| Summary: | Crash using RegularStructure with multiple BeamOn calls | ||
|---|---|---|---|
| Product: | Geant4 | Reporter: | Derek Dolney <Derek.Dolney> |
| Component: | run | Assignee: | asai |
| Status: | RESOLVED FIXED | ||
| Severity: | critical | CC: | John.Apostolakis |
| Priority: | P1 | ||
| Version: | 9.4 | ||
| Hardware: | All | ||
| OS: | All | ||
Thanks for the clear problem report. Indeed there is a problem when the geometry is changed and such a type of regular voxel volume exists in the geometry. From your clear report it was possible to create a simple protection for this, which was communicated to you privately. This or an improved fix will be tested further and should be included in the next round of fixes. Fix is now included in 9.4.p02. |
I have simulation code that uses a G4PhantomParameterisation (hence, IsRegularStructure() == true) with a rotating modulator wheel (somewhat based off the one in the advanced/hadrontherapy example). The modulator wheel could rotate as many as 256 times in a given simulation run. Each time it is rotated, I call GeometryHasBeenModified(), followed by another BeamOn. This code worked with 4.9.3 and older. With 4.9.4, my simulations terminate early with the following error: ---SNIP--------------------------------------------------------------- Beginning segment 93 of 156 Start of my run 94 My event 0 ERROR - G4SteppingManager::GetProcessNumber() SizeOfSelectedDoItVector= 100 ; is smaller then one of MAXofAtRestLoops= 0 or MAXofAlongStepLoops= 3 or MAXofPostStepLoops= 101 *** G4Exception : Tracking0012 issued by : G4SteppingManager::GetProcessNumber() The array size is smaller than the actual No of processes. *** Fatal Exception *** core dump *** *** G4Exception: Aborting execution *** ---SNIP--------------------------------------------------------------- Note that it is almost to segment 100, and apparently I have more than 100 processes with PostStep actions registered with the ProcessManager. Indeed, I traced this out and found that every call to G4RunManager::BeamOn() results in the following call stack: #0 G4RunManagerKernel::SetScoreSplitter (this=0xabf440) at src/G4RunManagerKernel.cc:556 #1 0x00007fffef9f1cf0 in G4RunManagerKernel::RunInitialization (this=0xabf440) at src/G4RunManagerKernel.cc:331 #2 0x00007fffef9f665d in G4RunManager::RunInitialization (this=0xabf6b0) at src/G4RunManager.cc:182 #3 0x00007fffef9f33d7 in G4RunManager::BeamOn (this=0xabf6b0, n_event=1, macroFile=0x0, n_select=-1) at src/G4RunManager.cc:142 Note that in G4RunManagerKernel::SetScoreSplitter(), the G4ScoreSplittingProcess is added to the ProcessManager. So the general problem is, if a RegularStructure is used, and the user wants to call BeamOn multiple times, eventually a process vector fills up and the simulation is terminated. Please help. I need other fixes that are in 4.9.4, but this issue is keeping me from upgrading. Thank you in advance for your help.