This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
https://perl5.git.perl.org / perl5.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
locale.c: Change some DEBUG_L to DEBUG_Lv
[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 25 *
7d4bcc4a
KW		 26 * All C programs have an underlying locale. Perl code generally doesn't pay
27 * any attention to it except within the scope of a 'use locale'. For most
0d071d52
KW		 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
f7416781 39#include "perl_langinfo.h"
98994639
HS		 40#include "perl.h"
41
a4af207c
JH		 42#include "reentr.h"
43
0dec74cd
KW		 44#ifdef I_WCHAR
45# include <wchar.h>
46#endif
47
2fcc0ca9
KW		 48/* If the environment says to, we can output debugging information during
49 * initialization. This is done before option parsing, and before any thread
50 * creation, so can be a file-level static */
5a4b0634
KW		 51#if ! defined(DEBUGGING) || defined(PERL_GLOBAL_STRUCT)
52# define debug_initialization 0
53# define DEBUG_INITIALIZATION_set(v)
54#else
2fcc0ca9 55static bool debug_initialization = FALSE;
5a4b0634 56# define DEBUG_INITIALIZATION_set(v) (debug_initialization = v)
2fcc0ca9
KW		 57#endif
58
291a84fb
KW		 59/* strlen() of a literal string constant. We might want this more general,
60 * but using it in just this file for now. A problem with more generality is
61 * the compiler warnings about comparing unlike signs */
ff1b739b
KW		 62#define STRLENs(s) (sizeof("" s "") - 1)
63
63e5b0d7
KW		 64/* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
65 * return of setlocale(), then this is extremely likely to be the C or POSIX
66 * locale. However, the output of setlocale() is documented to be opaque, but
67 * the odds are extremely small that it would return these two strings for some
68 * other locale. Note that VMS in these two locales includes many non-ASCII
69 * characters as controls and punctuation (below are hex bytes):
70 * cntrl: 84-97 9B-9F
71 * punct: A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
72 * Oddly, none there are listed as alphas, though some represent alphabetics
73 * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
74#define isNAME_C_OR_POSIX(name) \
75 ( (name) != NULL \
76 && (( *(name) == 'C' && (*(name + 1)) == '\0') \
77 || strEQ((name), "POSIX")))
78
8ef6e574
KW		 79#ifdef USE_LOCALE
80
47280b20
KW		 81/* This code keeps a LRU cache of the UTF-8ness of the locales it has so-far
82 * looked up. This is in the form of a C string: */
83
84#define UTF8NESS_SEP "\v"
85#define UTF8NESS_PREFIX "\f"
86
87/* So, the string looks like:
98994639 88 *
47280b20 89 * \vC\a0\vPOSIX\a0\vam_ET\a0\vaf_ZA.utf8\a1\ven_US.UTF-8\a1\0
98994639 90 *
47280b20
KW		 91 * where the digit 0 after the \a indicates that the locale starting just
92 * after the preceding \v is not UTF-8, and the digit 1 mean it is. */
93
94STATIC_ASSERT_DECL(STRLENs(UTF8NESS_SEP) == 1);
95STATIC_ASSERT_DECL(STRLENs(UTF8NESS_PREFIX) == 1);
96
97#define C_and_POSIX_utf8ness UTF8NESS_SEP "C" UTF8NESS_PREFIX "0" \
98 UTF8NESS_SEP "POSIX" UTF8NESS_PREFIX "0"
99
100/* The cache is initialized to C_and_POSIX_utf8ness at start up. These are
101 * kept there always. The remining portion of the cache is LRU, with the
102 * oldest looked-up locale at the tail end */
103
98994639
HS		 104STATIC char *
105S_stdize_locale(pTHX_ char *locs)
106{
47280b20
KW		 107 /* Standardize the locale name from a string returned by 'setlocale',
108 * possibly modifying that string.
109 *
110 * The typical return value of setlocale() is either
111 * (1) "xx_YY" if the first argument of setlocale() is not LC_ALL
112 * (2) "xa_YY xb_YY ..." if the first argument of setlocale() is LC_ALL
113 * (the space-separated values represent the various sublocales,
114 * in some unspecified order). This is not handled by this function.
115 *
116 * In some platforms it has a form like "LC_SOMETHING=Lang_Country.866\n",
117 * which is harmful for further use of the string in setlocale(). This
118 * function removes the trailing new line and everything up through the '='
119 * */
120
7452cf6a 121 const char * const s = strchr(locs, '=');
98994639
HS		 122 bool okay = TRUE;
123
7918f24d
NC		 124 PERL_ARGS_ASSERT_STDIZE_LOCALE;
125
8772537c
AL		 126 if (s) {
127 const char * const t = strchr(s, '.');
98994639 128 okay = FALSE;
8772537c
AL		 129 if (t) {
130 const char * const u = strchr(t, '\n');
131 if (u && (u[1] == 0)) {
132 const STRLEN len = u - s;
133 Move(s + 1, locs, len, char);
134 locs[len] = 0;
135 okay = TRUE;
98994639
HS		 136 }
137 }
138 }
139
140 if (!okay)
141 Perl_croak(aTHX_ "Can't fix broken locale name \"%s\"", locs);
142
143 return locs;
144}
145
e5f10d49
KW		 146/* Two parallel arrays; first the locale categories Perl uses on this system;
147 * the second array is their names. These arrays are in mostly arbitrary
148 * order. */
149
150const int categories[] = {
151
152# ifdef USE_LOCALE_NUMERIC
153 LC_NUMERIC,
154# endif
155# ifdef USE_LOCALE_CTYPE
156 LC_CTYPE,
157# endif
158# ifdef USE_LOCALE_COLLATE
159 LC_COLLATE,
160# endif
161# ifdef USE_LOCALE_TIME
162 LC_TIME,
163# endif
164# ifdef USE_LOCALE_MESSAGES
165 LC_MESSAGES,
166# endif
167# ifdef USE_LOCALE_MONETARY
168 LC_MONETARY,
169# endif
9821811f
KW		 170# ifdef USE_LOCALE_ADDRESS
171 LC_ADDRESS,
172# endif
173# ifdef USE_LOCALE_IDENTIFICATION
174 LC_IDENTIFICATION,
175# endif
176# ifdef USE_LOCALE_MEASUREMENT
177 LC_MEASUREMENT,
178# endif
179# ifdef USE_LOCALE_PAPER
180 LC_PAPER,
181# endif
182# ifdef USE_LOCALE_TELEPHONE
183 LC_TELEPHONE,
184# endif
e5f10d49
KW		 185# ifdef LC_ALL
186 LC_ALL,
187# endif
188 -1 /* Placeholder because C doesn't allow a
189 trailing comma, and it would get complicated
190 with all the #ifdef's */
191};
192
193/* The top-most real element is LC_ALL */
194
195const char * category_names[] = {
196
197# ifdef USE_LOCALE_NUMERIC
198 "LC_NUMERIC",
199# endif
200# ifdef USE_LOCALE_CTYPE
201 "LC_CTYPE",
202# endif
203# ifdef USE_LOCALE_COLLATE
204 "LC_COLLATE",
205# endif
206# ifdef USE_LOCALE_TIME
207 "LC_TIME",
208# endif
209# ifdef USE_LOCALE_MESSAGES
210 "LC_MESSAGES",
211# endif
212# ifdef USE_LOCALE_MONETARY
213 "LC_MONETARY",
214# endif
9821811f
KW		 215# ifdef USE_LOCALE_ADDRESS
216 "LC_ADDRESS",
217# endif
218# ifdef USE_LOCALE_IDENTIFICATION
219 "LC_IDENTIFICATION",
220# endif
221# ifdef USE_LOCALE_MEASUREMENT
222 "LC_MEASUREMENT",
223# endif
224# ifdef USE_LOCALE_PAPER
225 "LC_PAPER",
226# endif
227# ifdef USE_LOCALE_TELEPHONE
228 "LC_TELEPHONE",
229# endif
e5f10d49
KW		 230# ifdef LC_ALL
231 "LC_ALL",
232# endif
233 NULL /* Placeholder */
234 };
235
236# ifdef LC_ALL
237
238 /* On systems with LC_ALL, it is kept in the highest index position. (-2
239 * to account for the final unused placeholder element.) */
240# define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 2)
241
242# else
243
244 /* On systems without LC_ALL, we pretend it is there, one beyond the real
245 * top element, hence in the unused placeholder element. */
246# define NOMINAL_LC_ALL_INDEX (C_ARRAY_LENGTH(categories) - 1)
247
248# endif
249
250/* Pretending there is an LC_ALL element just above allows us to avoid most
251 * special cases. Most loops through these arrays in the code below are
252 * written like 'for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++)'. They will work
253 * on either type of system. But the code must be written to not access the
948523db
KW		 254 * element at 'LC_ALL_INDEX' except on platforms that have it. This can be
255 * checked for at compile time by using the #define LC_ALL_INDEX which is only
256 * defined if we do have LC_ALL. */
e5f10d49 257
b09aaf40
KW		 258STATIC const char *
259S_category_name(const int category)
260{
261 unsigned int i;
262
263#ifdef LC_ALL
264
265 if (category == LC_ALL) {
266 return "LC_ALL";
267 }
268
269#endif
270
271 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
272 if (category == categories[i]) {
273 return category_names[i];
274 }
275 }
276
277 {
278 const char suffix[] = " (unknown)";
279 int temp = category;
280 Size_t length = sizeof(suffix) + 1;
281 char * unknown;
282 dTHX;
283
284 if (temp < 0) {
285 length++;
286 temp = - temp;
287 }
288
289 /* Calculate the number of digits */
290 while (temp >= 10) {
291 temp /= 10;
292 length++;
293 }
294
295 Newx(unknown, length, char);
296 my_snprintf(unknown, length, "%d%s", category, suffix);
297 SAVEFREEPV(unknown);
298 return unknown;
299 }
300}
301
948523db
KW		 302/* Now create LC_foo_INDEX #defines for just those categories on this system */
303# ifdef USE_LOCALE_NUMERIC
304# define LC_NUMERIC_INDEX 0
305# define _DUMMY_NUMERIC LC_NUMERIC_INDEX
306# else
307# define _DUMMY_NUMERIC -1
308# endif
309# ifdef USE_LOCALE_CTYPE
310# define LC_CTYPE_INDEX _DUMMY_NUMERIC + 1
311# define _DUMMY_CTYPE LC_CTYPE_INDEX
312# else
313# define _DUMMY_CTYPE _DUMMY_NUMERIC
314# endif
315# ifdef USE_LOCALE_COLLATE
316# define LC_COLLATE_INDEX _DUMMY_CTYPE + 1
317# define _DUMMY_COLLATE LC_COLLATE_INDEX
318# else
319# define _DUMMY_COLLATE _DUMMY_COLLATE
320# endif
321# ifdef USE_LOCALE_TIME
322# define LC_TIME_INDEX _DUMMY_COLLATE + 1
323# define _DUMMY_TIME LC_TIME_INDEX
324# else
325# define _DUMMY_TIME _DUMMY_COLLATE
326# endif
327# ifdef USE_LOCALE_MESSAGES
328# define LC_MESSAGES_INDEX _DUMMY_TIME + 1
329# define _DUMMY_MESSAGES LC_MESSAGES_INDEX
330# else
331# define _DUMMY_MESSAGES _DUMMY_TIME
332# endif
333# ifdef USE_LOCALE_MONETARY
334# define LC_MONETARY_INDEX _DUMMY_MESSAGES + 1
335# define _DUMMY_MONETARY LC_MONETARY_INDEX
336# else
337# define _DUMMY_MONETARY _DUMMY_MESSAGES
338# endif
9821811f
KW		 339# ifdef USE_LOCALE_ADDRESS
340# define LC_ADDRESS_INDEX _DUMMY_MONETARY + 1
341# define _DUMMY_ADDRESS LC_ADDRESS_INDEX
342# else
343# define _DUMMY_ADDRESS _DUMMY_MONETARY
344# endif
345# ifdef USE_LOCALE_IDENTIFICATION
346# define LC_IDENTIFICATION_INDEX _DUMMY_ADDRESS + 1
347# define _DUMMY_IDENTIFICATION LC_IDENTIFICATION_INDEX
348# else
349# define _DUMMY_IDENTIFICATION _DUMMY_ADDRESS
350# endif
351# ifdef USE_LOCALE_MEASUREMENT
352# define LC_MEASUREMENT_INDEX _DUMMY_IDENTIFICATION + 1
353# define _DUMMY_MEASUREMENT LC_MEASUREMENT_INDEX
354# else
355# define _DUMMY_MEASUREMENT _DUMMY_IDENTIFICATION
356# endif
357# ifdef USE_LOCALE_PAPER
358# define LC_PAPER_INDEX _DUMMY_MEASUREMENT + 1
359# define _DUMMY_PAPER LC_PAPER_INDEX
360# else
361# define _DUMMY_PAPER _DUMMY_MEASUREMENT
362# endif
363# ifdef USE_LOCALE_TELEPHONE
364# define LC_TELEPHONE_INDEX _DUMMY_PAPER + 1
365# define _DUMMY_TELEPHONE LC_TELEPHONE_INDEX
366# else
367# define _DUMMY_TELEPHONE _DUMMY_PAPER
368# endif
948523db 369# ifdef LC_ALL
9821811f 370# define LC_ALL_INDEX _DUMMY_TELEPHONE + 1
948523db
KW		 371# endif
372#endif /* ifdef USE_LOCALE */
8ef6e574 373
d2b24094
KW		 374/* Windows requres a customized base-level setlocale() */
375# ifdef WIN32
376# define my_setlocale(cat, locale) win32_setlocale(cat, locale)
377# else
378# define my_setlocale(cat, locale) setlocale(cat, locale)
379# endif
380
837ce802
KW		 381/* Just placeholders for now. "_c" is intended to be called when the category
382 * is a constant known at compile time; "_r", not known until run time */
d2b24094
KW		 383# define do_setlocale_c(category, locale) my_setlocale(category, locale)
384# define do_setlocale_r(category, locale) my_setlocale(category, locale)
837ce802 385
a4f00dcc 386STATIC void
86799d2d 387S_set_numeric_radix(pTHX_ const bool use_locale)
98994639 388{
86799d2d
KW		 389 /* If 'use_locale' is FALSE, set to use a dot for the radix character. If
390 * TRUE, use the radix character derived from the current locale */
7d4bcc4a 391
86799d2d
KW		 392#if defined(USE_LOCALE_NUMERIC) && ( defined(HAS_LOCALECONV) \
393 || defined(HAS_NL_LANGINFO))
98994639 394
87f8e8e7
KW		 395 const char * radix = (use_locale)
396 ? my_nl_langinfo(PERL_RADIXCHAR, FALSE)
397 /* FALSE => already in dest locale */
398 : ".";
2213a3be 399
87f8e8e7 400 sv_setpv(PL_numeric_radix_sv, radix);
86799d2d 401
87f8e8e7
KW		 402 /* If this is valid UTF-8 that isn't totally ASCII, and we are in
403 * a UTF-8 locale, then mark the radix as being in UTF-8 */
404 if (is_utf8_non_invariant_string((U8 *) SvPVX(PL_numeric_radix_sv),
7a393424 405 SvCUR(PL_numeric_radix_sv))
87f8e8e7
KW		 406 && _is_cur_LC_category_utf8(LC_NUMERIC))
407 {
408 SvUTF8_on(PL_numeric_radix_sv);
409 }
86799d2d
KW		 410
411# ifdef DEBUGGING
7d4bcc4a 412
2fcc0ca9
KW		 413 if (DEBUG_L_TEST || debug_initialization) {
414 PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
3ca88433
KW		 415 SvPVX(PL_numeric_radix_sv),
416 cBOOL(SvUTF8(PL_numeric_radix_sv)));
2fcc0ca9 417 }
69014004 418
86799d2d
KW		 419# endif
420#endif /* USE_LOCALE_NUMERIC and can find the radix char */
7d4bcc4a 421
98994639
HS		 422}
423
a39edc4c 424
98994639 425void
8772537c 426Perl_new_numeric(pTHX_ const char *newnum)
98994639 427{
7d4bcc4a
KW		 428
429#ifndef USE_LOCALE_NUMERIC
430
431 PERL_UNUSED_ARG(newnum);
432
433#else
0d071d52 434
291a84fb 435 /* Called after each libc setlocale() call affecting LC_NUMERIC, to tell
0d071d52
KW		 436 * core Perl this and that 'newnum' is the name of the new locale.
437 * It installs this locale as the current underlying default.
438 *
439 * The default locale and the C locale can be toggled between by use of the
5792c642
KW		 440 * set_numeric_underlying() and set_numeric_standard() functions, which
441 * should probably not be called directly, but only via macros like
0d071d52
KW		 442 * SET_NUMERIC_STANDARD() in perl.h.
443 *
444 * The toggling is necessary mainly so that a non-dot radix decimal point
445 * character can be output, while allowing internal calculations to use a
446 * dot.
447 *
448 * This sets several interpreter-level variables:
bb304765 449 * PL_numeric_name The underlying locale's name: a copy of 'newnum'
892e6465 450 * PL_numeric_underlying A boolean indicating if the toggled state is such
7738054c
KW		 451 * that the current locale is the program's underlying
452 * locale
453 * PL_numeric_standard An int indicating if the toggled state is such
4c68b815
KW		 454 * that the current locale is the C locale or
455 * indistinguishable from the C locale. If non-zero, it
456 * is in C; if > 1, it means it may not be toggled away
7738054c 457 * from C.
4c68b815
KW		 458 * PL_numeric_underlying_is_standard A bool kept by this function
459 * indicating that the underlying locale and the standard
460 * C locale are indistinguishable for the purposes of
461 * LC_NUMERIC. This happens when both of the above two
462 * variables are true at the same time. (Toggling is a
463 * no-op under these circumstances.) This variable is
464 * used to avoid having to recalculate.
0d071d52
KW		 465 * Any code changing the locale (outside this file) should use
466 * POSIX::setlocale, which calls this function. Therefore this function
467 * should be called directly only from this file and from
468 * POSIX::setlocale() */
469
b03f34cf 470 char *save_newnum;
98994639
HS		 471
472 if (! newnum) {
43c5f42d
NC		 473 Safefree(PL_numeric_name);
474 PL_numeric_name = NULL;
98994639 475 PL_numeric_standard = TRUE;
892e6465 476 PL_numeric_underlying = TRUE;
4c68b815 477 PL_numeric_underlying_is_standard = TRUE;
98994639
HS		 478 return;
479 }
480
b03f34cf 481 save_newnum = stdize_locale(savepv(newnum));
892e6465 482 PL_numeric_underlying = TRUE;
4c68b815
KW		 483 PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
484
485 /* If its name isn't C nor POSIX, it could still be indistinguishable from
486 * them */
487 if (! PL_numeric_standard) {
488 PL_numeric_standard = cBOOL(strEQ(".", my_nl_langinfo(PERL_RADIXCHAR,
489 FALSE /* Don't toggle locale */ ))
490 && strEQ("", my_nl_langinfo(PERL_THOUSEP,
491 FALSE)));
492 }
abe1abcf 493
4c68b815 494 /* Save the new name if it isn't the same as the previous one, if any */
b03f34cf 495 if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
98994639 496 Safefree(PL_numeric_name);
b03f34cf 497 PL_numeric_name = save_newnum;
b03f34cf 498 }
abe1abcf
KW		 499 else {
500 Safefree(save_newnum);
501 }
4c28b29c 502
4c68b815
KW		 503 PL_numeric_underlying_is_standard = PL_numeric_standard;
504
7e5377f7 505# ifdef HAS_POSIX_2008_LOCALE
e1aa2579
KW		 506
507 PL_underlying_numeric_obj = newlocale(LC_NUMERIC_MASK,
508 PL_numeric_name,
509 PL_underlying_numeric_obj);
510
511#endif
512
4c68b815
KW		 513 if (DEBUG_L_TEST || debug_initialization) {
514 PerlIO_printf(Perl_debug_log, "Called new_numeric with %s, PL_numeric_name=%s\n", newnum, PL_numeric_name);
515 }
516
4c28b29c
KW		 517 /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
518 * have to worry about the radix being a non-dot. (Core operations that
519 * need the underlying locale change to it temporarily). */
520 set_numeric_standard();
521
98994639 522#endif /* USE_LOCALE_NUMERIC */
7d4bcc4a 523
98994639
HS		 524}
525
526void
527Perl_set_numeric_standard(pTHX)
528{
7d4bcc4a 529
98994639 530#ifdef USE_LOCALE_NUMERIC
7d4bcc4a 531
28c1bf33
KW		 532 /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
533 * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
534 * macro avoids calling this routine if toggling isn't necessary according
535 * to our records (which could be wrong if some XS code has changed the
536 * locale behind our back) */
0d071d52 537
837ce802 538 do_setlocale_c(LC_NUMERIC, "C");
a9b8c0d8 539 PL_numeric_standard = TRUE;
4c68b815 540 PL_numeric_underlying = PL_numeric_underlying_is_standard;
86799d2d 541 set_numeric_radix(0);
7d4bcc4a
KW		 542
543# ifdef DEBUGGING
544
2fcc0ca9
KW		 545 if (DEBUG_L_TEST || debug_initialization) {
546 PerlIO_printf(Perl_debug_log,
58e4a467 547 "LC_NUMERIC locale now is standard C\n");
2fcc0ca9 548 }
98994639 549
7d4bcc4a 550# endif
98994639 551#endif /* USE_LOCALE_NUMERIC */
7d4bcc4a 552
98994639
HS		 553}
554
555void
5792c642 556Perl_set_numeric_underlying(pTHX)
98994639 557{
7d4bcc4a 558
98994639 559#ifdef USE_LOCALE_NUMERIC
7d4bcc4a 560
28c1bf33 561 /* Toggle the LC_NUMERIC locale to the current underlying default. Most
7d4bcc4a
KW		 562 * code should use the macros like SET_NUMERIC_UNDERLYING() in perl.h
563 * instead of calling this directly. The macro avoids calling this routine
564 * if toggling isn't necessary according to our records (which could be
565 * wrong if some XS code has changed the locale behind our back) */
a9b8c0d8 566
837ce802 567 do_setlocale_c(LC_NUMERIC, PL_numeric_name);
4c68b815 568 PL_numeric_standard = PL_numeric_underlying_is_standard;
892e6465 569 PL_numeric_underlying = TRUE;
3c62fd46 570 set_numeric_radix(! PL_numeric_standard);
7d4bcc4a
KW		 571
572# ifdef DEBUGGING
573
2fcc0ca9
KW		 574 if (DEBUG_L_TEST || debug_initialization) {
575 PerlIO_printf(Perl_debug_log,
58e4a467 576 "LC_NUMERIC locale now is %s\n",
2fcc0ca9
KW		 577 PL_numeric_name);
578 }
98994639 579
7d4bcc4a 580# endif
98994639 581#endif /* USE_LOCALE_NUMERIC */
7d4bcc4a 582
98994639
HS		 583}
584
585/*
586 * Set up for a new ctype locale.
587 */
a4f00dcc
KW		 588STATIC void
589S_new_ctype(pTHX_ const char *newctype)
98994639 590{
7d4bcc4a
KW		 591
592#ifndef USE_LOCALE_CTYPE
593
594 PERL_ARGS_ASSERT_NEW_CTYPE;
595 PERL_UNUSED_ARG(newctype);
596 PERL_UNUSED_CONTEXT;
597
598#else
0d071d52 599
291a84fb 600 /* Called after each libc setlocale() call affecting LC_CTYPE, to tell
0d071d52
KW		 601 * core Perl this and that 'newctype' is the name of the new locale.
602 *
603 * This function sets up the folding arrays for all 256 bytes, assuming
604 * that tofold() is tolc() since fold case is not a concept in POSIX,
605 *
606 * Any code changing the locale (outside this file) should use
607 * POSIX::setlocale, which calls this function. Therefore this function
608 * should be called directly only from this file and from
609 * POSIX::setlocale() */
610
27da23d5 611 dVAR;
68067e4e 612 UV i;
98994639 613
8b7358b9
KW		 614 /* Don't check for problems if we are suppressing the warnings */
615 bool check_for_problems = ckWARN_d(WARN_LOCALE) || UNLIKELY(DEBUG_L_TEST);
616
7918f24d
NC		 617 PERL_ARGS_ASSERT_NEW_CTYPE;
618
215c5139
KW		 619 /* We will replace any bad locale warning with 1) nothing if the new one is
620 * ok; or 2) a new warning for the bad new locale */
621 if (PL_warn_locale) {
622 SvREFCNT_dec_NN(PL_warn_locale);
623 PL_warn_locale = NULL;
624 }
625
c1284011 626 PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
31f05a37
KW		 627
628 /* A UTF-8 locale gets standard rules. But note that code still has to
629 * handle this specially because of the three problematic code points */
630 if (PL_in_utf8_CTYPE_locale) {
631 Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
632 }
8b7358b9
KW		 633
634 /* We don't populate the other lists if a UTF-8 locale, but do check that
635 * everything works as expected, unless checking turned off */
636 if (check_for_problems || ! PL_in_utf8_CTYPE_locale) {
8c6180a9
KW		 637 /* Assume enough space for every character being bad. 4 spaces each
638 * for the 94 printable characters that are output like "'x' "; and 5
639 * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
640 * NUL */
8b7358b9 641 char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ] = { '\0' };
8c6180a9
KW		 642 bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
643 to start */
644 unsigned int bad_count = 0; /* Count of bad characters */
645
baa60164 646 for (i = 0; i < 256; i++) {
8b7358b9 647 if (! PL_in_utf8_CTYPE_locale) {
bd6d0898
KW		 648 if (isupper(i))
649 PL_fold_locale[i] = (U8) tolower(i);
650 else if (islower(i))
651 PL_fold_locale[i] = (U8) toupper(i);
652 else
653 PL_fold_locale[i] = (U8) i;
8b7358b9 654 }
8c6180a9
KW		 655
656 /* If checking for locale problems, see if the native ASCII-range
657 * printables plus \n and \t are in their expected categories in
658 * the new locale. If not, this could mean big trouble, upending
659 * Perl's and most programs' assumptions, like having a
660 * metacharacter with special meaning become a \w. Fortunately,
661 * it's very rare to find locales that aren't supersets of ASCII
662 * nowadays. It isn't a problem for most controls to be changed
663 * into something else; we check only \n and \t, though perhaps \r
664 * could be an issue as well. */
7d4bcc4a 665 if ( check_for_problems
8c6180a9
KW		 666 && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
667 {
8b7358b9
KW		 668 bool is_bad = FALSE;
669 char name[3] = { '\0' };
670
671 /* Convert the name into a string */
672 if (isPRINT_A(i)) {
673 name[0] = i;
674 name[1] = '\0';
675 }
676 else if (i == '\n') {
677 my_strlcpy(name, "\n", sizeof(name));
678 }
679 else {
680 my_strlcpy(name, "\t", sizeof(name));
681 }
682
683 /* Check each possibe class */
684 if (UNLIKELY(cBOOL(isalnum(i)) != cBOOL(isALPHANUMERIC_A(i)))) {
685 is_bad = TRUE;
686 DEBUG_L(PerlIO_printf(Perl_debug_log,
687 "isalnum('%s') unexpectedly is %d\n",
688 name, cBOOL(isalnum(i))));
689 }
690 if (UNLIKELY(cBOOL(isalpha(i)) != cBOOL(isALPHA_A(i)))) {
691 is_bad = TRUE;
692 DEBUG_L(PerlIO_printf(Perl_debug_log,
693 "isalpha('%s') unexpectedly is %d\n",
694 name, cBOOL(isalpha(i))));
695 }
696 if (UNLIKELY(cBOOL(isdigit(i)) != cBOOL(isDIGIT_A(i)))) {
697 is_bad = TRUE;
698 DEBUG_L(PerlIO_printf(Perl_debug_log,
699 "isdigit('%s') unexpectedly is %d\n",
700 name, cBOOL(isdigit(i))));
701 }
702 if (UNLIKELY(cBOOL(isgraph(i)) != cBOOL(isGRAPH_A(i)))) {
703 is_bad = TRUE;
704 DEBUG_L(PerlIO_printf(Perl_debug_log,
705 "isgraph('%s') unexpectedly is %d\n",
706 name, cBOOL(isgraph(i))));
707 }
708 if (UNLIKELY(cBOOL(islower(i)) != cBOOL(isLOWER_A(i)))) {
709 is_bad = TRUE;
710 DEBUG_L(PerlIO_printf(Perl_debug_log,
711 "islower('%s') unexpectedly is %d\n",
712 name, cBOOL(islower(i))));
713 }
714 if (UNLIKELY(cBOOL(isprint(i)) != cBOOL(isPRINT_A(i)))) {
715 is_bad = TRUE;
716 DEBUG_L(PerlIO_printf(Perl_debug_log,
717 "isprint('%s') unexpectedly is %d\n",
718 name, cBOOL(isprint(i))));
719 }
720 if (UNLIKELY(cBOOL(ispunct(i)) != cBOOL(isPUNCT_A(i)))) {
721 is_bad = TRUE;
722 DEBUG_L(PerlIO_printf(Perl_debug_log,
723 "ispunct('%s') unexpectedly is %d\n",
724 name, cBOOL(ispunct(i))));
725 }
726 if (UNLIKELY(cBOOL(isspace(i)) != cBOOL(isSPACE_A(i)))) {
727 is_bad = TRUE;
728 DEBUG_L(PerlIO_printf(Perl_debug_log,
729 "isspace('%s') unexpectedly is %d\n",
730 name, cBOOL(isspace(i))));
731 }
732 if (UNLIKELY(cBOOL(isupper(i)) != cBOOL(isUPPER_A(i)))) {
733 is_bad = TRUE;
734 DEBUG_L(PerlIO_printf(Perl_debug_log,
735 "isupper('%s') unexpectedly is %d\n",
736 name, cBOOL(isupper(i))));
737 }
738 if (UNLIKELY(cBOOL(isxdigit(i))!= cBOOL(isXDIGIT_A(i)))) {
739 is_bad = TRUE;
740 DEBUG_L(PerlIO_printf(Perl_debug_log,
741 "isxdigit('%s') unexpectedly is %d\n",
742 name, cBOOL(isxdigit(i))));
743 }
744 if (UNLIKELY(tolower(i) != (int) toLOWER_A(i))) {
745 is_bad = TRUE;
746 DEBUG_L(PerlIO_printf(Perl_debug_log,
747 "tolower('%s')=0x%x instead of the expected 0x%x\n",
748 name, tolower(i), (int) toLOWER_A(i)));
749 }
750 if (UNLIKELY(toupper(i) != (int) toUPPER_A(i))) {
751 is_bad = TRUE;
752 DEBUG_L(PerlIO_printf(Perl_debug_log,
753 "toupper('%s')=0x%x instead of the expected 0x%x\n",
754 name, toupper(i), (int) toUPPER_A(i)));
755 }
756 if (UNLIKELY((i == '\n' && ! isCNTRL_LC(i)))) {
757 is_bad = TRUE;
758 DEBUG_L(PerlIO_printf(Perl_debug_log,
759 "'\\n' (=%02X) is not a control\n", (int) i));
760 }
761
762 /* Add to the list; Separate multiple entries with a blank */
763 if (is_bad) {
764 if (bad_count) {
765 my_strlcat(bad_chars_list, " ", sizeof(bad_chars_list));
8c6180a9 766 }
8b7358b9
KW		 767 my_strlcat(bad_chars_list, name, sizeof(bad_chars_list));
768 bad_count++;
8c6180a9
KW		 769 }
770 }
771 }
772
7d4bcc4a
KW		 773# ifdef MB_CUR_MAX
774
8c6180a9 775 /* We only handle single-byte locales (outside of UTF-8 ones; so if
d35fca5f 776 * this locale requires more than one byte, there are going to be
8c6180a9 777 * problems. */
9c8a6dc2
KW		 778 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
779 "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n",
780 __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX));
781
8b7358b9
KW		 782 if ( check_for_problems && MB_CUR_MAX > 1
783 && ! PL_in_utf8_CTYPE_locale
ba1a4362
KW		 784
785 /* Some platforms return MB_CUR_MAX > 1 for even the "C"
786 * locale. Just assume that the implementation for them (plus
787 * for POSIX) is correct and the > 1 value is spurious. (Since
788 * these are specially handled to never be considered UTF-8
789 * locales, as long as this is the only problem, everything
790 * should work fine */
791 && strNE(newctype, "C") && strNE(newctype, "POSIX"))
792 {
8c6180a9
KW		 793 multi_byte_locale = TRUE;
794 }
7d4bcc4a
KW		 795
796# endif
8c6180a9 797
8b7358b9
KW		 798 if (UNLIKELY(bad_count) || UNLIKELY(multi_byte_locale)) {
799 if (UNLIKELY(bad_count) && PL_in_utf8_CTYPE_locale) {
800 PL_warn_locale = Perl_newSVpvf(aTHX_
801 "Locale '%s' contains (at least) the following characters"
802 " which have\nnon-standard meanings: %s\nThe Perl program"
803 " will use the standard meanings",
804 newctype, bad_chars_list);
805
806 }
807 else {
808 PL_warn_locale = Perl_newSVpvf(aTHX_
8c6180a9 809 "Locale '%s' may not work well.%s%s%s\n",
780fcc9f 810 newctype,
8c6180a9
KW		 811 (multi_byte_locale)
812 ? " Some characters in it are not recognized by"
813 " Perl."
814 : "",
815 (bad_count)
816 ? "\nThe following characters (and maybe others)"
817 " may not have the same meaning as the Perl"
818 " program expects:\n"
819 : "",
820 (bad_count)
821 ? bad_chars_list
822 : ""
823 );
8b7358b9
KW		 824 }
825
b119c2be
KW		 826# ifdef HAS_NL_LANGINFO
827
828 Perl_sv_catpvf(aTHX_ PL_warn_locale, "; codeset=%s",
829 /* parameter FALSE is a don't care here */
830 my_nl_langinfo(PERL_CODESET, FALSE));
831
832# endif
833
8b7358b9
KW		 834 Perl_sv_catpvf(aTHX_ PL_warn_locale, "\n");
835
cc9eaeb0 836 /* If we are actually in the scope of the locale or are debugging,
bddebb56
KW		 837 * output the message now. If not in that scope, we save the
838 * message to be output at the first operation using this locale,
839 * if that actually happens. Most programs don't use locales, so
840 * they are immune to bad ones. */
cc9eaeb0 841 if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) {
780fcc9f 842
780fcc9f
KW		 843 /* The '0' below suppresses a bogus gcc compiler warning */
844 Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
bddebb56 845
bddebb56
KW		 846 if (IN_LC(LC_CTYPE)) {
847 SvREFCNT_dec_NN(PL_warn_locale);
848 PL_warn_locale = NULL;
849 }
780fcc9f 850 }
baa60164 851 }
31f05a37 852 }
98994639
HS		 853
854#endif /* USE_LOCALE_CTYPE */
7d4bcc4a 855
98994639
HS		 856}
857
98994639 858void
2726666d
KW		 859Perl__warn_problematic_locale()
860{
2726666d
KW		 861
862#ifdef USE_LOCALE_CTYPE
863
5f04a188
KW		 864 dTHX;
865
866 /* Internal-to-core function that outputs the message in PL_warn_locale,
867 * and then NULLS it. Should be called only through the macro
868 * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
869
2726666d 870 if (PL_warn_locale) {
2726666d
KW		 871 Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
872 SvPVX(PL_warn_locale),
873 0 /* dummy to avoid compiler warning */ );
2726666d
KW		 874 SvREFCNT_dec_NN(PL_warn_locale);
875 PL_warn_locale = NULL;
876 }
877
878#endif
879
880}
881
a4f00dcc
KW		 882STATIC void
883S_new_collate(pTHX_ const char *newcoll)
98994639 884{
7d4bcc4a
KW		 885
886#ifndef USE_LOCALE_COLLATE
887
888 PERL_UNUSED_ARG(newcoll);
889 PERL_UNUSED_CONTEXT;
890
891#else
0d071d52 892
291a84fb 893 /* Called after each libc setlocale() call affecting LC_COLLATE, to tell
0d071d52
KW		 894 * core Perl this and that 'newcoll' is the name of the new locale.
895 *
d35fca5f
KW		 896 * The design of locale collation is that every locale change is given an
897 * index 'PL_collation_ix'. The first time a string particpates in an
898 * operation that requires collation while locale collation is active, it
899 * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
900 * magic includes the collation index, and the transformation of the string
901 * by strxfrm(), q.v. That transformation is used when doing comparisons,
902 * instead of the string itself. If a string changes, the magic is
903 * cleared. The next time the locale changes, the index is incremented,
904 * and so we know during a comparison that the transformation is not
905 * necessarily still valid, and so is recomputed. Note that if the locale
906 * changes enough times, the index could wrap (a U32), and it is possible
907 * that a transformation would improperly be considered valid, leading to
908 * an unlikely bug */
0d071d52 909
98994639
HS		 910 if (! newcoll) {
911 if (PL_collation_name) {
912 ++PL_collation_ix;
913 Safefree(PL_collation_name);
914 PL_collation_name = NULL;
915 }
916 PL_collation_standard = TRUE;
00bf60ca 917 is_standard_collation:
98994639
HS		 918 PL_collxfrm_base = 0;
919 PL_collxfrm_mult = 2;
165a1c52 920 PL_in_utf8_COLLATE_locale = FALSE;
f28f4d2a 921 PL_strxfrm_NUL_replacement = '\0';
a4a439fb 922 PL_strxfrm_max_cp = 0;
98994639
HS		 923 return;
924 }
925
d35fca5f 926 /* If this is not the same locale as currently, set the new one up */
98994639
HS		 927 if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
928 ++PL_collation_ix;
929 Safefree(PL_collation_name);
930 PL_collation_name = stdize_locale(savepv(newcoll));
a39edc4c 931 PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
00bf60ca
KW		 932 if (PL_collation_standard) {
933 goto is_standard_collation;
934 }
98994639 935
165a1c52 936 PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
f28f4d2a 937 PL_strxfrm_NUL_replacement = '\0';
a4a439fb 938 PL_strxfrm_max_cp = 0;
165a1c52 939
59c018b9
KW		 940 /* A locale collation definition includes primary, secondary, tertiary,
941 * etc. weights for each character. To sort, the primary weights are
942 * used, and only if they compare equal, then the secondary weights are
943 * used, and only if they compare equal, then the tertiary, etc.
944 *
945 * strxfrm() works by taking the input string, say ABC, and creating an
946 * output transformed string consisting of first the primary weights,
947 * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
948 * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
949 * may not have weights at every level. In our example, let's say B
950 * doesn't have a tertiary weight, and A doesn't have a secondary
951 * weight. The constructed string is then going to be
952 * A¹B¹C¹ B²C² A³C³ ....
953 * This has the desired effect that strcmp() will look at the secondary
954 * or tertiary weights only if the strings compare equal at all higher
955 * priority weights. The spaces shown here, like in
c342d20e 956 * "A¹B¹C¹ A²B²C² "
59c018b9
KW		 957 * are not just for readability. In the general case, these must
958 * actually be bytes, which we will call here 'separator weights'; and
959 * they must be smaller than any other weight value, but since these
960 * are C strings, only the terminating one can be a NUL (some
961 * implementations may include a non-NUL separator weight just before
962 * the NUL). Implementations tend to reserve 01 for the separator
963 * weights. They are needed so that a shorter string's secondary
964 * weights won't be misconstrued as primary weights of a longer string,
965 * etc. By making them smaller than any other weight, the shorter
966 * string will sort first. (Actually, if all secondary weights are
967 * smaller than all primary ones, there is no need for a separator
968 * weight between those two levels, etc.)
969 *
970 * The length of the transformed string is roughly a linear function of
971 * the input string. It's not exactly linear because some characters
972 * don't have weights at all levels. When we call strxfrm() we have to
973 * allocate some memory to hold the transformed string. The
974 * calculations below try to find coefficients 'm' and 'b' for this
975 * locale so that m*x + b equals how much space we need, given the size
976 * of the input string in 'x'. If we calculate too small, we increase
977 * the size as needed, and call strxfrm() again, but it is better to
978 * get it right the first time to avoid wasted expensive string
979 * transformations. */
980
98994639 981 {
79f120c8
KW		 982 /* We use the string below to find how long the tranformation of it
983 * is. Almost all locales are supersets of ASCII, or at least the
984 * ASCII letters. We use all of them, half upper half lower,
985 * because if we used fewer, we might hit just the ones that are
986 * outliers in a particular locale. Most of the strings being
987 * collated will contain a preponderance of letters, and even if
988 * they are above-ASCII, they are likely to have the same number of
989 * weight levels as the ASCII ones. It turns out that digits tend
990 * to have fewer levels, and some punctuation has more, but those
991 * are relatively sparse in text, and khw believes this gives a
992 * reasonable result, but it could be changed if experience so
993 * dictates. */
994 const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
995 char * x_longer; /* Transformed 'longer' */
996 Size_t x_len_longer; /* Length of 'x_longer' */
997
998 char * x_shorter; /* We also transform a substring of 'longer' */
999 Size_t x_len_shorter;
1000
a4a439fb 1001 /* _mem_collxfrm() is used get the transformation (though here we
79f120c8
KW		 1002 * are interested only in its length). It is used because it has
1003 * the intelligence to handle all cases, but to work, it needs some
1004 * values of 'm' and 'b' to get it started. For the purposes of
1005 * this calculation we use a very conservative estimate of 'm' and
1006 * 'b'. This assumes a weight can be multiple bytes, enough to
1007 * hold any UV on the platform, and there are 5 levels, 4 weight
1008 * bytes, and a trailing NUL. */
1009 PL_collxfrm_base = 5;
1010 PL_collxfrm_mult = 5 * sizeof(UV);
1011
1012 /* Find out how long the transformation really is */
a4a439fb
KW		 1013 x_longer = _mem_collxfrm(longer,
1014 sizeof(longer) - 1,
1015 &x_len_longer,
1016
1017 /* We avoid converting to UTF-8 in the
1018 * called function by telling it the
1019 * string is in UTF-8 if the locale is a
1020 * UTF-8 one. Since the string passed
1021 * here is invariant under UTF-8, we can
1022 * claim it's UTF-8 even though it isn't.
1023 * */
1024 PL_in_utf8_COLLATE_locale);
79f120c8
KW		 1025 Safefree(x_longer);
1026
1027 /* Find out how long the transformation of a substring of 'longer'
1028 * is. Together the lengths of these transformations are
1029 * sufficient to calculate 'm' and 'b'. The substring is all of
1030 * 'longer' except the first character. This minimizes the chances
1031 * of being swayed by outliers */
a4a439fb 1032 x_shorter = _mem_collxfrm(longer + 1,
79f120c8 1033 sizeof(longer) - 2,
a4a439fb
KW		 1034 &x_len_shorter,
1035 PL_in_utf8_COLLATE_locale);
79f120c8
KW		 1036 Safefree(x_shorter);
1037
1038 /* If the results are nonsensical for this simple test, the whole
1039 * locale definition is suspect. Mark it so that locale collation
1040 * is not active at all for it. XXX Should we warn? */
1041 if ( x_len_shorter == 0
1042 || x_len_longer == 0
1043 || x_len_shorter >= x_len_longer)
1044 {
1045 PL_collxfrm_mult = 0;
1046 PL_collxfrm_base = 0;
1047 }
1048 else {
1049 SSize_t base; /* Temporary */
1050
1051 /* We have both: m * strlen(longer) + b = x_len_longer
1052 * m * strlen(shorter) + b = x_len_shorter;
1053 * subtracting yields:
1054 * m * (strlen(longer) - strlen(shorter))
1055 * = x_len_longer - x_len_shorter
1056 * But we have set things up so that 'shorter' is 1 byte smaller
1057 * than 'longer'. Hence:
1058 * m = x_len_longer - x_len_shorter
1059 *
1060 * But if something went wrong, make sure the multiplier is at
1061 * least 1.
1062 */
1063 if (x_len_longer > x_len_shorter) {
1064 PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
1065 }
1066 else {
1067 PL_collxfrm_mult = 1;
1068 }
1069
1070 /* mx + b = len
1071 * so: b = len - mx
1072 * but in case something has gone wrong, make sure it is
1073 * non-negative */
1074 base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
1075 if (base < 0) {
1076 base = 0;
1077 }
1078
1079 /* Add 1 for the trailing NUL */
1080 PL_collxfrm_base = base + 1;
1081 }
58eebef2 1082
7d4bcc4a
KW		 1083# ifdef DEBUGGING
1084
58eebef2
KW		 1085 if (DEBUG_L_TEST || debug_initialization) {
1086 PerlIO_printf(Perl_debug_log,
b07929e4
KW		 1087 "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
1088 "x_len_longer=%zu,"
1089 " collate multipler=%zu, collate base=%zu\n",
58eebef2
KW		 1090 __FILE__, __LINE__,
1091 PL_in_utf8_COLLATE_locale,
1092 x_len_shorter, x_len_longer,
1093 PL_collxfrm_mult, PL_collxfrm_base);
1094 }
7d4bcc4a
KW		 1095# endif
1096
98994639
HS		 1097 }
1098 }
1099
1100#endif /* USE_LOCALE_COLLATE */
7d4bcc4a 1101
98994639
HS		 1102}
1103
d2b24094 1104#ifdef WIN32
b8cc575c 1105
a4f00dcc 1106STATIC char *
b8cc575c 1107S_win32_setlocale(pTHX_ int category, const char* locale)
b385bb4d
KW		 1108{
1109 /* This, for Windows, emulates POSIX setlocale() behavior. There is no
7d4bcc4a
KW		 1110 * difference between the two unless the input locale is "", which normally
1111 * means on Windows to get the machine default, which is set via the
1112 * computer's "Regional and Language Options" (or its current equivalent).
1113 * In POSIX, it instead means to find the locale from the user's
1114 * environment. This routine changes the Windows behavior to first look in
1115 * the environment, and, if anything is found, use that instead of going to
1116 * the machine default. If there is no environment override, the machine
1117 * default is used, by calling the real setlocale() with "".
1118 *
1119 * The POSIX behavior is to use the LC_ALL variable if set; otherwise to
1120 * use the particular category's variable if set; otherwise to use the LANG
1121 * variable. */
b385bb4d 1122
175c4cf9 1123 bool override_LC_ALL = FALSE;
89f7b9aa 1124 char * result;
e5f10d49 1125 unsigned int i;
89f7b9aa 1126
b385bb4d 1127 if (locale && strEQ(locale, "")) {
7d4bcc4a
KW		 1128
1129# ifdef LC_ALL
1130
b385bb4d
KW		 1131 locale = PerlEnv_getenv("LC_ALL");
1132 if (! locale) {
e5f10d49
KW		 1133 if (category == LC_ALL) {
1134 override_LC_ALL = TRUE;
1135 }
1136 else {
7d4bcc4a
KW		 1137
1138# endif
7d4bcc4a 1139
e5f10d49
KW		 1140 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
1141 if (category == categories[i]) {
1142 locale = PerlEnv_getenv(category_names[i]);
1143 goto found_locale;
1144 }
1145 }
7d4bcc4a 1146
b385bb4d 1147 locale = PerlEnv_getenv("LANG");
481465ea 1148 if (! locale) {
b385bb4d
KW		 1149 locale = "";
1150 }
e5f10d49
KW		 1151
1152 found_locale: ;
7d4bcc4a
KW		 1153
1154# ifdef LC_ALL
1155
e5f10d49 1156 }
b385bb4d 1157 }
7d4bcc4a
KW		 1158
1159# endif
1160
b385bb4d
KW		 1161 }
1162
89f7b9aa 1163 result = setlocale(category, locale);
bbc98134 1164 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
a4f00dcc 1165 setlocale_debug_string(category, locale, result)));
89f7b9aa 1166
481465ea 1167 if (! override_LC_ALL) {
89f7b9aa
KW		 1168 return result;
1169 }
1170
dfd77d7a 1171 /* Here the input category was LC_ALL, and we have set it to what is in the
481465ea
KW		 1172 * LANG variable or the system default if there is no LANG. But these have
1173 * lower priority than the other LC_foo variables, so override it for each
1174 * one that is set. (If they are set to "", it means to use the same thing
1175 * we just set LC_ALL to, so can skip) */
7d4bcc4a 1176
948523db 1177 for (i = 0; i < LC_ALL_INDEX; i++) {
e5f10d49
KW		 1178 result = PerlEnv_getenv(category_names[i]);
1179 if (result && strNE(result, "")) {
1180 setlocale(categories[i], result);
1181 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1182 __FILE__, __LINE__,
1183 setlocale_debug_string(categories[i], result, "not captured")));
1184 }
89f7b9aa 1185 }
7d4bcc4a 1186
bbc98134
KW		 1187 result = setlocale(LC_ALL, NULL);
1188 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
1189 __FILE__, __LINE__,
a4f00dcc 1190 setlocale_debug_string(LC_ALL, NULL, result)));
89f7b9aa 1191
bbc98134 1192 return result;
b385bb4d
KW		 1193}
1194
1195#endif
1196
a4f00dcc
KW		 1197char *
1198Perl_setlocale(int category, const char * locale)
1199{
1200 /* This wraps POSIX::setlocale() */
1201
1202 char * retval;
1159483a 1203 char * newlocale;
a4f00dcc
KW		 1204 dTHX;
1205
a4f00dcc
KW		 1206#ifdef USE_LOCALE_NUMERIC
1207
291a84fb
KW		 1208 /* A NULL locale means only query what the current one is. We have the
1209 * LC_NUMERIC name saved, because we are normally switched into the C
1210 * locale for it. For an LC_ALL query, switch back to get the correct
1211 * results. All other categories don't require special handling */
a4f00dcc
KW		 1212 if (locale == NULL) {
1213 if (category == LC_NUMERIC) {
1214 return savepv(PL_numeric_name);
1215 }
1216
7d4bcc4a 1217# ifdef LC_ALL
a4f00dcc 1218
c0d737a8
KW		 1219 else if (category == LC_ALL && ! PL_numeric_underlying) {
1220
a4f00dcc
KW		 1221 SET_NUMERIC_UNDERLYING();
1222 }
1223
7d4bcc4a 1224# endif
a4f00dcc
KW		 1225
1226 }
1227
1228#endif
1229
e5f10d49
KW		 1230 /* Save retval since subsequent setlocale() calls may overwrite it. */
1231 retval = savepv(do_setlocale_r(category, locale));
a4f00dcc
KW		 1232
1233 DEBUG_L(PerlIO_printf(Perl_debug_log,
1234 "%s:%d: %s\n", __FILE__, __LINE__,
1235 setlocale_debug_string(category, locale, retval)));
1236 if (! retval) {
1237 /* Should never happen that a query would return an error, but be
1238 * sure and reset to C locale */
1239 if (locale == 0) {
1240 SET_NUMERIC_STANDARD();
1241 }
7d4bcc4a 1242
a4f00dcc
KW		 1243 return NULL;
1244 }
1245
a4f00dcc
KW		 1246 /* If locale == NULL, we are just querying the state, but may have switched
1247 * to NUMERIC_UNDERLYING. Switch back before returning. */
1248 if (locale == NULL) {
1249 SET_NUMERIC_STANDARD();
1250 return retval;
1251 }
a4f00dcc 1252
1159483a
KW		 1253 /* Now that have switched locales, we have to update our records to
1254 * correspond. */
a4f00dcc 1255
1159483a 1256 switch (category) {
a4f00dcc 1257
1159483a 1258#ifdef USE_LOCALE_CTYPE
a4f00dcc 1259
1159483a
KW		 1260 case LC_CTYPE:
1261 new_ctype(retval);
1262 break;
a4f00dcc 1263
1159483a 1264#endif
a4f00dcc
KW		 1265#ifdef USE_LOCALE_COLLATE
1266
1159483a
KW		 1267 case LC_COLLATE:
1268 new_collate(retval);
1269 break;
a4f00dcc 1270
1159483a 1271#endif
a4f00dcc
KW		 1272#ifdef USE_LOCALE_NUMERIC
1273
1159483a
KW		 1274 case LC_NUMERIC:
1275 new_numeric(retval);
1276 break;
a4f00dcc 1277
1159483a
KW		 1278#endif
1279#ifdef LC_ALL
a4f00dcc 1280
1159483a 1281 case LC_ALL:
a4f00dcc 1282
1159483a
KW		 1283 /* LC_ALL updates all the things we care about. The values may not
1284 * be the same as 'retval', as the locale "" may have set things
1285 * individually */
a4f00dcc 1286
1159483a 1287# ifdef USE_LOCALE_CTYPE
a4f00dcc 1288
1159483a
KW		 1289 newlocale = do_setlocale_c(LC_CTYPE, NULL);
1290 new_ctype(newlocale);
a4f00dcc 1291
1159483a
KW		 1292# endif /* USE_LOCALE_CTYPE */
1293# ifdef USE_LOCALE_COLLATE
1294
1295 newlocale = do_setlocale_c(LC_COLLATE, NULL);
1296 new_collate(newlocale);
a4f00dcc 1297
7d4bcc4a 1298# endif
1159483a 1299# ifdef USE_LOCALE_NUMERIC
a4f00dcc 1300
1159483a
KW		 1301 newlocale = do_setlocale_c(LC_NUMERIC, NULL);
1302 new_numeric(newlocale);
a4f00dcc 1303
1159483a
KW		 1304# endif /* USE_LOCALE_NUMERIC */
1305#endif /* LC_ALL */
a4f00dcc 1306
1159483a
KW		 1307 default:
1308 break;
a4f00dcc
KW		 1309 }
1310
1311 return retval;
1312
f7416781
KW		 1313
1314}
1315
1316PERL_STATIC_INLINE const char *
1317S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset)
1318{
1319 /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size',
1320 * growing it if necessary */
1321
1322 const Size_t string_size = strlen(string) + offset + 1;
1323
1324 PERL_ARGS_ASSERT_SAVE_TO_BUFFER;
1325
1326 if (*buf_size == 0) {
1327 Newx(*buf, string_size, char);
1328 *buf_size = string_size;
1329 }
1330 else if (string_size > *buf_size) {
1331 Renew(*buf, string_size, char);
1332 *buf_size = string_size;
1333 }
1334
1335 Copy(string, *buf + offset, string_size - offset, char);
1336 return *buf;
1337}
1338
1339/*
1340
1341=head1 Locale-related functions and macros
1342
1343=for apidoc Perl_langinfo
1344
7d4bcc4a 1345This is an (almost ª) drop-in replacement for the system C<L<nl_langinfo(3)>>,
f7416781
KW		 1346taking the same C<item> parameter values, and returning the same information.
1347But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks
1348of Perl's locale handling from your code, and can be used on systems that lack
1349a native C<nl_langinfo>.
1350
1351Expanding on these:
1352
1353=over
1354
1355=item *
1356
1357It delivers the correct results for the C<RADIXCHAR> and C<THOUSESEP> items,
1358without you having to write extra code. The reason for the extra code would be
1359because these are from the C<LC_NUMERIC> locale category, which is normally
1360kept set to the C locale by Perl, no matter what the underlying locale is
1361supposed to be, and so to get the expected results, you have to temporarily
1362toggle into the underlying locale, and later toggle back. (You could use
1363plain C<nl_langinfo> and C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this
1364but then you wouldn't get the other advantages of C<Perl_langinfo()>; not
1365keeping C<LC_NUMERIC> in the C locale would break a lot of CPAN, which is
1366expecting the radix (decimal point) character to be a dot.)
1367
1368=item *
1369
1370Depending on C<item>, it works on systems that don't have C<nl_langinfo>, hence
1371makes your code more portable. Of the fifty-some possible items specified by
1372the POSIX 2008 standard,
1373L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>,
1374only two are completely unimplemented. It uses various techniques to recover
1375the other items, including calling C<L<localeconv(3)>>, and C<L<strftime(3)>>,
1376both of which are specified in C89, so should be always be available. Later
1377C<strftime()> versions have additional capabilities; C<""> is returned for
1378those not available on your system.
1379
9f0bc911
KW		 1380It is important to note that on such systems, this calls C<localeconv>, and so
1381overwrites the static buffer returned from previous explicit calls to that
1382function. Thus, if the program doesn't use or save the information from an
1383explicit C<localeconv> call (which good practice suggests should be done
1384anyway), use of this function can break it.
1385
f7416781
KW		 1386The details for those items which may differ from what this emulation returns
1387and what a native C<nl_langinfo()> would return are:
1388
1389=over
1390
1391=item C<CODESET>
1392
1393=item C<ERA>
1394
1395Unimplemented, so returns C<"">.
1396
1397=item C<YESEXPR>
1398
c1566110
KW		 1399=item C<YESSTR>
1400
f7416781
KW		 1401=item C<NOEXPR>
1402
c1566110
KW		 1403=item C<NOSTR>
1404
1405Only the values for English are returned. C<YESSTR> and C<NOSTR> have been
1406removed from POSIX 2008, and are retained for backwards compatibility. Your
1407platform's C<nl_langinfo> may not support them.
f7416781
KW		 1408
1409=item C<D_FMT>
1410
1411Always evaluates to C<%x>, the locale's appropriate date representation.
1412
1413=item C<T_FMT>
1414
1415Always evaluates to C<%X>, the locale's appropriate time representation.
1416
1417=item C<D_T_FMT>
1418
1419Always evaluates to C<%c>, the locale's appropriate date and time
1420representation.
1421
1422=item C<CRNCYSTR>
1423
1424The return may be incorrect for those rare locales where the currency symbol
1425replaces the radix character.
1426Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1427to work differently.
1428
1429=item C<ALT_DIGITS>
1430
1431Currently this gives the same results as Linux does.
1432Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
1433to work differently.
1434
1435=item C<ERA_D_FMT>
1436
1437=item C<ERA_T_FMT>
1438
1439=item C<ERA_D_T_FMT>
1440
1441=item C<T_FMT_AMPM>
1442
1443These are derived by using C<strftime()>, and not all versions of that function
1444know about them. C<""> is returned for these on such systems.
1445
1446=back
1447
1448When using C<Perl_langinfo> on systems that don't have a native
1449C<nl_langinfo()>, you must
1450
1451 #include "perl_langinfo.h"
1452
1453before the C<perl.h> C<#include>. You can replace your C<langinfo.h>
1454C<#include> with this one. (Doing it this way keeps out the symbols that plain
1455C<langinfo.h> imports into the namespace for code that doesn't need it.)
1456
1457You also should not use the bare C<langinfo.h> item names, but should preface
1458them with C<PERL_>, so use C<PERL_RADIXCHAR> instead of plain C<RADIXCHAR>.
1459The C<PERL_I<foo>> versions will also work for this function on systems that do
1460have a native C<nl_langinfo>.
1461
1462=item *
1463
1464It is thread-friendly, returning its result in a buffer that won't be
1465overwritten by another thread, so you don't have to code for that possibility.
1466The buffer can be overwritten by the next call to C<nl_langinfo> or
1467C<Perl_langinfo> in the same thread.
1468
1469=item *
1470
7d4bcc4a 1471ª It returns S<C<const char *>>, whereas plain C<nl_langinfo()> returns S<C<char
f7416781
KW		 1472*>>, but you are (only by documentation) forbidden to write into the buffer.
1473By declaring this C<const>, the compiler enforces this restriction. The extra
1474C<const> is why this isn't an unequivocal drop-in replacement for
1475C<nl_langinfo>.
1476
1477=back
1478
1479The original impetus for C<Perl_langinfo()> was so that code that needs to
1480find out the current currency symbol, floating point radix character, or digit
1481grouping separator can use, on all systems, the simpler and more
1482thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a
1483pain to make thread-friendly. For other fields returned by C<localeconv>, it
1484is better to use the methods given in L<perlcall> to call
1485L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly.
1486
1487=cut
1488
1489*/
1490
1491const char *
1492#ifdef HAS_NL_LANGINFO
1493Perl_langinfo(const nl_item item)
1494#else
1495Perl_langinfo(const int item)
1496#endif
1497{
f61748ac
KW		 1498 return my_nl_langinfo(item, TRUE);
1499}
1500
1501const char *
1502#ifdef HAS_NL_LANGINFO
1503S_my_nl_langinfo(const nl_item item, bool toggle)
1504#else
1505S_my_nl_langinfo(const int item, bool toggle)
1506#endif
1507{
ae74815b 1508 dTHX;
f7416781 1509
5a854ab3
KW		 1510 /* We only need to toggle into the underlying LC_NUMERIC locale for these
1511 * two items, and only if not already there */
1512 if (toggle && (( item != PERL_RADIXCHAR && item != PERL_THOUSEP)
1513 || PL_numeric_underlying))
1514 {
1515 toggle = FALSE;
1516 }
1517
ab340fff
KW		 1518#if defined(HAS_NL_LANGINFO) /* nl_langinfo() is available. */
1519#if ! defined(HAS_POSIX_2008_LOCALE)
f7416781 1520
ab340fff 1521 /* Here, use plain nl_langinfo(), switching to the underlying LC_NUMERIC
ae74815b
KW		 1522 * for those items dependent on it. This must be copied to a buffer before
1523 * switching back, as some systems destroy the buffer when setlocale() is
1524 * called */
f7416781 1525
038d3702
KW		 1526 {
1527 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
1528
b2fee59c 1529 if (toggle) {
038d3702 1530 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
b2fee59c 1531 }
f7416781 1532
5acc4454
KW		 1533 LOCALE_LOCK; /* Prevent interference from another thread executing
1534 this code section (the only call to nl_langinfo in
1535 the core) */
1536
b2fee59c
KW		 1537 save_to_buffer(nl_langinfo(item), &PL_langinfo_buf,
1538 &PL_langinfo_bufsize, 0);
f7416781 1539
5acc4454
KW		 1540 LOCALE_UNLOCK;
1541
b2fee59c 1542 if (toggle) {
038d3702 1543 RESTORE_LC_NUMERIC();
b2fee59c 1544 }
f7416781
KW		 1545 }
1546
ab340fff
KW		 1547# else /* Use nl_langinfo_l(), avoiding both a mutex and changing the locale */
1548
5a854ab3 1549 {
b2fee59c
KW		 1550 bool do_free = FALSE;
1551 locale_t cur = uselocale((locale_t) 0);
ab340fff 1552
b2fee59c
KW		 1553 if (cur == LC_GLOBAL_LOCALE) {
1554 cur = duplocale(LC_GLOBAL_LOCALE);
1555 do_free = TRUE;
1556 }
ab340fff 1557
b2fee59c 1558 if (toggle) {
e1aa2579
KW		 1559 if (PL_underlying_numeric_obj) {
1560 cur = PL_underlying_numeric_obj;
1561 }
1562 else {
1563 cur = newlocale(LC_NUMERIC_MASK, PL_numeric_name, cur);
1564 do_free = TRUE;
1565 }
b2fee59c 1566 }
ab340fff 1567
b2fee59c
KW		 1568 save_to_buffer(nl_langinfo_l(item, cur),
1569 &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
1570 if (do_free) {
1571 freelocale(cur);
1572 }
5a854ab3 1573 }
ab340fff 1574
c1566110
KW		 1575# endif
1576
1577 if (strEQ(PL_langinfo_buf, "")) {
1578 if (item == PERL_YESSTR) {
1579 return "yes";
1580 }
1581 if (item == PERL_NOSTR) {
1582 return "no";
1583 }
1584 }
1585
ab340fff
KW		 1586 return PL_langinfo_buf;
1587
f7416781 1588#else /* Below, emulate nl_langinfo as best we can */
43dd6b15
KW		 1589
1590 {
1591
f7416781
KW		 1592# ifdef HAS_LOCALECONV
1593
43dd6b15 1594 const struct lconv* lc;
038d3702 1595 DECLARATION_FOR_LC_NUMERIC_MANIPULATION;
f7416781
KW		 1596
1597# endif
1598# ifdef HAS_STRFTIME
1599
43dd6b15
KW		 1600 struct tm tm;
1601 bool return_format = FALSE; /* Return the %format, not the value */
1602 const char * format;
f7416781
KW		 1603
1604# endif
1605
43dd6b15
KW		 1606 /* We copy the results to a per-thread buffer, even if not
1607 * multi-threaded. This is in part to simplify this code, and partly
1608 * because we need a buffer anyway for strftime(), and partly because a
1609 * call of localeconv() could otherwise wipe out the buffer, and the
1610 * programmer would not be expecting this, as this is a nl_langinfo()
1611 * substitute after all, so s/he might be thinking their localeconv()
1612 * is safe until another localeconv() call. */
f7416781 1613
43dd6b15
KW		 1614 switch (item) {
1615 Size_t len;
1616 const char * retval;
f7416781 1617
43dd6b15
KW		 1618 /* These 2 are unimplemented */
1619 case PERL_CODESET:
1620 case PERL_ERA: /* For use with strftime() %E modifier */
f7416781 1621
43dd6b15
KW		 1622 default:
1623 return "";
f7416781 1624
43dd6b15
KW		 1625 /* We use only an English set, since we don't know any more */
1626 case PERL_YESEXPR: return "^[+1yY]";
1627 case PERL_YESSTR: return "yes";
1628 case PERL_NOEXPR: return "^[-0nN]";
1629 case PERL_NOSTR: return "no";
f7416781
KW		 1630
1631# ifdef HAS_LOCALECONV
1632
43dd6b15 1633 case PERL_CRNCYSTR:
f7416781 1634
43dd6b15
KW		 1635 /* We don't bother with localeconv_l() because any system that
1636 * has it is likely to also have nl_langinfo() */
291a84fb 1637
5acc4454
KW		 1638 LOCALE_LOCK; /* Prevent interference with other threads
1639 using localeconv() */
1640
43dd6b15
KW		 1641 lc = localeconv();
1642 if ( ! lc
1643 || ! lc->currency_symbol
1644 || strEQ("", lc->currency_symbol))
1645 {
1646 LOCALE_UNLOCK;
1647 return "";
1648 }
f7416781 1649
43dd6b15
KW		 1650 /* Leave the first spot empty to be filled in below */
1651 save_to_buffer(lc->currency_symbol, &PL_langinfo_buf,
1652 &PL_langinfo_bufsize, 1);
1653 if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, ""))
1654 { /* khw couldn't figure out how the localedef specifications
1655 would show that the $ should replace the radix; this is
1656 just a guess as to how it might work.*/
1657 *PL_langinfo_buf = '.';
1658 }
1659 else if (lc->p_cs_precedes) {
1660 *PL_langinfo_buf = '-';
1661 }
1662 else {
1663 *PL_langinfo_buf = '+';
1664 }
f7416781 1665
43dd6b15
KW		 1666 LOCALE_UNLOCK;
1667 break;
f7416781 1668
43dd6b15
KW		 1669 case PERL_RADIXCHAR:
1670 case PERL_THOUSEP:
f7416781 1671
43dd6b15 1672 if (toggle) {
038d3702 1673 STORE_LC_NUMERIC_FORCE_TO_UNDERLYING();
c0d737a8 1674 }
f7416781 1675
5acc4454
KW		 1676 LOCALE_LOCK; /* Prevent interference with other threads
1677 using localeconv() */
1678
43dd6b15
KW		 1679 lc = localeconv();
1680 if (! lc) {
33394adc
KW		 1681 retval = "";
1682 }
43dd6b15
KW		 1683 else {
1684 retval = (item == PERL_RADIXCHAR)
1685 ? lc->decimal_point
1686 : lc->thousands_sep;
1687 if (! retval) {
1688 retval = "";
1689 }
1690 }
f7416781 1691
43dd6b15
KW		 1692 save_to_buffer(retval, &PL_langinfo_buf,
1693 &PL_langinfo_bufsize, 0);
f7416781 1694
5acc4454
KW		 1695 LOCALE_UNLOCK;
1696
43dd6b15 1697 if (toggle) {
038d3702 1698 RESTORE_LC_NUMERIC();
43dd6b15 1699 }
f7416781 1700
43dd6b15 1701 break;
f7416781
KW		 1702
1703# endif
1704# ifdef HAS_STRFTIME
1705
43dd6b15
KW		 1706 /* These are defined by C89, so we assume that strftime supports
1707 * them, and so are returned unconditionally; they may not be what
1708 * the locale actually says, but should give good enough results
1709 * for someone using them as formats (as opposed to trying to parse
1710 * them to figure out what the locale says). The other format
1711 * items are actually tested to verify they work on the platform */
1712 case PERL_D_FMT: return "%x";
1713 case PERL_T_FMT: return "%X";
1714 case PERL_D_T_FMT: return "%c";
1715
1716 /* These formats are only available in later strfmtime's */
1717 case PERL_ERA_D_FMT: case PERL_ERA_T_FMT: case PERL_ERA_D_T_FMT:
1718 case PERL_T_FMT_AMPM:
1719
1720 /* The rest can be gotten from most versions of strftime(). */
1721 case PERL_ABDAY_1: case PERL_ABDAY_2: case PERL_ABDAY_3:
1722 case PERL_ABDAY_4: case PERL_ABDAY_5: case PERL_ABDAY_6:
1723 case PERL_ABDAY_7:
1724 case PERL_ALT_DIGITS:
1725 case PERL_AM_STR: case PERL_PM_STR:
1726 case PERL_ABMON_1: case PERL_ABMON_2: case PERL_ABMON_3:
1727 case PERL_ABMON_4: case PERL_ABMON_5: case PERL_ABMON_6:
1728 case PERL_ABMON_7: case PERL_ABMON_8: case PERL_ABMON_9:
1729 case PERL_ABMON_10: case PERL_ABMON_11: case PERL_ABMON_12:
1730 case PERL_DAY_1: case PERL_DAY_2: case PERL_DAY_3: case PERL_DAY_4:
1731 case PERL_DAY_5: case PERL_DAY_6: case PERL_DAY_7:
1732 case PERL_MON_1: case PERL_MON_2: case PERL_MON_3: case PERL_MON_4:
1733 case PERL_MON_5: case PERL_MON_6: case PERL_MON_7: case PERL_MON_8:
1734 case PERL_MON_9: case PERL_MON_10: case PERL_MON_11:
1735 case PERL_MON_12:
1736
1737 LOCALE_LOCK;
1738
1739 init_tm(&tm); /* Precaution against core dumps */
1740 tm.tm_sec = 30;
1741 tm.tm_min = 30;
1742 tm.tm_hour = 6;
1743 tm.tm_year = 2017 - 1900;
1744 tm.tm_wday = 0;
1745 tm.tm_mon = 0;
1746 switch (item) {
1747 default:
1748 LOCALE_UNLOCK;
1749 Perl_croak(aTHX_
1750 "panic: %s: %d: switch case: %d problem",
1751 __FILE__, __LINE__, item);
1752 NOT_REACHED; /* NOTREACHED */
1753
1754 case PERL_PM_STR: tm.tm_hour = 18;
1755 case PERL_AM_STR:
1756 format = "%p";
1757 break;
1758
1759 case PERL_ABDAY_7: tm.tm_wday++;
1760 case PERL_ABDAY_6: tm.tm_wday++;
1761 case PERL_ABDAY_5: tm.tm_wday++;
1762 case PERL_ABDAY_4: tm.tm_wday++;
1763 case PERL_ABDAY_3: tm.tm_wday++;
1764 case PERL_ABDAY_2: tm.tm_wday++;
1765 case PERL_ABDAY_1:
1766 format = "%a";
1767 break;
1768
1769 case PERL_DAY_7: tm.tm_wday++;
1770 case PERL_DAY_6: tm.tm_wday++;
1771 case PERL_DAY_5: tm.tm_wday++;
1772 case PERL_DAY_4: tm.tm_wday++;
1773 case PERL_DAY_3: tm.tm_wday++;
1774 case PERL_DAY_2: tm.tm_wday++;
1775 case PERL_DAY_1:
1776 format = "%A";
1777 break;
1778
1779 case PERL_ABMON_12: tm.tm_mon++;
1780 case PERL_ABMON_11: tm.tm_mon++;
1781 case PERL_ABMON_10: tm.tm_mon++;
1782 case PERL_ABMON_9: tm.tm_mon++;
1783 case PERL_ABMON_8: tm.tm_mon++;
1784 case PERL_ABMON_7: tm.tm_mon++;
1785 case PERL_ABMON_6: tm.tm_mon++;
1786 case PERL_ABMON_5: tm.tm_mon++;
1787 case PERL_ABMON_4: tm.tm_mon++;
1788 case PERL_ABMON_3: tm.tm_mon++;
1789 case PERL_ABMON_2: tm.tm_mon++;
1790 case PERL_ABMON_1:
1791 format = "%b";
1792 break;
1793
1794 case PERL_MON_12: tm.tm_mon++;
1795 case PERL_MON_11: tm.tm_mon++;
1796 case PERL_MON_10: tm.tm_mon++;
1797 case PERL_MON_9: tm.tm_mon++;
1798 case PERL_MON_8: tm.tm_mon++;
1799 case PERL_MON_7: tm.tm_mon++;
1800 case PERL_MON_6: tm.tm_mon++;
1801 case PERL_MON_5: tm.tm_mon++;
1802 case PERL_MON_4: tm.tm_mon++;
1803 case PERL_MON_3: tm.tm_mon++;
1804 case PERL_MON_2: tm.tm_mon++;
1805 case PERL_MON_1:
1806 format = "%B";
1807 break;
1808
1809 case PERL_T_FMT_AMPM:
1810 format = "%r";
1811 return_format = TRUE;
1812 break;
1813
1814 case PERL_ERA_D_FMT:
1815 format = "%Ex";
1816 return_format = TRUE;
1817 break;
1818
1819 case PERL_ERA_T_FMT:
1820 format = "%EX";
1821 return_format = TRUE;
1822 break;
1823
1824 case PERL_ERA_D_T_FMT:
1825 format = "%Ec";
1826 return_format = TRUE;
1827 break;
1828
1829 case PERL_ALT_DIGITS:
1830 tm.tm_wday = 0;
1831 format = "%Ow"; /* Find the alternate digit for 0 */
1832 break;
1833 }
f7416781 1834
43dd6b15
KW		 1835 /* We can't use my_strftime() because it doesn't look at
1836 * tm_wday */
1837 while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize,
1838 format, &tm))
1839 {
1840 /* A zero return means one of:
1841 * a) there wasn't enough space in PL_langinfo_buf
1842 * b) the format, like a plain %p, returns empty
1843 * c) it was an illegal format, though some
1844 * implementations of strftime will just return the
1845 * illegal format as a plain character sequence.
1846 *
1847 * To quickly test for case 'b)', try again but precede
1848 * the format with a plain character. If that result is
1849 * still empty, the problem is either 'a)' or 'c)' */
1850
1851 Size_t format_size = strlen(format) + 1;
1852 Size_t mod_size = format_size + 1;
1853 char * mod_format;
1854 char * temp_result;
1855
1856 Newx(mod_format, mod_size, char);
1857 Newx(temp_result, PL_langinfo_bufsize, char);
6873aa47 1858 *mod_format = ' ';
43dd6b15
KW		 1859 my_strlcpy(mod_format + 1, format, mod_size);
1860 len = strftime(temp_result,
1861 PL_langinfo_bufsize,
1862 mod_format, &tm);
1863 Safefree(mod_format);
1864 Safefree(temp_result);
1865
1866 /* If 'len' is non-zero, it means that we had a case like
1867 * %p which means the current locale doesn't use a.m. or
1868 * p.m., and that is valid */
1869 if (len == 0) {
1870
1871 /* Here, still didn't work. If we get well beyond a
1872 * reasonable size, bail out to prevent an infinite
1873 * loop. */
1874
1875 if (PL_langinfo_bufsize > 100 * format_size) {
1876 *PL_langinfo_buf = '\0';
1877 }
1878 else {
1879 /* Double the buffer size to retry; Add 1 in case
1880 * original was 0, so we aren't stuck at 0. */
1881 PL_langinfo_bufsize *= 2;
1882 PL_langinfo_bufsize++;
1883 Renew(PL_langinfo_buf, PL_langinfo_bufsize, char);
1884 continue;
1885 }
1886 }
f7416781 1887
f7416781 1888 break;
43dd6b15 1889 }
f7416781 1890
43dd6b15
KW		 1891 /* Here, we got a result.
1892 *
1893 * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the
1894 * alternate format for wday 0. If the value is the same as
1895 * the normal 0, there isn't an alternate, so clear the buffer.
1896 * */
1897 if ( item == PERL_ALT_DIGITS
1898 && strEQ(PL_langinfo_buf, "0"))
1899 {
1900 *PL_langinfo_buf = '\0';
1901 }
f7416781 1902
43dd6b15
KW		 1903 /* ALT_DIGITS is problematic. Experiments on it showed that
1904 * strftime() did not always work properly when going from
1905 * alt-9 to alt-10. Only a few locales have this item defined,
1906 * and in all of them on Linux that khw was able to find,
1907 * nl_langinfo() merely returned the alt-0 character, possibly
1908 * doubled. Most Unicode digits are in blocks of 10
1909 * consecutive code points, so that is sufficient information
1910 * for those scripts, as we can infer alt-1, alt-2, .... But
1911 * for a Japanese locale, a CJK ideographic 0 is returned, and
1912 * the CJK digits are not in code point order, so you can't
1913 * really infer anything. The localedef for this locale did
1914 * specify the succeeding digits, so that strftime() works
1915 * properly on them, without needing to infer anything. But
1916 * the nl_langinfo() return did not give sufficient information
1917 * for the caller to understand what's going on. So until
1918 * there is evidence that it should work differently, this
1919 * returns the alt-0 string for ALT_DIGITS.
1920 *
1921 * wday was chosen because its range is all a single digit.
1922 * Things like tm_sec have two digits as the minimum: '00' */
f7416781 1923
43dd6b15 1924 LOCALE_UNLOCK;
f7416781 1925
43dd6b15
KW		 1926 /* If to return the format, not the value, overwrite the buffer
1927 * with it. But some strftime()s will keep the original format
1928 * if illegal, so change those to "" */
1929 if (return_format) {
1930 if (strEQ(PL_langinfo_buf, format)) {
f7416781
KW		 1931 *PL_langinfo_buf = '\0';
1932 }
43dd6b15
KW		 1933 else {
1934 save_to_buffer(format, &PL_langinfo_buf,
1935 &PL_langinfo_bufsize, 0);
f7416781
KW		 1936 }
1937 }
1938
1939 break;
f7416781
KW		 1940
1941# endif
1942
43dd6b15 1943 }
f7416781
KW		 1944 }
1945
1946 return PL_langinfo_buf;
1947
1948#endif
1949
a4f00dcc 1950}
b385bb4d 1951
98994639
HS		 1952/*
1953 * Initialize locale awareness.
1954 */
1955int
1956Perl_init_i18nl10n(pTHX_ int printwarn)
1957{
0e92a118
KW		 1958 /* printwarn is
1959 *
1960 * 0 if not to output warning when setup locale is bad
1961 * 1 if to output warning based on value of PERL_BADLANG
1962 * >1 if to output regardless of PERL_BADLANG
1963 *
1964 * returns
98994639 1965 * 1 = set ok or not applicable,
0e92a118
KW		 1966 * 0 = fallback to a locale of lower priority
1967 * -1 = fallback to all locales failed, not even to the C locale
6b058d42
KW		 1968 *
1969 * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
1970 * set, debugging information is output.
1971 *
1972 * This looks more complicated than it is, mainly due to the #ifdefs.
1973 *
1974 * We try to set LC_ALL to the value determined by the environment. If
1975 * there is no LC_ALL on this platform, we try the individual categories we
1976 * know about. If this works, we are done.
1977 *
1978 * But if it doesn't work, we have to do something else. We search the
1979 * environment variables ourselves instead of relying on the system to do
1980 * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
1981 * think there is one), and the ultimate fallback "C". This is all done in
1982 * the same loop as above to avoid duplicating code, but it makes things
7d4bcc4a
KW		 1983 * more complex. The 'trial_locales' array is initialized with just one
1984 * element; it causes the behavior described in the paragraph above this to
1985 * happen. If that fails, we add elements to 'trial_locales', and do extra
1986 * loop iterations to cause the behavior described in this paragraph.
6b058d42
KW		 1987 *
1988 * On Ultrix, the locale MUST come from the environment, so there is
1989 * preliminary code to set it. I (khw) am not sure that it is necessary,
1990 * and that this couldn't be folded into the loop, but barring any real
1991 * platforms to test on, it's staying as-is
1992 *
1993 * A slight complication is that in embedded Perls, the locale may already
1994 * be set-up, and we don't want to get it from the normal environment
1995 * variables. This is handled by having a special environment variable
1996 * indicate we're in this situation. We simply set setlocale's 2nd
1997 * parameter to be a NULL instead of "". That indicates to setlocale that
1998 * it is not to change anything, but to return the current value,
1999 * effectively initializing perl's db to what the locale already is.
2000 *
2001 * We play the same trick with NULL if a LC_ALL succeeds. We call
2002 * setlocale() on the individual categores with NULL to get their existing
2003 * values for our db, instead of trying to change them.
2004 * */
98994639 2005
0e92a118
KW		 2006 int ok = 1;
2007
7d4bcc4a
KW		 2008#ifndef USE_LOCALE
2009
2010 PERL_UNUSED_ARG(printwarn);
2011
2012#else /* USE_LOCALE */
7d4bcc4a
KW		 2013# ifdef __GLIBC__
2014
175c4cf9 2015 const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
7d4bcc4a
KW		 2016
2017# endif
65ebb059 2018
ccd65d51
KW		 2019 /* NULL uses the existing already set up locale */
2020 const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
2021 ? NULL
2022 : "";
c3fcd832
KW		 2023 const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
2024 unsigned int trial_locales_count;
175c4cf9
KW		 2025 const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
2026 const char * const lang = savepv(PerlEnv_getenv("LANG"));
98994639 2027 bool setlocale_failure = FALSE;
65ebb059 2028 unsigned int i;
175c4cf9
KW		 2029
2030 /* A later getenv() could zap this, so only use here */
2031 const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
2032
2033 const bool locwarn = (printwarn > 1
e5f10d49
KW		 2034 || ( printwarn
2035 && ( ! bad_lang_use_once
22ff3130 2036 || (
e5f10d49
KW		 2037 /* disallow with "" or "0" */
2038 *bad_lang_use_once
2039 && strNE("0", bad_lang_use_once)))));
ea92aad8 2040
291a84fb 2041 /* setlocale() return vals; not copied so must be looked at immediately */
8de4332b 2042 const char * sl_result[NOMINAL_LC_ALL_INDEX + 1];
291a84fb
KW		 2043
2044 /* current locale for given category; should have been copied so aren't
2045 * volatile */
8de4332b 2046 const char * curlocales[NOMINAL_LC_ALL_INDEX + 1];
291a84fb 2047
7d4bcc4a
KW		 2048# ifdef WIN32
2049
6bce99ee
JH		 2050 /* In some systems you can find out the system default locale
2051 * and use that as the fallback locale. */
7d4bcc4a
KW		 2052# define SYSTEM_DEFAULT_LOCALE
2053# endif
2054# ifdef SYSTEM_DEFAULT_LOCALE
2055
65ebb059 2056 const char *system_default_locale = NULL;
98994639 2057
7d4bcc4a 2058# endif
948523db
KW		 2059
2060# ifndef DEBUGGING
2061# define DEBUG_LOCALE_INIT(a,b,c)
2062# else
7d4bcc4a 2063
8298454c 2064 DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT")));
7d4bcc4a
KW		 2065
2066# define DEBUG_LOCALE_INIT(category, locale, result) \
2fcc0ca9
KW		 2067 STMT_START { \
2068 if (debug_initialization) { \
2069 PerlIO_printf(Perl_debug_log, \
2070 "%s:%d: %s\n", \
2071 __FILE__, __LINE__, \
a4f00dcc 2072 setlocale_debug_string(category, \
2fcc0ca9
KW		 2073 locale, \
2074 result)); \
2075 } \
2076 } STMT_END
2fcc0ca9 2077
948523db
KW		 2078/* Make sure the parallel arrays are properly set up */
2079# ifdef USE_LOCALE_NUMERIC
2080 assert(categories[LC_NUMERIC_INDEX] == LC_NUMERIC);
2081 assert(strEQ(category_names[LC_NUMERIC_INDEX], "LC_NUMERIC"));
2082# endif
2083# ifdef USE_LOCALE_CTYPE
2084 assert(categories[LC_CTYPE_INDEX] == LC_CTYPE);
2085 assert(strEQ(category_names[LC_CTYPE_INDEX], "LC_CTYPE"));
2086# endif
2087# ifdef USE_LOCALE_COLLATE
2088 assert(categories[LC_COLLATE_INDEX] == LC_COLLATE);
2089 assert(strEQ(category_names[LC_COLLATE_INDEX], "LC_COLLATE"));
2090# endif
2091# ifdef USE_LOCALE_TIME
2092 assert(categories[LC_TIME_INDEX] == LC_TIME);
2093 assert(strEQ(category_names[LC_TIME_INDEX], "LC_TIME"));
2094# endif
2095# ifdef USE_LOCALE_MESSAGES
2096 assert(categories[LC_MESSAGES_INDEX] == LC_MESSAGES);
2097 assert(strEQ(category_names[LC_MESSAGES_INDEX], "LC_MESSAGES"));
2098# endif
2099# ifdef USE_LOCALE_MONETARY
2100 assert(categories[LC_MONETARY_INDEX] == LC_MONETARY);
2101 assert(strEQ(category_names[LC_MONETARY_INDEX], "LC_MONETARY"));
2102# endif
9821811f
KW		 2103# ifdef USE_LOCALE_ADDRESS
2104 assert(categories[LC_ADDRESS_INDEX] == LC_ADDRESS);
2105 assert(strEQ(category_names[LC_ADDRESS_INDEX], "LC_ADDRESS"));
2106# endif
2107# ifdef USE_LOCALE_IDENTIFICATION
2108 assert(categories[LC_IDENTIFICATION_INDEX] == LC_IDENTIFICATION);
2109 assert(strEQ(category_names[LC_IDENTIFICATION_INDEX], "LC_IDENTIFICATION"));
2110# endif
2111# ifdef USE_LOCALE_MEASUREMENT
2112 assert(categories[LC_MEASUREMENT_INDEX] == LC_MEASUREMENT);
2113 assert(strEQ(category_names[LC_MEASUREMENT_INDEX], "LC_MEASUREMENT"));
2114# endif
2115# ifdef USE_LOCALE_PAPER
2116 assert(categories[LC_PAPER_INDEX] == LC_PAPER);
2117 assert(strEQ(category_names[LC_PAPER_INDEX], "LC_PAPER"));
2118# endif
2119# ifdef USE_LOCALE_TELEPHONE
2120 assert(categories[LC_TELEPHONE_INDEX] == LC_TELEPHONE);
2121 assert(strEQ(category_names[LC_TELEPHONE_INDEX], "LC_TELEPHONE"));
2122# endif
948523db
KW		 2123# ifdef LC_ALL
2124 assert(categories[LC_ALL_INDEX] == LC_ALL);
2125 assert(strEQ(category_names[LC_ALL_INDEX], "LC_ALL"));
2126 assert(NOMINAL_LC_ALL_INDEX == LC_ALL_INDEX);
2127# endif
2128# endif /* DEBUGGING */
47280b20
KW		 2129
2130 /* Initialize the cache of the program's UTF-8ness for the always known
2131 * locales C and POSIX */
2132 my_strlcpy(PL_locale_utf8ness, C_and_POSIX_utf8ness,
2133 sizeof(PL_locale_utf8ness));
2134
3ca88433
KW		 2135 PL_numeric_radix_sv = newSVpvs(".");
2136
b56c4436 2137# ifdef LOCALE_ENVIRON_REQUIRED
98994639
HS		 2138
2139 /*
2140 * Ultrix setlocale(..., "") fails if there are no environment
2141 * variables from which to get a locale name.
2142 */
2143
b56c4436
KW		 2144# ifndef LC_ALL
2145# error Ultrix without LC_ALL not implemented
2146# else
7d4bcc4a 2147
b56c4436
KW		 2148 {
2149 bool done = FALSE;
ea92aad8
KW		 2150 if (lang) {
2151 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, setlocale_init);
2152 DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result[LC_ALL_INDEX]);
2153 if (sl_result[LC_ALL_INDEX])
2154 done = TRUE;
2155 else
e5f10d49 2156 setlocale_failure = TRUE;
ea92aad8
KW		 2157 }
2158 if (! setlocale_failure) {
2159 const char * locale_param;
2160 for (i = 0; i < LC_ALL_INDEX; i++) {
2161 locale_param = (! done && (lang || PerlEnv_getenv(category_names[i])))
2162 ? setlocale_init
2163 : NULL;
2164 sl_result[i] = do_setlocale_r(categories[i], locale_param);
2165 if (! sl_result[i]) {
2166 setlocale_failure = TRUE;
2167 }
2168 DEBUG_LOCALE_INIT(categories[i], locale_param, sl_result[i]);
e5f10d49 2169 }
c835d6be 2170 }
7d4bcc4a
KW		 2171 }
2172
2173# endif /* LC_ALL */
e5f10d49 2174# endif /* LOCALE_ENVIRON_REQUIRED */
98994639 2175
65ebb059 2176 /* We try each locale in the list until we get one that works, or exhaust
20a240df
KW		 2177 * the list. Normally the loop is executed just once. But if setting the
2178 * locale fails, inside the loop we add fallback trials to the array and so
2179 * will execute the loop multiple times */
c3fcd832
KW		 2180 trial_locales[0] = setlocale_init;
2181 trial_locales_count = 1;
7d4bcc4a 2182
65ebb059
KW		 2183 for (i= 0; i < trial_locales_count; i++) {
2184 const char * trial_locale = trial_locales[i];
2185
2186 if (i > 0) {
2187
2188 /* XXX This is to preserve old behavior for LOCALE_ENVIRON_REQUIRED
2189 * when i==0, but I (khw) don't think that behavior makes much
2190 * sense */
2191 setlocale_failure = FALSE;
2192
7d4bcc4a 2193# ifdef SYSTEM_DEFAULT_LOCALE
291a84fb 2194# ifdef WIN32 /* Note that assumes Win32 has LC_ALL */
7d4bcc4a 2195
65ebb059
KW		 2196 /* On Windows machines, an entry of "" after the 0th means to use
2197 * the system default locale, which we now proceed to get. */
2198 if (strEQ(trial_locale, "")) {
2199 unsigned int j;
2200
2201 /* Note that this may change the locale, but we are going to do
2202 * that anyway just below */
837ce802 2203 system_default_locale = do_setlocale_c(LC_ALL, "");
5d1187d1 2204 DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
65ebb059 2205
7d4bcc4a 2206 /* Skip if invalid or if it's already on the list of locales to
65ebb059
KW		 2207 * try */
2208 if (! system_default_locale) {
2209 goto next_iteration;
2210 }
2211 for (j = 0; j < trial_locales_count; j++) {
2212 if (strEQ(system_default_locale, trial_locales[j])) {
2213 goto next_iteration;
2214 }
2215 }
2216
2217 trial_locale = system_default_locale;
2218 }
ec0202b5
KW		 2219# else
2220# error SYSTEM_DEFAULT_LOCALE only implemented for Win32
2221# endif
7d4bcc4a 2222# endif /* SYSTEM_DEFAULT_LOCALE */
291a84fb
KW		 2223
2224 } /* For i > 0 */
65ebb059 2225
7d4bcc4a
KW		 2226# ifdef LC_ALL
2227
948523db
KW		 2228 sl_result[LC_ALL_INDEX] = do_setlocale_c(LC_ALL, trial_locale);
2229 DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result[LC_ALL_INDEX]);
2230 if (! sl_result[LC_ALL_INDEX]) {
49c85077 2231 setlocale_failure = TRUE;
7cd8b568
KW		 2232 }
2233 else {
2234 /* Since LC_ALL succeeded, it should have changed all the other
2235 * categories it can to its value; so we massage things so that the
2236 * setlocales below just return their category's current values.
2237 * This adequately handles the case in NetBSD where LC_COLLATE may
2238 * not be defined for a locale, and setting it individually will
7d4bcc4a 2239 * fail, whereas setting LC_ALL succeeds, leaving LC_COLLATE set to
7cd8b568
KW		 2240 * the POSIX locale. */
2241 trial_locale = NULL;
2242 }
7d4bcc4a
KW		 2243
2244# endif /* LC_ALL */
98994639 2245
e5f10d49
KW		 2246 if (! setlocale_failure) {
2247 unsigned int j;
2248 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2249 curlocales[j]
2250 = savepv(do_setlocale_r(categories[j], trial_locale));
2251 if (! curlocales[j]) {
2252 setlocale_failure = TRUE;
2253 }
2254 DEBUG_LOCALE_INIT(categories[j], trial_locale, curlocales[j]);
2255 }
c835d6be 2256
e5f10d49
KW		 2257 if (! setlocale_failure) { /* All succeeded */
2258 break; /* Exit trial_locales loop */
49c85077 2259 }
65ebb059 2260 }
98994639 2261
49c85077
KW		 2262 /* Here, something failed; will need to try a fallback. */
2263 ok = 0;
65ebb059 2264
49c85077
KW		 2265 if (i == 0) {
2266 unsigned int j;
98994639 2267
65ebb059 2268 if (locwarn) { /* Output failure info only on the first one */
7d4bcc4a
KW		 2269
2270# ifdef LC_ALL
98994639 2271
49c85077
KW		 2272 PerlIO_printf(Perl_error_log,
2273 "perl: warning: Setting locale failed.\n");
98994639 2274
7d4bcc4a 2275# else /* !LC_ALL */
98994639 2276
49c85077
KW		 2277 PerlIO_printf(Perl_error_log,
2278 "perl: warning: Setting locale failed for the categories:\n\t");
7d4bcc4a 2279
e5f10d49
KW		 2280 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2281 if (! curlocales[j]) {
2282 PerlIO_printf(Perl_error_log, category_names[j]);
2283 }
2284 else {
2285 Safefree(curlocales[j]);
2286 }
2287 }
7d4bcc4a 2288
7d4bcc4a 2289# endif /* LC_ALL */
98994639 2290
49c85077
KW		 2291 PerlIO_printf(Perl_error_log,
2292 "perl: warning: Please check that your locale settings:\n");
98994639 2293
7d4bcc4a
KW		 2294# ifdef __GLIBC__
2295
49c85077
KW		 2296 PerlIO_printf(Perl_error_log,
2297 "\tLANGUAGE = %c%s%c,\n",
2298 language ? '"' : '(',
2299 language ? language : "unset",
2300 language ? '"' : ')');
7d4bcc4a 2301# endif
98994639 2302
49c85077
KW		 2303 PerlIO_printf(Perl_error_log,
2304 "\tLC_ALL = %c%s%c,\n",
2305 lc_all ? '"' : '(',
2306 lc_all ? lc_all : "unset",
2307 lc_all ? '"' : ')');
98994639 2308
7d4bcc4a
KW		 2309# if defined(USE_ENVIRON_ARRAY)
2310
49c85077 2311 {
cd999af9 2312 char **e;
d5e32b93
KW		 2313
2314 /* Look through the environment for any variables of the
2315 * form qr/ ^ LC_ [A-Z]+ = /x, except LC_ALL which was
2316 * already handled above. These are assumed to be locale
2317 * settings. Output them and their values. */
cd999af9 2318 for (e = environ; *e; e++) {
d5e32b93
KW		 2319 const STRLEN prefix_len = sizeof("LC_") - 1;
2320 STRLEN uppers_len;
2321
cd999af9 2322 if ( strBEGINs(*e, "LC_")
c8b388b0 2323 && ! strBEGINs(*e, "LC_ALL=")
d5e32b93
KW		 2324 && (uppers_len = strspn(*e + prefix_len,
2325 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"))
2326 && ((*e)[prefix_len + uppers_len] == '='))
cd999af9
KW		 2327 {
2328 PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
d5e32b93
KW		 2329 (int) (prefix_len + uppers_len), *e,
2330 *e + prefix_len + uppers_len + 1);
cd999af9
KW		 2331 }
2332 }
49c85077 2333 }
7d4bcc4a
KW		 2334
2335# else
2336
49c85077
KW		 2337 PerlIO_printf(Perl_error_log,
2338 "\t(possibly more locale environment variables)\n");
7d4bcc4a
KW		 2339
2340# endif
98994639 2341
49c85077
KW		 2342 PerlIO_printf(Perl_error_log,
2343 "\tLANG = %c%s%c\n",
2344 lang ? '"' : '(',
2345 lang ? lang : "unset",
2346 lang ? '"' : ')');
98994639 2347
49c85077
KW		 2348 PerlIO_printf(Perl_error_log,
2349 " are supported and installed on your system.\n");
2350 }
98994639 2351
65ebb059 2352 /* Calculate what fallback locales to try. We have avoided this
f6bab5f6 2353 * until we have to, because failure is quite unlikely. This will
65ebb059
KW		 2354 * usually change the upper bound of the loop we are in.
2355 *
2356 * Since the system's default way of setting the locale has not
2357 * found one that works, We use Perl's defined ordering: LC_ALL,
2358 * LANG, and the C locale. We don't try the same locale twice, so
2359 * don't add to the list if already there. (On POSIX systems, the
2360 * LC_ALL element will likely be a repeat of the 0th element "",
6b058d42
KW		 2361 * but there's no harm done by doing it explicitly.
2362 *
2363 * Note that this tries the LC_ALL environment variable even on
2364 * systems which have no LC_ALL locale setting. This may or may
2365 * not have been originally intentional, but there's no real need
2366 * to change the behavior. */
65ebb059
KW		 2367 if (lc_all) {
2368 for (j = 0; j < trial_locales_count; j++) {
2369 if (strEQ(lc_all, trial_locales[j])) {
2370 goto done_lc_all;
2371 }
2372 }
2373 trial_locales[trial_locales_count++] = lc_all;
2374 }
2375 done_lc_all:
98994639 2376
65ebb059
KW		 2377 if (lang) {
2378 for (j = 0; j < trial_locales_count; j++) {
2379 if (strEQ(lang, trial_locales[j])) {
2380 goto done_lang;
2381 }
2382 }
2383 trial_locales[trial_locales_count++] = lang;
2384 }
2385 done_lang:
2386
7d4bcc4a
KW		 2387# if defined(WIN32) && defined(LC_ALL)
2388
65ebb059
KW		 2389 /* For Windows, we also try the system default locale before "C".
2390 * (If there exists a Windows without LC_ALL we skip this because
2391 * it gets too complicated. For those, the "C" is the next
2392 * fallback possibility). The "" is the same as the 0th element of
2393 * the array, but the code at the loop above knows to treat it
2394 * differently when not the 0th */
2395 trial_locales[trial_locales_count++] = "";
7d4bcc4a
KW		 2396
2397# endif
65ebb059
KW		 2398
2399 for (j = 0; j < trial_locales_count; j++) {
2400 if (strEQ("C", trial_locales[j])) {
2401 goto done_C;
2402 }
2403 }
2404 trial_locales[trial_locales_count++] = "C";
98994639 2405
65ebb059
KW		 2406 done_C: ;
2407 } /* end of first time through the loop */
98994639 2408
7d4bcc4a
KW		 2409# ifdef WIN32
2410
65ebb059 2411 next_iteration: ;
7d4bcc4a
KW		 2412
2413# endif
65ebb059
KW		 2414
2415 } /* end of looping through the trial locales */
2416
2417 if (ok < 1) { /* If we tried to fallback */
2418 const char* msg;
2419 if (! setlocale_failure) { /* fallback succeeded */
2420 msg = "Falling back to";
2421 }
2422 else { /* fallback failed */
e5f10d49 2423 unsigned int j;
98994639 2424
65ebb059
KW		 2425 /* We dropped off the end of the loop, so have to decrement i to
2426 * get back to the value the last time through */
2427 i--;
98994639 2428
65ebb059
KW		 2429 ok = -1;
2430 msg = "Failed to fall back to";
2431
2432 /* To continue, we should use whatever values we've got */
7d4bcc4a 2433
e5f10d49
KW		 2434 for (j = 0; j < NOMINAL_LC_ALL_INDEX; j++) {
2435 Safefree(curlocales[j]);
2436 curlocales[j] = savepv(do_setlocale_r(categories[j], NULL));
2437 DEBUG_LOCALE_INIT(categories[j], NULL, curlocales[j]);
2438 }
65ebb059
KW		 2439 }
2440
2441 if (locwarn) {
2442 const char * description;
2443 const char * name = "";
2444 if (strEQ(trial_locales[i], "C")) {
2445 description = "the standard locale";
2446 name = "C";
2447 }
7d4bcc4a
KW		 2448
2449# ifdef SYSTEM_DEFAULT_LOCALE
2450
65ebb059
KW		 2451 else if (strEQ(trial_locales[i], "")) {
2452 description = "the system default locale";
2453 if (system_default_locale) {
2454 name = system_default_locale;
2455 }
2456 }
7d4bcc4a
KW		 2457
2458# endif /* SYSTEM_DEFAULT_LOCALE */
2459
65ebb059
KW		 2460 else {
2461 description = "a fallback locale";
2462 name = trial_locales[i];
2463 }
2464 if (name && strNE(name, "")) {
2465 PerlIO_printf(Perl_error_log,
2466 "perl: warning: %s %s (\"%s\").\n", msg, description, name);
2467 }
2468 else {
2469 PerlIO_printf(Perl_error_log,
2470 "perl: warning: %s %s.\n", msg, description);
2471 }
2472 }
2473 } /* End of tried to fallback */
98994639 2474
e5f10d49
KW		 2475 /* Done with finding the locales; update our records */
2476
7d4bcc4a
KW		 2477# ifdef USE_LOCALE_CTYPE
2478
948523db 2479 new_ctype(curlocales[LC_CTYPE_INDEX]);
98994639 2480
e5f10d49 2481# endif
7d4bcc4a
KW		 2482# ifdef USE_LOCALE_COLLATE
2483
948523db 2484 new_collate(curlocales[LC_COLLATE_INDEX]);
98994639 2485
e5f10d49 2486# endif
7d4bcc4a
KW		 2487# ifdef USE_LOCALE_NUMERIC
2488
948523db 2489 new_numeric(curlocales[LC_NUMERIC_INDEX]);
e5f10d49
KW		 2490
2491# endif
2492
948523db 2493 for (i = 0; i < NOMINAL_LC_ALL_INDEX; i++) {
47280b20
KW		 2494
2495# if defined(USE_ITHREADS)
2496
2497 /* This caches whether each category's locale is UTF-8 or not. This
2498 * may involve changing the locale. It is ok to do this at
2499 * initialization time before any threads have started, but not later.
2500 * Caching means that if the program heeds our dictate not to change
2501 * locales in threaded applications, this data will remain valid, and
9de6fe47
KW		 2502 * it may get queried without having to change locales. If the
2503 * environment is such that all categories have the same locale, this
2504 * isn't needed, as the code will not change the locale; but this
2505 * handles the uncommon case where the environment has disparate
2506 * locales for the categories */
47280b20
KW		 2507 (void) _is_cur_LC_category_utf8(categories[i]);
2508
2509# endif
2510
e5f10d49
KW		 2511 Safefree(curlocales[i]);
2512 }
b310b053 2513
7d4bcc4a
KW		 2514# if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
2515
49c85077 2516 /* Set PL_utf8locale to TRUE if using PerlIO _and_ the current LC_CTYPE
50bf02bd
KW		 2517 * locale is UTF-8. The call to new_ctype() just above has already
2518 * calculated the latter value and saved it in PL_in_utf8_CTYPE_locale. If
2519 * both PL_utf8locale and PL_unicode (set by -C or by $ENV{PERL_UNICODE})
2520 * are true, perl.c:S_parse_body() will turn on the PerlIO :utf8 layer on
2521 * STDIN, STDOUT, STDERR, _and_ the default open discipline. */
2522 PL_utf8locale = PL_in_utf8_CTYPE_locale;
49c85077 2523
a05d7ebb 2524 /* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
fde18df1
JH		 2525 This is an alternative to using the -C command line switch
2526 (the -C if present will override this). */
2527 {
dd374669 2528 const char *p = PerlEnv_getenv("PERL_UNICODE");
a05d7ebb 2529 PL_unicode = p ? parse_unicode_opts(&p) : 0;
5a22a2bb
NC		 2530 if (PL_unicode & PERL_UNICODE_UTF8CACHEASSERT_FLAG)
2531 PL_utf8cache = -1;
b310b053
JH		 2532 }
2533
7d4bcc4a 2534# endif
7d4bcc4a
KW		 2535# ifdef __GLIBC__
2536
175c4cf9 2537 Safefree(language);
7d4bcc4a
KW		 2538
2539# endif
175c4cf9
KW		 2540
2541 Safefree(lc_all);
2542 Safefree(lang);
2543
e3305790 2544#endif /* USE_LOCALE */
2fcc0ca9 2545#ifdef DEBUGGING
7d4bcc4a 2546
2fcc0ca9 2547 /* So won't continue to output stuff */
27cdc72e 2548 DEBUG_INITIALIZATION_set(FALSE);
7d4bcc4a 2549
2fcc0ca9
KW		 2550#endif
2551
98994639
HS		 2552 return ok;
2553}
2554
98994639
HS		 2555#ifdef USE_LOCALE_COLLATE
2556
a4a439fb 2557char *
a4a439fb
KW		 2558Perl__mem_collxfrm(pTHX_ const char *input_string,
2559 STRLEN len, /* Length of 'input_string' */
2560 STRLEN *xlen, /* Set to length of returned string
2561 (not including the collation index
2562 prefix) */
2563 bool utf8 /* Is the input in UTF-8? */
6696cfa7 2564 )
98994639 2565{
a4a439fb
KW		 2566
2567 /* _mem_collxfrm() is a bit like strxfrm() but with two important
2568 * differences. First, it handles embedded NULs. Second, it allocates a bit
2569 * more memory than needed for the transformed data itself. The real
55e5378d 2570 * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
a4a439fb
KW		 2571 * the length of that, and doesn't include the collation index size.
2572 * Please see sv_collxfrm() to see how this is used. */
2573
55e5378d
KW		 2574#define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
2575
6696cfa7
KW		 2576 char * s = (char *) input_string;
2577 STRLEN s_strlen = strlen(input_string);
79f120c8 2578 char *xbuf = NULL;
55e5378d 2579 STRLEN xAlloc; /* xalloc is a reserved word in VC */
17f41037 2580 STRLEN length_in_chars;
c664130f 2581 bool first_time = TRUE; /* Cleared after first loop iteration */
98994639 2582
a4a439fb
KW		 2583 PERL_ARGS_ASSERT__MEM_COLLXFRM;
2584
2585 /* Must be NUL-terminated */
2586 assert(*(input_string + len) == '\0');
7918f24d 2587
79f120c8
KW		 2588 /* If this locale has defective collation, skip */
2589 if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
c7202dee
KW		 2590 DEBUG_L(PerlIO_printf(Perl_debug_log,
2591 "_mem_collxfrm: locale's collation is defective\n"));
79f120c8
KW		 2592 goto bad;
2593 }
2594
6696cfa7
KW		 2595 /* Replace any embedded NULs with the control that sorts before any others.
2596 * This will give as good as possible results on strings that don't
2597 * otherwise contain that character, but otherwise there may be
2598 * less-than-perfect results with that character and NUL. This is
fdc080f3 2599 * unavoidable unless we replace strxfrm with our own implementation. */
fd43f63c
KW		 2600 if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded
2601 NUL */
6696cfa7
KW		 2602 char * e = s + len;
2603 char * sans_nuls;
fdc080f3 2604 STRLEN sans_nuls_len;
94762aa0 2605 int try_non_controls;
afc4976f
KW		 2606 char this_replacement_char[] = "?\0"; /* Room for a two-byte string,
2607 making sure 2nd byte is NUL.
2608 */
2609 STRLEN this_replacement_len;
2610
1e4c9676
KW		 2611 /* If we don't know what non-NUL control character sorts lowest for
2612 * this locale, find it */
f28f4d2a 2613 if (PL_strxfrm_NUL_replacement == '\0') {
6696cfa7 2614 int j;
afc4976f 2615 char * cur_min_x = NULL; /* The min_char's xfrm, (except it also
6696cfa7
KW		 2616 includes the collation index
2617 prefixed. */
2618
91c0e2e0 2619 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
94762aa0
KW		 2620
2621 /* Unlikely, but it may be that no control will work to replace
1e4c9676
KW		 2622 * NUL, in which case we instead look for any character. Controls
2623 * are preferred because collation order is, in general, context
2624 * sensitive, with adjoining characters affecting the order, and
2625 * controls are less likely to have such interactions, allowing the
2626 * NUL-replacement to stand on its own. (Another way to look at it
2627 * is to imagine what would happen if the NUL were replaced by a
2628 * combining character; it wouldn't work out all that well.) */
94762aa0
KW		 2629 for (try_non_controls = 0;
2630 try_non_controls < 2;
2631 try_non_controls++)
2632 {
d4ff9586
KW		 2633 /* Look through all legal code points (NUL isn't) */
2634 for (j = 1; j < 256; j++) {
2635 char * x; /* j's xfrm plus collation index */
2636 STRLEN x_len; /* length of 'x' */
2637 STRLEN trial_len = 1;
736a4fed 2638 char cur_source[] = { '\0', '\0' };
d4ff9586 2639
736a4fed
KW		 2640 /* Skip non-controls the first time through the loop. The
2641 * controls in a UTF-8 locale are the L1 ones */
afc4976f
KW		 2642 if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
2643 ? ! isCNTRL_L1(j)
2644 : ! isCNTRL_LC(j))
2645 {
d4ff9586 2646 continue;
6696cfa7 2647 }
6696cfa7 2648
736a4fed
KW		 2649 /* Create a 1-char string of the current code point */
2650 cur_source[0] = (char) j;
2651
d4ff9586
KW		 2652 /* Then transform it */
2653 x = _mem_collxfrm(cur_source, trial_len, &x_len,
afc4976f 2654 0 /* The string is not in UTF-8 */);
6696cfa7 2655
1e4c9676 2656 /* Ignore any character that didn't successfully transform.
d4ff9586
KW		 2657 * */
2658 if (! x) {
2659 continue;
2660 }
6696cfa7 2661
d4ff9586
KW		 2662 /* If this character's transformation is lower than
2663 * the current lowest, this one becomes the lowest */
2664 if ( cur_min_x == NULL
2665 || strLT(x + COLLXFRM_HDR_LEN,
2666 cur_min_x + COLLXFRM_HDR_LEN))
2667 {
f28f4d2a 2668 PL_strxfrm_NUL_replacement = j;
d4ff9586 2669 cur_min_x = x;
d4ff9586
KW		 2670 }
2671 else {
2672 Safefree(x);
2673 }
1e4c9676 2674 } /* end of loop through all 255 characters */
6696cfa7 2675
1e4c9676 2676 /* Stop looking if found */
94762aa0
KW		 2677 if (cur_min_x) {
2678 break;
2679 }
2680
2681 /* Unlikely, but possible, if there aren't any controls that
2682 * work in the locale, repeat the loop, looking for any
2683 * character that works */
2684 DEBUG_L(PerlIO_printf(Perl_debug_log,
2685 "_mem_collxfrm: No control worked. Trying non-controls\n"));
1e4c9676 2686 } /* End of loop to try first the controls, then any char */
6696cfa7 2687
94762aa0
KW		 2688 if (! cur_min_x) {
2689 DEBUG_L(PerlIO_printf(Perl_debug_log,
2690 "_mem_collxfrm: Couldn't find any character to replace"
2691 " embedded NULs in locale %s with", PL_collation_name));
2692 goto bad;
58eebef2
KW		 2693 }
2694
94762aa0
KW		 2695 DEBUG_L(PerlIO_printf(Perl_debug_log,
2696 "_mem_collxfrm: Replacing embedded NULs in locale %s with "
f28f4d2a 2697 "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement));
94762aa0 2698
6696cfa7 2699 Safefree(cur_min_x);
1e4c9676 2700 } /* End of determining the character that is to replace NULs */
afc4976f
KW		 2701
2702 /* If the replacement is variant under UTF-8, it must match the
291a84fb 2703 * UTF8-ness of the original */
f28f4d2a
KW		 2704 if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) {
2705 this_replacement_char[0] =
2706 UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement);
2707 this_replacement_char[1] =
2708 UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement);
afc4976f
KW		 2709 this_replacement_len = 2;
2710 }
2711 else {
f28f4d2a 2712 this_replacement_char[0] = PL_strxfrm_NUL_replacement;
afc4976f
KW		 2713 /* this_replacement_char[1] = '\0' was done at initialization */
2714 this_replacement_len = 1;
6696cfa7
KW		 2715 }
2716
2717 /* The worst case length for the replaced string would be if every
2718 * character in it is NUL. Multiply that by the length of each
2719 * replacement, and allow for a trailing NUL */
afc4976f 2720 sans_nuls_len = (len * this_replacement_len) + 1;
fdc080f3 2721 Newx(sans_nuls, sans_nuls_len, char);
6696cfa7
KW		 2722 *sans_nuls = '\0';
2723
6696cfa7
KW		 2724 /* Replace each NUL with the lowest collating control. Loop until have
2725 * exhausted all the NULs */
2726 while (s + s_strlen < e) {
6069d6c5 2727 my_strlcat(sans_nuls, s, sans_nuls_len);
6696cfa7
KW		 2728
2729 /* Do the actual replacement */
6069d6c5 2730 my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len);
6696cfa7
KW		 2731
2732 /* Move past the input NUL */
2733 s += s_strlen + 1;
2734 s_strlen = strlen(s);
2735 }
2736
2737 /* And add anything that trails the final NUL */
6069d6c5 2738 my_strlcat(sans_nuls, s, sans_nuls_len);
6696cfa7
KW		 2739
2740 /* Switch so below we transform this modified string */
2741 s = sans_nuls;
2742 len = strlen(s);
1e4c9676 2743 } /* End of replacing NULs */
6696cfa7 2744
a4a439fb
KW		 2745 /* Make sure the UTF8ness of the string and locale match */
2746 if (utf8 != PL_in_utf8_COLLATE_locale) {
9de6fe47 2747 /* XXX convert above Unicode to 10FFFF? */
a4a439fb
KW		 2748 const char * const t = s; /* Temporary so we can later find where the
2749 input was */
2750
2751 /* Here they don't match. Change the string's to be what the locale is
2752 * expecting */
2753
2754 if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
2755 s = (char *) bytes_to_utf8((const U8 *) s, &len);
2756 utf8 = TRUE;
2757 }
2758 else { /* locale is not UTF-8; but input is; downgrade the input */
2759
2760 s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
2761
2762 /* If the downgrade was successful we are done, but if the input
2763 * contains things that require UTF-8 to represent, have to do
2764 * damage control ... */
2765 if (UNLIKELY(utf8)) {
2766
2767 /* What we do is construct a non-UTF-8 string with
2768 * 1) the characters representable by a single byte converted
2769 * to be so (if necessary);
2770 * 2) and the rest converted to collate the same as the
2771 * highest collating representable character. That makes
2772 * them collate at the end. This is similar to how we
2773 * handle embedded NULs, but we use the highest collating
2774 * code point instead of the smallest. Like the NUL case,
2775 * this isn't perfect, but is the best we can reasonably
2776 * do. Every above-255 code point will sort the same as
2777 * the highest-sorting 0-255 code point. If that code
2778 * point can combine in a sequence with some other code
2779 * points for weight calculations, us changing something to
2780 * be it can adversely affect the results. But in most
2781 * cases, it should work reasonably. And note that this is
2782 * really an illegal situation: using code points above 255
2783 * on a locale where only 0-255 are valid. If two strings
2784 * sort entirely equal, then the sort order for the
2785 * above-255 code points will be in code point order. */
2786
2787 utf8 = FALSE;
2788
2789 /* If we haven't calculated the code point with the maximum
2790 * collating order for this locale, do so now */
2791 if (! PL_strxfrm_max_cp) {
2792 int j;
2793
2794 /* The current transformed string that collates the
2795 * highest (except it also includes the prefixed collation
2796 * index. */
2797 char * cur_max_x = NULL;
2798
2799 /* Look through all legal code points (NUL isn't) */
2800 for (j = 1; j < 256; j++) {
2801 char * x;
2802 STRLEN x_len;
736a4fed 2803 char cur_source[] = { '\0', '\0' };
a4a439fb 2804
736a4fed
KW		 2805 /* Create a 1-char string of the current code point */
2806 cur_source[0] = (char) j;
a4a439fb
KW		 2807
2808 /* Then transform it */
2809 x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
2810
2811 /* If something went wrong (which it shouldn't), just
2812 * ignore this code point */
94762aa0 2813 if (! x) {
a4a439fb
KW		 2814 continue;
2815 }
2816
2817 /* If this character's transformation is higher than
2818 * the current highest, this one becomes the highest */
2819 if ( cur_max_x == NULL
55e5378d
KW		 2820 || strGT(x + COLLXFRM_HDR_LEN,
2821 cur_max_x + COLLXFRM_HDR_LEN))
a4a439fb
KW		 2822 {
2823 PL_strxfrm_max_cp = j;
2824 cur_max_x = x;
2825 }
2826 else {
2827 Safefree(x);
2828 }
2829 }
2830
94762aa0
KW		 2831 if (! cur_max_x) {
2832 DEBUG_L(PerlIO_printf(Perl_debug_log,
2833 "_mem_collxfrm: Couldn't find any character to"
2834 " replace above-Latin1 chars in locale %s with",
2835 PL_collation_name));
2836 goto bad;
2837 }
2838
58eebef2
KW		 2839 DEBUG_L(PerlIO_printf(Perl_debug_log,
2840 "_mem_collxfrm: highest 1-byte collating character"
2841 " in locale %s is 0x%02X\n",
2842 PL_collation_name,
2843 PL_strxfrm_max_cp));
58eebef2 2844
a4a439fb
KW		 2845 Safefree(cur_max_x);
2846 }
2847
2848 /* Here we know which legal code point collates the highest.
2849 * We are ready to construct the non-UTF-8 string. The length
2850 * will be at least 1 byte smaller than the input string
2851 * (because we changed at least one 2-byte character into a
2852 * single byte), but that is eaten up by the trailing NUL */
2853 Newx(s, len, char);
2854
2855 {
2856 STRLEN i;
2857 STRLEN d= 0;
042d9e50 2858 char * e = (char *) t + len;
a4a439fb
KW		 2859
2860 for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
2861 U8 cur_char = t[i];
2862 if (UTF8_IS_INVARIANT(cur_char)) {
2863 s[d++] = cur_char;
2864 }
042d9e50 2865 else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) {
a4a439fb
KW		 2866 s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
2867 }
2868 else { /* Replace illegal cp with highest collating
2869 one */
2870 s[d++] = PL_strxfrm_max_cp;
2871 }
2872 }
2873 s[d++] = '\0';
2874 Renew(s, d, char); /* Free up unused space */
2875 }
2876 }
2877 }
2878
2879 /* Here, we have constructed a modified version of the input. It could
2880 * be that we already had a modified copy before we did this version.
2881 * If so, that copy is no longer needed */
2882 if (t != input_string) {
2883 Safefree(t);
2884 }
2885 }
2886
17f41037
KW		 2887 length_in_chars = (utf8)
2888 ? utf8_length((U8 *) s, (U8 *) s + len)
2889 : len;
2890
59c018b9
KW		 2891 /* The first element in the output is the collation id, used by
2892 * sv_collxfrm(); then comes the space for the transformed string. The
2893 * equation should give us a good estimate as to how much is needed */
55e5378d 2894 xAlloc = COLLXFRM_HDR_LEN
a4a439fb 2895 + PL_collxfrm_base
17f41037 2896 + (PL_collxfrm_mult * length_in_chars);
a02a5408 2897 Newx(xbuf, xAlloc, char);
c7202dee
KW		 2898 if (UNLIKELY(! xbuf)) {
2899 DEBUG_L(PerlIO_printf(Perl_debug_log,
2900 "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
98994639 2901 goto bad;
c7202dee 2902 }
98994639 2903
d35fca5f 2904 /* Store the collation id */
98994639 2905 *(U32*)xbuf = PL_collation_ix;
d35fca5f
KW		 2906
2907 /* Then the transformation of the input. We loop until successful, or we
2908 * give up */
4ebeff16 2909 for (;;) {
1adab0a7 2910
55e5378d 2911 *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
4ebeff16
KW		 2912
2913 /* If the transformed string occupies less space than we told strxfrm()
2914 * was available, it means it successfully transformed the whole
2915 * string. */
55e5378d 2916 if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
17f41037 2917
1adab0a7
KW		 2918 /* Some systems include a trailing NUL in the returned length.
2919 * Ignore it, using a loop in case multiple trailing NULs are
2920 * returned. */
2921 while ( (*xlen) > 0
2922 && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
2923 {
2924 (*xlen)--;
2925 }
2926
17f41037
KW		 2927 /* If the first try didn't get it, it means our prediction was low.
2928 * Modify the coefficients so that we predict a larger value in any
2929 * future transformations */
2930 if (! first_time) {
2931 STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
2932 STRLEN computed_guess = PL_collxfrm_base
2933 + (PL_collxfrm_mult * length_in_chars);
e1c30f0c
KW		 2934
2935 /* On zero-length input, just keep current slope instead of
2936 * dividing by 0 */
2937 const STRLEN new_m = (length_in_chars != 0)
2938 ? needed / length_in_chars
2939 : PL_collxfrm_mult;
17f41037
KW		 2940
2941 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
b07929e4
KW		 2942 "%s: %d: initial size of %zu bytes for a length "
2943 "%zu string was insufficient, %zu needed\n",
17f41037 2944 __FILE__, __LINE__,
b07929e4 2945 computed_guess, length_in_chars, needed));
17f41037
KW		 2946
2947 /* If slope increased, use it, but discard this result for
2948 * length 1 strings, as we can't be sure that it's a real slope
2949 * change */
2950 if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
7d4bcc4a
KW		 2951
2952# ifdef DEBUGGING
2953
17f41037
KW		 2954 STRLEN old_m = PL_collxfrm_mult;
2955 STRLEN old_b = PL_collxfrm_base;
7d4bcc4a
KW		 2956
2957# endif
2958
17f41037
KW		 2959 PL_collxfrm_mult = new_m;
2960 PL_collxfrm_base = 1; /* +1 For trailing NUL */
2961 computed_guess = PL_collxfrm_base
2962 + (PL_collxfrm_mult * length_in_chars);
2963 if (computed_guess < needed) {
2964 PL_collxfrm_base += needed - computed_guess;
2965 }
2966
2967 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
b07929e4
KW		 2968 "%s: %d: slope is now %zu; was %zu, base "
2969 "is now %zu; was %zu\n",
17f41037 2970 __FILE__, __LINE__,
b07929e4
KW		 2971 PL_collxfrm_mult, old_m,
2972 PL_collxfrm_base, old_b));
17f41037
KW		 2973 }
2974 else { /* Slope didn't change, but 'b' did */
2975 const STRLEN new_b = needed
2976 - computed_guess
2977 + PL_collxfrm_base;
2978 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
b07929e4 2979 "%s: %d: base is now %zu; was %zu\n",
17f41037 2980 __FILE__, __LINE__,
b07929e4 2981 new_b, PL_collxfrm_base));
17f41037
KW		 2982 PL_collxfrm_base = new_b;
2983 }
2984 }
2985
4ebeff16
KW		 2986 break;
2987 }
bb0f664e 2988
c7202dee
KW		 2989 if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
2990 DEBUG_L(PerlIO_printf(Perl_debug_log,
2991 "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
2992 *xlen, PERL_INT_MAX));
4ebeff16 2993 goto bad;
c7202dee 2994 }
d35fca5f 2995
c664130f 2996 /* A well-behaved strxfrm() returns exactly how much space it needs
1adab0a7
KW		 2997 * (usually not including the trailing NUL) when it fails due to not
2998 * enough space being provided. Assume that this is the case unless
2999 * it's been proven otherwise */
c664130f 3000 if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
55e5378d 3001 xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
c664130f
KW		 3002 }
3003 else { /* Here, either:
3004 * 1) The strxfrm() has previously shown bad behavior; or
3005 * 2) It isn't the first time through the loop, which means
3006 * that the strxfrm() is now showing bad behavior, because
3007 * we gave it what it said was needed in the previous
3008 * iteration, and it came back saying it needed still more.
3009 * (Many versions of cygwin fit this. When the buffer size
3010 * isn't sufficient, they return the input size instead of
3011 * how much is needed.)
d4ff9586
KW		 3012 * Increase the buffer size by a fixed percentage and try again.
3013 * */
6ddd902c 3014 xAlloc += (xAlloc / 4) + 1;
c664130f 3015 PL_strxfrm_is_behaved = FALSE;
c664130f 3016
7d4bcc4a
KW		 3017# ifdef DEBUGGING
3018
58eebef2
KW		 3019 if (DEBUG_Lv_TEST || debug_initialization) {
3020 PerlIO_printf(Perl_debug_log,
3021 "_mem_collxfrm required more space than previously calculated"
b07929e4 3022 " for locale %s, trying again with new guess=%d+%zu\n",
58eebef2 3023 PL_collation_name, (int) COLLXFRM_HDR_LEN,
b07929e4 3024 xAlloc - COLLXFRM_HDR_LEN);
58eebef2 3025 }
7d4bcc4a
KW		 3026
3027# endif
3028
58eebef2 3029 }
c664130f 3030
4ebeff16 3031 Renew(xbuf, xAlloc, char);
c7202dee
KW		 3032 if (UNLIKELY(! xbuf)) {
3033 DEBUG_L(PerlIO_printf(Perl_debug_log,
3034 "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
4ebeff16 3035 goto bad;
c7202dee 3036 }
c664130f
KW		 3037
3038 first_time = FALSE;
4ebeff16 3039 }
98994639 3040
6696cfa7 3041
7d4bcc4a
KW		 3042# ifdef DEBUGGING
3043
58eebef2 3044 if (DEBUG_Lv_TEST || debug_initialization) {
c7202dee
KW		 3045
3046 print_collxfrm_input_and_return(s, s + len, xlen, utf8);
3047 PerlIO_printf(Perl_debug_log, "Its xfrm is:");
7e2f38b2
KW		 3048 PerlIO_printf(Perl_debug_log, "%s\n",
3049 _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN,
3050 *xlen, 1));
58eebef2 3051 }
7d4bcc4a
KW		 3052
3053# endif
58eebef2 3054
3c5f993e 3055 /* Free up unneeded space; retain ehough for trailing NUL */
55e5378d 3056 Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
98994639 3057
6696cfa7
KW		 3058 if (s != input_string) {
3059 Safefree(s);
98994639
HS		 3060 }
3061
98994639
HS		 3062 return xbuf;
3063
3064 bad:
3065 Safefree(xbuf);
6696cfa7
KW		 3066 if (s != input_string) {
3067 Safefree(s);
3068 }
98994639 3069 *xlen = 0;
7d4bcc4a
KW		 3070
3071# ifdef DEBUGGING
3072
58eebef2 3073 if (DEBUG_Lv_TEST || debug_initialization) {
c7202dee 3074 print_collxfrm_input_and_return(s, s + len, NULL, utf8);
58eebef2 3075 }
7d4bcc4a
KW		 3076
3077# endif
3078
98994639
HS		 3079 return NULL;
3080}
3081
7d4bcc4a 3082# ifdef DEBUGGING
c7202dee 3083
4cbaac56 3084STATIC void
c7202dee
KW		 3085S_print_collxfrm_input_and_return(pTHX_
3086 const char * const s,
3087 const char * const e,
3088 const STRLEN * const xlen,
3089 const bool is_utf8)
3090{
c7202dee
KW		 3091
3092 PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
3093
511e4ff7
DM		 3094 PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ",
3095 (UV)PL_collation_ix);
c7202dee 3096 if (xlen) {
08b6dc1d 3097 PerlIO_printf(Perl_debug_log, "%zu", *xlen);
c7202dee
KW		 3098 }
3099 else {
3100 PerlIO_printf(Perl_debug_log, "NULL");
3101 }
3102 PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
3103 PL_collation_name);
9c8a6dc2
KW		 3104 print_bytes_for_locale(s, e, is_utf8);
3105
3106 PerlIO_printf(Perl_debug_log, "'\n");
3107}
3108
3109STATIC void
3110S_print_bytes_for_locale(pTHX_
3111 const char * const s,
3112 const char * const e,
3113 const bool is_utf8)
3114{
3115 const char * t = s;
3116 bool prev_was_printable = TRUE;
3117 bool first_time = TRUE;
3118
3119 PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE;
c7202dee
KW		 3120
3121 while (t < e) {
3122 UV cp = (is_utf8)
3123 ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
3124 : * (U8 *) t;
3125 if (isPRINT(cp)) {
3126 if (! prev_was_printable) {
3127 PerlIO_printf(Perl_debug_log, " ");
3128 }
3129 PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
3130 prev_was_printable = TRUE;
3131 }
3132 else {
3133 if (! first_time) {
3134 PerlIO_printf(Perl_debug_log, " ");
3135 }
147e3846 3136 PerlIO_printf(Perl_debug_log, "%02" UVXf, cp);
c7202dee
KW		 3137 prev_was_printable = FALSE;
3138 }
3139 t += (is_utf8) ? UTF8SKIP(t) : 1;
3140 first_time = FALSE;
3141 }
c7202dee
KW		 3142}
3143
7d4bcc4a 3144# endif /* #ifdef DEBUGGING */
98994639 3145#endif /* USE_LOCALE_COLLATE */
58eebef2 3146
8ef6e574
KW		 3147#ifdef USE_LOCALE
3148
962aa53f
KW		 3149STATIC const char *
3150S_switch_category_locale_to_template(pTHX_ const int switch_category, const int template_category, const char * template_locale)
3151{
3152 /* Changes the locale for LC_'switch_category" to that of
3153 * LC_'template_category', if they aren't already the same. If not NULL,
3154 * 'template_locale' is the locale that 'template_category' is in.
3155 *
9de6fe47
KW		 3156 * Returns a copy of the name of the original locale for 'switch_category'
3157 * so can be switched back to with the companion function
3158 * restore_switched_locale(), (NULL if no restoral is necessary.) */
962aa53f
KW		 3159
3160 char * restore_to_locale = NULL;
3161
3162 if (switch_category == template_category) { /* No changes needed */
3163 return NULL;
3164 }
3165
3166 /* Find the original locale of the category we may need to change, so that
3167 * it can be restored to later */
d6189047
KW		 3168 restore_to_locale = stdize_locale(savepv(do_setlocale_r(switch_category,
3169 NULL)));
962aa53f
KW		 3170 if (! restore_to_locale) {
3171 Perl_croak(aTHX_
3172 "panic: %s: %d: Could not find current %s locale, errno=%d\n",
3173 __FILE__, __LINE__, category_name(switch_category), errno);
3174 }
962aa53f
KW		 3175
3176 /* If the locale of the template category wasn't passed in, find it now */
3177 if (template_locale == NULL) {
3178 template_locale = do_setlocale_r(template_category, NULL);
3179 if (! template_locale) {
3180 Perl_croak(aTHX_
3181 "panic: %s: %d: Could not find current %s locale, errno=%d\n",
3182 __FILE__, __LINE__, category_name(template_category), errno);
3183 }
3184 }
3185
3186 /* It the locales are the same, there's nothing to do */
3187 if (strEQ(restore_to_locale, template_locale)) {
3188 Safefree(restore_to_locale);
3189
3190 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s locale unchanged as %s\n",
3191 category_name(switch_category), restore_to_locale));
3192
3193 return NULL;
3194 }
3195
3196 /* Finally, change the locale to the template one */
3197 if (! do_setlocale_r(switch_category, template_locale)) {
3198 Perl_croak(aTHX_
3199 "panic: %s: %d: Could not change %s locale to %s, errno=%d\n",
3200 __FILE__, __LINE__, category_name(switch_category),
3201 template_locale, errno);
3202 }
3203
3204 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s locale switched to %s\n",
3205 category_name(switch_category), template_locale));
3206
3207 return restore_to_locale;
3208}
3209
3210STATIC void
3211S_restore_switched_locale(pTHX_ const int category, const char * const original_locale)
3212{
9de6fe47
KW		 3213 /* Restores the locale for LC_'category' to 'original_locale' (which is a
3214 * copy that will be freed by this function), or do nothing if the latter
3215 * parameter is NULL */
962aa53f
KW		 3216
3217 if (original_locale == NULL) {
3218 return;
3219 }
3220
3221 if (! do_setlocale_r(category, original_locale)) {
3222 Perl_croak(aTHX_
3223 "panic: %s: %d: setlocale %s restore to %s failed, errno=%d\n",
3224 __FILE__, __LINE__,
3225 category_name(category), original_locale, errno);
3226 }
3227
3228 Safefree(original_locale);
3229}
3230
c1284011
KW		 3231bool
3232Perl__is_cur_LC_category_utf8(pTHX_ int category)
7d74bb61
KW		 3233{
3234 /* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
3235 * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
119ee68b 3236 * nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
609548d2
KW		 3237 * could give the wrong result. The result will very likely be correct for
3238 * languages that have commonly used non-ASCII characters, but for notably
3239 * English, it comes down to if the locale's name ends in something like
9de6fe47
KW		 3240 * "UTF-8". It errs on the side of not being a UTF-8 locale.
3241 *
3242 * If the platform is early C89, not containing mbtowc(), or we are
3243 * compiled to not pay attention to LC_CTYPE, this employs heuristics.
3244 * These work very well for non-Latin locales or those whose currency
3245 * symbol isn't a '$' nor plain ASCII text. But without LC_CTYPE and at
3246 * least MB_CUR_MAX, English locales with an ASCII currency symbol depend
3247 * on the name containing UTF-8 or not. */
7d74bb61 3248
47280b20 3249 /* Name of current locale corresponding to the input category */
8de4332b 3250 const char *save_input_locale = NULL;
47280b20
KW		 3251
3252 bool is_utf8 = FALSE; /* The return value */
7d74bb61 3253
47280b20
KW		 3254 /* The variables below are for the cache of previous lookups using this
3255 * function. The cache is a C string, described at the definition for
3256 * 'C_and_POSIX_utf8ness'.
3257 *
3258 * The first part of the cache is fixed, for the C and POSIX locales. The
3259 * varying part starts just after them. */
3260 char * utf8ness_cache = PL_locale_utf8ness + STRLENs(C_and_POSIX_utf8ness);
3261
3262 Size_t utf8ness_cache_size; /* Size of the varying portion */
3263 Size_t input_name_len; /* Length in bytes of save_input_locale */
3264 Size_t input_name_len_with_overhead; /* plus extra chars used to store
3265 the name in the cache */
3266 char * delimited; /* The name plus the delimiters used to store
3267 it in the cache */
3268 char * name_pos; /* position of 'delimited' in the cache, or 0
3269 if not there */
3270
3271
7d4bcc4a
KW		 3272# ifdef LC_ALL
3273
7d74bb61 3274 assert(category != LC_ALL);
7d4bcc4a
KW		 3275
3276# endif
7d74bb61 3277
47280b20 3278 /* Get the desired category's locale */
d6189047 3279 save_input_locale = stdize_locale(savepv(do_setlocale_r(category, NULL)));
7d74bb61 3280 if (! save_input_locale) {
47280b20 3281 Perl_croak(aTHX_
d707d779
KW		 3282 "panic: %s: %d: Could not find current %s locale, errno=%d\n",
3283 __FILE__, __LINE__, category_name(category), errno);
7d74bb61 3284 }
47280b20 3285
47280b20
KW		 3286 DEBUG_L(PerlIO_printf(Perl_debug_log,
3287 "Current locale for %s is %s\n",
3288 category_name(category), save_input_locale));
3289
3290 input_name_len = strlen(save_input_locale);
3291
3292 /* In our cache, each name is accompanied by two delimiters and a single
3293 * utf8ness digit */
3294 input_name_len_with_overhead = input_name_len + 3;
3295
3296 /* Allocate and populate space for a copy of the name surrounded by the
3297 * delimiters */
3298 Newx(delimited, input_name_len_with_overhead, char);
3299 delimited[0] = UTF8NESS_SEP[0];
3300 Copy(save_input_locale, delimited + 1, input_name_len, char);
3301 delimited[input_name_len+1] = UTF8NESS_PREFIX[0];
3302 delimited[input_name_len+2] = '\0';
3303
3304 /* And see if that is in the cache */
3305 name_pos = instr(PL_locale_utf8ness, delimited);
3306 if (name_pos) {
3307 is_utf8 = *(name_pos + input_name_len_with_overhead - 1) - '0';
3308
3309# ifdef DEBUGGING
3310
3311 if (DEBUG_Lv_TEST || debug_initialization) {
3312 PerlIO_printf(Perl_debug_log, "UTF8ness for locale %s=%d, \n",
3313 save_input_locale, is_utf8);
3314 }
3315
3316# endif
3317
3318 /* And, if not already in that position, move it to the beginning of
3319 * the non-constant portion of the list, since it is the most recently
3320 * used. (We don't have to worry about overflow, since just moving
3321 * existing names around) */
3322 if (name_pos > utf8ness_cache) {
3323 Move(utf8ness_cache,
3324 utf8ness_cache + input_name_len_with_overhead,
3325 name_pos - utf8ness_cache, char);
3326 Copy(delimited,
3327 utf8ness_cache,
3328 input_name_len_with_overhead - 1, char);
3329 utf8ness_cache[input_name_len_with_overhead - 1] = is_utf8 + '0';
3330 }
3331
3332 Safefree(delimited);
b07fffd1 3333 Safefree(save_input_locale);
47280b20 3334 return is_utf8;
7d74bb61
KW		 3335 }
3336
47280b20
KW		 3337 /* Here we don't have stored the utf8ness for the input locale. We have to
3338 * calculate it */
3339
94646a69
KW		 3340# if defined(USE_LOCALE_CTYPE) \
3341 && ( (defined(HAS_NL_LANGINFO) && defined(CODESET)) \
3342 || (defined(HAS_MBTOWC) || defined(HAS_MBRTOWC)))
7d74bb61 3343
0dec74cd 3344 {
962aa53f
KW		 3345 const char *original_ctype_locale
3346 = switch_category_locale_to_template(LC_CTYPE,
3347 category,
3348 save_input_locale);
69014004 3349
7d74bb61 3350 /* Here the current LC_CTYPE is set to the locale of the category whose
0dec74cd 3351 * information is desired. This means that nl_langinfo() and mbtowc()
1d958db2 3352 * should give the correct results */
119ee68b 3353
0dec74cd
KW		 3354# ifdef MB_CUR_MAX /* But we can potentially rule out UTF-8ness, avoiding
3355 calling the functions if we have this */
3356
3357 /* Standard UTF-8 needs at least 4 bytes to represent the maximum
3358 * Unicode code point. */
3359
de3a3072
KW		 3360 DEBUG_L(PerlIO_printf(Perl_debug_log, "%s: %d: MB_CUR_MAX=%d\n",
3361 __FILE__, __LINE__, (int) MB_CUR_MAX));
0dec74cd
KW		 3362 if ((unsigned) MB_CUR_MAX < STRLENs(MAX_UNICODE_UTF8)) {
3363 is_utf8 = FALSE;
b5b2847c
KW		 3364 restore_switched_locale(LC_CTYPE, original_ctype_locale);
3365 goto finish_and_return;
0dec74cd
KW		 3366 }
3367
3368# endif
7d4bcc4a
KW		 3369# if defined(HAS_NL_LANGINFO) && defined(CODESET)
3370
0dec74cd
KW		 3371 { /* The task is easiest if the platform has this POSIX 2001 function.
3372 Except on some platforms it can wrongly return "", so have to have
3373 a fallback. And it can return that it's UTF-8, even if there are
3374 variances from that. For example, Turkish locales may use the
3375 alternate dotted I rules, and sometimes it appears to be a
3376 defective locale definition. XXX We should probably check for
9de6fe47
KW		 3377 these in the Latin1 range and warn (but on glibc, requires
3378 iswalnum() etc. due to their not handling 80-FF correctly */
c70a3e68
KW		 3379 const char *codeset = my_nl_langinfo(PERL_CODESET, FALSE);
3380 /* FALSE => already in dest locale */
119ee68b 3381
af84e886 3382 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
c70a3e68
KW		 3383 "\tnllanginfo returned CODESET '%s'\n", codeset));
3384
3385 if (codeset && strNE(codeset, "")) {
1d958db2 3386
1d075173
KW		 3387 /* If the implementation of foldEQ() somehow were
3388 * to change to not go byte-by-byte, this could
3389 * read past end of string, as only one length is
3390 * checked. But currently, a premature NUL will
3391 * compare false, and it will stop there */
3392 is_utf8 = cBOOL( foldEQ(codeset, STR_WITH_LEN("UTF-8"))
3393 || foldEQ(codeset, STR_WITH_LEN("UTF8")));
1d958db2 3394
69014004
KW		 3395 DEBUG_L(PerlIO_printf(Perl_debug_log,
3396 "\tnllanginfo returned CODESET '%s'; ?UTF8 locale=%d\n",
3397 codeset, is_utf8));
b5b2847c
KW		 3398 restore_switched_locale(LC_CTYPE, original_ctype_locale);
3399 goto finish_and_return;
1d958db2 3400 }
119ee68b
KW		 3401 }
3402
7d4bcc4a 3403# endif
94646a69 3404# if defined(HAS_MBTOWC) || defined(HAS_MBRTOWC)
0dec74cd
KW		 3405 /* We can see if this is a UTF-8-like locale if have mbtowc(). It was a
3406 * late adder to C89, so very likely to have it. However, testing has
3407 * shown that, like nl_langinfo() above, there are locales that are not
3408 * strictly UTF-8 that this will return that they are */
69014004 3409
0dec74cd 3410 {
119ee68b 3411 wchar_t wc;
51fc4b19
KW		 3412 int len;
3413
94646a69
KW		 3414# if defined(HAS_MBRTOWC) && defined(USE_ITHREADS)
3415
3416 mbstate_t ps;
3417
3418# endif
3419
3420 /* mbrtowc() and mbtowc() convert a byte string to a wide
3421 * character. Feed a byte string to one of them and check that the
3422 * result is the expected Unicode code point */
3423
3424# if defined(HAS_MBRTOWC) && defined(USE_ITHREADS)
3425 /* Prefer this function if available, as it's reentrant */
3426
3427 memset(&ps, 0, sizeof(ps));;
3428 PERL_UNUSED_RESULT(mbrtowc(&wc, NULL, 0, &ps)); /* Reset any shift
3429 state */
3430 errno = 0;
3431 len = mbrtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8), &ps);
3432
3433# else
0dec74cd 3434
856b881c 3435 PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
69014004 3436 errno = 0;
b1d4925c 3437 len = mbtowc(&wc, STR_WITH_LEN(REPLACEMENT_CHARACTER_UTF8));
51fc4b19 3438
94646a69
KW		 3439# endif
3440
af84e886 3441 DEBUG_Lv(PerlIO_printf(Perl_debug_log,
0dec74cd
KW		 3442 "\treturn from mbtowc; len=%d; code_point=%x; errno=%d\n",
3443 len, (unsigned int) wc, errno));
b1d4925c 3444
0dec74cd
KW		 3445 is_utf8 = cBOOL( len == STRLENs(REPLACEMENT_CHARACTER_UTF8)
3446 && wc == (wchar_t) UNICODE_REPLACEMENT);
119ee68b 3447 }
7d4bcc4a 3448
962aa53f 3449 restore_switched_locale(LC_CTYPE, original_ctype_locale);
47280b20 3450 goto finish_and_return;
7d74bb61 3451 }
119ee68b 3452
0dec74cd
KW		 3453# endif
3454# else
7d74bb61 3455
0dec74cd
KW		 3456 /* Here, we must have a C89 compiler that doesn't have mbtowc(). Next
3457 * try looking at the currency symbol to see if it disambiguates
3458 * things. Often that will be in the native script, and if the symbol
3459 * isn't in UTF-8, we know that the locale isn't. If it is non-ASCII
3460 * UTF-8, we infer that the locale is too, as the odds of a non-UTF8
3461 * string being valid UTF-8 are quite small */
fa9b773e 3462
660ee27b
KW		 3463# ifdef USE_LOCALE_MONETARY
3464
3465 /* If have LC_MONETARY, we can look at the currency symbol. Often that
3466 * will be in the native script. We do this one first because there is
3467 * just one string to examine, so potentially avoids work */
7d4bcc4a 3468
9db16864
KW		 3469 {
3470 const char *original_monetary_locale
660ee27b
KW		 3471 = switch_category_locale_to_template(LC_MONETARY,
3472 category,
3473 save_input_locale);
9db16864 3474 bool only_ascii = FALSE;
660ee27b
KW		 3475 const U8 * currency_string
3476 = (const U8 *) my_nl_langinfo(PERL_CRNCYSTR, FALSE);
3477 /* 2nd param not relevant for this item */
3478 const U8 * first_variant;
fa9b773e 3479
660ee27b
KW		 3480 assert( *currency_string == '-'
3481 || *currency_string == '+'
3482 || *currency_string == '.');
97f4de96 3483
660ee27b 3484 currency_string++;
fa9b773e 3485
660ee27b 3486 if (is_utf8_invariant_string_loc(currency_string, 0, &first_variant))
9db16864
KW		 3487 {
3488 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));
3489 only_ascii = TRUE;
3490 }
3491 else {
660ee27b 3492 is_utf8 = is_strict_utf8_string(first_variant, 0);
9db16864 3493 }
fa9b773e 3494
9db16864 3495 restore_switched_locale(LC_MONETARY, original_monetary_locale);
fa9b773e 3496
9db16864 3497 if (! only_ascii) {
fa9b773e 3498
9db16864
KW		 3499 /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
3500 * otherwise assume the locale is UTF-8 if and only if the symbol
3501 * is non-ascii UTF-8. */
af84e886 3502 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
9db16864
KW		 3503 save_input_locale, is_utf8));
3504 goto finish_and_return;
3505 }
13542a67 3506 }
fa9b773e 3507
660ee27b 3508# endif /* USE_LOCALE_MONETARY */
7d4bcc4a 3509# if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
15f7e74e 3510
9db16864
KW		 3511 /* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
3512 * the names of the months and weekdays, timezone, and am/pm indicator */
3513 {
3514 const char *original_time_locale
3515 = switch_category_locale_to_template(LC_TIME,
3516 category,
3517 save_input_locale);
3518 int hour = 10;
3519 bool is_dst = FALSE;
3520 int dom = 1;
3521 int month = 0;
3522 int i;
3523 char * formatted_time;
3524
3525 /* Here the current LC_TIME is set to the locale of the category
3526 * whose information is desired. Look at all the days of the week and
3527 * month names, and the timezone and am/pm indicator for UTF-8 variant
3528 * characters. The first such a one found will tell us if the locale
3529 * is UTF-8 or not */
3530
3531 for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
3532 formatted_time = my_strftime("%A %B %Z %p",
3533 0, 0, hour, dom, month, 2012 - 1900, 0, 0, is_dst);
3534 if ( ! formatted_time
3535 || is_utf8_invariant_string((U8 *) formatted_time, 0))
3536 {
15f7e74e 3537
9db16864
KW		 3538 /* Here, we didn't find a non-ASCII. Try the next time through
3539 * with the complemented dst and am/pm, and try with the next
3540 * weekday. After we have gotten all weekdays, try the next
3541 * month */
3542 is_dst = ! is_dst;
3543 hour = (hour + 12) % 24;
3544 dom++;
3545 if (i > 6) {
3546 month++;
3547 }
3548 continue;
15f7e74e 3549 }
9db16864
KW		 3550
3551 /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
3552 * false otherwise. But first, restore LC_TIME to its original
3553 * locale if we changed it */
3554 restore_switched_locale(LC_TIME, original_time_locale);
3555
af84e886 3556 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
9db16864
KW		 3557 save_input_locale,
3558 is_utf8_string((U8 *) formatted_time, 0)));
3559 is_utf8 = is_utf8_string((U8 *) formatted_time, 0);
3560 goto finish_and_return;
15f7e74e
KW		 3561 }
3562
9db16864
KW		 3563 /* Falling off the end of the loop indicates all the names were just
3564 * ASCII. Go on to the next test. If we changed it, restore LC_TIME
3565 * to its original locale */
962aa53f 3566 restore_switched_locale(LC_TIME, original_time_locale);
af84e886 3567 DEBUG_Lv(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));
15f7e74e
KW		 3568 }
3569
7d4bcc4a 3570# endif
15f7e74e 3571
7d4bcc4a 3572# if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
855aeb93 3573
9db16864
KW		 3574 /* This code is ifdefd out because it was found to not be necessary in testing
3575 * on our dromedary test machine, which has over 700 locales. There, this
3576 * added no value to looking at the currency symbol and the time strings. I
3577 * left it in so as to avoid rewriting it if real-world experience indicates
3578 * that dromedary is an outlier. Essentially, instead of returning abpve if we
3579 * haven't found illegal utf8, we continue on and examine all the strerror()
3580 * messages on the platform for utf8ness. If all are ASCII, we still don't
3581 * know the answer; but otherwise we have a pretty good indication of the
3582 * utf8ness. The reason this doesn't help much is that the messages may not
3583 * have been translated into the locale. The currency symbol and time strings
3584 * are much more likely to have been translated. */
3585 {
3586 int e;
3587 bool non_ascii = FALSE;
3588 const char *original_messages_locale
3589 = switch_category_locale_to_template(LC_MESSAGES,
3590 category,
3591 save_input_locale);
3592 const char * errmsg = NULL;
3593
3594 /* Here the current LC_MESSAGES is set to the locale of the category
3595 * whose information is desired. Look through all the messages. We
3596 * can't use Strerror() here because it may expand to code that
3597 * segfaults in miniperl */
3598
3599 for (e = 0; e <= sys_nerr; e++) {
3600 errno = 0;
3601 errmsg = sys_errlist[e];
3602 if (errno || !errmsg) {
3603 break;
3604 }
3605 errmsg = savepv(errmsg);
3606 if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
3607 non_ascii = TRUE;
3608 is_utf8 = is_utf8_string((U8 *) errmsg, 0);
3609 break;
3610 }
855aeb93 3611 }
9db16864 3612 Safefree(errmsg);
855aeb93 3613
9db16864 3614 restore_switched_locale(LC_MESSAGES, original_messages_locale);
855aeb93 3615
9db16864 3616 if (non_ascii) {
5857e934 3617
9db16864
KW		 3618 /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
3619 * any non-ascii means it is one; otherwise we assume it isn't */
af84e886 3620 DEBUG_Lv(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
9db16864
KW		 3621 save_input_locale,
3622 is_utf8));
3623 goto finish_and_return;
3624 }
855aeb93 3625
9db16864
KW		 3626 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));
3627 }
855aeb93 3628
7d4bcc4a 3629# endif
0dec74cd 3630# ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
92c0a900 3631 UTF-8 locale */
7d4bcc4a 3632
97f4de96
KW		 3633 /* As a last resort, look at the locale name to see if it matches
3634 * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
3635 * return of setlocale(), is actually defined to be opaque, so we can't
3636 * really rely on the absence of various substrings in the name to indicate
3637 * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
3638 * be a UTF-8 locale. Similarly for the other common names */
3639
0dec74cd 3640 {
c36d8df8 3641 const Size_t final_pos = strlen(save_input_locale) - 1;
0dec74cd 3642
c36d8df8
KW		 3643 if (final_pos >= 3) {
3644 const char *name = save_input_locale;
97f4de96 3645
c36d8df8
KW		 3646 /* Find next 'U' or 'u' and look from there */
3647 while ((name += strcspn(name, "Uu") + 1)
3648 <= save_input_locale + final_pos - 2)
97f4de96 3649 {
c36d8df8
KW		 3650 if ( isALPHA_FOLD_NE(*name, 't')
3651 || isALPHA_FOLD_NE(*(name + 1), 'f'))
3652 {
3653 continue;
3654 }
3655 name += 2;
3656 if (*(name) == '-') {
3657 if ((name > save_input_locale + final_pos - 1)) {
3658 break;
3659 }
3660 name++;
3661 }
3662 if (*(name) == '8') {
3663 DEBUG_L(PerlIO_printf(Perl_debug_log,
3664 "Locale %s ends with UTF-8 in name\n",
3665 save_input_locale));
3666 is_utf8 = TRUE;
3667 goto finish_and_return;
97f4de96 3668 }
97f4de96 3669 }
c36d8df8
KW		 3670 DEBUG_L(PerlIO_printf(Perl_debug_log,
3671 "Locale %s doesn't end with UTF-8 in name\n",
3672 save_input_locale));
97f4de96 3673 }
97f4de96 3674
4d8d465a 3675# ifdef WIN32
7d4bcc4a 3676
c36d8df8
KW		 3677 /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
3678 if (memENDs(save_input_locale, final_pos, "65001")) {
3679 DEBUG_L(PerlIO_printf(Perl_debug_log,
8a0832a1 3680 "Locale %s ends with 65001 in name, is UTF-8 locale\n",
97f4de96 3681 save_input_locale));
c36d8df8
KW		 3682 is_utf8 = TRUE;
3683 goto finish_and_return;
3684 }
0dec74cd 3685 }
7d4bcc4a 3686
4d8d465a
KW		 3687# endif
3688# endif
97f4de96
KW		 3689
3690 /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
3691 * since we are about to return FALSE anyway, there is no point in doing
3692 * this extra work */
7d4bcc4a 3693
0dec74cd 3694# if 0
97f4de96
KW		 3695 if (instr(save_input_locale, "8859")) {
3696 DEBUG_L(PerlIO_printf(Perl_debug_log,
3697 "Locale %s has 8859 in name, not UTF-8 locale\n",
3698 save_input_locale));
47280b20
KW		 3699 is_utf8 = FALSE;
3700 goto finish_and_return;
97f4de96 3701 }
0dec74cd 3702# endif
97f4de96 3703
69014004
KW		 3704 DEBUG_L(PerlIO_printf(Perl_debug_log,
3705 "Assuming locale %s is not a UTF-8 locale\n",
3706 save_input_locale));
47280b20
KW		 3707 is_utf8 = FALSE;
3708
0dec74cd
KW		 3709# endif /* the code that is compiled when no modern LC_CTYPE */
3710
47280b20
KW		 3711 finish_and_return:
3712
3713 /* Cache this result so we don't have to go through all this next time. */
3714 utf8ness_cache_size = sizeof(PL_locale_utf8ness)
3715 - (utf8ness_cache - PL_locale_utf8ness);
3716
3717 /* But we can't save it if it is too large for the total space available */
3718 if (LIKELY(input_name_len_with_overhead < utf8ness_cache_size)) {
3719 Size_t utf8ness_cache_len = strlen(utf8ness_cache);
3720
3721 /* Here it can fit, but we may need to clear out the oldest cached
3722 * result(s) to do so. Check */
3723 if (utf8ness_cache_len + input_name_len_with_overhead
3724 >= utf8ness_cache_size)
3725 {
3726 /* Here we have to clear something out to make room for this.
3727 * Start looking at the rightmost place where it could fit and find
3728 * the beginning of the entry that extends past that. */
3729 char * cutoff = (char *) my_memrchr(utf8ness_cache,
3730 UTF8NESS_SEP[0],
3731 utf8ness_cache_size
3732 - input_name_len_with_overhead);
3733
3734 assert(cutoff);
3735 assert(cutoff >= utf8ness_cache);
3736
3737 /* This and all subsequent entries must be removed */
3738 *cutoff = '\0';
3739 utf8ness_cache_len = strlen(utf8ness_cache);
3740 }
3741
3742 /* Make space for the new entry */
3743 Move(utf8ness_cache,
3744 utf8ness_cache + input_name_len_with_overhead,
3745 utf8ness_cache_len + 1 /* Incl. trailing NUL */, char);