This paper reconsiders the most basic scheduling problem, that of minimizing the makespan of a partially ordered set of activities, in the context of incomplete knowledge. While this problem is very easy in the deterministic case, its counterpart when durations are interval-valued is much trickier, as standard results and algorithms no longer apply. After positioning this paper in the scope of temporal networks under uncertainty, we provide a complete solution to the problem of finding the latest starting times and floats of activities, and of locating surely critical ones, as they are often isolated. The minimal float problem is NP-hard while the maximal float problem is polynomial. New complexity results and efficient algorithms are provided for the interval-valued makespan minimization problem.