#include "reentr.h"
+#ifdef USE_LOCALE
+
/*
* Standardize the locale name from a string returned by 'setlocale'.
*
return locs;
}
+#endif
+
void
Perl_set_numeric_radix(pTHX)
{
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)
+ && is_utf8_string((U8 *) lc->decimal_point, 0)
+ && is_cur_LC_category_utf8(LC_NUMERIC))
+ {
+ SvUTF8_on(PL_numeric_radix_sv);
+ }
}
}
else
{
#ifdef USE_LOCALE_CTYPE
dVAR;
- int i;
+ UV i;
PERL_ARGS_ASSERT_NEW_CTYPE;
for (i = 0; i < 256; i++) {
- if (isUPPER_LC(i))
- PL_fold_locale[i] = toLOWER_LC(i);
- else if (isLOWER_LC(i))
- PL_fold_locale[i] = toUPPER_LC(i);
+ if (isUPPER_LC((U8) i))
+ PL_fold_locale[i] = toLOWER_LC((U8) i);
+ else if (isLOWER_LC((U8) i))
+ PL_fold_locale[i] = toUPPER_LC((U8) i);
else
PL_fold_locale[i] = i;
}
#ifdef __GLIBC__
char * const language = PerlEnv_getenv("LANGUAGE");
#endif
+ /* NULL uses the existing already set up locale */
+ const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
+ ? NULL
+ : "";
char * const lc_all = PerlEnv_getenv("LC_ALL");
char * const lang = PerlEnv_getenv("LANG");
bool setlocale_failure = FALSE;
# ifdef LC_ALL
if (lang) {
- if (setlocale(LC_ALL, ""))
+ if (setlocale(LC_ALL, setlocale_init))
done = TRUE;
else
setlocale_failure = TRUE;
if (! (curctype =
setlocale(LC_CTYPE,
(!done && (lang || PerlEnv_getenv("LC_CTYPE")))
- ? "" : NULL)))
+ ? setlocale_init : NULL)))
setlocale_failure = TRUE;
else
curctype = savepv(curctype);
if (! (curcoll =
setlocale(LC_COLLATE,
(!done && (lang || PerlEnv_getenv("LC_COLLATE")))
- ? "" : NULL)))
+ ? setlocale_init : NULL)))
setlocale_failure = TRUE;
else
curcoll = savepv(curcoll);
if (! (curnum =
setlocale(LC_NUMERIC,
(!done && (lang || PerlEnv_getenv("LC_NUMERIC")))
- ? "" : NULL)))
+ ? setlocale_init : NULL)))
setlocale_failure = TRUE;
else
curnum = savepv(curnum);
#endif /* !LOCALE_ENVIRON_REQUIRED */
#ifdef LC_ALL
- if (! setlocale(LC_ALL, ""))
+ if (! setlocale(LC_ALL, setlocale_init))
setlocale_failure = TRUE;
#endif /* LC_ALL */
if (!setlocale_failure) {
#ifdef USE_LOCALE_CTYPE
Safefree(curctype);
- if (! (curctype = setlocale(LC_CTYPE, "")))
+ if (! (curctype = setlocale(LC_CTYPE, setlocale_init)))
setlocale_failure = TRUE;
else
curctype = savepv(curctype);
#endif /* USE_LOCALE_CTYPE */
#ifdef USE_LOCALE_COLLATE
Safefree(curcoll);
- if (! (curcoll = setlocale(LC_COLLATE, "")))
+ if (! (curcoll = setlocale(LC_COLLATE, setlocale_init)))
setlocale_failure = TRUE;
else
curcoll = savepv(curcoll);
#endif /* USE_LOCALE_COLLATE */
#ifdef USE_LOCALE_NUMERIC
Safefree(curnum);
- if (! (curnum = setlocale(LC_NUMERIC, "")))
+ if (! (curnum = setlocale(LC_NUMERIC, setlocale_init)))
setlocale_failure = TRUE;
else
curnum = savepv(curnum);
}
-#endif /* USE_LOCALE */
-
-#ifdef USE_PERLIO
+#if defined(USE_PERLIO) && defined(USE_LOCALE_CTYPE)
{
/* Set PL_utf8locale to TRUE if using PerlIO _and_
the current LC_CTYPE locale is UTF-8.
#ifdef USE_LOCALE_NUMERIC
Safefree(curnum);
#endif /* USE_LOCALE_NUMERIC */
+
+#endif /* USE_LOCALE */
+
return ok;
}
#endif /* USE_LOCALE_COLLATE */
-bool
+#ifdef USE_LOCALE
+
+STATIC bool
S_is_cur_LC_category_utf8(pTHX_ int category)
{
/* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
* wrong result. It errs on the side of not being a UTF-8 locale. */
char *save_input_locale = NULL;
- int has_hyphen;
STRLEN final_pos;
+#ifdef LC_ALL
assert(category != LC_ALL);
+#endif
/* First dispose of the trivial cases */
save_input_locale = stdize_locale(setlocale(category, NULL));
#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 */
+ * see if it matches qr/UTF -? 8 /ix */
final_pos = strlen(save_input_locale) - 1;
- if (final_pos >= 3
- && *(save_input_locale + final_pos) == '8')
- {
- has_hyphen = *(save_input_locale + final_pos - 1 ) == '-';
- if ((! has_hyphen || final_pos >= 4)
- && toFOLD(*(save_input_locale + final_pos - has_hyphen - 1)) == 'f'
- && toFOLD(*(save_input_locale + final_pos - has_hyphen - 2)) == 't'
- && toFOLD(*(save_input_locale + final_pos - has_hyphen - 3)) == 'u')
+ 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)
{
- Safefree(save_input_locale);
- return TRUE;
+ 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);
+ 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);
+ return FALSE;
+ }
+
+#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();
+
+ if (category != LC_MONETARY) {
+
+ save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY,
+ NULL)));
+ if (! save_monetary_locale) {
+ goto cant_use_monetary;
+ }
+
+ if (strNE(save_monetary_locale, save_input_locale)) {
+ if (! setlocale(LC_MONETARY, 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)) {
+ illegal_utf8 = TRUE;
+ }
+ else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) {
+ only_ascii = TRUE;
+ }
+ }
+
+ /* If we changed it, restore LC_MONETARY to its original locale */
+ if (save_monetary_locale) {
+ setlocale(LC_MONETARY, save_monetary_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)
+ * */
+ return (illegal_utf8)
+ ? FALSE
+ : ! only_ascii;
+
+ }
+ cant_use_monetary:
+
+# endif /* USE_LOCALE_MONETARY */
+#endif /* HAS_LOCALECONV */
+
+#if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES)
+
+/* 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
+ * 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. */
+ int e;
+ unsigned int failures = 0, non_ascii = 0;
+ char *save_messages_locale = NULL;
+
+ /* Like above for LC_CTYPE, 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)));
+ if (! save_messages_locale) {
+ goto cant_use_messages;
+ }
+
+ if (strEQ(save_messages_locale, save_input_locale)) {
+ Safefree(save_input_locale);
+ }
+ else if (! setlocale(LC_MESSAGES, 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 */
+
+ 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++;
+ break;
+ }
+ else if (! is_ascii_string(errmsg, 0)) {
+ non_ascii++;
+ }
+ }
+
+ /* And, if we changed it, restore LC_MESSAGES to its original locale */
+ if (save_messages_locale) {
+ setlocale(LC_MESSAGES, 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;
+
+ }
+ cant_use_messages:
+
+#endif
+
+ Safefree(save_input_locale);
return FALSE;
}
-
+#endif
/*
* Local variables: