POSIX::strftime: Add better fallback about UTF-8

If the function returns a valid string that isn't completely UTF-8
invariant, the function assumes it is UTF-8 if we are in a UTF-8 locale.
This works, but in the unlikely event that the system has no LC_TIME, we
can't tesll if it is in a UTF-8 locale.  As a better fallback position,
this commit adds the check that there is just a single  script of the
time string, adding a measure of reassurance that out call that it is
UTF-8 is correct.

This is unlikely to be used, but now that there is a function to call
that determines if this is a script run, it's easy to add, and unlikely
to actually get compiled.

diff --git a/ext/POSIX/POSIX.xs b/ext/POSIX/POSIX.xs
index 82fba1f..d3bbfd8 100644 (file)
--- a/ext/POSIX/POSIX.xs
+++ b/ext/POSIX/POSIX.xs
@@ -3571,6 +3571,12 @@ strftime(fmt, sec, min, hour, mday, mon, year, wday = -1, yday = -1, isdst = -1)
                     || (   is_utf8_non_invariant_string((U8*) buf, len)
 #ifdef USE_LOCALE_TIME
                         && _is_cur_LC_category_utf8(LC_TIME)
+#else   /* If can't check directly, at least can see if script is consistent,
+           under UTF-8, which gives us an extra measure of confidence. */
+
+                        && isSCRIPT_RUN((const U8 *) buf, buf + len,
+                                        TRUE, /* Means assume UTF-8 */
+                                        NULL)
 #endif
                 )) {
                    SvUTF8_on(sv);