Problem 1742

Created attachment 343 [details]
Test program to reproduce the issue

While simulating the passage of muons through an uranium target my simulation sometimes gets stuck. Investigating a bit deeper I found that the simulation actually runs, but that the event never finishes since many slow neutrons are generated inside the uranium, and they in turn induce some fission reactions which generate other neutrons and so on. The process looks similar to a "nuclear fission reactor". Typically, the reactor is ignited by a stopping muon or by a gamma. I see this behaviour very frequently compared to what one can reasonably expect: when running 10 instances of my simulation program, each one simulating 10^5 muons, typically one or two of them get stuck because of this mechanism

I have been able to put together a simple simulation to 
show off the problem. I attach a tarball containing the code and the CMakeLists
to reproduce the issue.

I tried this program on my 64 bit ArchLinux system with gcc 4.9.2. With Geant4 
10.00.p02 the simulation shows the "nuclear reactor" behaviour, while with 
Geant4 9.6.p03 the muon is stopped without producing any nuclear cascade. To 
better analyze what's going on during the simulation I applied this patch:

https://www.hightail.com/download/UlRUa3NVNkc3bUEwTWNUQw

to G4EventManager.cc to dump the secondaries. Here are the logs for the two
runs with 9.6.p03 and 10.00.p02 (I stopped the latter with Ctrl-C after a few 
seconds):

https://www.hightail.com/download/UlRUa3NVNkc3bURxYk1UQw
https://www.hightail.com/download/UlRUa3NVNkc3bUROUjhUQw

I don't know which behaviour is the correct one, and for what I know it could 
be due to changes in the code, changes in the models or  changes in the 
datasets (for each Geant4 version I used the datasets which are automatically 
installed when configuring the build with -DGEANT4_INSTALL_DATA=ON). Yet the 
two are really different so something must be wrong somewhere.