Created attachment 272 [details] Example code for reproducing the negative copyNo in NestedParameterisation of layered geometry. When the NestedParameterisation volume is used in a parallel world geometry, the CopyNo given to ComputeTransform/ComputeMaterial methods sometimes become negative (e.g. -1). It happens in both geant4 version 9.6.p03 and 10.00.p01. I attached the modified version of RE04 example (tested in 10.00.p01), which uses NestedParameteriation geometry for a layered mass geometry. I inserted G4cout in ComputeTransform/ComputeMaterial methods if the copyNo was -1. It shows the total counts of the error too. For example, --- snipped --- **** ComputeTransformation:: Nagative CopyNo -1 **** ComputeMaterial:: Nagative CopyNo Total: 351 >>> Summary of Event 9991 --- snipped --- The attached file is the tar ball of RE04Nested code for reproducing the problem. 1) $ tar zxf RE04Nested.tgz 2) $ mkdir RE04Nested-build 3) $ cd RE04Nested-build 4) $ cmake ../RE04Nested 5) $ make 6) $ ./RE04 7) Idle> /run/beamOn 10000
We are investigating the problem. ( Gabriele and I have worked on it.) We have been able to reproduce it, and to identify the condition that triggers it. At the moment we are not able to identify a problem in either the user code which creates the geometry or in the navigation code in Geant4. There is clearly an abnormal condition. A volume which has a nested parameterisation appears to be misidentified at some stage of Navigation (occasionally) as a replica - and the replica number is set to -1. It is not yet clear what causes this problem.
The problem has been finally identified in G4ReplicaNavigation::ComputeStep(), where the copy-number of the current volume for entering particles in replicas was occasionally wrongly set to -1. The fix is included in tag "geomnav-V10-02-01" and will be included in future patches to 10.1 and 10.2 series.
