Mark default interned strings as valid UTF-8 where appropriate

Aside from being used by the pcre extension, various functions in
mbstring are faster if the input strings are known to be valid UTF-8.
We might as well mark the default interned strings (which are
initialized when PHP starts up) as valid UTF-8 where appropriate.

Author: alexdowad

Zend/zend_string.c

@@ -102,18 +102,23 @@ ZEND_API void zend_interned_strings_init(void)
 	str = zend_string_alloc(sizeof("")-1, 1);
 	ZSTR_VAL(str)[0] = '\000';
 	zend_empty_string = zend_new_interned_string_permanent(str);
+	GC_ADD_FLAGS(zend_empty_string, IS_STR_VALID_UTF8);
 
 	s[1] = 0;
 	for (i = 0; i < 256; i++) {
 		s[0] = i;
 		zend_one_char_string[i] = zend_new_interned_string_permanent(zend_string_init(s, 1, 1));
+		if (i < 0x80) {
+			GC_ADD_FLAGS(zend_one_char_string[i], IS_STR_VALID_UTF8);
+		}
 	}
 
 	/* known strings */
 	zend_known_strings = pemalloc(sizeof(zend_string*) * ((sizeof(known_strings) / sizeof(known_strings[0]) - 1)), 1);
 	for (i = 0; i < (sizeof(known_strings) / sizeof(known_strings[0])) - 1; i++) {
 		str = zend_string_init(known_strings[i], strlen(known_strings[i]), 1);
 		zend_known_strings[i] = zend_new_interned_string_permanent(str);
+		GC_ADD_FLAGS(zend_known_strings[i], IS_STR_VALID_UTF8);
 	}
 }