URL filter does not convert a link inside a div

If you put in your body this kind of code : <div>http://www.example.com</div>, the URL is not converted into a hyperlink. Actually, there is more than this : in the following example, lines 3 and 4 fail.

1. URL alone (followed by a stop) : http://www.example.com.
2. URL in a p tag : <p>http://www.example.com</p>
3. URL in a p tag containing HTML attributes : <p style="border:1px solid black">http://www.example.com</p>
4. URL in a div tag : <div>http://www.example.com</div>
5. URL without nothing following : http://www.example.com

(By reading my post preview, I see a bug on line 1, but this is caused by <code> tag. Another issue to raise ? Or maybe changing the filters order ?)

Also, I saw that the regular expression used to match incomplete URLs (like www.example.com) was a little bit different than the one used for full URLs : it doesn't work when following a <br> tag :
6. www adress following a &lt;br&gt; tag : <br />www.example.com
won't convert the URL.

Finally, seeing the regular expressions used, a URL, with no following text at all, should not match the patters. But it does, because, for a reason I ignore, a space is added to the end of the text the filter is applied on.

I want to stress that there must be no confusion : what is on grey background in my post is exactly what is to be pasted in a node's textarea, to encounter the bugs I am explaining (try yourself ;) ). The fact that only one URL is converted here is just side-effect.

So I tried to simplify and fix the regular expressions used to match full URLs, e-mail addresses and "www domains/addresses". Here is a list of things I did :

• replacing the regular-expression surrounding double quotes for simple quotes (double quotes are useless and may be source of confusion : \n would be converted to a line-break with double quotes)
• using a positive look-behind to match what must precede the URL, this avoids parasite contain inside $match variable
• changing the index used to get $match content accordingly
• use \s , which is an "any space" entity, instead of " \n\r\t" inside of the character class (square brackets)
• remove the backslash escaping an open bracket, which is useless inside a character class
• remove the part matching exactly <p>, <li>, or any kind of <br>, and, instead, matching a closing angled bracket (or greater-than sign, > which can be a sign of a previous tag)
• repeating the three points above at the end of the regular expression
• merging the part that checks if the URL is followed by a punctuation sign, and the part that "looks ahead" for a space character, or closing bracket, or opening angled bracket (potential sign of a following tag)
• remove the now useless concatenations to $match[1] and $match[3], now that there are only "look arround" expressions around the part matching the URL
• I did not modify the part in the regular expressions that is actually matching a URL/e-mail

I would like to complete SimpleTest code for this, but I didn't find any place where to download Drupal 7 core SimpleTest files (anyone can help ?).

Please find my patch attached.

Regards,

David