_cc = $utf_combining_class[$utf_char];
					}

					$utf_sort[$utf_combining_class[$utf_char]][] = $utf_char;
				}
				else
				{
					// Non-decomposable starter, check out if it's a Hangul syllable
					if ($utf_char < UTF8_HANGUL_FIRST || $utf_char > UTF8_HANGUL_LAST)
					{
						// Nope, regular UTF char, check that we have the correct number of trailing bytes
						if (($utf_char & $utf_validation_mask[$utf_len]) != $utf_validation_check[$utf_len])
						{
							// Current char isn't well-formed or legal: either one or several trailing bytes are missing, or the Unicode char
							// has been encoded in a five- or six- byte sequence.
							// Move the cursor back to its original position then advance it to the position it should really be at
							$pos -= $utf_len;
							$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

							if (!empty($utf_sort))
							{
								ksort($utf_sort);

								foreach ($utf_sort as $utf_chars)
								{
									$tmp .= implode('', $utf_chars);
								}
								$utf_sort = array();
							}

							// Add a replacement char then another replacement char for every trailing byte.
							//
							// @todo I'm not entirely sure that's how we're supposed to mark invalidated byte sequences, check this
							$spn = strspn($str, UTF8_TRAILING_BYTES, ++$pos);
							$tmp .= str_repeat(UTF8_REPLACEMENT, $spn + 1);

							$dump = $sort = 0;

							$pos += $spn;
							$tmp_pos = $pos;
							continue;
						}

						if (isset($extra_check[$utf_char[0]]))
						{
							switch ($utf_char[0])
							{
								// Note: 0xED is quite common in Korean
								case "\xED":
									if ($utf_char >= "\xED\xA0\x80")
									{
										// Surrogates (U+D800..U+DFFF) are not allowed in UTF-8 (UTF sequence 0xEDA080..0xEDBFBF)
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;

								// Note: 0xEF is quite common in Japanese
								case "\xEF":
									if ($utf_char == "\xEF\xBF\xBE" || $utf_char == "\xEF\xBF\xBF")
									{
										// U+FFFE and U+FFFF are explicitly disallowed (UTF sequence 0xEFBFBE..0xEFBFBF)
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;

								case "\xC0":
								case "\xC1":
									if ($utf_char <= "\xC1\xBF")
									{
										// Overlong sequence: Unicode char U+0000..U+007F encoded as a double-byte UTF char
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;

								case "\xE0":
									if ($utf_char <= "\xE0\x9F\xBF")
									{
										// Unicode char U+0000..U+07FF encoded in 3 bytes
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;

								case "\xF0":
									if ($utf_char <= "\xF0\x8F\xBF\xBF")
									{
										// Unicode char U+0000..U+FFFF encoded in 4 bytes
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;

								default:
									if ($utf_char > UTF8_MAX)
									{
										// Out of the Unicode range
										$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

										if (!empty($utf_sort))
										{
											ksort($utf_sort);

											foreach ($utf_sort as $utf_chars)
											{
												$tmp .= implode('', $utf_chars);
											}
											$utf_sort = array();
										}

										$tmp .= UTF8_REPLACEMENT;
										$dump = $sort = 0;

										$tmp_pos = $starter_pos = $pos;
										continue 2;
									}
								break;
							}
						}
					}
					else
					{
						// Hangul syllabl