nextSupport function does not compute the right answer about libccd HOT 4 OPEN

@hongkai-dai: For point 2 above, the calculated dist ought to be a squared distance like all the other Dist2 calculations there (but of course the variable name shouldn't have been dist if it isn't a distance!). So it can still be calculated without a square root: d²=(v⋅w)²/(w⋅w).

from libccd.

Oh, I take it back. Combined with point 3 I see you need the actual distance there (and sqrt(el->dist)) to compute a meaningful termination at 975.

from libccd.

This also makes me wonder whether epa_tolerance is actually a distance squared, which is how it is being used. Looks like it is set to .0001m² by default which means the length tolerance is .01m. Is that intentional @danfis? That seems like a very loose tolerance.

from libccd.

@sherm1 for point 2, if we want to compute the squared distance, it should be d²=(v⋅w)²/(w⋅w), that is still not what the code dist = ccdVec3Dot(&out->v, &el->witness); does (it computes v⋅w), so the code is still wrong.

from libccd.