With large sample sizes, both tests are very sensitive to slightest violations and will usually have p-values below .05, although the data/residuals are sufficiently normally distributed.