Firstly I would like to suggest that even if you can only tackle the symptom, not the cause, it would still be worthwhile - a patient with untreatable cancer should nonetheless get painkillers. You are correct that the number of people in the second generation would be higher than the number you bring out of famine. But if you do nothing the second generation would be even larger still since the fertility rate of the extremely poor is so high. Therefore simply alleviating the symptoms and not the causes of poverty (a) delays the suffering of hunger by a generation, and (b) reduces the total number people suffering in the second generation.
Of course, it would be far better to eliminate the cancer. In many cases it's difficult to know how effective the long-term treatment is, but in some cases the outcome can be measured quite well. Here's a great talk by Esther Duflo about measuring the effectiveness of aid: http://www.youtube.com/watch?v=0zvrGiPkVcs In the book she co-authored called "Poor Economics" she mentions a study of a deworming program for children in Kenya. The study concluded that an extra year of deworming costing $1.36 USD PPP lead to a lifetime income gain of $3,269 USD PPP for the child. To me that is a slam-dunk evidence that aid can be effective in the long term.