Merge pull request #456 from RosettaCommons/vmullig/fix_rmsd
Vmullig/fix rmsd
Fixing an invalid assumption that was introduced in the RMSD calculator three years ago: if the "uu" matrix is the identity matrix, it is NOT necessarily true that the RMSD is 0. You get uu=I if pose1 and pose 2 are aligned, even if their RMSD is very large.
Tasks:
-- Adding a unit test that illustrates this aberrant behaviour (checks whether RMSD of pose1 to pose2 is the same whether pose2 is aligned to pose1 or not, for pose1 and pose2 with large RMSD).
-- Taking out the identity matrix check (along with two functions lacking prototypes that are only used for this purpose).
Expected test changes:
-- Several integration tests that were reporting 0 RMSDs before now report very tiny numbers, as one might expect.