173=cut
2174*/
2175
2176void
2177Perl_sync_locale(pTHX)
2178{
2179
2180#ifdef USE_LOCALE_CTYPE
2181 new_ctype(setlocale(LC_CTYPE, NULL));
2182#endif /* USE_LOCALE_CTYPE */
2183
2184#ifdef USE_LOCALE_COLLATE
2185 new_collate(setlocale(LC_COLLATE, NULL));
2186#endif
2187
2188#ifdef USE_LOCALE_NUMERIC
2189 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
2190 new_numeric(setlocale(LC_NUMERIC, NULL));
2191#endif /* USE_LOCALE_NUMERIC */
2192
2193}
2194
5d1187d1
KW
2195#if defined(DEBUGGING) && defined(USE_LOCALE)
2196
2197char *
2198Perl__setlocale_debug_string(const int category, /* category number,
2199 like LC_ALL */
2200 const char* const locale, /* locale name */
2201
2202 /* return value from setlocale() when attempting to
2203 * set 'category' to 'locale' */
2204 const char* const retval)
2205{
2206 /* Returns a pointer to a NUL-terminated string in static storage with
2207 * added text about the info passed in. This is not thread safe and will
2208 * be overwritten by the next call, so this should be used just to
fa07b8e5 2209 * formulate a string to immediately print or savepv() on. */
5d1187d1 2210
398a990f
DM
2211 /* initialise to a non-null value to keep it out of BSS and so keep
2212 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
2213 static char ret[128] = "x";
5d1187d1 2214
fa07b8e5 2215 my_strlcpy(ret, "setlocale(", sizeof(ret));
5d1187d1
KW
2216
2217 switch (category) {
2218 default:
fa07b8e5 2219 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
5d1187d1
KW
2220 break;
2221# ifdef LC_ALL
2222 case LC_ALL:
fa07b8e5 2223 my_strlcat(ret, "LC_ALL", sizeof(ret));
5d1187d1
KW
2224 break;
2225# endif
2226# ifdef LC_CTYPE
2227 case LC_CTYPE:
fa07b8e5 2228 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
5d1187d1
KW
2229 break;
2230# endif
2231# ifdef LC_NUMERIC
2232 case LC_NUMERIC:
fa07b8e5 2233 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
5d1187d1
KW
2234 break;
2235# endif
2236# ifdef LC_COLLATE
2237 case LC_COLLATE:
fa07b8e5 2238 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
5d1187d1
KW
2239 break;
2240# endif
2241# ifdef LC_TIME
2242 case LC_TIME:
fa07b8e5 2243 my_strlcat(ret, "LC_TIME", sizeof(ret));
5d1187d1
KW
2244 break;
2245# endif
2246# ifdef LC_MONETARY
2247 case LC_MONETARY:
fa07b8e5 2248 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
5d1187d1
KW
2249 break;
2250# endif
2251# ifdef LC_MESSAGES
2252 case LC_MESSAGES:
fa07b8e5 2253 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
5d1187d1
KW
2254 break;
2255# endif
2256 }
2257
fa07b8e5 2258 my_strlcat(ret, ", ", sizeof(ret));
5d1187d1
KW
2259
2260 if (locale) {
fa07b8e5
KW
2261 my_strlcat(ret, "\"", sizeof(ret));
2262 my_strlcat(ret, locale, sizeof(ret));
2263 my_strlcat(ret, "\"", sizeof(ret));
5d1187d1
KW
2264 }
2265 else {
fa07b8e5 2266 my_strlcat(ret, "NULL", sizeof(ret));
5d1187d1
KW
2267 }
2268
fa07b8e5 2269 my_strlcat(ret, ") returned ", sizeof(ret));
5d1187d1
KW
2270
2271 if (retval) {
fa07b8e5
KW
2272 my_strlcat(ret, "\"", sizeof(ret));
2273 my_strlcat(ret, retval, sizeof(ret));
2274 my_strlcat(ret, "\"", sizeof(ret));
5d1187d1
KW
2275 }
2276 else {
fa07b8e5 2277 my_strlcat(ret, "NULL", sizeof(ret));
5d1187d1
KW
2278 }
2279
2280 assert(strlen(ret) < sizeof(ret));
2281
2282 return ret;
2283}
2284
2285#endif
747c467a
KW
2286
2287
2288/*
14d04a33 2289 * ex: set ts=8 sts=4 sw=4 et:
37442d52 2290 */