This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
locale.c, sv.c: Add, fix some comments
[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;
485 PL_collxfrm_base = 0;
486 PL_collxfrm_mult = 2;
487 return;
488 }
489
d35fca5f 490 /* If this is not the same locale as currently, set the new one up */
98994639
HS
491 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
492 ++PL_collation_ix;
493 Safefree(PL_collation_name);
494 PL_collation_name = stdize_locale(savepv(newcoll));
a39edc4c 495 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
98994639
HS
496
497 {
d35fca5f
KW
498 /* A locale collation definition includes primary, secondary,
499 * tertiary, etc. weights for each character. To sort, the primary
500 * weights are used, and only if they compare equal, then the
501 * secondary weights are used, and only if they compare equal, then
502 * the tertiary, etc. strxfrm() works by taking the input string,
503 * say ABC, and creating an output string consisting of first the
504 * primary weights, A¹B¹C¹ followed by the secondary ones, A²B²C²;
505 * and then the tertiary, etc, yielding A¹B¹C¹A²B²C²A³B³C³....
506 * Some characters may not have weights at every level. In our
507 * example, let's say B doesn't have a tertiary weight, and A
508 * doesn't have a secondary weight. The constructed string is then
509 * going to be A¹B¹C¹B²C²A³C³.... This has the desired
510 * characteristics that strcmp() will look at the secondary or
511 * tertiary weights only if the strings compare equal at all higher
512 * priority weights. The length of the transformed string is
513 * roughly a linear function of the input string. It's not exactly
514 * linear because some characters don't have weights at all levels,
515 * and there are some complications, so there is often per-string
516 * overhead. When we call strxfrm() we have to allocate some
517 * memory to hold the transformed string. The calculations below
518 * try to find constants for this locale 'm' and 'b' so that m*x +
519 * b equals how much space we need given the size of the input
520 * string in 'x'. If we calculate too small, we increase the size
521 * as needed, and call strxfrm() again, but it is better to get it
522 * right the first time to avoid wasted expensive string
523 * transformations. */
98994639
HS
524 /* 2: at most so many chars ('a', 'b'). */
525 /* 50: surely no system expands a char more. */
526#define XFRMBUFSIZE (2 * 50)
527 char xbuf[XFRMBUFSIZE];
8772537c
AL
528 const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE);
529 const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE);
530 const SSize_t mult = fb - fa;
e0601d3c 531 if (mult < 1 && !(fa == 0 && fb == 0))
ad49ad39
NC
532 Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
533 (UV) fa, (UV) fb);
eb160463 534 PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
98994639
HS
535 PL_collxfrm_mult = mult;
536 }
537 }
538
f2ce9e1c
JH
539#else
540 PERL_UNUSED_ARG(newcoll);
98994639
HS
541#endif /* USE_LOCALE_COLLATE */
542}
543
b385bb4d
KW
544#ifdef WIN32
545
546char *
547Perl_my_setlocale(pTHX_ int category, const char* locale)
548{
549 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
550 * difference unless the input locale is "", which means on Windows to get
551 * the machine default, which is set via the computer's "Regional and
552 * Language Options" (or its current equivalent). In POSIX, it instead
553 * means to find the locale from the user's environment. This routine
554 * looks in the environment, and, if anything is found, uses that instead
555 * of going to the machine default. If there is no environment override,
556 * the machine default is used, as normal, by calling the real setlocale()
557 * with "". The POSIX behavior is to use the LC_ALL variable if set;
558 * otherwise to use the particular category's variable if set; otherwise to
559 * use the LANG variable. */
560
175c4cf9 561 bool override_LC_ALL = FALSE;
89f7b9aa
KW
562 char * result;
563
b385bb4d
KW
564 if (locale && strEQ(locale, "")) {
565# ifdef LC_ALL
566 locale = PerlEnv_getenv("LC_ALL");
567 if (! locale) {
568#endif
569 switch (category) {
570# ifdef LC_ALL
571 case LC_ALL:
481465ea 572 override_LC_ALL = TRUE;
b385bb4d
KW
573 break; /* We already know its variable isn't set */
574# endif
575# ifdef USE_LOCALE_TIME
576 case LC_TIME:
577 locale = PerlEnv_getenv("LC_TIME");
578 break;
579# endif
580# ifdef USE_LOCALE_CTYPE
581 case LC_CTYPE:
582 locale = PerlEnv_getenv("LC_CTYPE");
583 break;
584# endif
585# ifdef USE_LOCALE_COLLATE
586 case LC_COLLATE:
587 locale = PerlEnv_getenv("LC_COLLATE");
588 break;
589# endif
590# ifdef USE_LOCALE_MONETARY
591 case LC_MONETARY:
592 locale = PerlEnv_getenv("LC_MONETARY");
593 break;
594# endif
595# ifdef USE_LOCALE_NUMERIC
596 case LC_NUMERIC:
597 locale = PerlEnv_getenv("LC_NUMERIC");
598 break;
599# endif
600# ifdef USE_LOCALE_MESSAGES
601 case LC_MESSAGES:
602 locale = PerlEnv_getenv("LC_MESSAGES");
603 break;
604# endif
605 default:
606 /* This is a category, like PAPER_SIZE that we don't
607 * know about; and so can't provide a wrapper. */
608 break;
609 }
610 if (! locale) {
611 locale = PerlEnv_getenv("LANG");
481465ea 612 if (! locale) {
b385bb4d
KW
613 locale = "";
614 }
615 }
616# ifdef LC_ALL
617 }
618# endif
619 }
620
89f7b9aa 621 result = setlocale(category, locale);
bbc98134
KW
622 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
623 _setlocale_debug_string(category, locale, result)));
89f7b9aa 624
481465ea 625 if (! override_LC_ALL) {
89f7b9aa
KW
626 return result;
627 }
628
dfd77d7a 629 /* Here the input category was LC_ALL, and we have set it to what is in the
481465ea
KW
630 * LANG variable or the system default if there is no LANG. But these have
631 * lower priority than the other LC_foo variables, so override it for each
632 * one that is set. (If they are set to "", it means to use the same thing
633 * we just set LC_ALL to, so can skip) */
89f7b9aa
KW
634# ifdef USE_LOCALE_TIME
635 result = PerlEnv_getenv("LC_TIME");
730252b2 636 if (result && strNE(result, "")) {
89f7b9aa 637 setlocale(LC_TIME, result);
bbc98134
KW
638 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
639 __FILE__, __LINE__,
640 _setlocale_debug_string(LC_TIME, result, "not captured")));
89f7b9aa
KW
641 }
642# endif
643# ifdef USE_LOCALE_CTYPE
644 result = PerlEnv_getenv("LC_CTYPE");
730252b2 645 if (result && strNE(result, "")) {
89f7b9aa 646 setlocale(LC_CTYPE, result);
bbc98134
KW
647 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
648 __FILE__, __LINE__,
649 _setlocale_debug_string(LC_CTYPE, result, "not captured")));
89f7b9aa
KW
650 }
651# endif
652# ifdef USE_LOCALE_COLLATE
653 result = PerlEnv_getenv("LC_COLLATE");
730252b2 654 if (result && strNE(result, "")) {
89f7b9aa 655 setlocale(LC_COLLATE, result);
bbc98134
KW
656 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
657 __FILE__, __LINE__,
658 _setlocale_debug_string(LC_COLLATE, result, "not captured")));
89f7b9aa
KW
659 }
660# endif
661# ifdef USE_LOCALE_MONETARY
662 result = PerlEnv_getenv("LC_MONETARY");
730252b2 663 if (result && strNE(result, "")) {
89f7b9aa 664 setlocale(LC_MONETARY, result);
bbc98134
KW
665 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
666 __FILE__, __LINE__,
667 _setlocale_debug_string(LC_MONETARY, result, "not captured")));
89f7b9aa
KW
668 }
669# endif
670# ifdef USE_LOCALE_NUMERIC
671 result = PerlEnv_getenv("LC_NUMERIC");
730252b2 672 if (result && strNE(result, "")) {
89f7b9aa 673 setlocale(LC_NUMERIC, result);
bbc98134
KW
674 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
675 __FILE__, __LINE__,
676 _setlocale_debug_string(LC_NUMERIC, result, "not captured")));
89f7b9aa
KW
677 }
678# endif
679# ifdef USE_LOCALE_MESSAGES
680 result = PerlEnv_getenv("LC_MESSAGES");
730252b2 681 if (result && strNE(result, "")) {
89f7b9aa 682 setlocale(LC_MESSAGES, result);
bbc98134
KW
683 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
684 __FILE__, __LINE__,
685 _setlocale_debug_string(LC_MESSAGES, result, "not captured")));
89f7b9aa
KW
686 }
687# endif
688
bbc98134
KW
689 result = setlocale(LC_ALL, NULL);
690 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
691 __FILE__, __LINE__,
692 _setlocale_debug_string(LC_ALL, NULL, result)));
89f7b9aa 693
bbc98134 694 return result;
b385bb4d
KW
695}
696
697#endif
698
699
98994639
HS
700/*
701 * Initialize locale awareness.
702 */
703int
704Perl_init_i18nl10n(pTHX_ int printwarn)
705{
0e92a118
KW
706 /* printwarn is
707 *
708 * 0 if not to output warning when setup locale is bad
709 * 1 if to output warning based on value of PERL_BADLANG
710 * >1 if to output regardless of PERL_BADLANG
711 *
712 * returns
98994639 713 * 1 = set ok or not applicable,
0e92a118
KW
714 * 0 = fallback to a locale of lower priority
715 * -1 = fallback to all locales failed, not even to the C locale
6b058d42
KW
716 *
717 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
718 * set, debugging information is output.
719 *
720 * This looks more complicated than it is, mainly due to the #ifdefs.
721 *
722 * We try to set LC_ALL to the value determined by the environment. If
723 * there is no LC_ALL on this platform, we try the individual categories we
724 * know about. If this works, we are done.
725 *
726 * But if it doesn't work, we have to do something else. We search the
727 * environment variables ourselves instead of relying on the system to do
728 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
729 * think there is one), and the ultimate fallback "C". This is all done in
730 * the same loop as above to avoid duplicating code, but it makes things
731 * more complex. After the original failure, we add the fallback
732 * possibilities to the list of locales to try, and iterate the loop
733 * through them all until one succeeds.
734 *
735 * On Ultrix, the locale MUST come from the environment, so there is
736 * preliminary code to set it. I (khw) am not sure that it is necessary,
737 * and that this couldn't be folded into the loop, but barring any real
738 * platforms to test on, it's staying as-is
739 *
740 * A slight complication is that in embedded Perls, the locale may already
741 * be set-up, and we don't want to get it from the normal environment
742 * variables. This is handled by having a special environment variable
743 * indicate we're in this situation. We simply set setlocale's 2nd
744 * parameter to be a NULL instead of "". That indicates to setlocale that
745 * it is not to change anything, but to return the current value,
746 * effectively initializing perl's db to what the locale already is.
747 *
748 * We play the same trick with NULL if a LC_ALL succeeds. We call
749 * setlocale() on the individual categores with NULL to get their existing
750 * values for our db, instead of trying to change them.
751 * */
98994639 752
0e92a118
KW
753 int ok = 1;
754
98994639 755#if defined(USE_LOCALE)
98994639
HS
756#ifdef USE_LOCALE_CTYPE
757 char *curctype = NULL;
758#endif /* USE_LOCALE_CTYPE */
759#ifdef USE_LOCALE_COLLATE
760 char *curcoll = NULL;
761#endif /* USE_LOCALE_COLLATE */
762#ifdef USE_LOCALE_NUMERIC
763 char *curnum = NULL;
764#endif /* USE_LOCALE_NUMERIC */
765#ifdef __GLIBC__
175c4cf9 766 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
98994639 767#endif
65ebb059 768
ccd65d51
KW
769 /* NULL uses the existing already set up locale */
770 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
771 ? NULL
772 : "";
5d1187d1
KW
773#ifdef DEBUGGING
774 const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
775 ? TRUE
776 : FALSE;
777# define DEBUG_LOCALE_INIT(category, locale, result) \
778 STMT_START { \
779 if (debug) { \
780 PerlIO_printf(Perl_debug_log, \
781 "%s:%d: %s\n", \
782 __FILE__, __LINE__, \
783 _setlocale_debug_string(category, \
784 locale, \
785 result)); \
786 } \
787 } STMT_END
788#else
789# define DEBUG_LOCALE_INIT(a,b,c)
790#endif
c3fcd832
KW
791 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
792 unsigned int trial_locales_count;
175c4cf9
KW
793 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
794 const char * const lang = savepv(PerlEnv_getenv("LANG"));
98994639 795 bool setlocale_failure = FALSE;
65ebb059
KW
796 unsigned int i;
797 char *p;
175c4cf9
KW
798
799 /* A later getenv() could zap this, so only use here */
800 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
801
802 const bool locwarn = (printwarn > 1
803 || (printwarn
804 && (! bad_lang_use_once
22ff3130
HS
805 || (
806 /* disallow with "" or "0" */
807 *bad_lang_use_once
808 && strNE("0", bad_lang_use_once)))));
0e92a118 809 bool done = FALSE;
5d1187d1
KW
810 char * sl_result; /* return from setlocale() */
811 char * locale_param;
6bce99ee
JH
812#ifdef WIN32
813 /* In some systems you can find out the system default locale
814 * and use that as the fallback locale. */
815# define SYSTEM_DEFAULT_LOCALE
816#endif
817#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059 818 const char *system_default_locale = NULL;
6bce99ee 819#endif
98994639 820
0e92a118
KW
821#ifndef LOCALE_ENVIRON_REQUIRED
822 PERL_UNUSED_VAR(done);
5d1187d1 823 PERL_UNUSED_VAR(locale_param);
0e92a118 824#else
98994639
HS
825
826 /*
827 * Ultrix setlocale(..., "") fails if there are no environment
828 * variables from which to get a locale name.
829 */
830
b3e384bf 831# ifdef LC_ALL
98994639 832 if (lang) {
5d1187d1
KW
833 sl_result = my_setlocale(LC_ALL, setlocale_init);
834 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
835 if (sl_result)
98994639
HS
836 done = TRUE;
837 else
838 setlocale_failure = TRUE;
839 }
5d1187d1 840 if (! setlocale_failure) {
b3e384bf 841# ifdef USE_LOCALE_CTYPE
5d1187d1
KW
842 locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
843 ? setlocale_init
844 : NULL;
845 curctype = my_setlocale(LC_CTYPE, locale_param);
846 DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
847 if (! curctype)
98994639
HS
848 setlocale_failure = TRUE;
849 else
850 curctype = savepv(curctype);
b3e384bf
KW
851# endif /* USE_LOCALE_CTYPE */
852# ifdef USE_LOCALE_COLLATE
5d1187d1
KW
853 locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
854 ? setlocale_init
855 : NULL;
856 curcoll = my_setlocale(LC_COLLATE, locale_param);
857 DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
858 if (! curcoll)
98994639
HS
859 setlocale_failure = TRUE;
860 else
861 curcoll = savepv(curcoll);
b3e384bf
KW
862# endif /* USE_LOCALE_COLLATE */
863# ifdef USE_LOCALE_NUMERIC
5d1187d1
KW
864 locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
865 ? setlocale_init
866 : NULL;
867 curnum = my_setlocale(LC_NUMERIC, locale_param);
868 DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
869 if (! curnum)
98994639
HS
870 setlocale_failure = TRUE;
871 else
872 curnum = savepv(curnum);
b3e384bf 873# endif /* USE_LOCALE_NUMERIC */
a782673d 874# ifdef USE_LOCALE_MESSAGES
5d1187d1
KW
875 locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
876 ? setlocale_init
877 : NULL;
878 sl_result = my_setlocale(LC_MESSAGES, locale_param);
879 DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
880 if (! sl_result)
a782673d
KW
881 setlocale_failure = TRUE;
882 }
883# endif /* USE_LOCALE_MESSAGES */
c835d6be 884# ifdef USE_LOCALE_MONETARY
5d1187d1
KW
885 locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
886 ? setlocale_init
887 : NULL;
888 sl_result = my_setlocale(LC_MONETARY, locale_param);
889 DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
890 if (! sl_result) {
c835d6be
KW
891 setlocale_failure = TRUE;
892 }
893# endif /* USE_LOCALE_MONETARY */
98994639
HS
894 }
895
b3e384bf 896# endif /* LC_ALL */
98994639
HS
897
898#endif /* !LOCALE_ENVIRON_REQUIRED */
899
65ebb059 900 /* We try each locale in the list until we get one that works, or exhaust
20a240df
KW
901 * the list. Normally the loop is executed just once. But if setting the
902 * locale fails, inside the loop we add fallback trials to the array and so
903 * will execute the loop multiple times */
c3fcd832
KW
904 trial_locales[0] = setlocale_init;
905 trial_locales_count = 1;
65ebb059
KW
906 for (i= 0; i < trial_locales_count; i++) {
907 const char * trial_locale = trial_locales[i];
908
909 if (i > 0) {
910
911 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
912 * when i==0, but I (khw) don't think that behavior makes much
913 * sense */
914 setlocale_failure = FALSE;
915
6bce99ee
JH
916#ifdef SYSTEM_DEFAULT_LOCALE
917# ifdef WIN32
65ebb059
KW
918 /* On Windows machines, an entry of "" after the 0th means to use
919 * the system default locale, which we now proceed to get. */
920 if (strEQ(trial_locale, "")) {
921 unsigned int j;
922
923 /* Note that this may change the locale, but we are going to do
924 * that anyway just below */
925 system_default_locale = setlocale(LC_ALL, "");
5d1187d1 926 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
65ebb059
KW
927
928 /* Skip if invalid or it's already on the list of locales to
929 * try */
930 if (! system_default_locale) {
931 goto next_iteration;
932 }
933 for (j = 0; j < trial_locales_count; j++) {
934 if (strEQ(system_default_locale, trial_locales[j])) {
935 goto next_iteration;
936 }
937 }
938
939 trial_locale = system_default_locale;
940 }
6bce99ee
JH
941# endif /* WIN32 */
942#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
943 }
944
98994639 945#ifdef LC_ALL
5d1187d1
KW
946 sl_result = my_setlocale(LC_ALL, trial_locale);
947 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
948 if (! sl_result) {
49c85077 949 setlocale_failure = TRUE;
7cd8b568
KW
950 }
951 else {
952 /* Since LC_ALL succeeded, it should have changed all the other
953 * categories it can to its value; so we massage things so that the
954 * setlocales below just return their category's current values.
955 * This adequately handles the case in NetBSD where LC_COLLATE may
956 * not be defined for a locale, and setting it individually will
957 * fail, whereas setting LC_ALL suceeds, leaving LC_COLLATE set to
958 * the POSIX locale. */
959 trial_locale = NULL;
960 }
98994639
HS
961#endif /* LC_ALL */
962
49c85077 963 if (!setlocale_failure) {
98994639 964#ifdef USE_LOCALE_CTYPE
49c85077 965 Safefree(curctype);
5d1187d1
KW
966 curctype = my_setlocale(LC_CTYPE, trial_locale);
967 DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
968 if (! curctype)
49c85077
KW
969 setlocale_failure = TRUE;
970 else
971 curctype = savepv(curctype);
98994639
HS
972#endif /* USE_LOCALE_CTYPE */
973#ifdef USE_LOCALE_COLLATE
49c85077 974 Safefree(curcoll);
5d1187d1
KW
975 curcoll = my_setlocale(LC_COLLATE, trial_locale);
976 DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
977 if (! curcoll)
49c85077
KW
978 setlocale_failure = TRUE;
979 else
980 curcoll = savepv(curcoll);
98994639
HS
981#endif /* USE_LOCALE_COLLATE */
982#ifdef USE_LOCALE_NUMERIC
49c85077 983 Safefree(curnum);
5d1187d1
KW
984 curnum = my_setlocale(LC_NUMERIC, trial_locale);
985 DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
986 if (! curnum)
49c85077
KW
987 setlocale_failure = TRUE;
988 else
989 curnum = savepv(curnum);
98994639 990#endif /* USE_LOCALE_NUMERIC */
a782673d 991#ifdef USE_LOCALE_MESSAGES
5d1187d1
KW
992 sl_result = my_setlocale(LC_MESSAGES, trial_locale);
993 DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
994 if (! (sl_result))
a782673d
KW
995 setlocale_failure = TRUE;
996#endif /* USE_LOCALE_MESSAGES */
c835d6be 997#ifdef USE_LOCALE_MONETARY
5d1187d1
KW
998 sl_result = my_setlocale(LC_MONETARY, trial_locale);
999 DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
1000 if (! (sl_result))
c835d6be
KW
1001 setlocale_failure = TRUE;
1002#endif /* USE_LOCALE_MONETARY */
1003
49c85077
KW
1004 if (! setlocale_failure) { /* Success */
1005 break;
1006 }
65ebb059 1007 }
98994639 1008
49c85077
KW
1009 /* Here, something failed; will need to try a fallback. */
1010 ok = 0;
65ebb059 1011
49c85077
KW
1012 if (i == 0) {
1013 unsigned int j;
98994639 1014
65ebb059 1015 if (locwarn) { /* Output failure info only on the first one */
98994639
HS
1016#ifdef LC_ALL
1017
49c85077
KW
1018 PerlIO_printf(Perl_error_log,
1019 "perl: warning: Setting locale failed.\n");
98994639
HS
1020
1021#else /* !LC_ALL */
1022
49c85077
KW
1023 PerlIO_printf(Perl_error_log,
1024 "perl: warning: Setting locale failed for the categories:\n\t");
20a240df 1025# ifdef USE_LOCALE_CTYPE
49c85077
KW
1026 if (! curctype)
1027 PerlIO_printf(Perl_error_log, "LC_CTYPE ");
20a240df
KW
1028# endif /* USE_LOCALE_CTYPE */
1029# ifdef USE_LOCALE_COLLATE
49c85077
KW
1030 if (! curcoll)
1031 PerlIO_printf(Perl_error_log, "LC_COLLATE ");
20a240df
KW
1032# endif /* USE_LOCALE_COLLATE */
1033# ifdef USE_LOCALE_NUMERIC
49c85077
KW
1034 if (! curnum)
1035 PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
20a240df 1036# endif /* USE_LOCALE_NUMERIC */
a782673d 1037 PerlIO_printf(Perl_error_log, "and possibly others\n");
98994639
HS
1038
1039#endif /* LC_ALL */
1040
49c85077
KW
1041 PerlIO_printf(Perl_error_log,
1042 "perl: warning: Please check that your locale settings:\n");
98994639
HS
1043
1044#ifdef __GLIBC__
49c85077
KW
1045 PerlIO_printf(Perl_error_log,
1046 "\tLANGUAGE = %c%s%c,\n",
1047 language ? '"' : '(',
1048 language ? language : "unset",
1049 language ? '"' : ')');
98994639
HS
1050#endif
1051
49c85077
KW
1052 PerlIO_printf(Perl_error_log,
1053 "\tLC_ALL = %c%s%c,\n",
1054 lc_all ? '"' : '(',
1055 lc_all ? lc_all : "unset",
1056 lc_all ? '"' : ')');
98994639
HS
1057
1058#if defined(USE_ENVIRON_ARRAY)
49c85077
KW
1059 {
1060 char **e;
1061 for (e = environ; *e; e++) {
1062 if (strnEQ(*e, "LC_", 3)
1063 && strnNE(*e, "LC_ALL=", 7)
1064 && (p = strchr(*e, '=')))
1065 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
1066 (int)(p - *e), *e, p + 1);
1067 }
1068 }
98994639 1069#else
49c85077
KW
1070 PerlIO_printf(Perl_error_log,
1071 "\t(possibly more locale environment variables)\n");
98994639
HS
1072#endif
1073
49c85077
KW
1074 PerlIO_printf(Perl_error_log,
1075 "\tLANG = %c%s%c\n",
1076 lang ? '"' : '(',
1077 lang ? lang : "unset",
1078 lang ? '"' : ')');
98994639 1079
49c85077
KW
1080 PerlIO_printf(Perl_error_log,
1081 " are supported and installed on your system.\n");
1082 }
98994639 1083
65ebb059 1084 /* Calculate what fallback locales to try. We have avoided this
f6bab5f6 1085 * until we have to, because failure is quite unlikely. This will
65ebb059
KW
1086 * usually change the upper bound of the loop we are in.
1087 *
1088 * Since the system's default way of setting the locale has not
1089 * found one that works, We use Perl's defined ordering: LC_ALL,
1090 * LANG, and the C locale. We don't try the same locale twice, so
1091 * don't add to the list if already there. (On POSIX systems, the
1092 * LC_ALL element will likely be a repeat of the 0th element "",
6b058d42
KW
1093 * but there's no harm done by doing it explicitly.
1094 *
1095 * Note that this tries the LC_ALL environment variable even on
1096 * systems which have no LC_ALL locale setting. This may or may
1097 * not have been originally intentional, but there's no real need
1098 * to change the behavior. */
65ebb059
KW
1099 if (lc_all) {
1100 for (j = 0; j < trial_locales_count; j++) {
1101 if (strEQ(lc_all, trial_locales[j])) {
1102 goto done_lc_all;
1103 }
1104 }
1105 trial_locales[trial_locales_count++] = lc_all;
1106 }
1107 done_lc_all:
98994639 1108
65ebb059
KW
1109 if (lang) {
1110 for (j = 0; j < trial_locales_count; j++) {
1111 if (strEQ(lang, trial_locales[j])) {
1112 goto done_lang;
1113 }
1114 }
1115 trial_locales[trial_locales_count++] = lang;
1116 }
1117 done_lang:
1118
1119#if defined(WIN32) && defined(LC_ALL)
1120 /* For Windows, we also try the system default locale before "C".
1121 * (If there exists a Windows without LC_ALL we skip this because
1122 * it gets too complicated. For those, the "C" is the next
1123 * fallback possibility). The "" is the same as the 0th element of
1124 * the array, but the code at the loop above knows to treat it
1125 * differently when not the 0th */
1126 trial_locales[trial_locales_count++] = "";
1127#endif
1128
1129 for (j = 0; j < trial_locales_count; j++) {
1130 if (strEQ("C", trial_locales[j])) {
1131 goto done_C;
1132 }
1133 }
1134 trial_locales[trial_locales_count++] = "C";
98994639 1135
65ebb059
KW
1136 done_C: ;
1137 } /* end of first time through the loop */
98994639 1138
65ebb059
KW
1139#ifdef WIN32
1140 next_iteration: ;
1141#endif
1142
1143 } /* end of looping through the trial locales */
1144
1145 if (ok < 1) { /* If we tried to fallback */
1146 const char* msg;
1147 if (! setlocale_failure) { /* fallback succeeded */
1148 msg = "Falling back to";
1149 }
1150 else { /* fallback failed */
98994639 1151
65ebb059
KW
1152 /* We dropped off the end of the loop, so have to decrement i to
1153 * get back to the value the last time through */
1154 i--;
98994639 1155
65ebb059
KW
1156 ok = -1;
1157 msg = "Failed to fall back to";
1158
1159 /* To continue, we should use whatever values we've got */
98994639 1160#ifdef USE_LOCALE_CTYPE
49c85077
KW
1161 Safefree(curctype);
1162 curctype = savepv(setlocale(LC_CTYPE, NULL));
5d1187d1 1163 DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
98994639
HS
1164#endif /* USE_LOCALE_CTYPE */
1165#ifdef USE_LOCALE_COLLATE
49c85077
KW
1166 Safefree(curcoll);
1167 curcoll = savepv(setlocale(LC_COLLATE, NULL));
5d1187d1 1168 DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
98994639
HS
1169#endif /* USE_LOCALE_COLLATE */
1170#ifdef USE_LOCALE_NUMERIC
49c85077
KW
1171 Safefree(curnum);
1172 curnum = savepv(setlocale(LC_NUMERIC, NULL));
5d1187d1 1173 DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
98994639 1174#endif /* USE_LOCALE_NUMERIC */
65ebb059
KW
1175 }
1176
1177 if (locwarn) {
1178 const char * description;
1179 const char * name = "";
1180 if (strEQ(trial_locales[i], "C")) {
1181 description = "the standard locale";
1182 name = "C";
1183 }
6bce99ee 1184#ifdef SYSTEM_DEFAULT_LOCALE
65ebb059
KW
1185 else if (strEQ(trial_locales[i], "")) {
1186 description = "the system default locale";
1187 if (system_default_locale) {
1188 name = system_default_locale;
1189 }
1190 }
6bce99ee 1191#endif /* SYSTEM_DEFAULT_LOCALE */
65ebb059
KW
1192 else {
1193 description = "a fallback locale";
1194 name = trial_locales[i];
1195 }
1196 if (name && strNE(name, "")) {
1197 PerlIO_printf(Perl_error_log,
1198 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
1199 }
1200 else {
1201 PerlIO_printf(Perl_error_log,
1202 "perl: warning: %s %s.\n", msg, description);
1203 }
1204 }
1205 } /* End of tried to fallback */
98994639
HS
1206
1207#ifdef USE_LOCALE_CTYPE
1208 new_ctype(curctype);
1209#endif /* USE_LOCALE_CTYPE */
1210
1211#ifdef USE_LOCALE_COLLATE
1212 new_collate(curcoll);
1213#endif /* USE_LOCALE_COLLATE */
1214
1215#ifdef USE_LOCALE_NUMERIC
1216 new_numeric(curnum);
1217#endif /* USE_LOCALE_NUMERIC */
b310b053 1218
8ef6e574 1219#if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
49c85077
KW
1220 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
1221 * locale is UTF-8. If PL_utf8locale and PL_unicode (set by -C or by
1222 * $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
1223 * PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
1224 * discipline. */
c1284011 1225 PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
49c85077 1226
a05d7ebb 1227 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
fde18df1
JH
1228 This is an alternative to using the -C command line switch
1229 (the -C if present will override this). */
1230 {
dd374669 1231 const char *p = PerlEnv_getenv("PERL_UNICODE");
a05d7ebb 1232 PL_unicode = p ? parse_unicode_opts(&p) : 0;
5a22a2bb
NC
1233 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
1234 PL_utf8cache = -1;
b310b053 1235 }
ec71e770 1236#endif
b310b053 1237
98994639 1238#ifdef USE_LOCALE_CTYPE
43c5f42d 1239 Safefree(curctype);
98994639
HS
1240#endif /* USE_LOCALE_CTYPE */
1241#ifdef USE_LOCALE_COLLATE
43c5f42d 1242 Safefree(curcoll);
98994639
HS
1243#endif /* USE_LOCALE_COLLATE */
1244#ifdef USE_LOCALE_NUMERIC
43c5f42d 1245 Safefree(curnum);
98994639 1246#endif /* USE_LOCALE_NUMERIC */
8ef6e574 1247
175c4cf9
KW
1248#ifdef __GLIBC__
1249 Safefree(language);
1250#endif
1251
1252 Safefree(lc_all);
1253 Safefree(lang);
1254
e3305790
KW
1255#else /* !USE_LOCALE */
1256 PERL_UNUSED_ARG(printwarn);
1257#endif /* USE_LOCALE */
1258
98994639
HS
1259 return ok;
1260}
1261
49c85077 1262
98994639
HS
1263#ifdef USE_LOCALE_COLLATE
1264
1265/*
1266 * mem_collxfrm() is a bit like strxfrm() but with two important
1267 * differences. First, it handles embedded NULs. Second, it allocates
1268 * a bit more memory than needed for the transformed data itself.
1269 * The real transformed data begins at offset sizeof(collationix).
d35fca5f
KW
1270 * *xlen is set to the length of that, and doesn't include the collation index
1271 * size.
98994639
HS
1272 * Please see sv_collxfrm() to see how this is used.
1273 */
166f8a29 1274
98994639
HS
1275char *
1276Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
1277{
1278 char *xbuf;
1279 STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
1280
7918f24d
NC
1281 PERL_ARGS_ASSERT_MEM_COLLXFRM;
1282
98994639
HS
1283 /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
1284 /* the +1 is for the terminating NUL. */
1285
1286 xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
a02a5408 1287 Newx(xbuf, xAlloc, char);
bb0f664e 1288 if (UNLIKELY(! xbuf))
98994639
HS
1289 goto bad;
1290
d35fca5f 1291 /* Store the collation id */
98994639
HS
1292 *(U32*)xbuf = PL_collation_ix;
1293 xout = sizeof(PL_collation_ix);
d35fca5f
KW
1294
1295 /* Then the transformation of the input. We loop until successful, or we
1296 * give up */
98994639 1297 for (xin = 0; xin < len; ) {
224e8ef5 1298 Size_t xused;
98994639
HS
1299
1300 for (;;) {
1301 xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
d35fca5f
KW
1302
1303 /* If the transformed string occupies less space than we told
1304 * strxfrm() was available, it means it successfully transformed
1305 * the whole string. */
eb160463 1306 if ((STRLEN)xused < xAlloc - xout)
98994639 1307 break;
bb0f664e
KW
1308
1309 if (UNLIKELY(xused >= PERL_INT_MAX))
1310 goto bad;
d35fca5f
KW
1311
1312 /* Otherwise it should be that the transformation stopped in the
1313 * middle because it ran out of space. Malloc more, and try again.
1314 * */
98994639
HS
1315 xAlloc = (2 * xAlloc) + 1;
1316 Renew(xbuf, xAlloc, char);
bb0f664e 1317 if (UNLIKELY(! xbuf))
98994639
HS
1318 goto bad;
1319 }
1320
1321 xin += strlen(s + xin) + 1;
1322 xout += xused;
1323
1324 /* Embedded NULs are understood but silently skipped
1325 * because they make no sense in locale collation. */
1326 }
1327
1328 xbuf[xout] = '\0';
1329 *xlen = xout - sizeof(PL_collation_ix);
1330 return xbuf;
1331
1332 bad:
1333 Safefree(xbuf);
1334 *xlen = 0;
1335 return NULL;
1336}
1337
1338#endif /* USE_LOCALE_COLLATE */
1339
8ef6e574
KW
1340#ifdef USE_LOCALE
1341
c1284011
KW
1342bool
1343Perl__is_cur_LC_category_utf8(pTHX_ int category)
7d74bb61
KW
1344{
1345 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
1346 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
119ee68b 1347 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
609548d2
KW
1348 * could give the wrong result. The result will very likely be correct for
1349 * languages that have commonly used non-ASCII characters, but for notably
1350 * English, it comes down to if the locale's name ends in something like
1351 * "UTF-8". It errs on the side of not being a UTF-8 locale. */
7d74bb61
KW
1352
1353 char *save_input_locale = NULL;
7d74bb61
KW
1354 STRLEN final_pos;
1355
8ef6e574 1356#ifdef LC_ALL
7d74bb61 1357 assert(category != LC_ALL);
8ef6e574 1358#endif
7d74bb61
KW
1359
1360 /* First dispose of the trivial cases */
b07fffd1 1361 save_input_locale = setlocale(category, NULL);
7d74bb61 1362 if (! save_input_locale) {
69014004
KW
1363 DEBUG_L(PerlIO_printf(Perl_debug_log,
1364 "Could not find current locale for category %d\n",
1365 category));
7d74bb61
KW
1366 return FALSE; /* XXX maybe should croak */
1367 }
b07fffd1 1368 save_input_locale = stdize_locale(savepv(save_input_locale));
a39edc4c 1369 if (isNAME_C_OR_POSIX(save_input_locale)) {
69014004
KW
1370 DEBUG_L(PerlIO_printf(Perl_debug_log,
1371 "Current locale for category %d is %s\n",
1372 category, save_input_locale));
b07fffd1 1373 Safefree(save_input_locale);
7d74bb61
KW
1374 return FALSE;
1375 }
1376
1d958db2
KW
1377#if defined(USE_LOCALE_CTYPE) \
1378 && (defined(MB_CUR_MAX) || (defined(HAS_NL_LANGINFO) && defined(CODESET)))
7d74bb61 1379
1d958db2 1380 { /* Next try nl_langinfo or MB_CUR_MAX if available */
7d74bb61
KW
1381
1382 char *save_ctype_locale = NULL;
119ee68b 1383 bool is_utf8;
7d74bb61 1384
119ee68b 1385 if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
7d74bb61
KW
1386
1387 /* Get the current LC_CTYPE locale */
4f72bb37 1388 save_ctype_locale = setlocale(LC_CTYPE, NULL);
7d74bb61 1389 if (! save_ctype_locale) {
69014004
KW
1390 DEBUG_L(PerlIO_printf(Perl_debug_log,
1391 "Could not find current locale for LC_CTYPE\n"));
7d74bb61
KW
1392 goto cant_use_nllanginfo;
1393 }
4f72bb37 1394 save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
7d74bb61
KW
1395
1396 /* If LC_CTYPE and the desired category use the same locale, this
1397 * means that finding the value for LC_CTYPE is the same as finding
1398 * the value for the desired category. Otherwise, switch LC_CTYPE
1399 * to the desired category's locale */
1400 if (strEQ(save_ctype_locale, save_input_locale)) {
1401 Safefree(save_ctype_locale);
1402 save_ctype_locale = NULL;
1403 }
1404 else if (! setlocale(LC_CTYPE, save_input_locale)) {
69014004
KW
1405 DEBUG_L(PerlIO_printf(Perl_debug_log,
1406 "Could not change LC_CTYPE locale to %s\n",
1407 save_input_locale));
7d74bb61
KW
1408 Safefree(save_ctype_locale);
1409 goto cant_use_nllanginfo;
1410 }
1411 }
1412
69014004
KW
1413 DEBUG_L(PerlIO_printf(Perl_debug_log, "Current LC_CTYPE locale=%s\n",
1414 save_input_locale));
1415
7d74bb61 1416 /* Here the current LC_CTYPE is set to the locale of the category whose
1d958db2
KW
1417 * information is desired. This means that nl_langinfo() and MB_CUR_MAX
1418 * should give the correct results */
119ee68b 1419
1d958db2
KW
1420# if defined(HAS_NL_LANGINFO) && defined(CODESET)
1421 {
4f72bb37 1422 char *codeset = nl_langinfo(CODESET);
1d958db2 1423 if (codeset && strNE(codeset, "")) {
4f72bb37 1424 codeset = savepv(codeset);
119ee68b 1425
1d958db2
KW
1426 /* If we switched LC_CTYPE, switch back */
1427 if (save_ctype_locale) {
1428 setlocale(LC_CTYPE, save_ctype_locale);
1429 Safefree(save_ctype_locale);
1430 }
1431
1432 is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
1433 || foldEQ(codeset, STR_WITH_LEN("UTF8"));
1434
69014004
KW
1435 DEBUG_L(PerlIO_printf(Perl_debug_log,
1436 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
1437 codeset, is_utf8));
1d958db2
KW
1438 Safefree(codeset);
1439 Safefree(save_input_locale);
1440 return is_utf8;
1441 }
119ee68b
KW
1442 }
1443
1d958db2
KW
1444# endif
1445# ifdef MB_CUR_MAX
1446
1447 /* Here, either we don't have nl_langinfo, or it didn't return a
1448 * codeset. Try MB_CUR_MAX */
1449
119ee68b
KW
1450 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
1451 * Unicode code point. Since UTF-8 is the only non-single byte
1452 * encoding we handle, we just say any such encoding is UTF-8, and if
1453 * turns out to be wrong, other things will fail */
1454 is_utf8 = MB_CUR_MAX >= 4;
1455
69014004
KW
1456 DEBUG_L(PerlIO_printf(Perl_debug_log,
1457 "\tMB_CUR_MAX=%d; ?UTF8 locale=%d\n",
1458 (int) MB_CUR_MAX, is_utf8));
1459
119ee68b
KW
1460 Safefree(save_input_locale);
1461
1462# ifdef HAS_MBTOWC
1463
1464 /* ... But, most system that have MB_CUR_MAX will also have mbtowc(),
1465 * since they are both in the C99 standard. We can feed a known byte
1466 * string to the latter function, and check that it gives the expected
1467 * result */
1468 if (is_utf8) {
1469 wchar_t wc;
856b881c 1470 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
69014004 1471 errno = 0;
f019f68f 1472 if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
119ee68b
KW
1473 != strlen(HYPHEN_UTF8)
1474 || wc != (wchar_t) 0x2010)
1475 {
1476 is_utf8 = FALSE;
abdcbdb8 1477 DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
69014004
KW
1478 DEBUG_L(PerlIO_printf(Perl_debug_log,
1479 "\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
1480 mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
119ee68b
KW
1481 }
1482 }
119ee68b
KW
1483# endif
1484
1d958db2
KW
1485 /* If we switched LC_CTYPE, switch back */
1486 if (save_ctype_locale) {
1487 setlocale(LC_CTYPE, save_ctype_locale);
1488 Safefree(save_ctype_locale);
119ee68b 1489 }
7d74bb61 1490
1d958db2 1491 return is_utf8;
119ee68b 1492# endif
7d74bb61 1493 }
119ee68b 1494
7d74bb61
KW
1495 cant_use_nllanginfo:
1496
0080c90a
KW
1497#else /* nl_langinfo should work if available, so don't bother compiling this
1498 fallback code. The final fallback of looking at the name is
1499 compiled, and will be executed if nl_langinfo fails */
7d74bb61 1500
97f4de96
KW
1501 /* nl_langinfo not available or failed somehow. Next try looking at the
1502 * currency symbol to see if it disambiguates things. Often that will be
1503 * in the native script, and if the symbol isn't in UTF-8, we know that the
1504 * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
609548d2
KW
1505 * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
1506 * */
fa9b773e
KW
1507
1508#ifdef HAS_LOCALECONV
fa9b773e 1509# ifdef USE_LOCALE_MONETARY
fa9b773e
KW
1510 {
1511 char *save_monetary_locale = NULL;
fa9b773e 1512 bool only_ascii = FALSE;
13542a67
KW
1513 bool is_utf8 = FALSE;
1514 struct lconv* lc;
fa9b773e 1515
97f4de96
KW
1516 /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
1517 * the desired category, if it isn't that locale already */
1518
fa9b773e
KW
1519 if (category != LC_MONETARY) {
1520
4f72bb37 1521 save_monetary_locale = setlocale(LC_MONETARY, NULL);
fa9b773e 1522 if (! save_monetary_locale) {
69014004
KW
1523 DEBUG_L(PerlIO_printf(Perl_debug_log,
1524 "Could not find current locale for LC_MONETARY\n"));
fa9b773e
KW
1525 goto cant_use_monetary;
1526 }
4f72bb37 1527 save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
fa9b773e 1528
13542a67
KW
1529 if (strEQ(save_monetary_locale, save_input_locale)) {
1530 Safefree(save_monetary_locale);
1531 save_monetary_locale = NULL;
1532 }
1533 else if (! setlocale(LC_MONETARY, save_input_locale)) {
59c234b4
KW
1534 DEBUG_L(PerlIO_printf(Perl_debug_log,
1535 "Could not change LC_MONETARY locale to %s\n",
1536 save_input_locale));
1537 Safefree(save_monetary_locale);
1538 goto cant_use_monetary;
fa9b773e
KW
1539 }
1540 }
1541
1542 /* Here the current LC_MONETARY is set to the locale of the category
1543 * whose information is desired. */
1544
13542a67
KW
1545 lc = localeconv();
1546 if (! lc
1547 || ! lc->currency_symbol
9f10db87 1548 || is_invariant_string((U8 *) lc->currency_symbol, 0))
13542a67
KW
1549 {
1550 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));
1551 only_ascii = TRUE;
1552 }
1553 else {
1554 is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
fa9b773e
KW
1555 }
1556
1557 /* If we changed it, restore LC_MONETARY to its original locale */
1558 if (save_monetary_locale) {
1559 setlocale(LC_MONETARY, save_monetary_locale);
1560 Safefree(save_monetary_locale);
1561 }
1562
13542a67 1563 if (! only_ascii) {
fa9b773e 1564
59c234b4
KW
1565 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
1566 * otherwise assume the locale is UTF-8 if and only if the symbol
1567 * is non-ascii UTF-8. */
1568 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
1569 save_input_locale, is_utf8));
1570 Safefree(save_input_locale);
1571 return is_utf8;
13542a67 1572 }
fa9b773e
KW
1573 }
1574 cant_use_monetary:
1575
1576# endif /* USE_LOCALE_MONETARY */
1577#endif /* HAS_LOCALECONV */
1578
15f7e74e
KW
1579#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
1580
1581/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
1582 * the names of the months and weekdays, timezone, and am/pm indicator */
1583 {
1584 char *save_time_locale = NULL;
1585 int hour = 10;
1586 bool is_dst = FALSE;
1587 int dom = 1;
1588 int month = 0;
1589 int i;
1590 char * formatted_time;
1591
1592
1593 /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
1594 * desired category, if it isn't that locale already */
1595
1596 if (category != LC_TIME) {
1597
1598 save_time_locale = setlocale(LC_TIME, NULL);
1599 if (! save_time_locale) {
1600 DEBUG_L(PerlIO_printf(Perl_debug_log,
1601 "Could not find current locale for LC_TIME\n"));
1602 goto cant_use_time;
1603 }
1604 save_time_locale = stdize_locale(savepv(save_time_locale));
1605
1606 if (strEQ(save_time_locale, save_input_locale)) {
1607 Safefree(save_time_locale);
1608 save_time_locale = NULL;
1609 }
1610 else if (! setlocale(LC_TIME, save_input_locale)) {
1611 DEBUG_L(PerlIO_printf(Perl_debug_log,
1612 "Could not change LC_TIME locale to %s\n",
1613 save_input_locale));
1614 Safefree(save_time_locale);
1615 goto cant_use_time;
1616 }
1617 }
1618
1619 /* Here the current LC_TIME is set to the locale of the category
1620 * whose information is desired. Look at all the days of the week and
9f10db87 1621 * month names, and the timezone and am/pm indicator for UTF-8 variant
15f7e74e
KW
1622 * characters. The first such a one found will tell us if the locale
1623 * is UTF-8 or not */
1624
1625 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
1626 formatted_time = my_strftime("%A %B %Z %p",
1627 0, 0, hour, dom, month, 112, 0, 0, is_dst);
9f10db87 1628 if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
15f7e74e
KW
1629
1630 /* Here, we didn't find a non-ASCII. Try the next time through
1631 * with the complemented dst and am/pm, and try with the next
1632 * weekday. After we have gotten all weekdays, try the next
1633 * month */
1634 is_dst = ! is_dst;
1635 hour = (hour + 12) % 24;
1636 dom++;
1637 if (i > 6) {
1638 month++;
1639 }
1640 continue;
1641 }
1642
1643 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
1644 * false otherwise. But first, restore LC_TIME to its original
1645 * locale if we changed it */
1646 if (save_time_locale) {
1647 setlocale(LC_TIME, save_time_locale);
1648 Safefree(save_time_locale);
1649 }
1650
1651 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
1652 save_input_locale,
1653 is_utf8_string((U8 *) formatted_time, 0)));
1654 Safefree(save_input_locale);
1655 return is_utf8_string((U8 *) formatted_time, 0);
1656 }
1657
1658 /* Falling off the end of the loop indicates all the names were just
1659 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
1660 * to its original locale */
1661 if (save_time_locale) {
1662 setlocale(LC_TIME, save_time_locale);
1663 Safefree(save_time_locale);
1664 }
1665 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));
1666 }
1667 cant_use_time:
1668
1669#endif
1670
1671#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
855aeb93
JH
1672
1673/* This code is ifdefd out because it was found to not be necessary in testing
5857e934
KW
1674 * on our dromedary test machine, which has over 700 locales. There, this
1675 * added no value to looking at the currency symbol and the time strings. I
1676 * left it in so as to avoid rewriting it if real-world experience indicates
1677 * that dromedary is an outlier. Essentially, instead of returning abpve if we
855aeb93
JH
1678 * haven't found illegal utf8, we continue on and examine all the strerror()
1679 * messages on the platform for utf8ness. If all are ASCII, we still don't
1680 * know the answer; but otherwise we have a pretty good indication of the
5857e934
KW
1681 * utf8ness. The reason this doesn't help much is that the messages may not
1682 * have been translated into the locale. The currency symbol and time strings
1683 * are much more likely to have been translated. */
1684 {
855aeb93 1685 int e;
5857e934
KW
1686 bool is_utf8 = FALSE;
1687 bool non_ascii = FALSE;
855aeb93 1688 char *save_messages_locale = NULL;
5857e934 1689 const char * errmsg = NULL;
855aeb93 1690
5857e934
KW
1691 /* Like above, we set LC_MESSAGES to the locale of the desired
1692 * category, if it isn't that locale already */
855aeb93
JH
1693
1694 if (category != LC_MESSAGES) {
1695
5857e934 1696 save_messages_locale = setlocale(LC_MESSAGES, NULL);
855aeb93 1697 if (! save_messages_locale) {
5857e934
KW
1698 DEBUG_L(PerlIO_printf(Perl_debug_log,
1699 "Could not find current locale for LC_MESSAGES\n"));
855aeb93
JH
1700 goto cant_use_messages;
1701 }
5857e934 1702 save_messages_locale = stdize_locale(savepv(save_messages_locale));
855aeb93
JH
1703
1704 if (strEQ(save_messages_locale, save_input_locale)) {
5857e934
KW
1705 Safefree(save_messages_locale);
1706 save_messages_locale = NULL;
855aeb93
JH
1707 }
1708 else if (! setlocale(LC_MESSAGES, save_input_locale)) {
5857e934
KW
1709 DEBUG_L(PerlIO_printf(Perl_debug_log,
1710 "Could not change LC_MESSAGES locale to %s\n",
1711 save_input_locale));
855aeb93
JH
1712 Safefree(save_messages_locale);
1713 goto cant_use_messages;
1714 }
1715 }
1716
1717 /* Here the current LC_MESSAGES is set to the locale of the category
5857e934
KW
1718 * whose information is desired. Look through all the messages. We
1719 * can't use Strerror() here because it may expand to code that
1720 * segfaults in miniperl */
855aeb93 1721
5857e934
KW
1722 for (e = 0; e <= sys_nerr; e++) {
1723 errno = 0;
1724 errmsg = sys_errlist[e];
1725 if (errno || !errmsg) {
855aeb93
JH
1726 break;
1727 }
5857e934 1728 errmsg = savepv(errmsg);
9f10db87 1729 if (! is_invariant_string((U8 *) errmsg, 0)) {
5857e934
KW
1730 non_ascii = TRUE;
1731 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
1732 break;
855aeb93
JH
1733 }
1734 }
5857e934 1735 Safefree(errmsg);
855aeb93
JH
1736
1737 /* And, if we changed it, restore LC_MESSAGES to its original locale */
1738 if (save_messages_locale) {
1739 setlocale(LC_MESSAGES, save_messages_locale);
1740 Safefree(save_messages_locale);
1741 }
1742
5857e934
KW
1743 if (non_ascii) {
1744
1745 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
1746 * any non-ascii means it is one; otherwise we assume it isn't */
1747 DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
1748 save_input_locale,
1749 is_utf8));
1750 Safefree(save_input_locale);
1751 return is_utf8;
1752 }
855aeb93 1753
5857e934 1754 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
1755 }
1756 cant_use_messages:
1757
1758#endif
fa9b773e 1759
0080c90a
KW
1760#endif /* the code that is compiled when no nl_langinfo */
1761
92c0a900
KW
1762#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
1763 UTF-8 locale */
97f4de96
KW
1764 /* As a last resort, look at the locale name to see if it matches
1765 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
1766 * return of setlocale(), is actually defined to be opaque, so we can't
1767 * really rely on the absence of various substrings in the name to indicate
1768 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
1769 * be a UTF-8 locale. Similarly for the other common names */
1770
1771 final_pos = strlen(save_input_locale) - 1;
1772 if (final_pos >= 3) {
1773 char *name = save_input_locale;
1774
1775 /* Find next 'U' or 'u' and look from there */
1776 while ((name += strcspn(name, "Uu") + 1)
1777 <= save_input_locale + final_pos - 2)
1778 {
305b8651
KW
1779 if (!isALPHA_FOLD_NE(*name, 't')
1780 || isALPHA_FOLD_NE(*(name + 1), 'f'))
97f4de96
KW
1781 {
1782 continue;
1783 }
1784 name += 2;
1785 if (*(name) == '-') {
1786 if ((name > save_input_locale + final_pos - 1)) {
1787 break;
1788 }
1789 name++;
1790 }
1791 if (*(name) == '8') {
97f4de96
KW
1792 DEBUG_L(PerlIO_printf(Perl_debug_log,
1793 "Locale %s ends with UTF-8 in name\n",
1794 save_input_locale));
00c54b9c 1795 Safefree(save_input_locale);
97f4de96
KW
1796 return TRUE;
1797 }
1798 }
1799 DEBUG_L(PerlIO_printf(Perl_debug_log,
1800 "Locale %s doesn't end with UTF-8 in name\n",
1801 save_input_locale));
1802 }
92c0a900 1803#endif
97f4de96
KW
1804
1805#ifdef WIN32
1806 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
1807 if (final_pos >= 4
1808 && *(save_input_locale + final_pos - 0) == '1'
1809 && *(save_input_locale + final_pos - 1) == '0'
1810 && *(save_input_locale + final_pos - 2) == '0'
1811 && *(save_input_locale + final_pos - 3) == '5'
1812 && *(save_input_locale + final_pos - 4) == '6')
1813 {
1814 DEBUG_L(PerlIO_printf(Perl_debug_log,
1815 "Locale %s ends with 10056 in name, is UTF-8 locale\n",
1816 save_input_locale));
1817 Safefree(save_input_locale);
1818 return TRUE;
1819 }
1820#endif
1821
1822 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
1823 * since we are about to return FALSE anyway, there is no point in doing
1824 * this extra work */
1825#if 0
1826 if (instr(save_input_locale, "8859")) {
1827 DEBUG_L(PerlIO_printf(Perl_debug_log,
1828 "Locale %s has 8859 in name, not UTF-8 locale\n",
1829 save_input_locale));
1830 Safefree(save_input_locale);
1831 return FALSE;
1832 }
1833#endif
1834
69014004
KW
1835 DEBUG_L(PerlIO_printf(Perl_debug_log,
1836 "Assuming locale %s is not a UTF-8 locale\n",
1837 save_input_locale));
fa9b773e 1838 Safefree(save_input_locale);
7d74bb61
KW
1839 return FALSE;
1840}
1841
8ef6e574 1842#endif
7d74bb61 1843
d6ded950
KW
1844
1845bool
1846Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
1847{
1a4f13e1 1848 dVAR;
d6ded950
KW
1849 /* Internal function which returns if we are in the scope of a pragma that
1850 * enables the locale category 'category'. 'compiling' should indicate if
1851 * this is during the compilation phase (TRUE) or not (FALSE). */
1852
1853 const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
1854
1855 SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
1856 if (! categories || categories == &PL_sv_placeholder) {
1857 return FALSE;
1858 }
1859
1860 /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
1861 * a valid unsigned */
1862 assert(category >= -1);
1863 return cBOOL(SvUV(categories) & (1U << (category + 1)));
1864}
1865
2c6ee1a7
KW
1866char *
1867Perl_my_strerror(pTHX_ const int errnum) {
a0b53297 1868 dVAR;
2c6ee1a7
KW
1869
1870 /* Uses C locale for the error text unless within scope of 'use locale' for
1871 * LC_MESSAGES */
1872
1873#ifdef USE_LOCALE_MESSAGES
8ee4b769 1874 if (! IN_LC(LC_MESSAGES)) {
a0b53297
KW
1875 char * save_locale;
1876
1877 /* We have a critical section to prevent another thread from changing
1878 * the locale out from under us (or zapping the buffer returned from
1879 * setlocale() ) */
1880 LOCALE_LOCK;
1881
1882 save_locale = setlocale(LC_MESSAGES, NULL);
a39edc4c 1883 if (! isNAME_C_OR_POSIX(save_locale)) {
2c6ee1a7
KW
1884 char *errstr;
1885
1886 /* The next setlocale likely will zap this, so create a copy */
1887 save_locale = savepv(save_locale);
1888
1889 setlocale(LC_MESSAGES, "C");
1890
1891 /* This points to the static space in Strerror, with all its
1892 * limitations */
1893 errstr = Strerror(errnum);
1894
1895 setlocale(LC_MESSAGES, save_locale);
1896 Safefree(save_locale);
a0b53297
KW
1897
1898 LOCALE_UNLOCK;
1899
2c6ee1a7
KW
1900 return errstr;
1901 }
a0b53297
KW
1902
1903 LOCALE_UNLOCK;
2c6ee1a7
KW
1904 }
1905#endif
1906
1907 return Strerror(errnum);
1908}
1909
66610fdd 1910/*
747c467a
KW
1911
1912=head1 Locale-related functions and macros
1913
1914=for apidoc sync_locale
1915
1916Changing the program's locale should be avoided by XS code. Nevertheless,
1917certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
1918happens, Perl needs to be told that the locale has changed. Use this function
1919to do so, before returning to Perl.
1920
1921=cut
1922*/
1923
1924void
1925Perl_sync_locale(pTHX)
1926{
1927
1928#ifdef USE_LOCALE_CTYPE
1929 new_ctype(setlocale(LC_CTYPE, NULL));
1930#endif /* USE_LOCALE_CTYPE */
1931
1932#ifdef USE_LOCALE_COLLATE
1933 new_collate(setlocale(LC_COLLATE, NULL));
1934#endif
1935
1936#ifdef USE_LOCALE_NUMERIC
1937 set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
1938 new_numeric(setlocale(LC_NUMERIC, NULL));
1939#endif /* USE_LOCALE_NUMERIC */
1940
1941}
1942
5d1187d1
KW
1943#if defined(DEBUGGING) && defined(USE_LOCALE)
1944
1945char *
1946Perl__setlocale_debug_string(const int category, /* category number,
1947 like LC_ALL */
1948 const char* const locale, /* locale name */
1949
1950 /* return value from setlocale() when attempting to
1951 * set 'category' to 'locale' */
1952 const char* const retval)
1953{
1954 /* Returns a pointer to a NUL-terminated string in static storage with
1955 * added text about the info passed in. This is not thread safe and will
1956 * be overwritten by the next call, so this should be used just to
fa07b8e5 1957 * formulate a string to immediately print or savepv() on. */
5d1187d1 1958
398a990f
DM
1959 /* initialise to a non-null value to keep it out of BSS and so keep
1960 * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
1961 static char ret[128] = "x";
5d1187d1 1962
fa07b8e5 1963 my_strlcpy(ret, "setlocale(", sizeof(ret));
5d1187d1
KW
1964
1965 switch (category) {
1966 default:
fa07b8e5 1967 my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
5d1187d1
KW
1968 break;
1969# ifdef LC_ALL
1970 case LC_ALL:
fa07b8e5 1971 my_strlcat(ret, "LC_ALL", sizeof(ret));
5d1187d1
KW
1972 break;
1973# endif
1974# ifdef LC_CTYPE
1975 case LC_CTYPE:
fa07b8e5 1976 my_strlcat(ret, "LC_CTYPE", sizeof(ret));
5d1187d1
KW
1977 break;
1978# endif
1979# ifdef LC_NUMERIC
1980 case LC_NUMERIC:
fa07b8e5 1981 my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
5d1187d1
KW
1982 break;
1983# endif
1984# ifdef LC_COLLATE
1985 case LC_COLLATE:
fa07b8e5 1986 my_strlcat(ret, "LC_COLLATE", sizeof(ret));
5d1187d1
KW
1987 break;
1988# endif
1989# ifdef LC_TIME
1990 case LC_TIME:
fa07b8e5 1991 my_strlcat(ret, "LC_TIME", sizeof(ret));
5d1187d1
KW
1992 break;
1993# endif
1994# ifdef LC_MONETARY
1995 case LC_MONETARY:
fa07b8e5 1996 my_strlcat(ret, "LC_MONETARY", sizeof(ret));
5d1187d1
KW
1997 break;
1998# endif
1999# ifdef LC_MESSAGES
2000 case LC_MESSAGES:
fa07b8e5 2001 my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
5d1187d1
KW
2002 break;
2003# endif
2004 }
2005
fa07b8e5 2006 my_strlcat(ret, ", ", sizeof(ret));
5d1187d1
KW
2007
2008 if (locale) {
fa07b8e5
KW
2009 my_strlcat(ret, "\"", sizeof(ret));
2010 my_strlcat(ret, locale, sizeof(ret));
2011 my_strlcat(ret, "\"", sizeof(ret));
5d1187d1
KW
2012 }
2013 else {
fa07b8e5 2014 my_strlcat(ret, "NULL", sizeof(ret));
5d1187d1
KW
2015 }
2016
fa07b8e5 2017 my_strlcat(ret, ") returned ", sizeof(ret));
5d1187d1
KW
2018
2019 if (retval) {
fa07b8e5
KW
2020 my_strlcat(ret, "\"", sizeof(ret));
2021 my_strlcat(ret, retval, sizeof(ret));
2022 my_strlcat(ret, "\"", sizeof(ret));
5d1187d1
KW
2023 }
2024 else {
fa07b8e5 2025 my_strlcat(ret, "NULL", sizeof(ret));
5d1187d1
KW
2026 }
2027
2028 assert(strlen(ret) < sizeof(ret));
2029
2030 return ret;
2031}
2032
2033#endif
747c467a
KW
2034
2035
2036/*
14d04a33 2037 * ex: set ts=8 sts=4 sw=4 et:
37442d52 2038 */