Zero length steps during LArG4
Running Dune's standard_g4_dune10kt_1x2x6.fcl on /pnfs/dune/scratch/users/jstock/GeneratedBugSample.root
produces a divide by zero error.
To reproduce, run the above with a breakpoint at larsim/LArG4/ISCalculationSeparate.cxx: 106
Using Allinea I have tracked the bug as far as larsim/LArG4/ISCalculationSeparate.cxx line 106. dx=0. The issue is that the g4 step it is looking at is apparently length 0 (G4Step Step->GetPreStepPoint->GetPosition()==ste->GetPostStepPoint()->GetPosition() ).
#2 Updated by Jason Stock over 2 years ago
I find the following will reliably produce the bug on a dunegpvm.
setup larsoft v06_40_01 -q e14:prof
setup dunetpc v06_40_01 -q e14:prof
lar -c prod_Ar39.fcl -n 10 -o gen.root ###(Available from dunetpc feature/JStock_Ar39Test
lar -c standard_g4_dune10kt_1x2x6.fcl -n 10 -o g4.root gen.root
#5 Updated by Hans-Joachim Wenzel about 1 year ago
- Status changed from Assigned to Resolved
- % Done changed from 0 to 100
I tried if I could reproduce the 0 step length problem using the new larg4 module where voxel read out is replaced with a simple step limiter. I ran a million single particle event and never ran into the problem. So my recommendation is to move to the new larg4