While taking a quick peek at Bug 2506, I get an error in cmucl in
get-one-factor-pollard. It's computing (float n) where n is
135*2^1204-1, which doesn't fit into a double float. In fact,
get-one-factor-pollard is really computing (ceiling (log (float n))).
Is there a reason why we try to float n before taking the log? Plain
(log n) should work. (Although, it might be that some implementations
also fail in this case, but that's an implementation bug.)
Ray