+#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
+ * otherwise. 'category' may not be LC_ALL. If the platform doesn't have
+ * nl_langinfo(), this employs a heuristic, which hence could give the
+ * wrong result. It errs on the side of not being a UTF-8 locale. */
+
+ char *save_input_locale = NULL;
+ 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));
+ if (! save_input_locale) {
+ return FALSE; /* XXX maybe should croak */
+ }
+ if ((*save_input_locale == 'C' && save_input_locale[1] == '\0')
+ || strEQ(save_input_locale, "POSIX"))
+ {
+ return FALSE;
+ }
+
+ save_input_locale = savepv(save_input_locale);
+
+#if defined(HAS_NL_LANGINFO) && defined(CODESET) && defined(USE_LOCALE_CTYPE)
+
+ { /* Next try nl_langinfo if available */
+
+ char *save_ctype_locale = NULL;
+ char *codeset = NULL;
+
+ if (category != LC_CTYPE) { /* nl_langinfo works only on LC_CTYPE */
+
+ /* Get the current LC_CTYPE locale */
+ save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL)));
+ if (! save_ctype_locale) {
+ goto cant_use_nllanginfo;
+ }
+
+ /* 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
+ * the value for the desired category. Otherwise, switch LC_CTYPE
+ * to the desired category's locale */
+ if (strEQ(save_ctype_locale, save_input_locale)) {
+ Safefree(save_ctype_locale);
+ save_ctype_locale = NULL;
+ }
+ else if (! setlocale(LC_CTYPE, save_input_locale)) {
+ Safefree(save_ctype_locale);
+ goto cant_use_nllanginfo;
+ }
+ }
+
+ /* Here the current LC_CTYPE is set to the locale of the category whose
+ * information is desired. This means that nl_langinfo() should give
+ * the correct results */
+ codeset = savepv(nl_langinfo(CODESET));
+ if (codeset) {
+ bool is_utf8;
+
+ /* If we switched LC_CTYPE, switch back */
+ if (save_ctype_locale) {
+ setlocale(LC_CTYPE, save_ctype_locale);
+ Safefree(save_ctype_locale);
+ }
+
+ is_utf8 = foldEQ(codeset, STR_WITH_LEN("UTF-8"))
+ || foldEQ(codeset, STR_WITH_LEN("UTF8"));
+
+ Safefree(codeset);
+ Safefree(save_input_locale);
+ return is_utf8;
+ }
+
+ }
+ 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 */
+
+ 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);
+ return TRUE;
+ }
+ }
+ }
+
+#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);
+ 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
+