BDSIM segfaulting on first event for AWAKE spectrometer example in 10.3

[Former user]

Running the AWAKE spectrometer.gmad example. BDSIM hangs at the first event. The problem does not occur at head of branch awake-develop-old-3.

[Jochem Snuverink]

@ldeacon : in which branch is this? I don't get that in develop for examples/AWAKE/awake.gmad or for examples/features/components/awakespectrometer.gmad

[Jochem Snuverink]

I checked with the most recent commit b36e92f. @nevay very recently fixed a few issues with the develop branch. However, there are still occasional hanging events, e.g. issue #178. Could you also try with this commit perhaps and also with a different seed? (--seed=123)

[Laurie Nevay]

I'm trying to make our dipole integrator more robust. In the case of the AWAKE spectrometer, you could safely use:

 option, integratorSet="geant4";

to use all G4ClassicalRK4. The required accuracy is still relatively high and I rarely see anything fail with the RK4 integrator. I avoid this as a default as it's not good for tracking through longer lattices. This is a bit of a presumption though until I reproduce the warnings / hanging. It's a commonly observed failure mode though.

[Jochem Snuverink]

@ldeacon : which Geant4 version was used? With Geant4.10.3 I get following segfault:

53: -------- WWWW ------- G4Exception-START -------- WWWW -------
53: *** G4Exception : GeomNav0003
53:       issued by : G4Navigator::GetLocalExitNormal()
53: Function called when *NOT* at a Boundary.
53: Exit Normal not calculated.
53: 
53: *** This is just a warning message. ***
53: -------- WWWW -------- G4Exception-END --------- WWWW -------
53: 
53: 
53: -------- EEEE ------- G4Exception-START -------- EEEE -------
53: *** G4Exception : OpBoun01
53:       issued by : G4OpBoundaryProcess::PostStepDoIt
53:  G4OpBoundaryProcess/PostStepDoIt():  The Navigator reports that it returned an invalid normal
53: Invalid Surface Normal - Geometry must return valid surface normal
53: 
53: *** Event Must Be Aborted ***
53: -------- EEEE -------- G4Exception-END --------- EEEE -------
53: 

[Lawrence Deacon]

I am using Geant4.10.03, and I get the same result when running in batch mode.

[Jochem Snuverink]

Okay. So one temporary solution could be to use 4.10.2, where I haven't seen this.

[Laurie Nevay]

In 10.3 I see the same segfault. This is definitely to do with optical boundaries. It happens in mySpectrometer_0_pv volume - or at least that's the last volume with --verbose_G4stepping=5.

The error only happens when "optical" physics is used.

[Laurie Nevay]

I would suggest we try to find a minimal example of this fault to understand it. We have a simple optical physics test and a simple screen example. Perhaps it would be good to make a multilayer screen example.

We have bdsim/examples/features/components/screen.gmad and bdsim/examples/processes/4_optical/optical.gmad. The last one though is 6MeV gamma rays, so it may be good to add a visible wavelength example, perhaps also with the multilayer screen.

For the AWAKE example it would be good to isolate which element this is happening in - perhaps it's an implementation problem or geometry (optical boundary) problem.

[Laurie Nevay]

Resolved along with Issue #186 by fixing BDSTrajectory::MergeTrajectory implementation.