Description:
------------
preg_match fails for some patterns when the searched string exceeds around 100300 bytes. It matches for same strings if enough text is removed from the middle part which gets captured by the pattern when working correctly. Also the exact same regexp works fine for perl on unshortened output.

Reproduce code:
---------------
<?php
//foo.html >= 100 kB
$contents = file_get_contents("foo.html");
$matches = array();
preg_match('%<BODY\s*>.*?</H[1-6]>(.+?)</BODY>%si', $contents, $matches);
//$matches is empty
var_dump($matches);
?>


Expected result:
----------------
array(1) {
  [0]=>
  string(12935) "/*lots of html*/
}


Actual result:
--------------
array(0) {
}