As @Jeremy Rickard has mentioned, the impossibility of the problem haswas shown by Osofsky in 1968. But it is interesting that there is another article dateddating back to 1968 (again !) that proves the same thing. See Cateforis, Sandomerski, The singular submodule spiltssplits off, J. Algebra, 10 (1968), 149-165, Theorem 4.1. This article haswas received by J. Algebra aton November 20, 1967 while Osofsky's paper haswas received by Proc. Amer. Math. Soc. aton July 10, 1967. I hope this information meets the expectations.