When I run my G4 11.4.1 job, those WARNING messages are showing up via rm->ReinitializeGeomtry(true) on the master thread. Apparently the geometry isn’t being reopened, despite the call to G4GeometryManager::OpenGeometry() at G4RunManager.cc:1038. I guess that Dmitri is seeing this as well.
The only difference between the G4 10.07.p04 and 11.4.1 versions of OpenGeometry() is that the latter checks for IsMasterThread(). But my job shows that it is is being called on the master thread, so that should succeed.
I think I traced the 11.4 problem to the MR which introduced parallel voxelisation optimisation. In my small reproducer, the store-clean warnings disappear if I disable this optimisation before the first geometry initialisation.
Could you try this in your code, before the first Initialize() / first geometry build?
That appears to be successful, Dmitri. I was able to do three runs, with geometry mods in between, and got no errors of any kind (including no GeomNav0002).
In my code, I put that line before the rm->InitializeGeometry() command, without any sort of if-first check. I presume that’s just redundant – once it’s false, it would just stay false. Also, I’ve wrapped it with #if G4VERSION_NUMBER >= 1130 (so we can maintain our framework against various G4 installations).
Would you expect this to be needed moving forward, or is there likely to be a “fix” introduced to 11.4.2?
Hey, Dmitri and Gabriele. It appears that we’ve gotten away from the GeomNav0002 error that was my original topic (I’m not getting that complaint, without changing how we build our world volumes at all). I’m happy to close this topic, marking one of Dmitri’s posts as Solution.
At the same time, we’re still having trouble with multiple geometry builds. My collaborator @BirgitZatschler is seeing an exception when we use importance biasing in our framework. Shall I create a new forum topic for that error (in case there are other users who run into it)? Or would you prefer that we move this discussion to e-mail?
Hello Michael, next days I will open a bugzilla ticket for this issue. If it is confirmed and fixed, I assume the fix will be considered for a future patch release. Once I have the ticket number, I will link it here.
For the remaining/new issue, I will let @gcosmo decide whether it is better to open a new thread or move it to email
@mkelsey, if you have a reproducer of the problem, you should directly open a ticket in Bugzilla so the information is kept in one place and can be tracked.
