<?php

function cleanup_utf8_end($s) {
   // Supposition: Only the last character can be broken.
   // Exception: My unpolishable Engrish.
   $slen = strlen($s);
   $cutlen = $slen;
   for ($i = 1; $i < $slen; $i++) { // Scan backward
      $b = ord($s{$slen - $i});
      //
      //   $i  ...  3  2  1
      // ------   ----------+
      //  |$b| ... |  |  |  | $s
      // ------   ----------+
      //
      if (($b & 0x80) == 0) { // or ($b <= 0x7f)
	 // $b == 0xxx xxxx (binary) : ASCII
	 if ($i == 1) return $s;
	 $cutlen = $i - 1; // To survive
	 break;
      } elseif (($b & 0x40) != 0) { // or (0xc0 <= $b)
	 // $b == 11xx xxxx (binary) : Non-ASCII, 1st
	 if ($i <= 6 && ($b & 0xff80 >> $i) == (0xff & 0xff00 >> $i)) {
	    //
	    // $i   $b (binary) : The first byte of a multi-byte sequence
	    // ---------------   (See utf-8(7))
	    // 2    110x xxxx
	    // :      :   :
	    // 6    1111 110x
	    //
	    return $s;
	    //$cutlen = 0; // To survive (all the bytes)
	 } else {
	    $cutlen = $i; // Otherwise it is considered BROKEN; to cut off
	 }
	 break;
      } else {
	 // $b == 10xx xxxx (binary) : Non-ASCII, non-1st
      }
   }
   return substr($s, 0, $slen - $cutlen);
}


?>