Multiplication by Zero and Infinity

[zenna]

Some quantile functions such as the normal family's return positive of negative infinity. When refining, this will happen, for example quantile(Normal(0,1), Omega) = [-Inf, Inf]. Typically this doesn't cause a problem, but if we multiply by zero, we can NaNs.

Example:

prob(0.0 * Gamma(1,1) + Normal(0,1) > 5)

The result is [1,1] which is clearly wrong. If we just look at the Gamma part, applying the whole of Omega to its quantile results in an Inf at the upper bound (I think). Multiplication then results in an interval with a NaN. After that, all bets are off.