Description:
------------
preg_match and preg_replace return false and set the last error to PREG_BAD_UTF8_ERROR when the u modifier is specified and the subject is invalid UTF-8. However, preg_split does not behave consistently in this regard.

Test script:
---------------
var_dump(preg_split("/a/u", "a\xff"));

Expected result:
----------------
preg_split should return false and set the last error PREG_BAD_UTF8_ERROR just like the other preg_ functions.

Actual result:
--------------
preg_split returns an array with the subject string.