Problem 1105

** Bug Description:

DistanceToIn(G4ThreeVector&, G4ThreeVector&) returns kInfinity when point on facet. Should return zero.

Problem causes wrong navigation in case of boundary processes (eg. total internal reflection of optical photons).

** Affected geant4 versions:

4.9.3, 4.9.2 (approved) and most likely older ones. 

** How to reproduce the problem:
 
- Enable Cherenkov and optical processes in physics list.
- Create an extruded volume with optical properties (MPT, eg. Fused Silica) and shoot a charged particle, e.g 4GeV/c pion with an angle of 20 deg (relative to the extrusion direction / local z) through this volume.

The charged pion will cause Cherenkov radiation. Some of those photons will be totally reflected at the surface of the extruded volume. After this boundary process, the photons will not reenter the radiator. 

This is caused by the buggy DistanceToIn computation in G4TesselatedSolid and G4TriangleFacet.  


** Quick fix:

I added a quick and dirty fix to my geant version by changing "distFromSurface > 0.5*kCarTolerance" to "distFromSurface >= -0.5*kCarTolerance"  in G4TessellatedSolid::DistanceToIn(G4ThreeVector&, G4ThreeVector&)


PS:
Instead of using my example including Cherenkov photons you should be able to shoot optical-photons with the particle gun.