Hello,
As shaihulub told me on irc, it seems the parser using tokenizer.php have been disabled. It looks like we are now using an old parser instead.
Lot of the recent bug reports on sourceforge are related to this swicth: lot of features and bugfixing are no more implemented / fixed :-(
It seems there is a performance issue on the tokenizer, so should we :
1/ get ride of the tokenizer (current situation) and fix bug / reimplement features
2/ improve the tokenizer performances
;)