We report a computational study for the O-17 NMR tensors (electric field gradient and chemical shielding tensors) in crystalline uracil. We found that N-(HO)-O-... and C-H center dot center dot center dot O hydrogen bonds around the uracil molecule in the crystal lattice have quite different influences on the O-17 NMR tensors for the two C=O groups. The computed O-17 NMR tensors on O4, which is involved in two strong N-H center dot center dot center dot O hydrogen bonds, show remarkable sensitivity toward the choice of cluster model, whereas the O-17 NMR tensors on O2, which is involved in two weak C-(HO)-O-... hydrogen bonds, show much smaller improvement when the cluster model includes the C-H center dot center dot center dot O hydrogen bonds. Our results demonstrate that it is important to have accurate hydrogen atom positions in the molecular models used for O-17 NMR tensor calculations. In the absence of low-temperature neutron diffraction data, an effective way to generate reliable hydrogen atom positions in the molecular cluster model is to employ partial geometry optimization for hydrogen atom positions using a cluster model that includes all neighboring hydrogen-bonded molecules. Using an optimized seven-molecule model (a total of 84 atoms), we were able to reproduce the experimental O-17 NMR tensors to a reasonably good degree of accuracy. However, we also found that the accuracy for the calculated O-17 NMR tensors at O-2 is not as good as that found for the corresponding tensors at O4. In particular, at the B3LYP/6-311++G(d,p) level of theory, the individual O-17 chemical shielding tensor components differ by less than 10 and 30 ppm from the experimental values for O4 and O2, respectively. For the O-17 quadrupole coupling constant, the calculated values differ by 0.30 and 0.87 MHz from the experimental values for O4 and O2, respectively.