This is only a slight modification of the argument already given, but
I like it enough to type it in.

Since $W$ acts with no stabilizer on the open Weyl chamber $K$,
for $K$ to contain a root, it cannot be perpendicular to any other root.
In particular the Dynkin diagram has to be a complete graph
(but the triangle $K_3 = \widehat A_2$ is already infinite type),
and by inspection of $B_2,G_2$ we see that 
every root is perpendicular to some other root.
So the only possibilities are $A_1,A_2$.