OK, I think I am missing something very basic... When regularizing phi^3 theory in six dimensions, Srednicki comes to eq 14.30, which shows that the 1-loop contribution to the propagator diverges (the gamma function has a pole). This is good. OK. Now let d=5 (or epsilon=1). Actually, go ahead and let d=1001 or any odd number. Then 14.30 is finite, which definitely seems wrong. "Superficially," the diagram should diverge... What am I screwing up?