Created attachment 7 [details]
Verbose tracking and valgrind output where the bug was observed

I am an app developer on G4MICE, simulation of muon ionisation cooling for a neutrino factory or muon collider.

I get a segmentation fault in the method:

G4Torus::SolveNumericJT(CLHEP::Hep3Vector const&, CLHEP::Hep3Vector const&, double, bool)

Valgrind reports "Invalid read of size 8" which means that Geant4 is trying to read an uninitialised bit of memory (e.g. reading past the end of an array).

Geant4 tracks happily through the torus for about 70 particles, so I think everything is initialised okay and the problem is with geant4. But when it makes the first step on ~ the 70-80th particle I hit the segmentation fault.

I am using G4 v4.8.2p01, running on a 64 bit linux (SuSe) machine. The Torus is initialised with pRMin 1.0  mm, pRMax 50.0 mm, pRTor 110.0 mm, pSPhi 0.0   degree, pDPhi 360.0  degree.

I attach the valgrind output. This is with Geant4 running in a highly verbose mode, so you can see all of the tracking details. I truncated the file so you only see the last two events (to save you having a massive file to read).

Hoping you can resolve the issue! Let me know if you need more details, or if I'm doing something obviously wrong.

Regards,
Chris