Problem 1976 - Overlap Check wrongly issues warning for nested cones.
Summary: Overlap Check wrongly issues warning for nested cones.
Status: RESOLVED WORKSFORME
Alias: None
Product: Geant4
Classification: Unclassified
Component: geometry (show other problems)
Version: 10.2
Hardware: PC Linux
: P4 minor
Assignee: Gabriele Cosmo
URL:
Depends on:
Blocks:
 
Reported: 2017-05-18 17:31 CEST by Valentin Volkl
Modified: 2017-05-19 10:40 CEST (History)
1 user (show)

See Also:


Attachments
Example constructing two nested cones and running the overlap checker. (40.00 KB, application/x-tar)
2017-05-18 17:31 CEST, Valentin Volkl
Details

Note You need to log in before you can comment on or make changes to this problem.
Description Valentin Volkl 2017-05-18 17:31:03 CEST
Created attachment 458 [details]
Example constructing two nested cones and running the overlap checker.

In a geometry configuration of two nested cones, the overlap checker ("geometry/test/run") gives different results depending on whether one of the cones is rotated by pi + R*1 / R*2 are interchanged or not (which should in principle be equivalent). See the attachment for a complete testcase.

Note that this problem seems to be version specific. It seems to be present in 10.2 / the version that is in LCG release 88, but not in 10.1. I didn't test any other versions.
Comment 1 Gabriele Cosmo 2017-05-19 10:30:00 CEST
We cannot reproduce the problem, using a vanilla installation of Geant4 (either 10.2 or more recent). The overlap check behaves as expected.
Looking at the setup being used, based on the LCG build of Geant4-10.2.p02 at CERN, I noticed the LCG_88 build is making use of the VecGeom library, to replace the original shapes implementation in Geant4. The problem is therefore affecting VecGeom itself.
I'm closing this ticket and transferring it to VecGeom JIRA.
Comment 2 Gabriele Cosmo 2017-05-19 10:40:21 CEST
JIRA ticket open for VecGeom:
https://its.cern.ch/jira/browse/VECGEOM-433