Perl_set_numeric_radix(pTHX)
{
#ifdef USE_LOCALE_NUMERIC
- dVAR;
# ifdef HAS_LOCALECONV
const struct lconv* const lc = localeconv();
sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
else
PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
- if (! is_ascii_string((U8 *) lc->decimal_point, 0)
+ if (! is_invariant_string((U8 *) lc->decimal_point, 0)
&& is_utf8_string((U8 *) lc->decimal_point, 0)
- && is_cur_LC_category_utf8(LC_NUMERIC))
+ && _is_cur_LC_category_utf8(LC_NUMERIC))
{
SvUTF8_on(PL_numeric_radix_sv);
}
#endif /* USE_LOCALE_NUMERIC */
}
+/* Is the C string input 'name' "C" or "POSIX"? If so, and 'name' is the
+ * return of setlocale(), then this is extremely likely to be the C or POSIX
+ * locale. However, the output of setlocale() is documented to be opaque, but
+ * the odds are extremely small that it would return these two strings for some
+ * other locale. Note that VMS in these two locales includes many non-ASCII
+ * characters as controls and punctuation (below are hex bytes):
+ * cntrl: 00-1F 7F 84-97 9B-9F
+ * punct: 21-2F 3A-40 5B-60 7B-7E A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
+ * Oddly, none there are listed as alphas, though some represent alphabetics
+ * http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
+#define isNAME_C_OR_POSIX(name) ((name) != NULL \
+ && ((*(name) == 'C' && (*(name + 1)) == '\0') \
+ || strEQ((name), "POSIX")))
+
void
Perl_new_numeric(pTHX_ const char *newnum)
{
* dot.
*
* This sets several interpreter-level variables:
- * PL_numeric_name The default locale's name: a copy of 'newnum'
+ * PL_numeric_name The underlying locale's name: a copy of 'newnum'
* PL_numeric_local A boolean indicating if the toggled state is such
- * that the current locale is the default locale
- * PL_numeric_standard A boolean indicating if the toggled state is such
- * that the current locale is the C locale
+ * that the current locale is the program's underlying
+ * locale
+ * PL_numeric_standard An int indicating if the toggled state is such
+ * that the current locale is the C locale. If non-zero,
+ * it is in C; if > 1, it means it may not be toggled away
+ * from C.
* Note that both of the last two variables can be true at the same time,
* if the underlying locale is C. (Toggling is a no-op under these
* circumstances.)
* POSIX::setlocale() */
char *save_newnum;
- dVAR;
if (! newnum) {
Safefree(PL_numeric_name);
PL_numeric_name = save_newnum;
}
- PL_numeric_standard = ((*save_newnum == 'C' && save_newnum[1] == '\0')
- || strEQ(save_newnum, "POSIX"));
+ PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
PL_numeric_local = TRUE;
+
+ /* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
+ * have to worry about the radix being a non-dot. (Core operations that
+ * need the underlying locale change to it temporarily). */
+ set_numeric_standard();
+
set_numeric_radix();
+#else
+ PERL_UNUSED_ARG(newnum);
#endif /* USE_LOCALE_NUMERIC */
}
Perl_set_numeric_standard(pTHX)
{
#ifdef USE_LOCALE_NUMERIC
- dVAR;
-
- /* Toggle the LC_NUMERIC locale to C, if not already there. Probably
- * should use the macros like SET_NUMERIC_STANDARD() in perl.h instead of
- * calling this directly. */
-
- if (! PL_numeric_standard) {
- setlocale(LC_NUMERIC, "C");
- PL_numeric_standard = TRUE;
- PL_numeric_local = FALSE;
- set_numeric_radix();
- }
+ /* Toggle the LC_NUMERIC locale to C. Most code should use the macros like
+ * SET_NUMERIC_STANDARD() in perl.h instead of calling this directly. The
+ * macro avoids calling this routine if toggling isn't necessary according
+ * to our records (which could be wrong if some XS code has changed the
+ * locale behind our back) */
+
+ setlocale(LC_NUMERIC, "C");
+ PL_numeric_standard = TRUE;
+ PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
+ set_numeric_radix();
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Underlying LC_NUMERIC locale now is C\n"));
Perl_set_numeric_local(pTHX)
{
#ifdef USE_LOCALE_NUMERIC
- dVAR;
-
- /* Toggle the LC_NUMERIC locale to the current underlying default, if not
- * already there. Probably should use the macros like SET_NUMERIC_LOCAL()
- * in perl.h instead of calling this directly. */
-
- if (! PL_numeric_local) {
- setlocale(LC_NUMERIC, PL_numeric_name);
- PL_numeric_standard = FALSE;
- PL_numeric_local = TRUE;
- set_numeric_radix();
- }
+ /* Toggle the LC_NUMERIC locale to the current underlying default. Most
+ * code should use the macros like SET_NUMERIC_LOCAL() in perl.h instead of
+ * calling this directly. The macro avoids calling this routine if
+ * toggling isn't necessary according to our records (which could be wrong
+ * if some XS code has changed the locale behind our back) */
+
+ setlocale(LC_NUMERIC, PL_numeric_name);
+ PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
+ PL_numeric_local = TRUE;
+ set_numeric_radix();
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Underlying LC_NUMERIC locale now is %s\n",
PL_numeric_name));
PERL_ARGS_ASSERT_NEW_CTYPE;
- PL_in_utf8_CTYPE_locale = is_cur_LC_category_utf8(LC_CTYPE);
+ PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
/* A UTF-8 locale gets standard rules. But note that code still has to
* handle this specially because of the three problematic code points */
Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
}
else {
+ /* Assume enough space for every character being bad. 4 spaces each
+ * for the 94 printable characters that are output like "'x' "; and 5
+ * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
+ * NUL */
+ char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
+
+ bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
+ no check */
+ bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
+ to start */
+ unsigned int bad_count = 0; /* Count of bad characters */
+
for (i = 0; i < 256; i++) {
if (isUPPER_LC((U8) i))
PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
else
PL_fold_locale[i] = (U8) i;
+
+ /* If checking for locale problems, see if the native ASCII-range
+ * printables plus \n and \t are in their expected categories in
+ * the new locale. If not, this could mean big trouble, upending
+ * Perl's and most programs' assumptions, like having a
+ * metacharacter with special meaning become a \w. Fortunately,
+ * it's very rare to find locales that aren't supersets of ASCII
+ * nowadays. It isn't a problem for most controls to be changed
+ * into something else; we check only \n and \t, though perhaps \r
+ * could be an issue as well. */
+ if (check_for_problems
+ && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
+ {
+ if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
+ || (isPUNCT_A(i) && ! isPUNCT_LC(i))
+ || (isBLANK_A(i) && ! isBLANK_LC(i))
+ || (i == '\n' && ! isCNTRL_LC(i)))
+ {
+ if (bad_count) { /* Separate multiple entries with a
+ blank */
+ bad_chars_list[bad_count++] = ' ';
+ }
+ bad_chars_list[bad_count++] = '\'';
+ if (isPRINT_A(i)) {
+ bad_chars_list[bad_count++] = (char) i;
+ }
+ else {
+ bad_chars_list[bad_count++] = '\\';
+ if (i == '\n') {
+ bad_chars_list[bad_count++] = 'n';
+ }
+ else {
+ assert(i == '\t');
+ bad_chars_list[bad_count++] = 't';
+ }
+ }
+ bad_chars_list[bad_count++] = '\'';
+ bad_chars_list[bad_count] = '\0';
+ }
+ }
+ }
+
+#ifdef MB_CUR_MAX
+ /* We only handle single-byte locales (outside of UTF-8 ones; so if
+ * this locale requires than one byte, there are going to be
+ * problems. */
+ if (check_for_problems && MB_CUR_MAX > 1) {
+ multi_byte_locale = TRUE;
+ }
+#endif
+
+ if (bad_count || multi_byte_locale) {
+
+ /* We have to save 'newctype' because the setlocale() just below
+ * may destroy it. The next setlocale() further down should
+ * restore it properly so that the intermediate change here is
+ * transparent to this function's caller */
+ const char * const badlocale = savepv(newctype);
+
+ setlocale(LC_CTYPE, "C");
+ Perl_warner(aTHX_ packWARN(WARN_LOCALE),
+ "Locale '%s' may not work well.%s%s%s\n",
+ badlocale,
+ (multi_byte_locale)
+ ? " Some characters in it are not recognized by"
+ " Perl."
+ : "",
+ (bad_count)
+ ? "\nThe following characters (and maybe others)"
+ " may not have the same meaning as the Perl"
+ " program expects:\n"
+ : "",
+ (bad_count)
+ ? bad_chars_list
+ : ""
+ );
+ setlocale(LC_CTYPE, badlocale);
}
}
* should be called directly only from this file and from
* POSIX::setlocale() */
- dVAR;
-
if (! newcoll) {
if (PL_collation_name) {
++PL_collation_ix;
++PL_collation_ix;
Safefree(PL_collation_name);
PL_collation_name = stdize_locale(savepv(newcoll));
- PL_collation_standard = ((*newcoll == 'C' && newcoll[1] == '\0')
- || strEQ(newcoll, "POSIX"));
+ PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
{
/* 2: at most so many chars ('a', 'b'). */
}
}
+#else
+ PERL_UNUSED_ARG(newcoll);
#endif /* USE_LOCALE_COLLATE */
}
* otherwise to use the particular category's variable if set; otherwise to
* use the LANG variable. */
+ bool override_LC_ALL = 0;
+ char * result;
+
if (locale && strEQ(locale, "")) {
# ifdef LC_ALL
locale = PerlEnv_getenv("LC_ALL");
switch (category) {
# ifdef LC_ALL
case LC_ALL:
+ override_LC_ALL = TRUE;
break; /* We already know its variable isn't set */
# endif
# ifdef USE_LOCALE_TIME
# endif
}
- return setlocale(category, locale);
+ result = setlocale(category, locale);
+
+ if (! override_LC_ALL) {
+ return result;
+ }
+
+ /* Here the input locale was LC_ALL, and we have set it to what is in the
+ * LANG variable or the system default if there is no LANG. But these have
+ * lower priority than the other LC_foo variables, so override it for each
+ * one that is set. (If they are set to "", it means to use the same thing
+ * we just set LC_ALL to, so can skip) */
+# ifdef USE_LOCALE_TIME
+ result = PerlEnv_getenv("LC_TIME");
+ if (result && strNE(result, "")) {
+ setlocale(LC_TIME, result);
+ }
+# endif
+# ifdef USE_LOCALE_CTYPE
+ result = PerlEnv_getenv("LC_CTYPE");
+ if (result && strNE(result, "")) {
+ setlocale(LC_CTYPE, result);
+ }
+# endif
+# ifdef USE_LOCALE_COLLATE
+ result = PerlEnv_getenv("LC_COLLATE");
+ if (result && strNE(result, "")) {
+ setlocale(LC_COLLATE, result);
+ }
+# endif
+# ifdef USE_LOCALE_MONETARY
+ result = PerlEnv_getenv("LC_MONETARY");
+ if (result && strNE(result, "")) {
+ setlocale(LC_MONETARY, result);
+ }
+# endif
+# ifdef USE_LOCALE_NUMERIC
+ result = PerlEnv_getenv("LC_NUMERIC");
+ if (result && strNE(result, "")) {
+ setlocale(LC_NUMERIC, result);
+ }
+# endif
+# ifdef USE_LOCALE_MESSAGES
+ result = PerlEnv_getenv("LC_MESSAGES");
+ if (result && strNE(result, "")) {
+ setlocale(LC_MESSAGES, result);
+ }
+# endif
+
+ return setlocale(LC_ALL, NULL);
+
}
#endif
int ok = 1;
#if defined(USE_LOCALE)
- dVAR;
-
#ifdef USE_LOCALE_CTYPE
char *curctype = NULL;
#endif /* USE_LOCALE_CTYPE */
#endif /* USE_LOCALE_NUMERIC */
#ifdef __GLIBC__
char * const language = PerlEnv_getenv("LANGUAGE");
-#else
- const char * const language = NULL;
#endif
/* NULL uses the existing already set up locale */
const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
? NULL
: "";
- const char* trial_locales[5] = { setlocale_init }; /* 5 = 1 each for "",
- LC_ALL, LANG, "", C
- */
- unsigned int trial_locales_count = 1;
+ const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
+ unsigned int trial_locales_count;
char * const lc_all = PerlEnv_getenv("LC_ALL");
char * const lang = PerlEnv_getenv("LANG");
bool setlocale_failure = FALSE;
char *p;
const bool locwarn = (printwarn > 1 ||
(printwarn &&
- (!(p = PerlEnv_getenv("PERL_BADLANG")) || atoi(p))));
+ (!(p = PerlEnv_getenv("PERL_BADLANG")) ||
+ grok_atou(p, NULL))));
bool done = FALSE;
- const char *description;
+#ifdef WIN32
+ /* In some systems you can find out the system default locale
+ * and use that as the fallback locale. */
+# define SYSTEM_DEFAULT_LOCALE
+#endif
+#ifdef SYSTEM_DEFAULT_LOCALE
const char *system_default_locale = NULL;
-
+#endif
#ifndef LOCALE_ENVIRON_REQUIRED
PERL_UNUSED_VAR(done);
/* We try each locale in the list until we get one that works, or exhaust
* the list */
+ trial_locales[0] = setlocale_init;
+ trial_locales_count = 1;
for (i= 0; i < trial_locales_count; i++) {
const char * trial_locale = trial_locales[i];
* sense */
setlocale_failure = FALSE;
-#ifdef WIN32
-
+#ifdef SYSTEM_DEFAULT_LOCALE
+# ifdef WIN32
/* On Windows machines, an entry of "" after the 0th means to use
* the system default locale, which we now proceed to get. */
if (strEQ(trial_locale, "")) {
trial_locale = system_default_locale;
}
-#endif
+# endif /* WIN32 */
+#endif /* SYSTEM_DEFAULT_LOCALE */
}
#ifdef LC_ALL
description = "the standard locale";
name = "C";
}
+#ifdef SYSTEM_DEFAULT_LOCALE
else if (strEQ(trial_locales[i], "")) {
description = "the system default locale";
if (system_default_locale) {
name = system_default_locale;
}
}
+#endif /* SYSTEM_DEFAULT_LOCALE */
else {
description = "a fallback locale";
name = trial_locales[i];
* $ENV{PERL_UNICODE}) are true, perl.c:S_parse_body() will turn on the
* PerlIO :utf8 layer on STDIN, STDOUT, STDERR, _and_ the default open
* discipline. */
- PL_utf8locale = is_cur_LC_category_utf8(LC_CTYPE);
+ PL_utf8locale = _is_cur_LC_category_utf8(LC_CTYPE);
/* Set PL_unicode to $ENV{PERL_UNICODE} if using PerlIO.
This is an alternative to using the -C command line switch
Safefree(curnum);
#endif /* USE_LOCALE_NUMERIC */
+#else /* !USE_LOCALE */
+ PERL_UNUSED_ARG(printwarn);
#endif /* USE_LOCALE */
return ok;
char *
Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
{
- dVAR;
char *xbuf;
STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
#ifdef USE_LOCALE
-STATIC bool
-S_is_cur_LC_category_utf8(pTHX_ int category)
+bool
+Perl__is_cur_LC_category_utf8(pTHX_ int category)
{
/* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
* otherwise. 'category' may not be LC_ALL. If the platform doesn't have
* nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
- * could give the wrong result. It errs on the side of not being a UTF-8
- * locale. */
+ * could give the wrong result. The result will very likely be correct for
+ * languages that have commonly used non-ASCII characters, but for notably
+ * English, it comes down to if the locale's name ends in something like
+ * "UTF-8". It errs on the side of not being a UTF-8 locale. */
char *save_input_locale = NULL;
STRLEN final_pos;
return FALSE; /* XXX maybe should croak */
}
save_input_locale = stdize_locale(savepv(save_input_locale));
- if ((*save_input_locale == 'C' && save_input_locale[1] == '\0')
- || strEQ(save_input_locale, "POSIX"))
- {
+ if (isNAME_C_OR_POSIX(save_input_locale)) {
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Current locale for category %d is %s\n",
category, save_input_locale));
if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
/* Get the current LC_CTYPE locale */
- save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL)));
+ save_ctype_locale = setlocale(LC_CTYPE, NULL);
if (! save_ctype_locale) {
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Could not find current locale for LC_CTYPE\n"));
goto cant_use_nllanginfo;
}
+ save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
/* If LC_CTYPE and the desired category use the same locale, this
* means that finding the value for LC_CTYPE is the same as finding
# if defined(HAS_NL_LANGINFO) && defined(CODESET)
{
- char *codeset = savepv(nl_langinfo(CODESET));
+ char *codeset = nl_langinfo(CODESET);
if (codeset && strNE(codeset, "")) {
+ codeset = savepv(codeset);
/* If we switched LC_CTYPE, switch back */
if (save_ctype_locale) {
* result */
if (is_utf8) {
wchar_t wc;
- GCC_DIAG_IGNORE(-Wunused-result);
- (void) mbtowc(&wc, NULL, 0); /* Reset any shift state */
- GCC_DIAG_RESTORE;
+ PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
errno = 0;
- if (mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
+ if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
!= strlen(HYPHEN_UTF8)
|| wc != (wchar_t) 0x2010)
{
is_utf8 = FALSE;
- DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", wc));
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
DEBUG_L(PerlIO_printf(Perl_debug_log,
"\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
cant_use_nllanginfo:
-#endif /* HAS_NL_LANGINFO etc */
-
- /* nl_langinfo not available or failed somehow. Look at the locale name to
- * see if it matches qr/UTF -? 8 /ix */
+#else /* nl_langinfo should work if available, so don't bother compiling this
+ fallback code. The final fallback of looking at the name is
+ compiled, and will be executed if nl_langinfo fails */
- final_pos = strlen(save_input_locale) - 1;
- if (final_pos >= 3) {
- char *name = save_input_locale;
-
- /* Find next 'U' or 'u' and look from there */
- while ((name += strcspn(name, "Uu") + 1)
- <= save_input_locale + final_pos - 2)
- {
- if (toFOLD(*(name)) != 't'
- || toFOLD(*(name + 1)) != 'f')
- {
- continue;
- }
- name += 2;
- if (*(name) == '-') {
- if ((name > save_input_locale + final_pos - 1)) {
- break;
- }
- name++;
- }
- if (*(name) == '8') {
- Safefree(save_input_locale);
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s ends with UTF-8 in name\n",
- save_input_locale));
- return TRUE;
- }
- }
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s doesn't end with UTF-8 in name\n",
- save_input_locale));
- }
-
-#ifdef WIN32
- /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
- if (final_pos >= 4
- && *(save_input_locale + final_pos - 0) == '1'
- && *(save_input_locale + final_pos - 1) == '0'
- && *(save_input_locale + final_pos - 2) == '0'
- && *(save_input_locale + final_pos - 3) == '5'
- && *(save_input_locale + final_pos - 4) == '6')
- {
- Safefree(save_input_locale);
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s ends with 10056 in name, is UTF-8 locale\n",
- save_input_locale));
- return TRUE;
- }
-#endif
-
- /* Other common encodings are the ISO 8859 series, which aren't UTF-8 */
- if (instr(save_input_locale, "8859")) {
- Safefree(save_input_locale);
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s has 8859 in name, not UTF-8 locale\n",
- save_input_locale));
- return FALSE;
- }
+ /* nl_langinfo not available or failed somehow. Next try looking at the
+ * currency symbol to see if it disambiguates things. Often that will be
+ * in the native script, and if the symbol isn't in UTF-8, we know that the
+ * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
+ * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
+ * */
#ifdef HAS_LOCALECONV
-
# ifdef USE_LOCALE_MONETARY
-
- /* Here, there is nothing in the locale name to indicate whether the locale
- * is UTF-8 or not. This "name", the return of setlocale(), is actually
- * defined to be opaque, so we can't really rely on the absence of various
- * substrings in the name to indicate its UTF-8ness. Look at the locale's
- * currency symbol. Often that will be in the native script, and if the
- * symbol isn't in UTF-8, we know that the locale isn't. If it is
- * non-ASCII UTF-8, we infer that the locale is too.
- * To do this, like above for LC_CTYPE, we first set LC_MONETARY to the
- * locale of the desired category, if it isn't that locale already */
-
{
char *save_monetary_locale = NULL;
- bool illegal_utf8 = FALSE;
bool only_ascii = FALSE;
- const struct lconv* const lc = localeconv();
+ bool is_utf8 = FALSE;
+ struct lconv* lc;
+
+ /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
+ * the desired category, if it isn't that locale already */
if (category != LC_MONETARY) {
- save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY,
- NULL)));
+ save_monetary_locale = setlocale(LC_MONETARY, NULL);
if (! save_monetary_locale) {
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Could not find current locale for LC_MONETARY\n"));
goto cant_use_monetary;
}
+ save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
- if (strNE(save_monetary_locale, save_input_locale)) {
- if (! setlocale(LC_MONETARY, save_input_locale)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Could not change LC_MONETARY locale to %s\n",
- save_input_locale));
- Safefree(save_monetary_locale);
- goto cant_use_monetary;
- }
+ if (strEQ(save_monetary_locale, save_input_locale)) {
+ Safefree(save_monetary_locale);
+ save_monetary_locale = NULL;
+ }
+ else if (! setlocale(LC_MONETARY, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_MONETARY locale to %s\n",
+ save_input_locale));
+ Safefree(save_monetary_locale);
+ goto cant_use_monetary;
}
}
/* Here the current LC_MONETARY is set to the locale of the category
* whose information is desired. */
- if (lc && lc->currency_symbol) {
- if (! is_utf8_string((U8 *) lc->currency_symbol, 0)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Currency symbol for %s is not legal UTF-8\n",
- save_input_locale));
- illegal_utf8 = TRUE;
- }
- else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log, "Currency symbol for %s contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
- only_ascii = TRUE;
- }
+ lc = localeconv();
+ if (! lc
+ || ! lc->currency_symbol
+ || is_invariant_string((U8 *) lc->currency_symbol, 0))
+ {
+ 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));
+ only_ascii = TRUE;
+ }
+ else {
+ is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
}
/* If we changed it, restore LC_MONETARY to its original locale */
Safefree(save_monetary_locale);
}
- Safefree(save_input_locale);
-
- /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; otherwise
- * assume the locale is UTF-8 if and only if the symbol is non-ascii
- * UTF-8. (We can't really tell if the locale is UTF-8 or not if the
- * symbol is just a '$', so we err on the side of it not being UTF-8)
- * */
- DEBUG_L(PerlIO_printf(Perl_debug_log, "\tis_utf8=%d\n", (illegal_utf8)
- ? FALSE
- : ! only_ascii));
- return (illegal_utf8)
- ? FALSE
- : ! only_ascii;
+ if (! only_ascii) {
+ /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
+ * otherwise assume the locale is UTF-8 if and only if the symbol
+ * is non-ascii UTF-8. */
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
+ save_input_locale, is_utf8));
+ Safefree(save_input_locale);
+ return is_utf8;
+ }
}
cant_use_monetary:
# endif /* USE_LOCALE_MONETARY */
#endif /* HAS_LOCALECONV */
-#if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES)
+#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
+
+/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
+ * the names of the months and weekdays, timezone, and am/pm indicator */
+ {
+ char *save_time_locale = NULL;
+ int hour = 10;
+ bool is_dst = FALSE;
+ int dom = 1;
+ int month = 0;
+ int i;
+ char * formatted_time;
+
+
+ /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
+ * desired category, if it isn't that locale already */
+
+ if (category != LC_TIME) {
+
+ save_time_locale = setlocale(LC_TIME, NULL);
+ if (! save_time_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not find current locale for LC_TIME\n"));
+ goto cant_use_time;
+ }
+ save_time_locale = stdize_locale(savepv(save_time_locale));
+
+ if (strEQ(save_time_locale, save_input_locale)) {
+ Safefree(save_time_locale);
+ save_time_locale = NULL;
+ }
+ else if (! setlocale(LC_TIME, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_TIME locale to %s\n",
+ save_input_locale));
+ Safefree(save_time_locale);
+ goto cant_use_time;
+ }
+ }
+
+ /* Here the current LC_TIME is set to the locale of the category
+ * whose information is desired. Look at all the days of the week and
+ * month names, and the timezone and am/pm indicator for UTF-8 variant
+ * characters. The first such a one found will tell us if the locale
+ * is UTF-8 or not */
+
+ for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
+ formatted_time = my_strftime("%A %B %Z %p",
+ 0, 0, hour, dom, month, 112, 0, 0, is_dst);
+ if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
+
+ /* Here, we didn't find a non-ASCII. Try the next time through
+ * with the complemented dst and am/pm, and try with the next
+ * weekday. After we have gotten all weekdays, try the next
+ * month */
+ is_dst = ! is_dst;
+ hour = (hour + 12) % 24;
+ dom++;
+ if (i > 6) {
+ month++;
+ }
+ continue;
+ }
+
+ /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
+ * false otherwise. But first, restore LC_TIME to its original
+ * locale if we changed it */
+ if (save_time_locale) {
+ setlocale(LC_TIME, save_time_locale);
+ Safefree(save_time_locale);
+ }
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
+ save_input_locale,
+ is_utf8_string((U8 *) formatted_time, 0)));
+ Safefree(save_input_locale);
+ return is_utf8_string((U8 *) formatted_time, 0);
+ }
+
+ /* Falling off the end of the loop indicates all the names were just
+ * ASCII. Go on to the next test. If we changed it, restore LC_TIME
+ * to its original locale */
+ if (save_time_locale) {
+ setlocale(LC_TIME, save_time_locale);
+ Safefree(save_time_locale);
+ }
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "All time-related words for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
+ }
+ cant_use_time:
+
+#endif
+
+#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
/* This code is ifdefd out because it was found to not be necessary in testing
- * on our dromedary test machine, which has over 700 locales. There, looking
- * at just the currency symbol gave essentially the same results as doing this
- * extra work. Executing this also caused segfaults in miniperl. I left it in
- * so as to avoid rewriting it if real-world experience indicates that
- * dromedary is an outlier. Essentially, instead of returning abpve if we
+ * on our dromedary test machine, which has over 700 locales. There, this
+ * added no value to looking at the currency symbol and the time strings. I
+ * left it in so as to avoid rewriting it if real-world experience indicates
+ * that dromedary is an outlier. Essentially, instead of returning abpve if we
* haven't found illegal utf8, we continue on and examine all the strerror()
* messages on the platform for utf8ness. If all are ASCII, we still don't
* know the answer; but otherwise we have a pretty good indication of the
- * utf8ness. The reason this doesn't necessarily help much is that the
- * messages may not have been translated into the locale. The currency symbol
- * is much more likely to have been translated. The code below would need to
- * be altered somewhat to just be a continuation of testing the currency
- * symbol. */
+ * utf8ness. The reason this doesn't help much is that the messages may not
+ * have been translated into the locale. The currency symbol and time strings
+ * are much more likely to have been translated. */
+ {
int e;
- unsigned int failures = 0, non_ascii = 0;
+ bool is_utf8 = FALSE;
+ bool non_ascii = FALSE;
char *save_messages_locale = NULL;
+ const char * errmsg = NULL;
- /* Like above for LC_CTYPE, we set LC_MESSAGES to the locale of the
- * desired category, if it isn't that locale already */
+ /* Like above, we set LC_MESSAGES to the locale of the desired
+ * category, if it isn't that locale already */
if (category != LC_MESSAGES) {
- save_messages_locale = stdize_locale(savepv(setlocale(LC_MESSAGES,
- NULL)));
+ save_messages_locale = setlocale(LC_MESSAGES, NULL);
if (! save_messages_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not find current locale for LC_MESSAGES\n"));
goto cant_use_messages;
}
+ save_messages_locale = stdize_locale(savepv(save_messages_locale));
if (strEQ(save_messages_locale, save_input_locale)) {
- Safefree(save_input_locale);
+ Safefree(save_messages_locale);
+ save_messages_locale = NULL;
}
else if (! setlocale(LC_MESSAGES, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_MESSAGES locale to %s\n",
+ save_input_locale));
Safefree(save_messages_locale);
goto cant_use_messages;
}
}
/* Here the current LC_MESSAGES is set to the locale of the category
- * whose information is desired. Look through all the messages */
+ * whose information is desired. Look through all the messages. We
+ * can't use Strerror() here because it may expand to code that
+ * segfaults in miniperl */
- for (e = 0;
-#ifdef HAS_SYS_ERRLIST
- e <= sys_nerr
-#endif
- ; e++)
- {
- const U8* const errmsg = (U8 *) Strerror(e) ;
- if (!errmsg)
- break;
- if (! is_utf8_string(errmsg, 0)) {
- failures++;
+ for (e = 0; e <= sys_nerr; e++) {
+ errno = 0;
+ errmsg = sys_errlist[e];
+ if (errno || !errmsg) {
break;
}
- else if (! is_ascii_string(errmsg, 0)) {
- non_ascii++;
+ errmsg = savepv(errmsg);
+ if (! is_invariant_string((U8 *) errmsg, 0)) {
+ non_ascii = TRUE;
+ is_utf8 = is_utf8_string((U8 *) errmsg, 0);
+ break;
}
}
+ Safefree(errmsg);
/* And, if we changed it, restore LC_MESSAGES to its original locale */
if (save_messages_locale) {
Safefree(save_messages_locale);
}
- /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
- * any non-ascii means it is one; otherwise we assume it isn't */
- return (failures) ? FALSE : non_ascii;
+ if (non_ascii) {
+
+ /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
+ * any non-ascii means it is one; otherwise we assume it isn't */
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
+ save_input_locale,
+ is_utf8));
+ Safefree(save_input_locale);
+ return is_utf8;
+ }
+ 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));
}
cant_use_messages:
#endif
+#endif /* the code that is compiled when no nl_langinfo */
+
+#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
+ UTF-8 locale */
+ /* As a last resort, look at the locale name to see if it matches
+ * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
+ * return of setlocale(), is actually defined to be opaque, so we can't
+ * really rely on the absence of various substrings in the name to indicate
+ * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
+ * be a UTF-8 locale. Similarly for the other common names */
+
+ final_pos = strlen(save_input_locale) - 1;
+ if (final_pos >= 3) {
+ char *name = save_input_locale;
+
+ /* Find next 'U' or 'u' and look from there */
+ while ((name += strcspn(name, "Uu") + 1)
+ <= save_input_locale + final_pos - 2)
+ {
+ if (!isALPHA_FOLD_NE(*name, 't')
+ || isALPHA_FOLD_NE(*(name + 1), 'f'))
+ {
+ continue;
+ }
+ name += 2;
+ if (*(name) == '-') {
+ if ((name > save_input_locale + final_pos - 1)) {
+ break;
+ }
+ name++;
+ }
+ if (*(name) == '8') {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s ends with UTF-8 in name\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return TRUE;
+ }
+ }
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s doesn't end with UTF-8 in name\n",
+ save_input_locale));
+ }
+#endif
+
+#ifdef WIN32
+ /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
+ if (final_pos >= 4
+ && *(save_input_locale + final_pos - 0) == '1'
+ && *(save_input_locale + final_pos - 1) == '0'
+ && *(save_input_locale + final_pos - 2) == '0'
+ && *(save_input_locale + final_pos - 3) == '5'
+ && *(save_input_locale + final_pos - 4) == '6')
+ {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s ends with 10056 in name, is UTF-8 locale\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return TRUE;
+ }
+#endif
+
+ /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
+ * since we are about to return FALSE anyway, there is no point in doing
+ * this extra work */
+#if 0
+ if (instr(save_input_locale, "8859")) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s has 8859 in name, not UTF-8 locale\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return FALSE;
+ }
+#endif
+
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Assuming locale %s is not a UTF-8 locale\n",
save_input_locale));
#endif
+
+bool
+Perl__is_in_locale_category(pTHX_ const bool compiling, const int category)
+{
+ dVAR;
+ /* Internal function which returns if we are in the scope of a pragma that
+ * enables the locale category 'category'. 'compiling' should indicate if
+ * this is during the compilation phase (TRUE) or not (FALSE). */
+
+ const COP * const cop = (compiling) ? &PL_compiling : PL_curcop;
+
+ SV *categories = cop_hints_fetch_pvs(cop, "locale", 0);
+ if (! categories || categories == &PL_sv_placeholder) {
+ return FALSE;
+ }
+
+ /* The pseudo-category 'not_characters' is -1, so just add 1 to each to get
+ * a valid unsigned */
+ assert(category >= -1);
+ return cBOOL(SvUV(categories) & (1U << (category + 1)));
+}
+
+char *
+Perl_my_strerror(pTHX_ const int errnum) {
+
+ /* Uses C locale for the error text unless within scope of 'use locale' for
+ * LC_MESSAGES */
+
+#ifdef USE_LOCALE_MESSAGES
+ if (! IN_LC(LC_MESSAGES)) {
+ char * save_locale = setlocale(LC_MESSAGES, NULL);
+ if (! isNAME_C_OR_POSIX(save_locale)) {
+ char *errstr;
+
+ /* The next setlocale likely will zap this, so create a copy */
+ save_locale = savepv(save_locale);
+
+ setlocale(LC_MESSAGES, "C");
+
+ /* This points to the static space in Strerror, with all its
+ * limitations */
+ errstr = Strerror(errnum);
+
+ setlocale(LC_MESSAGES, save_locale);
+ Safefree(save_locale);
+ return errstr;
+ }
+ }
+#endif
+
+ return Strerror(errnum);
+}
+
+/*
+
+=head1 Locale-related functions and macros
+
+=for apidoc sync_locale
+
+Changing the program's locale should be avoided by XS code. Nevertheless,
+certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
+happens, Perl needs to be told that the locale has changed. Use this function
+to do so, before returning to Perl.
+
+=cut
+*/
+
+void
+Perl_sync_locale(pTHX)
+{
+
+#ifdef USE_LOCALE_CTYPE
+ new_ctype(setlocale(LC_CTYPE, NULL));
+#endif /* USE_LOCALE_CTYPE */
+
+#ifdef USE_LOCALE_COLLATE
+ new_collate(setlocale(LC_COLLATE, NULL));
+#endif
+
+#ifdef USE_LOCALE_NUMERIC
+ set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
+ new_numeric(setlocale(LC_NUMERIC, NULL));
+#endif /* USE_LOCALE_NUMERIC */
+
+}
+
+
+
/*
* Local variables:
* c-indentation-style: bsd