validateNearestFeatureOfPolytopeBeingEdge triggered in latest master

[wxmerkt]

Issue #390 still happens on the latest master:

  Original error message: [path]/include/fcl/narrowphase/detail/convexity_based_algorithm/gjk_libccd-inl.h:(1623): validateNearestFeatureOfPolytopeBeingEdge(): The origin is outside of the polytope. This should already have been identified as separating.
  Shape 1: Sphere(0.05999999999999999778)
  X_FS1
   -0.99999999999999955591 -4.4637642593504144998e-09                          0  1.7855056639081962376e-10
 4.4637642593504144998e-09    -0.99999999999999955591                          0    0.039999999999999993894
                         0                          0      1.0000000000000008882     0.33000000000000012657
                         0                          0                          0                          1
  Shape 2: Box0.10000000000000000555 0.10000000000000000555 0.10000000000000000555
  X_FS2
                      1                       0                       0 0.050000000000000002776
                      0                       1                       0   0.1500000000000000222
                      0                       0                       1   0.3499999999999999778
                      0                       0                       0                       1
  Solver: GjkSolver_libccd
    collision_tolerance:      2.0097183471152322134e-14
    max collision iterations: 500
    distance tolerance:       9.9999999999999995475e-07
    max distance iterations:  1000

and another one (almost the same):

  Original error message: [path]/include/fcl/narrowphase/detail/convexity_based_algorithm/gjk_libccd-inl.h:(1623): validateNearestFeatureOfPolytopeBeingEdge(): The origin is outside of the polytope. This should already have been identified as separating.
  Shape 1: Sphere(0.05999999999999999778)
  X_FS1
  -0.99999999999999777955 -5.994344502469753097e-08                         0 2.3977377970047051304e-09
 5.994344502469753097e-08   -0.99999999999999777955                         0   0.039999999999999924505
                        0                         0     1.0000000000000008882    0.33000000000000012657
                        0                         0                         0                         1
  Shape 2: Box0.10000000000000000555 0.10000000000000000555 0.10000000000000000555
  X_FS2
                      1                       0                       0 0.050000000000000002776
                      0                       1                       0   0.1500000000000000222
                      0                       0                       1   0.3499999999999999778
                      0                       0                       0                       1
  Solver: GjkSolver_libccd
    collision_tolerance:      2.0097183471152322134e-14
    max collision iterations: 500
    distance tolerance:       9.9999999999999995475e-07
    max distance iterations:  1000

[sherm1]

cc @SeanCurtis-TRI @hongkai-dai

[hongkai-dai]

This is another case of touching contact being identified as penetration (due to small roundoff errors in __ccdGJK), I have seen about 10 of these cases in the last month or so.

[sadraddini]

We encountered this message in our internal code. Is there remedy for this or should we consider it as a very rare event?