Where is numpy log overflow documented?

Question

Consider these issues of numerical precision:

import numpy as np
np.log10(1e-324)   # -inf & divide-by-zero
np.log10(1e+309)   # +inf
np.round(x,  +308) #