Line data Source code
1 : /*-----------------------------------------------------------------------
2 : *
3 : * PostgreSQL locale utilities
4 : *
5 : * Portions Copyright (c) 2002-2026, PostgreSQL Global Development Group
6 : *
7 : * src/backend/utils/adt/pg_locale.c
8 : *
9 : *-----------------------------------------------------------------------
10 : */
11 :
12 : /*----------
13 : * Here is how the locale stuff is handled: LC_COLLATE and LC_CTYPE
14 : * are fixed at CREATE DATABASE time, stored in pg_database, and cannot
15 : * be changed. Thus, the effects of strcoll(), strxfrm(), isupper(),
16 : * toupper(), etc. are always in the same fixed locale.
17 : *
18 : * LC_MESSAGES is settable at run time and will take effect
19 : * immediately.
20 : *
21 : * The other categories, LC_MONETARY, LC_NUMERIC, and LC_TIME are
22 : * permanently set to "C", and then we use temporary locale_t
23 : * objects when we need to look up locale data based on the GUCs
24 : * of the same name. Information is cached when the GUCs change.
25 : * The cached information is only used by the formatting functions
26 : * (to_char, etc.) and the money type. For the user, this should all be
27 : * transparent.
28 : *----------
29 : */
30 :
31 :
32 : #include "postgres.h"
33 :
34 : #include <time.h>
35 : #ifdef USE_ICU
36 : #include <unicode/ucol.h>
37 : #endif
38 :
39 : #include "access/htup_details.h"
40 : #include "catalog/pg_collation.h"
41 : #include "catalog/pg_database.h"
42 : #include "common/hashfn.h"
43 : #include "common/string.h"
44 : #include "mb/pg_wchar.h"
45 : #include "miscadmin.h"
46 : #include "utils/builtins.h"
47 : #include "utils/guc_hooks.h"
48 : #include "utils/lsyscache.h"
49 : #include "utils/memutils.h"
50 : #include "utils/pg_locale.h"
51 : #include "utils/pg_locale_c.h"
52 : #include "utils/relcache.h"
53 : #include "utils/syscache.h"
54 :
55 : #ifdef WIN32
56 : #include <shlwapi.h>
57 : #endif
58 :
59 : /* Error triggered for locale-sensitive subroutines */
60 : #define PGLOCALE_SUPPORT_ERROR(provider) \
61 : elog(ERROR, "unsupported collprovider for %s: %c", __func__, provider)
62 :
63 : /*
64 : * This should be large enough that most strings will fit, but small enough
65 : * that we feel comfortable putting it on the stack
66 : */
67 : #define TEXTBUFLEN 1024
68 :
69 : #define MAX_L10N_DATA 80
70 :
71 : /* pg_locale_builtin.c */
72 : extern pg_locale_t create_pg_locale_builtin(Oid collid, MemoryContext context);
73 : extern char *get_collation_actual_version_builtin(const char *collcollate);
74 :
75 : /* pg_locale_icu.c */
76 : #ifdef USE_ICU
77 : extern UCollator *pg_ucol_open(const char *loc_str);
78 : extern char *get_collation_actual_version_icu(const char *collcollate);
79 : #endif
80 : extern pg_locale_t create_pg_locale_icu(Oid collid, MemoryContext context);
81 :
82 : /* pg_locale_libc.c */
83 : extern pg_locale_t create_pg_locale_libc(Oid collid, MemoryContext context);
84 : extern char *get_collation_actual_version_libc(const char *collcollate);
85 :
86 : /* GUC settings */
87 : char *locale_messages;
88 : char *locale_monetary;
89 : char *locale_numeric;
90 : char *locale_time;
91 :
92 : int icu_validation_level = WARNING;
93 :
94 : /*
95 : * lc_time localization cache.
96 : *
97 : * We use only the first 7 or 12 entries of these arrays. The last array
98 : * element is left as NULL for the convenience of outside code that wants
99 : * to sequentially scan these arrays.
100 : */
101 : char *localized_abbrev_days[7 + 1];
102 : char *localized_full_days[7 + 1];
103 : char *localized_abbrev_months[12 + 1];
104 : char *localized_full_months[12 + 1];
105 :
106 : static pg_locale_t default_locale = NULL;
107 :
108 : /* indicates whether locale information cache is valid */
109 : static bool CurrentLocaleConvValid = false;
110 : static bool CurrentLCTimeValid = false;
111 :
112 : static struct pg_locale_struct c_locale = {
113 : .deterministic = true,
114 : .collate_is_c = true,
115 : .ctype_is_c = true,
116 : };
117 :
118 : /* Cache for collation-related knowledge */
119 :
120 : typedef struct
121 : {
122 : Oid collid; /* hash key: pg_collation OID */
123 : pg_locale_t locale; /* locale_t struct, or 0 if not valid */
124 :
125 : /* needed for simplehash */
126 : uint32 hash;
127 : char status;
128 : } collation_cache_entry;
129 :
130 : #define SH_PREFIX collation_cache
131 : #define SH_ELEMENT_TYPE collation_cache_entry
132 : #define SH_KEY_TYPE Oid
133 : #define SH_KEY collid
134 : #define SH_HASH_KEY(tb, key) murmurhash32((uint32) key)
135 : #define SH_EQUAL(tb, a, b) (a == b)
136 : #define SH_GET_HASH(tb, a) a->hash
137 : #define SH_SCOPE static inline
138 : #define SH_STORE_HASH
139 : #define SH_DECLARE
140 : #define SH_DEFINE
141 : #include "lib/simplehash.h"
142 :
143 : static MemoryContext CollationCacheContext = NULL;
144 : static collation_cache_hash *CollationCache = NULL;
145 :
146 : /*
147 : * The collation cache is often accessed repeatedly for the same collation, so
148 : * remember the last one used.
149 : */
150 : static Oid last_collation_cache_oid = InvalidOid;
151 : static pg_locale_t last_collation_cache_locale = NULL;
152 :
153 : #if defined(WIN32) && defined(LC_MESSAGES)
154 : static char *IsoLocaleName(const char *);
155 : #endif
156 :
157 : /*
158 : * pg_perm_setlocale
159 : *
160 : * This wraps the libc function setlocale(), with two additions. First, when
161 : * changing LC_CTYPE, update gettext's encoding for the current message
162 : * domain. GNU gettext automatically tracks LC_CTYPE on most platforms, but
163 : * not on Windows. Second, if the operation is successful, the corresponding
164 : * LC_XXX environment variable is set to match. By setting the environment
165 : * variable, we ensure that any subsequent use of setlocale(..., "") will
166 : * preserve the settings made through this routine. Of course, LC_ALL must
167 : * also be unset to fully ensure that, but that has to be done elsewhere after
168 : * all the individual LC_XXX variables have been set correctly. (Thank you
169 : * Perl for making this kluge necessary.)
170 : */
171 : char *
172 76252 : pg_perm_setlocale(int category, const char *locale)
173 : {
174 : char *result;
175 : const char *envvar;
176 :
177 : #ifndef WIN32
178 76252 : result = setlocale(category, locale);
179 : #else
180 :
181 : /*
182 : * On Windows, setlocale(LC_MESSAGES) does not work, so just assume that
183 : * the given value is good and set it in the environment variables. We
184 : * must ignore attempts to set to "", which means "keep using the old
185 : * environment value".
186 : */
187 : #ifdef LC_MESSAGES
188 : if (category == LC_MESSAGES)
189 : {
190 : result = (char *) locale;
191 : if (locale == NULL || locale[0] == '\0')
192 : return result;
193 : }
194 : else
195 : #endif
196 : result = setlocale(category, locale);
197 : #endif /* WIN32 */
198 :
199 76252 : if (result == NULL)
200 0 : return result; /* fall out immediately on failure */
201 :
202 : /*
203 : * Use the right encoding in translated messages. Under ENABLE_NLS, let
204 : * pg_bind_textdomain_codeset() figure it out. Under !ENABLE_NLS, message
205 : * format strings are ASCII, but database-encoding strings may enter the
206 : * message via %s. This makes the overall message encoding equal to the
207 : * database encoding.
208 : */
209 76252 : if (category == LC_CTYPE)
210 : {
211 : static char save_lc_ctype[LOCALE_NAME_BUFLEN];
212 :
213 : /* copy setlocale() return value before callee invokes it again */
214 37568 : strlcpy(save_lc_ctype, result, sizeof(save_lc_ctype));
215 37568 : result = save_lc_ctype;
216 :
217 : #ifdef ENABLE_NLS
218 37568 : SetMessageEncoding(pg_bind_textdomain_codeset(textdomain(NULL)));
219 : #else
220 : SetMessageEncoding(GetDatabaseEncoding());
221 : #endif
222 : }
223 :
224 76252 : switch (category)
225 : {
226 3964 : case LC_COLLATE:
227 3964 : envvar = "LC_COLLATE";
228 3964 : break;
229 37568 : case LC_CTYPE:
230 37568 : envvar = "LC_CTYPE";
231 37568 : break;
232 : #ifdef LC_MESSAGES
233 22828 : case LC_MESSAGES:
234 22828 : envvar = "LC_MESSAGES";
235 : #ifdef WIN32
236 : result = IsoLocaleName(locale);
237 : if (result == NULL)
238 : result = (char *) locale;
239 : elog(DEBUG3, "IsoLocaleName() executed; locale: \"%s\"", result);
240 : #endif /* WIN32 */
241 22828 : break;
242 : #endif /* LC_MESSAGES */
243 3964 : case LC_MONETARY:
244 3964 : envvar = "LC_MONETARY";
245 3964 : break;
246 3964 : case LC_NUMERIC:
247 3964 : envvar = "LC_NUMERIC";
248 3964 : break;
249 3964 : case LC_TIME:
250 3964 : envvar = "LC_TIME";
251 3964 : break;
252 0 : default:
253 0 : elog(FATAL, "unrecognized LC category: %d", category);
254 : return NULL; /* keep compiler quiet */
255 : }
256 :
257 76252 : if (setenv(envvar, result, 1) != 0)
258 0 : return NULL;
259 :
260 76252 : return result;
261 : }
262 :
263 :
264 : /*
265 : * Is the locale name valid for the locale category?
266 : *
267 : * If successful, and canonname isn't NULL, a palloc'd copy of the locale's
268 : * canonical name is stored there. This is especially useful for figuring out
269 : * what locale name "" means (ie, the server environment value). (Actually,
270 : * it seems that on most implementations that's the only thing it's good for;
271 : * we could wish that setlocale gave back a canonically spelled version of
272 : * the locale name, but typically it doesn't.)
273 : */
274 : bool
275 72596 : check_locale(int category, const char *locale, char **canonname)
276 : {
277 : char *save;
278 : char *res;
279 :
280 : /* Don't let Windows' non-ASCII locale names in. */
281 72596 : if (!pg_is_ascii(locale))
282 : {
283 0 : ereport(WARNING,
284 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
285 : errmsg("locale name \"%s\" contains non-ASCII characters",
286 : locale)));
287 0 : return false;
288 : }
289 :
290 72596 : if (canonname)
291 1614 : *canonname = NULL; /* in case of failure */
292 :
293 72596 : save = setlocale(category, NULL);
294 72596 : if (!save)
295 0 : return false; /* won't happen, we hope */
296 :
297 : /* save may be pointing at a modifiable scratch variable, see above. */
298 72596 : save = pstrdup(save);
299 :
300 : /* set the locale with setlocale, to see if it accepts it. */
301 72596 : res = setlocale(category, locale);
302 :
303 : /* save canonical name if requested. */
304 72596 : if (res && canonname)
305 1610 : *canonname = pstrdup(res);
306 :
307 : /* restore old value. */
308 72596 : if (!setlocale(category, save))
309 0 : elog(WARNING, "failed to restore old locale \"%s\"", save);
310 72596 : pfree(save);
311 :
312 : /* Don't let Windows' non-ASCII locale names out. */
313 72596 : if (canonname && *canonname && !pg_is_ascii(*canonname))
314 : {
315 0 : ereport(WARNING,
316 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
317 : errmsg("locale name \"%s\" contains non-ASCII characters",
318 : *canonname)));
319 0 : pfree(*canonname);
320 0 : *canonname = NULL;
321 0 : return false;
322 : }
323 :
324 72596 : return (res != NULL);
325 : }
326 :
327 :
328 : /*
329 : * GUC check/assign hooks
330 : *
331 : * For most locale categories, the assign hook doesn't actually set the locale
332 : * permanently, just reset flags so that the next use will cache the
333 : * appropriate values. (See explanation at the top of this file.)
334 : *
335 : * Note: we accept value = "" as selecting the postmaster's environment
336 : * value, whatever it was (so long as the environment setting is legal).
337 : * This will have been locked down by an earlier call to pg_perm_setlocale.
338 : */
339 : bool
340 19034 : check_locale_monetary(char **newval, void **extra, GucSource source)
341 : {
342 19034 : return check_locale(LC_MONETARY, *newval, NULL);
343 : }
344 :
345 : void
346 18820 : assign_locale_monetary(const char *newval, void *extra)
347 : {
348 18820 : CurrentLocaleConvValid = false;
349 18820 : }
350 :
351 : bool
352 19040 : check_locale_numeric(char **newval, void **extra, GucSource source)
353 : {
354 19040 : return check_locale(LC_NUMERIC, *newval, NULL);
355 : }
356 :
357 : void
358 18832 : assign_locale_numeric(const char *newval, void *extra)
359 : {
360 18832 : CurrentLocaleConvValid = false;
361 18832 : }
362 :
363 : bool
364 19034 : check_locale_time(char **newval, void **extra, GucSource source)
365 : {
366 19034 : return check_locale(LC_TIME, *newval, NULL);
367 : }
368 :
369 : void
370 18820 : assign_locale_time(const char *newval, void *extra)
371 : {
372 18820 : CurrentLCTimeValid = false;
373 18820 : }
374 :
375 : /*
376 : * We allow LC_MESSAGES to actually be set globally.
377 : *
378 : * Note: we normally disallow value = "" because it wouldn't have consistent
379 : * semantics (it'd effectively just use the previous value). However, this
380 : * is the value passed for PGC_S_DEFAULT, so don't complain in that case,
381 : * not even if the attempted setting fails due to invalid environment value.
382 : * The idea there is just to accept the environment setting *if possible*
383 : * during startup, until we can read the proper value from postgresql.conf.
384 : */
385 : bool
386 19092 : check_locale_messages(char **newval, void **extra, GucSource source)
387 : {
388 19092 : if (**newval == '\0')
389 : {
390 5218 : if (source == PGC_S_DEFAULT)
391 5218 : return true;
392 : else
393 0 : return false;
394 : }
395 :
396 : /*
397 : * LC_MESSAGES category does not exist everywhere, but accept it anyway
398 : *
399 : * On Windows, we can't even check the value, so accept blindly
400 : */
401 : #if defined(LC_MESSAGES) && !defined(WIN32)
402 13874 : return check_locale(LC_MESSAGES, *newval, NULL);
403 : #else
404 : return true;
405 : #endif
406 : }
407 :
408 : void
409 18864 : assign_locale_messages(const char *newval, void *extra)
410 : {
411 : /*
412 : * LC_MESSAGES category does not exist everywhere, but accept it anyway.
413 : * We ignore failure, as per comment above.
414 : */
415 : #ifdef LC_MESSAGES
416 18864 : (void) pg_perm_setlocale(LC_MESSAGES, newval);
417 : #endif
418 18864 : }
419 :
420 :
421 : /*
422 : * Frees the malloced content of a struct lconv. (But not the struct
423 : * itself.) It's important that this not throw elog(ERROR).
424 : */
425 : static void
426 6 : free_struct_lconv(struct lconv *s)
427 : {
428 6 : free(s->decimal_point);
429 6 : free(s->thousands_sep);
430 6 : free(s->grouping);
431 6 : free(s->int_curr_symbol);
432 6 : free(s->currency_symbol);
433 6 : free(s->mon_decimal_point);
434 6 : free(s->mon_thousands_sep);
435 6 : free(s->mon_grouping);
436 6 : free(s->positive_sign);
437 6 : free(s->negative_sign);
438 6 : }
439 :
440 : /*
441 : * Check that all fields of a struct lconv (or at least, the ones we care
442 : * about) are non-NULL. The field list must match free_struct_lconv().
443 : */
444 : static bool
445 56 : struct_lconv_is_valid(struct lconv *s)
446 : {
447 56 : if (s->decimal_point == NULL)
448 0 : return false;
449 56 : if (s->thousands_sep == NULL)
450 0 : return false;
451 56 : if (s->grouping == NULL)
452 0 : return false;
453 56 : if (s->int_curr_symbol == NULL)
454 0 : return false;
455 56 : if (s->currency_symbol == NULL)
456 0 : return false;
457 56 : if (s->mon_decimal_point == NULL)
458 0 : return false;
459 56 : if (s->mon_thousands_sep == NULL)
460 0 : return false;
461 56 : if (s->mon_grouping == NULL)
462 0 : return false;
463 56 : if (s->positive_sign == NULL)
464 0 : return false;
465 56 : if (s->negative_sign == NULL)
466 0 : return false;
467 56 : return true;
468 : }
469 :
470 :
471 : /*
472 : * Convert the strdup'd string at *str from the specified encoding to the
473 : * database encoding.
474 : */
475 : static void
476 448 : db_encoding_convert(int encoding, char **str)
477 : {
478 : char *pstr;
479 : char *mstr;
480 :
481 : /* convert the string to the database encoding */
482 448 : pstr = pg_any_to_server(*str, strlen(*str), encoding);
483 448 : if (pstr == *str)
484 448 : return; /* no conversion happened */
485 :
486 : /* need it malloc'd not palloc'd */
487 0 : mstr = strdup(pstr);
488 0 : if (mstr == NULL)
489 0 : ereport(ERROR,
490 : (errcode(ERRCODE_OUT_OF_MEMORY),
491 : errmsg("out of memory")));
492 :
493 : /* replace old string */
494 0 : free(*str);
495 0 : *str = mstr;
496 :
497 0 : pfree(pstr);
498 : }
499 :
500 :
501 : /*
502 : * Return the POSIX lconv struct (contains number/money formatting
503 : * information) with locale information for all categories.
504 : */
505 : struct lconv *
506 2966 : PGLC_localeconv(void)
507 : {
508 : static struct lconv CurrentLocaleConv;
509 : static bool CurrentLocaleConvAllocated = false;
510 : struct lconv *extlconv;
511 : struct lconv tmp;
512 2966 : struct lconv worklconv = {0};
513 :
514 : /* Did we do it already? */
515 2966 : if (CurrentLocaleConvValid)
516 2910 : return &CurrentLocaleConv;
517 :
518 : /* Free any already-allocated storage */
519 56 : if (CurrentLocaleConvAllocated)
520 : {
521 6 : free_struct_lconv(&CurrentLocaleConv);
522 6 : CurrentLocaleConvAllocated = false;
523 : }
524 :
525 : /*
526 : * Use thread-safe method of obtaining a copy of lconv from the operating
527 : * system.
528 : */
529 56 : if (pg_localeconv_r(locale_monetary,
530 : locale_numeric,
531 : &tmp) != 0)
532 0 : elog(ERROR,
533 : "could not get lconv for LC_MONETARY = \"%s\", LC_NUMERIC = \"%s\": %m",
534 : locale_monetary, locale_numeric);
535 :
536 : /* Must copy data now so we can re-encode it. */
537 56 : extlconv = &tmp;
538 56 : worklconv.decimal_point = strdup(extlconv->decimal_point);
539 56 : worklconv.thousands_sep = strdup(extlconv->thousands_sep);
540 56 : worklconv.grouping = strdup(extlconv->grouping);
541 56 : worklconv.int_curr_symbol = strdup(extlconv->int_curr_symbol);
542 56 : worklconv.currency_symbol = strdup(extlconv->currency_symbol);
543 56 : worklconv.mon_decimal_point = strdup(extlconv->mon_decimal_point);
544 56 : worklconv.mon_thousands_sep = strdup(extlconv->mon_thousands_sep);
545 56 : worklconv.mon_grouping = strdup(extlconv->mon_grouping);
546 56 : worklconv.positive_sign = strdup(extlconv->positive_sign);
547 56 : worklconv.negative_sign = strdup(extlconv->negative_sign);
548 : /* Copy scalar fields as well */
549 56 : worklconv.int_frac_digits = extlconv->int_frac_digits;
550 56 : worklconv.frac_digits = extlconv->frac_digits;
551 56 : worklconv.p_cs_precedes = extlconv->p_cs_precedes;
552 56 : worklconv.p_sep_by_space = extlconv->p_sep_by_space;
553 56 : worklconv.n_cs_precedes = extlconv->n_cs_precedes;
554 56 : worklconv.n_sep_by_space = extlconv->n_sep_by_space;
555 56 : worklconv.p_sign_posn = extlconv->p_sign_posn;
556 56 : worklconv.n_sign_posn = extlconv->n_sign_posn;
557 :
558 : /* Free the contents of the object populated by pg_localeconv_r(). */
559 56 : pg_localeconv_free(&tmp);
560 :
561 : /* If any of the preceding strdup calls failed, complain now. */
562 56 : if (!struct_lconv_is_valid(&worklconv))
563 0 : ereport(ERROR,
564 : (errcode(ERRCODE_OUT_OF_MEMORY),
565 : errmsg("out of memory")));
566 :
567 56 : PG_TRY();
568 : {
569 : int encoding;
570 :
571 : /*
572 : * Now we must perform encoding conversion from whatever's associated
573 : * with the locales into the database encoding. If we can't identify
574 : * the encoding implied by LC_NUMERIC or LC_MONETARY (ie we get -1),
575 : * use PG_SQL_ASCII, which will result in just validating that the
576 : * strings are OK in the database encoding.
577 : */
578 56 : encoding = pg_get_encoding_from_locale(locale_numeric, true);
579 56 : if (encoding < 0)
580 0 : encoding = PG_SQL_ASCII;
581 :
582 56 : db_encoding_convert(encoding, &worklconv.decimal_point);
583 56 : db_encoding_convert(encoding, &worklconv.thousands_sep);
584 : /* grouping is not text and does not require conversion */
585 :
586 56 : encoding = pg_get_encoding_from_locale(locale_monetary, true);
587 56 : if (encoding < 0)
588 0 : encoding = PG_SQL_ASCII;
589 :
590 56 : db_encoding_convert(encoding, &worklconv.int_curr_symbol);
591 56 : db_encoding_convert(encoding, &worklconv.currency_symbol);
592 56 : db_encoding_convert(encoding, &worklconv.mon_decimal_point);
593 56 : db_encoding_convert(encoding, &worklconv.mon_thousands_sep);
594 : /* mon_grouping is not text and does not require conversion */
595 56 : db_encoding_convert(encoding, &worklconv.positive_sign);
596 56 : db_encoding_convert(encoding, &worklconv.negative_sign);
597 : }
598 0 : PG_CATCH();
599 : {
600 0 : free_struct_lconv(&worklconv);
601 0 : PG_RE_THROW();
602 : }
603 56 : PG_END_TRY();
604 :
605 : /*
606 : * Everything is good, so save the results.
607 : */
608 56 : CurrentLocaleConv = worklconv;
609 56 : CurrentLocaleConvAllocated = true;
610 56 : CurrentLocaleConvValid = true;
611 56 : return &CurrentLocaleConv;
612 : }
613 :
614 : #ifdef WIN32
615 : /*
616 : * On Windows, strftime() returns its output in encoding CP_ACP (the default
617 : * operating system codepage for the computer), which is likely different
618 : * from SERVER_ENCODING. This is especially important in Japanese versions
619 : * of Windows which will use SJIS encoding, which we don't support as a
620 : * server encoding.
621 : *
622 : * So, instead of using strftime(), use wcsftime() to return the value in
623 : * wide characters (internally UTF16) and then convert to UTF8, which we
624 : * know how to handle directly.
625 : *
626 : * Note that this only affects the calls to strftime() in this file, which are
627 : * used to get the locale-aware strings. Other parts of the backend use
628 : * pg_strftime(), which isn't locale-aware and does not need to be replaced.
629 : */
630 : static size_t
631 : strftime_l_win32(char *dst, size_t dstlen,
632 : const char *format, const struct tm *tm, locale_t locale)
633 : {
634 : size_t len;
635 : wchar_t wformat[8]; /* formats used below need 3 chars */
636 : wchar_t wbuf[MAX_L10N_DATA];
637 :
638 : /*
639 : * Get a wchar_t version of the format string. We only actually use
640 : * plain-ASCII formats in this file, so we can say that they're UTF8.
641 : */
642 : len = MultiByteToWideChar(CP_UTF8, 0, format, -1,
643 : wformat, lengthof(wformat));
644 : if (len == 0)
645 : elog(ERROR, "could not convert format string from UTF-8: error code %lu",
646 : GetLastError());
647 :
648 : len = _wcsftime_l(wbuf, MAX_L10N_DATA, wformat, tm, locale);
649 : if (len == 0)
650 : {
651 : /*
652 : * wcsftime failed, possibly because the result would not fit in
653 : * MAX_L10N_DATA. Return 0 with the contents of dst unspecified.
654 : */
655 : return 0;
656 : }
657 :
658 : len = WideCharToMultiByte(CP_UTF8, 0, wbuf, len, dst, dstlen - 1,
659 : NULL, NULL);
660 : if (len == 0)
661 : elog(ERROR, "could not convert string to UTF-8: error code %lu",
662 : GetLastError());
663 :
664 : dst[len] = '\0';
665 :
666 : return len;
667 : }
668 :
669 : /* redefine strftime_l() */
670 : #define strftime_l(a,b,c,d,e) strftime_l_win32(a,b,c,d,e)
671 : #endif /* WIN32 */
672 :
673 : /*
674 : * Subroutine for cache_locale_time().
675 : * Convert the given string from encoding "encoding" to the database
676 : * encoding, and store the result at *dst, replacing any previous value.
677 : */
678 : static void
679 1900 : cache_single_string(char **dst, const char *src, int encoding)
680 : {
681 : char *ptr;
682 : char *olddst;
683 :
684 : /* Convert the string to the database encoding, or validate it's OK */
685 1900 : ptr = pg_any_to_server(src, strlen(src), encoding);
686 :
687 : /* Store the string in long-lived storage, replacing any previous value */
688 1900 : olddst = *dst;
689 1900 : *dst = MemoryContextStrdup(TopMemoryContext, ptr);
690 1900 : if (olddst)
691 0 : pfree(olddst);
692 :
693 : /* Might as well clean up any palloc'd conversion result, too */
694 1900 : if (ptr != src)
695 0 : pfree(ptr);
696 1900 : }
697 :
698 : /*
699 : * Update the lc_time localization cache variables if needed.
700 : */
701 : void
702 50034 : cache_locale_time(void)
703 : {
704 : char buf[(2 * 7 + 2 * 12) * MAX_L10N_DATA];
705 : char *bufptr;
706 : time_t timenow;
707 : struct tm *timeinfo;
708 : struct tm timeinfobuf;
709 50034 : bool strftimefail = false;
710 : int encoding;
711 : int i;
712 : locale_t locale;
713 :
714 : /* did we do this already? */
715 50034 : if (CurrentLCTimeValid)
716 49984 : return;
717 :
718 50 : elog(DEBUG3, "cache_locale_time() executed; locale: \"%s\"", locale_time);
719 :
720 50 : errno = ENOENT;
721 : #ifdef WIN32
722 : locale = _create_locale(LC_ALL, locale_time);
723 : if (locale == (locale_t) 0)
724 : _dosmaperr(GetLastError());
725 : #else
726 50 : locale = newlocale(LC_ALL_MASK, locale_time, (locale_t) 0);
727 : #endif
728 50 : if (!locale)
729 0 : report_newlocale_failure(locale_time);
730 :
731 : /* We use times close to current time as data for strftime(). */
732 50 : timenow = time(NULL);
733 50 : timeinfo = gmtime_r(&timenow, &timeinfobuf);
734 :
735 : /* Store the strftime results in MAX_L10N_DATA-sized portions of buf[] */
736 50 : bufptr = buf;
737 :
738 : /*
739 : * MAX_L10N_DATA is sufficient buffer space for every known locale, and
740 : * POSIX defines no strftime() errors. (Buffer space exhaustion is not an
741 : * error.) An implementation might report errors (e.g. ENOMEM) by
742 : * returning 0 (or, less plausibly, a negative value) and setting errno.
743 : * Report errno just in case the implementation did that, but clear it in
744 : * advance of the calls so we don't emit a stale, unrelated errno.
745 : */
746 50 : errno = 0;
747 :
748 : /* localized days */
749 400 : for (i = 0; i < 7; i++)
750 : {
751 350 : timeinfo->tm_wday = i;
752 350 : if (strftime_l(bufptr, MAX_L10N_DATA, "%a", timeinfo, locale) <= 0)
753 0 : strftimefail = true;
754 350 : bufptr += MAX_L10N_DATA;
755 350 : if (strftime_l(bufptr, MAX_L10N_DATA, "%A", timeinfo, locale) <= 0)
756 0 : strftimefail = true;
757 350 : bufptr += MAX_L10N_DATA;
758 : }
759 :
760 : /* localized months */
761 650 : for (i = 0; i < 12; i++)
762 : {
763 600 : timeinfo->tm_mon = i;
764 600 : timeinfo->tm_mday = 1; /* make sure we don't have invalid date */
765 600 : if (strftime_l(bufptr, MAX_L10N_DATA, "%b", timeinfo, locale) <= 0)
766 0 : strftimefail = true;
767 600 : bufptr += MAX_L10N_DATA;
768 600 : if (strftime_l(bufptr, MAX_L10N_DATA, "%B", timeinfo, locale) <= 0)
769 0 : strftimefail = true;
770 600 : bufptr += MAX_L10N_DATA;
771 : }
772 :
773 : #ifdef WIN32
774 : _free_locale(locale);
775 : #else
776 50 : freelocale(locale);
777 : #endif
778 :
779 : /*
780 : * At this point we've done our best to clean up, and can throw errors, or
781 : * call functions that might throw errors, with a clean conscience.
782 : */
783 50 : if (strftimefail)
784 0 : elog(ERROR, "strftime_l() failed");
785 :
786 : #ifndef WIN32
787 :
788 : /*
789 : * As in PGLC_localeconv(), we must convert strftime()'s output from the
790 : * encoding implied by LC_TIME to the database encoding. If we can't
791 : * identify the LC_TIME encoding, just perform encoding validation.
792 : */
793 50 : encoding = pg_get_encoding_from_locale(locale_time, true);
794 50 : if (encoding < 0)
795 0 : encoding = PG_SQL_ASCII;
796 :
797 : #else
798 :
799 : /*
800 : * On Windows, strftime_win32() always returns UTF8 data, so convert from
801 : * that if necessary.
802 : */
803 : encoding = PG_UTF8;
804 :
805 : #endif /* WIN32 */
806 :
807 50 : bufptr = buf;
808 :
809 : /* localized days */
810 400 : for (i = 0; i < 7; i++)
811 : {
812 350 : cache_single_string(&localized_abbrev_days[i], bufptr, encoding);
813 350 : bufptr += MAX_L10N_DATA;
814 350 : cache_single_string(&localized_full_days[i], bufptr, encoding);
815 350 : bufptr += MAX_L10N_DATA;
816 : }
817 50 : localized_abbrev_days[7] = NULL;
818 50 : localized_full_days[7] = NULL;
819 :
820 : /* localized months */
821 650 : for (i = 0; i < 12; i++)
822 : {
823 600 : cache_single_string(&localized_abbrev_months[i], bufptr, encoding);
824 600 : bufptr += MAX_L10N_DATA;
825 600 : cache_single_string(&localized_full_months[i], bufptr, encoding);
826 600 : bufptr += MAX_L10N_DATA;
827 : }
828 50 : localized_abbrev_months[12] = NULL;
829 50 : localized_full_months[12] = NULL;
830 :
831 50 : CurrentLCTimeValid = true;
832 : }
833 :
834 :
835 : #if defined(WIN32) && defined(LC_MESSAGES)
836 : /*
837 : * Convert a Windows setlocale() argument to a Unix-style one.
838 : *
839 : * Regardless of platform, we install message catalogs under a Unix-style
840 : * LL[_CC][.ENCODING][@VARIANT] naming convention. Only LC_MESSAGES settings
841 : * following that style will elicit localized interface strings.
842 : *
843 : * Before Visual Studio 2012 (msvcr110.dll), Windows setlocale() accepted "C"
844 : * (but not "c") and strings of the form <Language>[_<Country>][.<CodePage>],
845 : * case-insensitive. setlocale() returns the fully-qualified form; for
846 : * example, setlocale("thaI") returns "Thai_Thailand.874". Internally,
847 : * setlocale() and _create_locale() select a "locale identifier"[1] and store
848 : * it in an undocumented _locale_t field. From that LCID, we can retrieve the
849 : * ISO 639 language and the ISO 3166 country. Character encoding does not
850 : * matter, because the server and client encodings govern that.
851 : *
852 : * Windows Vista introduced the "locale name" concept[2], closely following
853 : * RFC 4646. Locale identifiers are now deprecated. Starting with Visual
854 : * Studio 2012, setlocale() accepts locale names in addition to the strings it
855 : * accepted historically. It does not standardize them; setlocale("Th-tH")
856 : * returns "Th-tH". setlocale(category, "") still returns a traditional
857 : * string. Furthermore, msvcr110.dll changed the undocumented _locale_t
858 : * content to carry locale names instead of locale identifiers.
859 : *
860 : * Visual Studio 2015 should still be able to do the same as Visual Studio
861 : * 2012, but the declaration of locale_name is missing in _locale_t, causing
862 : * this code compilation to fail, hence this falls back instead on to
863 : * enumerating all system locales by using EnumSystemLocalesEx to find the
864 : * required locale name. If the input argument is in Unix-style then we can
865 : * get ISO Locale name directly by using GetLocaleInfoEx() with LCType as
866 : * LOCALE_SNAME.
867 : *
868 : * This function returns a pointer to a static buffer bearing the converted
869 : * name or NULL if conversion fails.
870 : *
871 : * [1] https://docs.microsoft.com/en-us/windows/win32/intl/locale-identifiers
872 : * [2] https://docs.microsoft.com/en-us/windows/win32/intl/locale-names
873 : */
874 :
875 : /*
876 : * Callback function for EnumSystemLocalesEx() in get_iso_localename().
877 : *
878 : * This function enumerates all system locales, searching for one that matches
879 : * an input with the format: <Language>[_<Country>], e.g.
880 : * English[_United States]
881 : *
882 : * The input is a three wchar_t array as an LPARAM. The first element is the
883 : * locale_name we want to match, the second element is an allocated buffer
884 : * where the Unix-style locale is copied if a match is found, and the third
885 : * element is the search status, 1 if a match was found, 0 otherwise.
886 : */
887 : static BOOL CALLBACK
888 : search_locale_enum(LPWSTR pStr, DWORD dwFlags, LPARAM lparam)
889 : {
890 : wchar_t test_locale[LOCALE_NAME_MAX_LENGTH];
891 : wchar_t **argv;
892 :
893 : (void) (dwFlags);
894 :
895 : argv = (wchar_t **) lparam;
896 : *argv[2] = (wchar_t) 0;
897 :
898 : memset(test_locale, 0, sizeof(test_locale));
899 :
900 : /* Get the name of the <Language> in English */
901 : if (GetLocaleInfoEx(pStr, LOCALE_SENGLISHLANGUAGENAME,
902 : test_locale, LOCALE_NAME_MAX_LENGTH))
903 : {
904 : /*
905 : * If the enumerated locale does not have a hyphen ("en") OR the
906 : * locale_name input does not have an underscore ("English"), we only
907 : * need to compare the <Language> tags.
908 : */
909 : if (wcsrchr(pStr, '-') == NULL || wcsrchr(argv[0], '_') == NULL)
910 : {
911 : if (_wcsicmp(argv[0], test_locale) == 0)
912 : {
913 : wcscpy(argv[1], pStr);
914 : *argv[2] = (wchar_t) 1;
915 : return FALSE;
916 : }
917 : }
918 :
919 : /*
920 : * We have to compare a full <Language>_<Country> tag, so we append
921 : * the underscore and name of the country/region in English, e.g.
922 : * "English_United States".
923 : */
924 : else
925 : {
926 : size_t len;
927 :
928 : wcscat(test_locale, L"_");
929 : len = wcslen(test_locale);
930 : if (GetLocaleInfoEx(pStr, LOCALE_SENGLISHCOUNTRYNAME,
931 : test_locale + len,
932 : LOCALE_NAME_MAX_LENGTH - len))
933 : {
934 : if (_wcsicmp(argv[0], test_locale) == 0)
935 : {
936 : wcscpy(argv[1], pStr);
937 : *argv[2] = (wchar_t) 1;
938 : return FALSE;
939 : }
940 : }
941 : }
942 : }
943 :
944 : return TRUE;
945 : }
946 :
947 : /*
948 : * This function converts a Windows locale name to an ISO formatted version
949 : * for Visual Studio 2015 or greater.
950 : *
951 : * Returns NULL, if no valid conversion was found.
952 : */
953 : static char *
954 : get_iso_localename(const char *winlocname)
955 : {
956 : wchar_t wc_locale_name[LOCALE_NAME_MAX_LENGTH];
957 : wchar_t buffer[LOCALE_NAME_MAX_LENGTH];
958 : static char iso_lc_messages[LOCALE_NAME_MAX_LENGTH];
959 : const char *period;
960 : int len;
961 : int ret_val;
962 :
963 : /*
964 : * Valid locales have the following syntax:
965 : * <Language>[_<Country>[.<CodePage>]]
966 : *
967 : * GetLocaleInfoEx can only take locale name without code-page and for the
968 : * purpose of this API the code-page doesn't matter.
969 : */
970 : period = strchr(winlocname, '.');
971 : if (period != NULL)
972 : len = period - winlocname;
973 : else
974 : len = pg_mbstrlen(winlocname);
975 :
976 : memset(wc_locale_name, 0, sizeof(wc_locale_name));
977 : memset(buffer, 0, sizeof(buffer));
978 : MultiByteToWideChar(CP_ACP, 0, winlocname, len, wc_locale_name,
979 : LOCALE_NAME_MAX_LENGTH);
980 :
981 : /*
982 : * If the lc_messages is already a Unix-style string, we have a direct
983 : * match with LOCALE_SNAME, e.g. en-US, en_US.
984 : */
985 : ret_val = GetLocaleInfoEx(wc_locale_name, LOCALE_SNAME, (LPWSTR) &buffer,
986 : LOCALE_NAME_MAX_LENGTH);
987 : if (!ret_val)
988 : {
989 : /*
990 : * Search for a locale in the system that matches language and country
991 : * name.
992 : */
993 : wchar_t *argv[3];
994 :
995 : argv[0] = wc_locale_name;
996 : argv[1] = buffer;
997 : argv[2] = (wchar_t *) &ret_val;
998 : EnumSystemLocalesEx(search_locale_enum, LOCALE_WINDOWS, (LPARAM) argv,
999 : NULL);
1000 : }
1001 :
1002 : if (ret_val)
1003 : {
1004 : size_t rc;
1005 : char *hyphen;
1006 :
1007 : /* Locale names use only ASCII, any conversion locale suffices. */
1008 : rc = wchar2char(iso_lc_messages, buffer, sizeof(iso_lc_messages), NULL);
1009 : if (rc == -1 || rc == sizeof(iso_lc_messages))
1010 : return NULL;
1011 :
1012 : /*
1013 : * Since the message catalogs sit on a case-insensitive filesystem, we
1014 : * need not standardize letter case here. So long as we do not ship
1015 : * message catalogs for which it would matter, we also need not
1016 : * translate the script/variant portion, e.g. uz-Cyrl-UZ to
1017 : * uz_UZ@cyrillic. Simply replace the hyphen with an underscore.
1018 : */
1019 : hyphen = strchr(iso_lc_messages, '-');
1020 : if (hyphen)
1021 : *hyphen = '_';
1022 : return iso_lc_messages;
1023 : }
1024 :
1025 : return NULL;
1026 : }
1027 :
1028 : static char *
1029 : IsoLocaleName(const char *winlocname)
1030 : {
1031 : static char iso_lc_messages[LOCALE_NAME_MAX_LENGTH];
1032 :
1033 : if (pg_strcasecmp("c", winlocname) == 0 ||
1034 : pg_strcasecmp("posix", winlocname) == 0)
1035 : {
1036 : strcpy(iso_lc_messages, "C");
1037 : return iso_lc_messages;
1038 : }
1039 : else
1040 : return get_iso_localename(winlocname);
1041 : }
1042 :
1043 : #endif /* WIN32 && LC_MESSAGES */
1044 :
1045 : /*
1046 : * Create a new pg_locale_t struct for the given collation oid.
1047 : */
1048 : static pg_locale_t
1049 326 : create_pg_locale(Oid collid, MemoryContext context)
1050 : {
1051 : HeapTuple tp;
1052 : Form_pg_collation collform;
1053 : pg_locale_t result;
1054 : Datum datum;
1055 : bool isnull;
1056 :
1057 326 : tp = SearchSysCache1(COLLOID, ObjectIdGetDatum(collid));
1058 326 : if (!HeapTupleIsValid(tp))
1059 0 : elog(ERROR, "cache lookup failed for collation %u", collid);
1060 326 : collform = (Form_pg_collation) GETSTRUCT(tp);
1061 :
1062 326 : if (collform->collprovider == COLLPROVIDER_BUILTIN)
1063 52 : result = create_pg_locale_builtin(collid, context);
1064 274 : else if (collform->collprovider == COLLPROVIDER_ICU)
1065 184 : result = create_pg_locale_icu(collid, context);
1066 90 : else if (collform->collprovider == COLLPROVIDER_LIBC)
1067 90 : result = create_pg_locale_libc(collid, context);
1068 : else
1069 : /* shouldn't happen */
1070 0 : PGLOCALE_SUPPORT_ERROR(collform->collprovider);
1071 :
1072 320 : result->is_default = false;
1073 :
1074 : Assert((result->collate_is_c && result->collate == NULL) ||
1075 : (!result->collate_is_c && result->collate != NULL));
1076 :
1077 : Assert((result->ctype_is_c && result->ctype == NULL) ||
1078 : (!result->ctype_is_c && result->ctype != NULL));
1079 :
1080 320 : datum = SysCacheGetAttr(COLLOID, tp, Anum_pg_collation_collversion,
1081 : &isnull);
1082 320 : if (!isnull)
1083 : {
1084 : char *actual_versionstr;
1085 : char *collversionstr;
1086 :
1087 230 : collversionstr = TextDatumGetCString(datum);
1088 :
1089 230 : if (collform->collprovider == COLLPROVIDER_LIBC)
1090 0 : datum = SysCacheGetAttrNotNull(COLLOID, tp, Anum_pg_collation_collcollate);
1091 : else
1092 230 : datum = SysCacheGetAttrNotNull(COLLOID, tp, Anum_pg_collation_colllocale);
1093 :
1094 230 : actual_versionstr = get_collation_actual_version(collform->collprovider,
1095 230 : TextDatumGetCString(datum));
1096 230 : if (!actual_versionstr)
1097 : {
1098 : /*
1099 : * This could happen when specifying a version in CREATE COLLATION
1100 : * but the provider does not support versioning, or manually
1101 : * creating a mess in the catalogs.
1102 : */
1103 0 : ereport(ERROR,
1104 : (errmsg("collation \"%s\" has no actual version, but a version was recorded",
1105 : NameStr(collform->collname))));
1106 : }
1107 :
1108 230 : if (strcmp(actual_versionstr, collversionstr) != 0)
1109 0 : ereport(WARNING,
1110 : (errmsg("collation \"%s\" has version mismatch",
1111 : NameStr(collform->collname)),
1112 : errdetail("The collation in the database was created using version %s, "
1113 : "but the operating system provides version %s.",
1114 : collversionstr, actual_versionstr),
1115 : errhint("Rebuild all objects affected by this collation and run "
1116 : "ALTER COLLATION %s REFRESH VERSION, "
1117 : "or build PostgreSQL with the right library version.",
1118 : quote_qualified_identifier(get_namespace_name(collform->collnamespace),
1119 : NameStr(collform->collname)))));
1120 : }
1121 :
1122 320 : ReleaseSysCache(tp);
1123 :
1124 320 : return result;
1125 : }
1126 :
1127 : /*
1128 : * Initialize default_locale with database locale settings.
1129 : */
1130 : void
1131 33604 : init_database_collation(void)
1132 : {
1133 : HeapTuple tup;
1134 : Form_pg_database dbform;
1135 : pg_locale_t result;
1136 :
1137 : Assert(default_locale == NULL);
1138 :
1139 : /* Fetch our pg_database row normally, via syscache */
1140 33604 : tup = SearchSysCache1(DATABASEOID, ObjectIdGetDatum(MyDatabaseId));
1141 33604 : if (!HeapTupleIsValid(tup))
1142 0 : elog(ERROR, "cache lookup failed for database %u", MyDatabaseId);
1143 33604 : dbform = (Form_pg_database) GETSTRUCT(tup);
1144 :
1145 33604 : if (dbform->datlocprovider == COLLPROVIDER_BUILTIN)
1146 1788 : result = create_pg_locale_builtin(DEFAULT_COLLATION_OID,
1147 : TopMemoryContext);
1148 31816 : else if (dbform->datlocprovider == COLLPROVIDER_ICU)
1149 26 : result = create_pg_locale_icu(DEFAULT_COLLATION_OID,
1150 : TopMemoryContext);
1151 31790 : else if (dbform->datlocprovider == COLLPROVIDER_LIBC)
1152 31790 : result = create_pg_locale_libc(DEFAULT_COLLATION_OID,
1153 : TopMemoryContext);
1154 : else
1155 : /* shouldn't happen */
1156 0 : PGLOCALE_SUPPORT_ERROR(dbform->datlocprovider);
1157 :
1158 33600 : result->is_default = true;
1159 :
1160 : Assert((result->collate_is_c && result->collate == NULL) ||
1161 : (!result->collate_is_c && result->collate != NULL));
1162 :
1163 : Assert((result->ctype_is_c && result->ctype == NULL) ||
1164 : (!result->ctype_is_c && result->ctype != NULL));
1165 :
1166 33600 : ReleaseSysCache(tup);
1167 :
1168 33600 : default_locale = result;
1169 33600 : }
1170 :
1171 : /*
1172 : * Get database default locale.
1173 : */
1174 : pg_locale_t
1175 3091548 : pg_database_locale(void)
1176 : {
1177 3091548 : return pg_newlocale_from_collation(DEFAULT_COLLATION_OID);
1178 : }
1179 :
1180 : /*
1181 : * Create a pg_locale_t from a collation OID. Results are cached for the
1182 : * lifetime of the backend. Thus, do not free the result with freelocale().
1183 : *
1184 : * For simplicity, we always generate COLLATE + CTYPE even though we
1185 : * might only need one of them. Since this is called only once per session,
1186 : * it shouldn't cost much.
1187 : */
1188 : pg_locale_t
1189 32738218 : pg_newlocale_from_collation(Oid collid)
1190 : {
1191 : collation_cache_entry *cache_entry;
1192 : bool found;
1193 :
1194 32738218 : if (collid == DEFAULT_COLLATION_OID)
1195 28489242 : return default_locale;
1196 :
1197 : /*
1198 : * Some callers expect C_COLLATION_OID to succeed even without catalog
1199 : * access.
1200 : */
1201 4248976 : if (collid == C_COLLATION_OID)
1202 4212962 : return &c_locale;
1203 :
1204 36014 : if (!OidIsValid(collid))
1205 0 : elog(ERROR, "cache lookup failed for collation %u", collid);
1206 :
1207 36014 : AssertCouldGetRelation();
1208 :
1209 36014 : if (last_collation_cache_oid == collid)
1210 34980 : return last_collation_cache_locale;
1211 :
1212 1034 : if (CollationCache == NULL)
1213 : {
1214 78 : CollationCacheContext = AllocSetContextCreate(TopMemoryContext,
1215 : "collation cache",
1216 : ALLOCSET_DEFAULT_SIZES);
1217 78 : CollationCache = collation_cache_create(CollationCacheContext,
1218 : 16, NULL);
1219 : }
1220 :
1221 1034 : cache_entry = collation_cache_insert(CollationCache, collid, &found);
1222 1034 : if (!found)
1223 : {
1224 : /*
1225 : * Make sure cache entry is marked invalid, in case we fail before
1226 : * setting things.
1227 : */
1228 326 : cache_entry->locale = NULL;
1229 : }
1230 :
1231 1034 : if (cache_entry->locale == NULL)
1232 : {
1233 326 : cache_entry->locale = create_pg_locale(collid, CollationCacheContext);
1234 : }
1235 :
1236 1028 : last_collation_cache_oid = collid;
1237 1028 : last_collation_cache_locale = cache_entry->locale;
1238 :
1239 1028 : return cache_entry->locale;
1240 : }
1241 :
1242 : /*
1243 : * Get provider-specific collation version string for the given collation from
1244 : * the operating system/library.
1245 : */
1246 : char *
1247 112554 : get_collation_actual_version(char collprovider, const char *collcollate)
1248 : {
1249 112554 : char *collversion = NULL;
1250 :
1251 112554 : if (collprovider == COLLPROVIDER_BUILTIN)
1252 1908 : collversion = get_collation_actual_version_builtin(collcollate);
1253 : #ifdef USE_ICU
1254 110646 : else if (collprovider == COLLPROVIDER_ICU)
1255 79454 : collversion = get_collation_actual_version_icu(collcollate);
1256 : #endif
1257 31192 : else if (collprovider == COLLPROVIDER_LIBC)
1258 31192 : collversion = get_collation_actual_version_libc(collcollate);
1259 :
1260 112554 : return collversion;
1261 : }
1262 :
1263 : /* lowercasing/casefolding in C locale */
1264 : static size_t
1265 6817766 : strlower_c(char *dst, size_t dstsize, const char *src, ssize_t srclen)
1266 : {
1267 : int i;
1268 :
1269 6817766 : srclen = (srclen >= 0) ? srclen : strlen(src);
1270 55228014 : for (i = 0; i < srclen && i < dstsize; i++)
1271 48410248 : dst[i] = pg_ascii_tolower(src[i]);
1272 6817766 : if (i < dstsize)
1273 6817766 : dst[i] = '\0';
1274 6817766 : return srclen;
1275 : }
1276 :
1277 : /* titlecasing in C locale */
1278 : static size_t
1279 0 : strtitle_c(char *dst, size_t dstsize, const char *src, ssize_t srclen)
1280 : {
1281 0 : bool wasalnum = false;
1282 : int i;
1283 :
1284 0 : srclen = (srclen >= 0) ? srclen : strlen(src);
1285 0 : for (i = 0; i < srclen && i < dstsize; i++)
1286 : {
1287 0 : char c = src[i];
1288 :
1289 0 : if (wasalnum)
1290 0 : dst[i] = pg_ascii_tolower(c);
1291 : else
1292 0 : dst[i] = pg_ascii_toupper(c);
1293 :
1294 0 : wasalnum = ((c >= '0' && c <= '9') ||
1295 0 : (c >= 'A' && c <= 'Z') ||
1296 0 : (c >= 'a' && c <= 'z'));
1297 : }
1298 0 : if (i < dstsize)
1299 0 : dst[i] = '\0';
1300 0 : return srclen;
1301 : }
1302 :
1303 : /* uppercasing in C locale */
1304 : static size_t
1305 0 : strupper_c(char *dst, size_t dstsize, const char *src, ssize_t srclen)
1306 : {
1307 : int i;
1308 :
1309 0 : srclen = (srclen >= 0) ? srclen : strlen(src);
1310 0 : for (i = 0; i < srclen && i < dstsize; i++)
1311 0 : dst[i] = pg_ascii_toupper(src[i]);
1312 0 : if (i < dstsize)
1313 0 : dst[i] = '\0';
1314 0 : return srclen;
1315 : }
1316 :
1317 : size_t
1318 437468 : pg_strlower(char *dst, size_t dstsize, const char *src, ssize_t srclen,
1319 : pg_locale_t locale)
1320 : {
1321 437468 : if (locale->ctype == NULL)
1322 0 : return strlower_c(dst, dstsize, src, srclen);
1323 : else
1324 437468 : return locale->ctype->strlower(dst, dstsize, src, srclen, locale);
1325 : }
1326 :
1327 : size_t
1328 232 : pg_strtitle(char *dst, size_t dstsize, const char *src, ssize_t srclen,
1329 : pg_locale_t locale)
1330 : {
1331 232 : if (locale->ctype == NULL)
1332 0 : return strtitle_c(dst, dstsize, src, srclen);
1333 : else
1334 232 : return locale->ctype->strtitle(dst, dstsize, src, srclen, locale);
1335 : }
1336 :
1337 : size_t
1338 1038312 : pg_strupper(char *dst, size_t dstsize, const char *src, ssize_t srclen,
1339 : pg_locale_t locale)
1340 : {
1341 1038312 : if (locale->ctype == NULL)
1342 0 : return strupper_c(dst, dstsize, src, srclen);
1343 : else
1344 1038312 : return locale->ctype->strupper(dst, dstsize, src, srclen, locale);
1345 : }
1346 :
1347 : size_t
1348 436960 : pg_strfold(char *dst, size_t dstsize, const char *src, ssize_t srclen,
1349 : pg_locale_t locale)
1350 : {
1351 : /* in the C locale, casefolding is the same as lowercasing */
1352 436960 : if (locale->ctype == NULL)
1353 0 : return strlower_c(dst, dstsize, src, srclen);
1354 : else
1355 436960 : return locale->ctype->strfold(dst, dstsize, src, srclen, locale);
1356 : }
1357 :
1358 : /*
1359 : * Lowercase an identifier using the database default locale.
1360 : *
1361 : * For historical reasons, does not use ordinary locale behavior. Should only
1362 : * be used for identifiers. XXX: can we make this equivalent to
1363 : * pg_strfold(..., default_locale)?
1364 : */
1365 : size_t
1366 6842918 : pg_downcase_ident(char *dst, size_t dstsize, const char *src, ssize_t srclen)
1367 : {
1368 6842918 : pg_locale_t locale = default_locale;
1369 :
1370 6842918 : if (locale == NULL || locale->ctype == NULL ||
1371 6413406 : locale->ctype->downcase_ident == NULL)
1372 6817766 : return strlower_c(dst, dstsize, src, srclen);
1373 : else
1374 25152 : return locale->ctype->downcase_ident(dst, dstsize, src, srclen,
1375 : locale);
1376 : }
1377 :
1378 : /*
1379 : * pg_strcoll
1380 : *
1381 : * Like pg_strncoll for NUL-terminated input strings.
1382 : */
1383 : int
1384 25712824 : pg_strcoll(const char *arg1, const char *arg2, pg_locale_t locale)
1385 : {
1386 25712824 : return locale->collate->strncoll(arg1, -1, arg2, -1, locale);
1387 : }
1388 :
1389 : /*
1390 : * pg_strncoll
1391 : *
1392 : * Call ucol_strcollUTF8(), ucol_strcoll(), strcoll_l() or wcscoll_l() as
1393 : * appropriate for the given locale, platform, and database encoding. If the
1394 : * locale is not specified, use the database collation.
1395 : *
1396 : * The input strings must be encoded in the database encoding. If an input
1397 : * string is NUL-terminated, its length may be specified as -1.
1398 : *
1399 : * The caller is responsible for breaking ties if the collation is
1400 : * deterministic; this maintains consistency with pg_strnxfrm(), which cannot
1401 : * easily account for deterministic collations.
1402 : */
1403 : int
1404 4451306 : pg_strncoll(const char *arg1, ssize_t len1, const char *arg2, ssize_t len2,
1405 : pg_locale_t locale)
1406 : {
1407 4451306 : return locale->collate->strncoll(arg1, len1, arg2, len2, locale);
1408 : }
1409 :
1410 : /*
1411 : * Return true if the collation provider supports pg_strxfrm() and
1412 : * pg_strnxfrm(); otherwise false.
1413 : *
1414 : *
1415 : * No similar problem is known for the ICU provider.
1416 : */
1417 : bool
1418 46016 : pg_strxfrm_enabled(pg_locale_t locale)
1419 : {
1420 : /*
1421 : * locale->collate->strnxfrm is still a required method, even if it may
1422 : * have the wrong behavior, because the planner uses it for estimates in
1423 : * some cases.
1424 : */
1425 46016 : return locale->collate->strxfrm_is_safe;
1426 : }
1427 :
1428 : /*
1429 : * pg_strxfrm
1430 : *
1431 : * Like pg_strnxfrm for a NUL-terminated input string.
1432 : */
1433 : size_t
1434 144 : pg_strxfrm(char *dest, const char *src, size_t destsize, pg_locale_t locale)
1435 : {
1436 144 : return locale->collate->strnxfrm(dest, destsize, src, -1, locale);
1437 : }
1438 :
1439 : /*
1440 : * pg_strnxfrm
1441 : *
1442 : * Transforms 'src' to a nul-terminated string stored in 'dest' such that
1443 : * ordinary strcmp() on transformed strings is equivalent to pg_strcoll() on
1444 : * untransformed strings.
1445 : *
1446 : * The input string must be encoded in the database encoding. If the input
1447 : * string is NUL-terminated, its length may be specified as -1. If 'destsize'
1448 : * is zero, 'dest' may be NULL.
1449 : *
1450 : * Not all providers support pg_strnxfrm() safely. The caller should check
1451 : * pg_strxfrm_enabled() first, otherwise this function may return wrong
1452 : * results or an error.
1453 : *
1454 : * Returns the number of bytes needed (or more) to store the transformed
1455 : * string, excluding the terminating nul byte. If the value returned is
1456 : * 'destsize' or greater, the resulting contents of 'dest' are undefined.
1457 : */
1458 : size_t
1459 5748 : pg_strnxfrm(char *dest, size_t destsize, const char *src, ssize_t srclen,
1460 : pg_locale_t locale)
1461 : {
1462 5748 : return locale->collate->strnxfrm(dest, destsize, src, srclen, locale);
1463 : }
1464 :
1465 : /*
1466 : * Return true if the collation provider supports pg_strxfrm_prefix() and
1467 : * pg_strnxfrm_prefix(); otherwise false.
1468 : */
1469 : bool
1470 1668 : pg_strxfrm_prefix_enabled(pg_locale_t locale)
1471 : {
1472 1668 : return (locale->collate->strnxfrm_prefix != NULL);
1473 : }
1474 :
1475 : /*
1476 : * pg_strxfrm_prefix
1477 : *
1478 : * Like pg_strnxfrm_prefix for a NUL-terminated input string.
1479 : */
1480 : size_t
1481 1668 : pg_strxfrm_prefix(char *dest, const char *src, size_t destsize,
1482 : pg_locale_t locale)
1483 : {
1484 1668 : return locale->collate->strnxfrm_prefix(dest, destsize, src, -1, locale);
1485 : }
1486 :
1487 : /*
1488 : * pg_strnxfrm_prefix
1489 : *
1490 : * Transforms 'src' to a byte sequence stored in 'dest' such that ordinary
1491 : * memcmp() on the byte sequence is equivalent to pg_strncoll() on
1492 : * untransformed strings. The result is not nul-terminated.
1493 : *
1494 : * The input string must be encoded in the database encoding. If the input
1495 : * string is NUL-terminated, its length may be specified as -1.
1496 : *
1497 : * Not all providers support pg_strnxfrm_prefix() safely. The caller should
1498 : * check pg_strxfrm_prefix_enabled() first, otherwise this function may return
1499 : * wrong results or an error.
1500 : *
1501 : * If destsize is not large enough to hold the resulting byte sequence, stores
1502 : * only the first destsize bytes in 'dest'. Returns the number of bytes
1503 : * actually copied to 'dest'.
1504 : */
1505 : size_t
1506 0 : pg_strnxfrm_prefix(char *dest, size_t destsize, const char *src,
1507 : ssize_t srclen, pg_locale_t locale)
1508 : {
1509 0 : return locale->collate->strnxfrm_prefix(dest, destsize, src, srclen, locale);
1510 : }
1511 :
1512 : bool
1513 37132 : pg_iswdigit(pg_wchar wc, pg_locale_t locale)
1514 : {
1515 37132 : if (locale->ctype == NULL)
1516 0 : return (wc <= (pg_wchar) 127 &&
1517 0 : (pg_char_properties[wc] & PG_ISDIGIT));
1518 : else
1519 37132 : return locale->ctype->wc_isdigit(wc, locale);
1520 : }
1521 :
1522 : bool
1523 104140 : pg_iswalpha(pg_wchar wc, pg_locale_t locale)
1524 : {
1525 104140 : if (locale->ctype == NULL)
1526 0 : return (wc <= (pg_wchar) 127 &&
1527 0 : (pg_char_properties[wc] & PG_ISALPHA));
1528 : else
1529 104140 : return locale->ctype->wc_isalpha(wc, locale);
1530 : }
1531 :
1532 : bool
1533 2804950 : pg_iswalnum(pg_wchar wc, pg_locale_t locale)
1534 : {
1535 2804950 : if (locale->ctype == NULL)
1536 0 : return (wc <= (pg_wchar) 127 &&
1537 0 : (pg_char_properties[wc] & PG_ISALNUM));
1538 : else
1539 2804950 : return locale->ctype->wc_isalnum(wc, locale);
1540 : }
1541 :
1542 : bool
1543 0 : pg_iswupper(pg_wchar wc, pg_locale_t locale)
1544 : {
1545 0 : if (locale->ctype == NULL)
1546 0 : return (wc <= (pg_wchar) 127 &&
1547 0 : (pg_char_properties[wc] & PG_ISUPPER));
1548 : else
1549 0 : return locale->ctype->wc_isupper(wc, locale);
1550 : }
1551 :
1552 : bool
1553 0 : pg_iswlower(pg_wchar wc, pg_locale_t locale)
1554 : {
1555 0 : if (locale->ctype == NULL)
1556 0 : return (wc <= (pg_wchar) 127 &&
1557 0 : (pg_char_properties[wc] & PG_ISLOWER));
1558 : else
1559 0 : return locale->ctype->wc_islower(wc, locale);
1560 : }
1561 :
1562 : bool
1563 0 : pg_iswgraph(pg_wchar wc, pg_locale_t locale)
1564 : {
1565 0 : if (locale->ctype == NULL)
1566 0 : return (wc <= (pg_wchar) 127 &&
1567 0 : (pg_char_properties[wc] & PG_ISGRAPH));
1568 : else
1569 0 : return locale->ctype->wc_isgraph(wc, locale);
1570 : }
1571 :
1572 : bool
1573 0 : pg_iswprint(pg_wchar wc, pg_locale_t locale)
1574 : {
1575 0 : if (locale->ctype == NULL)
1576 0 : return (wc <= (pg_wchar) 127 &&
1577 0 : (pg_char_properties[wc] & PG_ISPRINT));
1578 : else
1579 0 : return locale->ctype->wc_isprint(wc, locale);
1580 : }
1581 :
1582 : bool
1583 0 : pg_iswpunct(pg_wchar wc, pg_locale_t locale)
1584 : {
1585 0 : if (locale->ctype == NULL)
1586 0 : return (wc <= (pg_wchar) 127 &&
1587 0 : (pg_char_properties[wc] & PG_ISPUNCT));
1588 : else
1589 0 : return locale->ctype->wc_ispunct(wc, locale);
1590 : }
1591 :
1592 : bool
1593 678 : pg_iswspace(pg_wchar wc, pg_locale_t locale)
1594 : {
1595 678 : if (locale->ctype == NULL)
1596 0 : return (wc <= (pg_wchar) 127 &&
1597 0 : (pg_char_properties[wc] & PG_ISSPACE));
1598 : else
1599 678 : return locale->ctype->wc_isspace(wc, locale);
1600 : }
1601 :
1602 : bool
1603 18 : pg_iswxdigit(pg_wchar wc, pg_locale_t locale)
1604 : {
1605 18 : if (locale->ctype == NULL)
1606 0 : return (wc <= (pg_wchar) 127 &&
1607 0 : ((pg_char_properties[wc] & PG_ISDIGIT) ||
1608 0 : ((wc >= 'A' && wc <= 'F') ||
1609 0 : (wc >= 'a' && wc <= 'f'))));
1610 : else
1611 18 : return locale->ctype->wc_isxdigit(wc, locale);
1612 : }
1613 :
1614 : bool
1615 266 : pg_iswcased(pg_wchar wc, pg_locale_t locale)
1616 : {
1617 : /* for the C locale, Cased and Alpha are equivalent */
1618 266 : if (locale->ctype == NULL)
1619 280 : return (wc <= (pg_wchar) 127 &&
1620 140 : (pg_char_properties[wc] & PG_ISALPHA));
1621 : else
1622 126 : return locale->ctype->wc_iscased(wc, locale);
1623 : }
1624 :
1625 : pg_wchar
1626 0 : pg_towupper(pg_wchar wc, pg_locale_t locale)
1627 : {
1628 0 : if (locale->ctype == NULL)
1629 : {
1630 0 : if (wc <= (pg_wchar) 127)
1631 0 : return pg_ascii_toupper((unsigned char) wc);
1632 0 : return wc;
1633 : }
1634 : else
1635 0 : return locale->ctype->wc_toupper(wc, locale);
1636 : }
1637 :
1638 : pg_wchar
1639 0 : pg_towlower(pg_wchar wc, pg_locale_t locale)
1640 : {
1641 0 : if (locale->ctype == NULL)
1642 : {
1643 0 : if (wc <= (pg_wchar) 127)
1644 0 : return pg_ascii_tolower((unsigned char) wc);
1645 0 : return wc;
1646 : }
1647 : else
1648 0 : return locale->ctype->wc_tolower(wc, locale);
1649 : }
1650 :
1651 : /* version of Unicode used by ICU */
1652 : const char *
1653 2 : pg_icu_unicode_version()
1654 : {
1655 : #ifdef USE_ICU
1656 2 : return U_UNICODE_VERSION;
1657 : #else
1658 : return NULL;
1659 : #endif
1660 : }
1661 :
1662 : /*
1663 : * Return required encoding ID for the given locale, or -1 if any encoding is
1664 : * valid for the locale.
1665 : */
1666 : int
1667 1964 : builtin_locale_encoding(const char *locale)
1668 : {
1669 1964 : if (strcmp(locale, "C") == 0)
1670 76 : return -1;
1671 1888 : else if (strcmp(locale, "C.UTF-8") == 0)
1672 1858 : return PG_UTF8;
1673 30 : else if (strcmp(locale, "PG_UNICODE_FAST") == 0)
1674 30 : return PG_UTF8;
1675 :
1676 :
1677 0 : ereport(ERROR,
1678 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
1679 : errmsg("invalid locale name \"%s\" for builtin provider",
1680 : locale)));
1681 :
1682 : return 0; /* keep compiler quiet */
1683 : }
1684 :
1685 :
1686 : /*
1687 : * Validate the locale and encoding combination, and return the canonical form
1688 : * of the locale name.
1689 : */
1690 : const char *
1691 1948 : builtin_validate_locale(int encoding, const char *locale)
1692 : {
1693 1948 : const char *canonical_name = NULL;
1694 : int required_encoding;
1695 :
1696 1948 : if (strcmp(locale, "C") == 0)
1697 64 : canonical_name = "C";
1698 1884 : else if (strcmp(locale, "C.UTF-8") == 0 || strcmp(locale, "C.UTF8") == 0)
1699 1844 : canonical_name = "C.UTF-8";
1700 40 : else if (strcmp(locale, "PG_UNICODE_FAST") == 0)
1701 22 : canonical_name = "PG_UNICODE_FAST";
1702 :
1703 1948 : if (!canonical_name)
1704 18 : ereport(ERROR,
1705 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
1706 : errmsg("invalid locale name \"%s\" for builtin provider",
1707 : locale)));
1708 :
1709 1930 : required_encoding = builtin_locale_encoding(canonical_name);
1710 1930 : if (required_encoding >= 0 && encoding != required_encoding)
1711 2 : ereport(ERROR,
1712 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
1713 : errmsg("encoding \"%s\" does not match locale \"%s\"",
1714 : pg_encoding_to_char(encoding), locale)));
1715 :
1716 1928 : return canonical_name;
1717 : }
1718 :
1719 :
1720 :
1721 : /*
1722 : * Return the BCP47 language tag representation of the requested locale.
1723 : *
1724 : * This function should be called before passing the string to ucol_open(),
1725 : * because conversion to a language tag also performs "level 2
1726 : * canonicalization". In addition to producing a consistent format, level 2
1727 : * canonicalization is able to more accurately interpret different input
1728 : * locale string formats, such as POSIX and .NET IDs.
1729 : */
1730 : char *
1731 79146 : icu_language_tag(const char *loc_str, int elevel)
1732 : {
1733 : #ifdef USE_ICU
1734 : UErrorCode status;
1735 : char *langtag;
1736 79146 : size_t buflen = 32; /* arbitrary starting buffer size */
1737 79146 : const bool strict = true;
1738 :
1739 : /*
1740 : * A BCP47 language tag doesn't have a clearly-defined upper limit (cf.
1741 : * RFC5646 section 4.4). Additionally, in older ICU versions,
1742 : * uloc_toLanguageTag() doesn't always return the ultimate length on the
1743 : * first call, necessitating a loop.
1744 : */
1745 79146 : langtag = palloc(buflen);
1746 : while (true)
1747 : {
1748 79146 : status = U_ZERO_ERROR;
1749 79146 : uloc_toLanguageTag(loc_str, langtag, buflen, strict, &status);
1750 :
1751 : /* try again if the buffer is not large enough */
1752 79146 : if ((status == U_BUFFER_OVERFLOW_ERROR ||
1753 79146 : status == U_STRING_NOT_TERMINATED_WARNING) &&
1754 : buflen < MaxAllocSize)
1755 : {
1756 0 : buflen = Min(buflen * 2, MaxAllocSize);
1757 0 : langtag = repalloc(langtag, buflen);
1758 0 : continue;
1759 : }
1760 :
1761 79146 : break;
1762 : }
1763 :
1764 79146 : if (U_FAILURE(status))
1765 : {
1766 18 : pfree(langtag);
1767 :
1768 18 : if (elevel > 0)
1769 14 : ereport(elevel,
1770 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1771 : errmsg("could not convert locale name \"%s\" to language tag: %s",
1772 : loc_str, u_errorName(status))));
1773 12 : return NULL;
1774 : }
1775 :
1776 79128 : return langtag;
1777 : #else /* not USE_ICU */
1778 : ereport(ERROR,
1779 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1780 : errmsg("ICU is not supported in this build")));
1781 : return NULL; /* keep compiler quiet */
1782 : #endif /* not USE_ICU */
1783 : }
1784 :
1785 : /*
1786 : * Perform best-effort check that the locale is a valid one.
1787 : */
1788 : void
1789 166 : icu_validate_locale(const char *loc_str)
1790 : {
1791 : #ifdef USE_ICU
1792 : UCollator *collator;
1793 : UErrorCode status;
1794 : char lang[ULOC_LANG_CAPACITY];
1795 166 : bool found = false;
1796 166 : int elevel = icu_validation_level;
1797 :
1798 : /* no validation */
1799 166 : if (elevel < 0)
1800 12 : return;
1801 :
1802 : /* downgrade to WARNING during pg_upgrade */
1803 154 : if (IsBinaryUpgrade && elevel > WARNING)
1804 0 : elevel = WARNING;
1805 :
1806 : /* validate that we can extract the language */
1807 154 : status = U_ZERO_ERROR;
1808 154 : uloc_getLanguage(loc_str, lang, ULOC_LANG_CAPACITY, &status);
1809 154 : if (U_FAILURE(status) || status == U_STRING_NOT_TERMINATED_WARNING)
1810 : {
1811 0 : ereport(elevel,
1812 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1813 : errmsg("could not get language from ICU locale \"%s\": %s",
1814 : loc_str, u_errorName(status)),
1815 : errhint("To disable ICU locale validation, set the parameter \"%s\" to \"%s\".",
1816 : "icu_validation_level", "disabled")));
1817 0 : return;
1818 : }
1819 :
1820 : /* check for special language name */
1821 154 : if (strcmp(lang, "") == 0 ||
1822 46 : strcmp(lang, "root") == 0 || strcmp(lang, "und") == 0)
1823 108 : found = true;
1824 :
1825 : /* search for matching language within ICU */
1826 15704 : for (int32_t i = 0; !found && i < uloc_countAvailable(); i++)
1827 : {
1828 15550 : const char *otherloc = uloc_getAvailable(i);
1829 : char otherlang[ULOC_LANG_CAPACITY];
1830 :
1831 15550 : status = U_ZERO_ERROR;
1832 15550 : uloc_getLanguage(otherloc, otherlang, ULOC_LANG_CAPACITY, &status);
1833 15550 : if (U_FAILURE(status) || status == U_STRING_NOT_TERMINATED_WARNING)
1834 0 : continue;
1835 :
1836 15550 : if (strcmp(lang, otherlang) == 0)
1837 32 : found = true;
1838 : }
1839 :
1840 154 : if (!found)
1841 14 : ereport(elevel,
1842 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1843 : errmsg("ICU locale \"%s\" has unknown language \"%s\"",
1844 : loc_str, lang),
1845 : errhint("To disable ICU locale validation, set the parameter \"%s\" to \"%s\".",
1846 : "icu_validation_level", "disabled")));
1847 :
1848 : /* check that it can be opened */
1849 148 : collator = pg_ucol_open(loc_str);
1850 140 : ucol_close(collator);
1851 : #else /* not USE_ICU */
1852 : /* could get here if a collation was created by a build with ICU */
1853 : ereport(ERROR,
1854 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1855 : errmsg("ICU is not supported in this build")));
1856 : #endif /* not USE_ICU */
1857 : }
|
