Problem 1991

Summary:	G4IStore::GetImportance() crashes with null G4VPhysicalVolume
Product:	Geant4	Reporter:	mstoeckl
Component:	geometry/biasing	Assignee:	alexander.howard
Status:	RESOLVED FIXED
Severity:	normal
Priority:	P4
Version:	10.3
Hardware:	All
OS:	All

Description mstoeckl 2017-07-11 23:42:16 CEST

The biasing process G4ImportanceBiasing can end up calling
G4IStore::GetImportance with a null physical volume; this 
causes a crash because one cannot add a null G4VPhysicalVolume
to the store.

G4ImportanceProcess::PostStepDoIt will call G4IStore::GetImportance
with a G4GeometryCell constructed from the pre/post point volumes
of a G4Step; and the G4SteppingManager uses a null volume
to signify a point being on the world boundary.

[The fix involving the least lines of code: default G4IStore::GetImportance
to 1./world volume importance when the volume is null. Alternatively,
test for a null volume in everything that might run after hitting the world boundary error messages, as they crash when trying to access properties 
of the null volume, for instance the G4IStore::GetImportance error message. 
Alternatively, constrain when the biasing is run to not happen
on exiting the world volume.]

Comment 1 alexander.howard 2023-11-21 16:40:05 CET

https://geant4-forum.web.cern.ch/t/g4istore-getimportance-crashes-with-null-g4vphysicalvolume/11132
Finds this problem manifested if a volume is placed outside of the world volume. Giving an anonymous segmentation fault is equally unhelpful, therefore a check should be introduced to throw a more controlled G4Exception instead. Will be added to the fix.

Comment 2 alexander.howard 2023-11-21 21:38:36 CET

I couldn't reproduce a crash for this, however, I can see the hazard of accessing the pointer to the PostStep volume, therefore I have returned no change if the next volume is null. Will be included in release v11.2.