FSet's test suite uses a PRNG to generate test cases. I recently switched it from using MT19937 to Yukari Hafner's random-state, not for any important reason; I just think Sebastiano Vigna and David Blackman have done some cool work on their new PRNG algorithms. (They're faster and their state vector is much smaller than MT's.) Alas, random-state, apparently through no fault of its own, tickles a bug in Allegro 11. So I've switched FSet back to MT19937.
Also, I just noticed and fixed a couple of corner-case bugs in the seq internals.