1 # Common tools for test files to find the locales which exist on the
2 # system. Caller should have verified that this isn't miniperl before calling
5 # Note that it's okay that some languages have their native names
6 # capitalized here even though that's not "right". They are lowercased
7 # anyway later during the scanning process (and besides, some clueless
8 # vendor might have them capitalized erroneously anyway).
10 # Functions whose names begin with underscore are internal helper functions
11 # for this file, and are not to be used by outside callers.
18 eval { require POSIX; import POSIX 'locale_h'; };
19 my $has_locale_h = ! $@;
21 my @known_categories = ( qw(LC_ALL LC_COLLATE LC_CTYPE LC_MESSAGES LC_MONETARY
22 LC_NUMERIC LC_TIME LC_ADDRESS LC_IDENTIFICATION
23 LC_MEASUREMENT LC_PAPER LC_TELEPHONE LC_SYNTAX
25 my @platform_categories;
27 sub is_category_valid($) {
28 my $cat_name = shift =~ s/^LC_//r;
30 # Recognize Configure option to exclude a category
31 return $Config{ccflags} !~ /\bD?NO_LOCALE_$cat_name\b/;
34 # LC_ALL can be -1 on some platforms. And, in fact the implementors could
35 # legally use any integer to represent any category. But it makes the most
36 # sense for them to have used small integers. Below, we create new locale
37 # numbers for ones missing from this machine. We make them very negative,
38 # hopefully more negative than anything likely to be a valid category on the
39 # platform, but also below is a check to be sure that our guess is valid.
40 my $max_bad_category_number = -1000000;
42 # Initialize this hash so that it looks like e.g.,
44 # where 6 is the value of &POSIX::LC_CTYPE
48 my $number_for_missing_category = $max_bad_category_number;
49 foreach my $name (@known_categories) {
50 my $number = eval "&POSIX::$name";
52 # Use a negative number (smaller than any legitimate category
53 # number) if the platform doesn't support this category, so we
54 # have an entry for all the ones that might be specified in calls
56 $number = $number_for_missing_category--;
58 elsif ( $number !~ / ^ -? \d+ $ /x
59 || $number <= $max_bad_category_number)
61 # We think this should be an int. And it has to be larger than
62 # any of our synthetic numbers.
63 die "Unexpected locale category number '$number' for $name"
66 push @platform_categories, $name;
70 $category_name{$number} = "$name";
71 $category_number{$name} = $number;
77 if (defined &main::diag) {
81 local($\, $", $,) = (undef, ' ', '');
82 print STDERR $message, "\n";
88 if (defined &main::fail) {
92 local($\, $", $,) = (undef, ' ', '');
93 print "not ok 0 $message\n";
97 sub _trylocale ($$$$) { # For use only by other functions in this file!
99 # Adds the locale given by the first parameter to the list given by the
100 # 3rd iff the platform supports the locale in each of the category numbers
101 # given by the 2nd parameter, which is either a single category or a
102 # reference to a list of categories.
104 # The 4th parameter is true if to accept locales that aren't apparently
105 # fully compatible with Perl.
108 my $categories = shift;
110 my $allow_incompatible = shift;
112 return if ! $locale || grep { $locale eq $_ } @$list;
114 # This is a toy (pig latin) locale that is not fully implemented on some
116 return if $locale =~ / ^ pig $ /ix;
118 # Certain platforms have a crippled locale system in which setlocale
119 # returns success for just about any possible locale name, but if anything
120 # actually happens as a result of the call, it is that the underlying
121 # locale is set to a system default, likely C or C.UTF-8. We can't test
122 # such systems fully, but we shouldn't disable the user from using
123 # locales, as it may work out for them (or not).
124 return if defined $Config{d_setlocale_accepts_any_locale_name}
125 && $locale !~ / ^ (?: C | POSIX | C\.UTF-8 ) $/ix;
127 $categories = [ $categories ] unless ref $categories;
132 use warnings 'locale';
134 local $SIG{__WARN__} = sub {
135 $badutf8 = 1 if grep { /Malformed UTF-8/ } @_;
136 $plays_well = 0 if grep {
137 /Locale .* may not work well(?#
138 )|The Perl program will use the expected meanings/i
142 # Incompatible locales aren't warned about unless using locales.
145 # Sort the input so CTYPE is first, COLLATE comes after all but ALL. This
146 # is because locale.c detects bad locales only with CTYPE, and COLLATE on
147 # some platforms can core dump if it is a bad locale.
152 foreach my $category (@$categories) {
153 die "category '$category' must instead be a number"
154 unless $category =~ / ^ -? \d+ $ /x;
155 if ($category_name{$category} eq 'CTYPE') {
158 elsif ($category_name{$category} eq 'ALL') {
161 elsif ($category_name{$category} eq 'COLLATE') {
165 push @sorted, $category unless grep { $_ == $category } @sorted;
168 push @sorted, $category_number{'COLLATE'} if $has_collate;
169 push @sorted, $category_number{'ALL'} if $has_all;
170 unshift @sorted, $category_number{'CTYPE'} if $has_ctype || ! $allow_incompatible;
172 foreach my $category (@sorted) {
173 return unless setlocale($category, $locale);
174 last if $badutf8 || ! $plays_well;
178 _my_fail("Verify locale name doesn't contain malformed utf8");
181 push @$list, $locale if $plays_well || $allow_incompatible;
184 sub _decode_encodings { # For use only by other functions in this file!
187 foreach (split(/ /, shift)) {
189 push @enc, "ISO8859-$1";
190 push @enc, "iso8859$1"; # HP
192 push @enc, "roman8"; # HP
195 push @enc, "$_.UTF-8";
196 push @enc, "$_.65001"; # Windows UTF-8
197 push @enc, "$_.ACP"; # Windows ANSI code page
198 push @enc, "$_.OCP"; # Windows OEM code page
199 push @enc, "$_.1252"; # Windows
202 if ($^O eq 'os390') {
203 push @enc, qw(IBM-037 IBM-819 IBM-1047);
206 push @enc, "65001"; # Windows UTF-8
211 sub valid_locale_categories() {
212 # Returns a list of the locale categories (expressed as strings, like
213 # "LC_ALL) known to this program that are available on this platform.
215 return grep { is_category_valid($_) } @platform_categories;
218 sub locales_enabled(;$) {
219 # If no parameter is specified, the function returns 1 if there is any
220 # "safe" locale handling available to the caller; otherwise 0. Safeness
221 # is defined here as the caller operating in the main thread of a program,
222 # or if threaded locales are safe on the platform and Configured to be
223 # used. This sub is used for testing purposes, and for those, this
224 # definition of safety is sufficient, and necessary to get some tests to
225 # run on certain configurations on certain platforms. But beware that the
226 # main thread can change the locale of any subthreads unless
227 # ${^SAFE_LOCALES} is non-zero.
229 # Use the optional parameter to discover if a particular category or
230 # categories are available on the system. 1 is returned if the global
231 # criteria described in the previous paragraph are true, AND if all the
232 # specified categories are available on the platform and Configured to be
233 # used. Otherwise 0 is returned. The parameter is either a single POSIX
234 # locale category or a reference to a list of them. Each category must be
235 # its name as a string, like 'LC_TIME' (the initial 'LC_' is optional), or
236 # the number this platform uses to signify the category (e.g.,
237 # 'locales_enabled(&POSIX::LC_CTYPE)'
239 # When the function returns 1 and a parameter was specified as a list
240 # reference, the reference will be altered on return to point to an
241 # equivalent list such that the categories are numeric instead of strings
242 # and sorted to meet the input expectations of _trylocale().
244 # It is a fatal error to call this with something that isn't a known
245 # category to this file. If this happens, look first for a typo, and
246 # second if you are using a category unknown to Perl. In the latter case
247 # a bug report should be submitted.
249 # khw cargo-culted the '?' in the pattern on the next line.
250 return 0 if $Config{ccflags} =~ /\bD?NO_LOCALE\b/;
252 # If we can't load the POSIX XS module, we can't have locales even if they
253 # normally would be available
254 return 0 if ! defined &DynaLoader::boot_DynaLoader;
256 # Don't test locales where they aren't safe. On systems with unsafe
257 # threads, for the purposes of testing, we consider the main thread safe,
258 # and all other threads unsafe.
259 if (! ${^SAFE_LOCALES}) {
261 return 0 if threads->tid() != 0;
264 # If no setlocale, we need the POSIX 2008 alternatives
265 if (! $Config{d_setlocale}) {
266 return 0 if $Config{ccflags} =~ /\bD?NO_POSIX_2008_LOCALE\b/;
267 return 0 unless $Config{d_newlocale};
268 return 0 unless $Config{d_uselocale};
269 return 0 unless $Config{d_duplocale};
270 return 0 unless $Config{d_freelocale};
273 # Done with the global possibilities. Now check if any passed in category
276 my $categories_ref = $_[0];
277 my $return_categories_numbers = 0;
278 my @categories_numbers;
280 my $has_LC_COLLATE = 0;
282 if (defined $categories_ref) {
283 my @local_categories_copy;
285 my $reftype = ref $categories_ref;
286 if ($reftype eq 'ARRAY') {
287 @local_categories_copy = @$categories_ref;
288 $return_categories_numbers = 1;
290 elsif ($reftype ne "") {
291 die "Parameter to locales_enabled() must be an ARRAY;"
292 . " instead you used a $reftype";
294 else { # Single category passed in
295 @local_categories_copy = $categories_ref;
298 for my $category_name_or_number (@local_categories_copy) {
301 if ($category_name_or_number =~ / ^ -? \d+ $ /x) {
302 $number = $category_name_or_number;
303 die "Invalid locale category number '$number'"
304 unless grep { $number == $_ } keys %category_name;
305 $name = $category_name{$number};
308 $name = $category_name_or_number;
309 $name =~ s/ ^ LC_ //x;
310 foreach my $trial (keys %category_name) {
311 if ($category_name{$trial} eq $name) {
316 die "Invalid locale category name '$name'"
317 unless defined $number;
320 return 0 if $number <= $max_bad_category_number
321 || ! is_category_valid($name);
324 eval "defined &POSIX::LC_$name";
327 if ($return_categories_numbers) {
328 if ($name eq 'CTYPE') {
329 unshift @categories_numbers, $number; # Always first
331 elsif ($name eq 'ALL') {
334 elsif ($name eq 'COLLATE') {
338 push @categories_numbers, $number;
344 if ($return_categories_numbers) {
346 # COLLATE comes after all other locales except ALL, which comes last
347 if ($has_LC_COLLATE) {
348 push @categories_numbers, $category_number{'COLLATE'};
351 push @categories_numbers, $category_number{'ALL'};
354 @$categories_ref = @categories_numbers;
361 sub find_locales ($;$) {
363 # Returns an array of all the locales we found on the system. If the
364 # optional 2nd parameter is non-zero, the list includes all found locales;
365 # otherwise it is restricted to those locales that play well with Perl, as
366 # far as we can easily determine.
368 # The first parameter is either a single locale category or a reference to
369 # a list of categories to find valid locales for it (or in the case of
370 # multiple) for all of them. Each category can be a name (like 'LC_ALL'
371 # or simply 'ALL') or the C enum value for the category.
373 my $input_categories = shift;
374 my $allow_incompatible = shift // 0;
376 my @categories = (ref $input_categories) ? $input_categories->@* : $input_categories;
377 return unless locales_enabled(\@categories);
379 # Note, the subroutine call above converts the $categories into a form
380 # suitable for _trylocale().
382 # Visual C's CRT goes silly on strings of the form "en_US.ISO8859-1"
383 # and mingw32 uses said silly CRT
384 # This doesn't seem to be an issue any more, at least on Windows XP,
385 # so re-enable the tests for Windows XP onwards.
386 my $winxp = ($^O eq 'MSWin32' && defined &Win32::GetOSVersion &&
387 join('.', (Win32::GetOSVersion())[1..2]) >= 5.1);
388 return if (($^O eq 'MSWin32' && !$winxp)
389 && $Config{cc} =~ /^(cl|gcc|g\+\+|ici)/i);
392 _trylocale("C", \@categories, \@Locale, $allow_incompatible);
393 _trylocale("POSIX", \@categories, \@Locale, $allow_incompatible);
395 if ($Config{d_has_C_UTF8} && $Config{d_has_C_UTF8} eq 'true') {
396 _trylocale("C.UTF-8", \@categories, \@Locale, $allow_incompatible);
399 # There's no point in looking at anything more if we know that setlocale
400 # will return success on any garbage or non-garbage name.
401 return sort @Locale if defined $Config{d_setlocale_accepts_any_locale_name};
404 _trylocale("ISO8859-$_", \@categories, \@Locale, $allow_incompatible);
405 _trylocale("iso8859$_", \@categories, \@Locale, $allow_incompatible);
406 _trylocale("iso8859-$_", \@categories, \@Locale, $allow_incompatible);
407 _trylocale("iso_8859_$_", \@categories, \@Locale, $allow_incompatible);
408 _trylocale("isolatin$_", \@categories, \@Locale, $allow_incompatible);
409 _trylocale("isolatin-$_", \@categories, \@Locale, $allow_incompatible);
410 _trylocale("iso_latin_$_", \@categories, \@Locale, $allow_incompatible);
413 # Sanitize the environment so that we can run the external 'locale'
414 # program without the taint mode getting grumpy.
416 # $ENV{PATH} is special in VMS.
417 delete local $ENV{PATH} if $^O ne 'VMS' or $Config{d_setenv};
419 # Other subversive stuff.
420 delete local @ENV{qw(IFS CDPATH ENV BASH_ENV)};
422 if (-x "/usr/bin/locale"
423 && open(LOCALES, '-|', "/usr/bin/locale -a 2>/dev/null"))
426 # It seems that /usr/bin/locale steadfastly outputs 8 bit data, which
427 # ain't great when we're running this testPERL_UNICODE= so that utf8
428 # locales will cause all IO hadles to default to (assume) utf8
429 next unless utf8::valid($_);
431 _trylocale($_, \@categories, \@Locale, $allow_incompatible);
434 } elsif ($^O eq 'VMS'
435 && defined($ENV{'SYS$I18N_LOCALE'})
436 && -d 'SYS$I18N_LOCALE')
438 # The SYS$I18N_LOCALE logical name search list was not present on
439 # VAX VMS V5.5-12, but was on AXP && VAX VMS V6.2 as well as later versions.
440 opendir(LOCALES, "SYS\$I18N_LOCALE:");
441 while ($_ = readdir(LOCALES)) {
443 _trylocale($_, \@categories, \@Locale, $allow_incompatible);
446 } elsif (($^O eq 'openbsd' || $^O eq 'bitrig' ) && -e '/usr/share/locale') {
448 # OpenBSD doesn't have a locale executable, so reading
449 # /usr/share/locale is much easier and faster than the last resort
452 opendir(LOCALES, '/usr/share/locale');
453 while ($_ = readdir(LOCALES)) {
455 _trylocale($_, \@categories, \@Locale, $allow_incompatible);
458 } else { # Final fallback. Try our list of locales hard-coded here
460 # This is going to be slow.
463 # Locales whose name differs if the utf8 bit is on are stored in these
464 # two files with appropriate encodings.
465 my $data_file = ($^H & 0x08 || (${^OPEN} || "") =~ /:utf8/)
466 ? _source_location() . "/lib/locale/utf8"
467 : _source_location() . "/lib/locale/latin1";
469 @Data = do $data_file;
472 _my_diag(__FILE__ . ":" . __LINE__ . ": '$data_file' doesn't exist");
475 # The rest of the locales are in this file.
476 state @my_data = <DATA>; close DATA if fileno DATA;
477 push @Data, @my_data;
479 foreach my $line (@Data) {
481 my ($locale_name, $language_codes, $country_codes, $encodings) =
483 _my_diag(__FILE__ . ":" . __LINE__ . ": Unexpected syntax in '$line'")
484 unless defined $locale_name;
485 my @enc = _decode_encodings($encodings);
486 foreach my $loc (split(/ /, $locale_name)) {
487 _trylocale($loc, \@categories, \@Locale, $allow_incompatible);
488 foreach my $enc (@enc) {
489 _trylocale("$loc.$enc", \@categories, \@Locale,
490 $allow_incompatible);
493 foreach my $enc (@enc) {
494 _trylocale("$loc.$enc", \@categories, \@Locale,
495 $allow_incompatible);
498 foreach my $lang (split(/ /, $language_codes)) {
499 _trylocale($lang, \@categories, \@Locale, $allow_incompatible);
500 foreach my $country (split(/ /, $country_codes)) {
501 my $lc = "${lang}_${country}";
502 _trylocale($lc, \@categories, \@Locale, $allow_incompatible);
503 foreach my $enc (@enc) {
504 _trylocale("$lc.$enc", \@categories, \@Locale,
505 $allow_incompatible);
507 my $lC = "${lang}_\U${country}";
508 _trylocale($lC, \@categories, \@Locale, $allow_incompatible);
509 foreach my $enc (@enc) {
510 _trylocale("$lC.$enc", \@categories, \@Locale,
511 $allow_incompatible);
518 @Locale = sort @Locale;
523 sub is_locale_utf8 ($) { # Return a boolean as to if core Perl thinks the input
526 # On z/OS, even locales marked as UTF-8 aren't.
527 return 0 if ord "A" != 65;
529 return 0 unless locales_enabled('LC_CTYPE');
534 no warnings 'locale'; # We may be trying out a weird locale
536 my $save_locale = setlocale(&POSIX::LC_CTYPE());
537 if (! $save_locale) {
538 ok(0, "Verify could save previous locale");
542 if (! setlocale(&POSIX::LC_CTYPE(), $locale)) {
543 ok(0, "Verify could setlocale to $locale");
549 # Use an op that gives different results for UTF-8 than any other locale.
550 # If a platform has UTF-8 locales, there should be at least one locale on
551 # most platforms with UTF-8 in its name, so if there is a bug in the op
552 # giving a false negative, we should get a failure for those locales as we
553 # go through testing all the locales on the platform.
554 if (CORE::fc(chr utf8::unicode_to_native(0xdf)) ne "ss") {
555 if ($locale =~ /UTF-?8/i) {
556 ok (0, "Verify $locale with UTF-8 in name is a UTF-8 locale");
563 die "Couldn't restore locale '$save_locale'"
564 unless setlocale(&POSIX::LC_CTYPE(), $save_locale);
569 sub find_utf8_ctype_locales (;$) { # Return the names of the locales that core
570 # Perl thinks are UTF-8 LC_CTYPE locales.
571 # Optional parameter is a reference to a
572 # list of locales to try; if omitted, this
573 # tries all locales it can find on the
575 return unless locales_enabled('LC_CTYPE');
577 my $locales_ref = shift;
580 if (! defined $locales_ref) {
582 my @locales = find_locales(&POSIX::LC_CTYPE());
583 $locales_ref = \@locales;
586 foreach my $locale (@$locales_ref) {
587 push @return, $locale if is_locale_utf8($locale);
594 sub find_utf8_ctype_locale (;$) { # Return the name of a locale that core Perl
595 # thinks is a UTF-8 LC_CTYPE non-turkic
597 # Optional parameter is a reference to a
598 # list of locales to try; if omitted, this
599 # tries all locales it can find on the
601 my $try_locales_ref = shift;
603 my @utf8_locales = find_utf8_ctype_locales($try_locales_ref);
604 my @turkic_locales = find_utf8_turkic_locales($try_locales_ref);
608 # Create undef elements in the hash for turkic locales
609 @seen_turkic{@turkic_locales} = ();
611 foreach my $locale (@utf8_locales) {
612 return $locale unless exists $seen_turkic{$locale};
618 sub find_utf8_turkic_locales (;$) {
620 # Return the name of all the locales that core Perl thinks are UTF-8
621 # Turkic LC_CTYPE. Optional parameter is a reference to a list of locales
622 # to try; if omitted, this tries all locales it can find on the platform
626 return unless locales_enabled('LC_CTYPE');
628 my $save_locale = setlocale(&POSIX::LC_CTYPE());
629 foreach my $locale (find_utf8_ctype_locales(shift)) {
631 setlocale(&POSIX::LC_CTYPE(), $locale);
632 push @return, $locale if uc('i') eq "\x{130}";
634 setlocale(&POSIX::LC_CTYPE(), $save_locale);
639 sub find_utf8_turkic_locale (;$) {
640 my @turkics = find_utf8_turkic_locales(shift);
642 return unless @turkics;
647 # returns full path to the directory containing the current source
648 # file, inspired by mauke's Dir::Self
649 sub _source_location {
652 my $caller_filename = (caller)[1];
654 my $loc = File::Spec->rel2abs(
656 (File::Spec->splitpath($caller_filename))[0, 1], ''
660 return ($loc =~ /^(.*)$/)[0]; # untaint
665 # Format of data is: locale_name, language_codes, country_codes, encodings
668 Arabic:ar:dz eg sa:6 arabic8
669 Brezhoneg Breton:br:fr:1 15
670 Bulgarski Bulgarian:bg:bg:5
671 Chinese:zh:cn tw:cn.EUC eucCN eucTW euc.CN euc.TW Big5 GB2312 tw.EUC
672 Hrvatski Croatian:hr:hr:2
673 Cymraeg Welsh:cy:cy:1 14 15
675 Dansk Danish:da:dk:1 15
676 Nederlands Dutch:nl:be nl:1 15
677 English American British:en:au ca gb ie nz us uk zw:1 15 cp850
679 Eesti Estonian:et:ee:4 6 13
680 Suomi Finnish:fi:fi:1 15
682 Deutsch German:de:at be ch de lu:1 15
683 Euskaraz Basque:eu:es fr:1 15
684 Galego Galician:gl:es:1 15
685 Ellada Greek:el:gr:7 g8
687 Greenlandic:kl:gl:4 6
688 Hebrew:iw:il:8 hebrew8
690 Indonesian:id:id:1 15
691 Gaeilge Irish:ga:IE:1 14 15
692 Italiano Italian:it:ch it:1 15
693 Nihongo Japanese:ja:jp:euc eucJP jp.EUC sjis
695 Latine Latin:la:va:1 15
697 Lithuanian:lt:lt:4 6 13
698 Macedonian:mk:mk:1 15
701 Norsk Norwegian:no no\@nynorsk nb nn:no:1 15
703 Polski Polish:pl:pl:2
705 Russki Russian:ru:ru su ua:5 koi8 koi8r KOI8-R koi8u cp1251 cp866
706 Serbski Serbian:sr:yu:5
708 Slovene Slovenian:sl:si:2
709 Sqhip Albanian:sq:sq:1 15
710 Svenska Swedish:sv:fi se:1 15
712 Turkish:tr:tr:9 turkish8