Problem is exactly what you pointed out. A recursive function without a "break condition". In this case if getLeft() or/and getRight() is NULL then you shouldn't call .iPathLength() on it, thus breaking the recursion.
Theoretically it might still be possible to make this code itself work (using specialized derived classes to act as leaf nodes without any children, which have a different impl for iPathLength() , instead of NULL). But it would be over-engineering IMO and especially irrelevant in classroom.
No, it's not right. In most cases it will generate an exception (null reference) unless you have a specialized 'leaf' node that implements different code for iPathLength (for example, some implementations of Red-Black trees use a specialized leaf node).