I can easily break this code by feeding it a valid html string:

$s = '
<p>something something</p>
<h2>blah <<< blah blah</h2>
<p>something something</p>
';
[download]

Update:

BUU, although parsing is not the purpose, you have to realize that implicitly it is done any way. Regexp is obviously a kind of parsing, and it has been repeatedly mentioned by many monks here that, regexp is not a good way to deal with html.

It is not an easy task to come up with perfect regexp for dealing with html.

On the other hand, although the purpose is not parsing, parsing is still a valid tool, isn't it? In this case, actually a better tool.