Overlaps found with 6.32 no longer found in 6.34+

[olantwin]

Check duplicate issues.

• Checked for duplicates

Description

Known overlaps visible by eye in the OGL viewer are not detected in new versions (6.34.08+?) when they were detected correctly in 6.32.08.

Reproducer

File with geometry: https://cernbox.cern.ch/s/RkfJhNsWlj3tkRT

Reproducible easily in interactive ROOT session:

auto f = TFile::Open("geofile_full.conical.Pythia8-TGeant4.root");
FAIRGeom->CheckOverlaps();
FAIRGeom->PrintOverlaps();

Expected output:

Info in <TGeoNodeMatrix::CheckOverlaps>: Checking overlaps for cave and daughters within 0.1
Warning in <TGeoMatrix::dtor>: Registered matrix  was removed00:00
Warning in <TGeoMatrix::dtor>: Registered matrix Identity was removed
Warning in <TGeoMatrix::dtor>: Registered matrix  was removed0:08
Warning in <TGeoMatrix::dtor>: Registered matrix  was removed
Check overlaps:      [==========] 45747 [100.00 %]           0:11
Info in <TGeoNodeMatrix::CheckOverlaps>: Number of illegal overlaps/extrusions : 2

=== Overlaps for FAIRGeom ===
 = Overlap ov00000: cave/Cavern_1 overlapping cave/SHiPMagnet_1/magyoke_1 ovlp=94
 = Overlap ov00001: cave/Cavern_1 overlapping cave/SHiPMagnet_1/MCoil3_1 ovlp=69

Actual output:

Info in <TGeoNodeMatrix::CheckOverlaps>: Checking overlaps for cave and daughters within 0.1
Info in <TGeoNodeMatrix::CheckOverlaps>: Number of illegal overlaps/extrusions : 0

Overlap also visible using FAIRGeom->GetMasterVolume()->Draw("ogl")

ROOT version

Working:

   ------------------------------------------------------------------
  | Welcome to ROOT 6.32.08                        https://root.cern |
  | (c) 1995-2024, The ROOT Team; conception: R. Brun, F. Rademakers |
  | Built for linuxx8664gcc on Nov 14 2024, 09:59:57                 |
  | From tags/v6-32-08@v6-32-08                                      |
  | With c++ (GCC) 11.4.1 20231218 (Red Hat 11.4.1-3)                |
  | Try '.help'/'.?', '.demo', '.license', '.credits', '.quit'/'.q'  |
   ------------------------------------------------------------------

Broken:

   ------------------------------------------------------------------
  | Welcome to ROOT 6.34.08                        https://root.cern |
  | (c) 1995-2024, The ROOT Team; conception: R. Brun, F. Rademakers |
  | Built for linuxx8664gcc on Apr 07 2025, 20:28:44                 |
  | From tags/v6-34-08@v6-34-08                                      |
  | With c++ (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5)                |
  | Try '.help'/'.?', '.demo', '.license', '.credits', '.quit'/'.q'  |
   ------------------------------------------------------------------

6.30.* also works, 6.36 also demonstrates the issue.

Installation method

Both taken from LCG builds on CVMFS, but also reproducible using ROOT built in the SHiP software stack.

Operating system

Almalinux 9 (LXPLUS)

Additional context

No response

[olantwin]

An additional observation: The overlap checks with ROOT 6.34+ run extremely fast, regardless of precision chosen. It seems that maybe something is skipped?

If it matters, the test geometry was created with ROOT 6.30/08 as part of the SHiP stack:

   ------------------------------------------------------------------
  | Welcome to ROOT 6.30/08                        https://root.cern |
  | (c) 1995-2024, The ROOT Team; conception: R. Brun, F. Rademakers |
  | Built for linuxx8664gcc on Jan 09 2025, 09:37:27                 |
  | From tags/v6-30-08@v6-30-08                                      |
  | With c++ (GCC) 11.4.1 20231218 (Red Hat 11.4.1-3)                |
  | Try '.help'/'.?', '.demo', '.license', '.credits', '.quit'/'.q'  |
   ------------------------------------------------------------------

[ferdymercury]

Could the problem be that 6.34 was somehow built without "geom" support?

[olantwin]

Could the problem be that 6.34 was somehow built without "geom" support?

In our stack build options definitely haven't changed, for the LCG builds I'd need to check.
Do you happen to know whether there is a way to check this using e.g. root-config? Looking at the --features list, I didn't notice any differences that might be related.

If this is the case, shouldn't ROOT raise at least a warning?

[olantwin]

I confirm that the LCG version is built with geom for 6.34, and the feature wasn't yet exposed via root-config for 6.32

[dpiparo]

It is indeed built with all options available, and geom is ON by default anyway. @agheata perhaps has a suggestion?

[ferdymercury]

There was a change in GetGeomPainter less than 1 year ago, not sure if it might have affected:

An attempt could be to run with root --web=off

[olantwin]

There was a change in GetGeomPainter less than 1 year ago, not sure if it might have affected:

An attempt could be to run with root --web=off

This is indeed the culprit! With root --web=off, the overlap is found again:

[olantwin@lxplus984 SHiP_Software]$ root --web=off
   ------------------------------------------------------------------
  | Welcome to ROOT 6.34.08                        https://root.cern |
  | (c) 1995-2024, The ROOT Team; conception: R. Brun, F. Rademakers |
  | Built for linuxx8664gcc on Apr 07 2025, 20:28:44                 |
  | From tags/v6-34-08@v6-34-08                                      |
  | With c++ (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5)                |
  | Try '.help'/'.?', '.demo', '.license', '.credits', '.quit'/'.q'  |
   ------------------------------------------------------------------

root [0] auto f = TFile::Open("geofile_full.conical.Pythia8-TGeant4.root")
(TFile *) 0x32d6560
root [1] FAIRGeom->CheckOverlaps(0.1)
Info in <TGeoManager::CloseGeometry>: Geometry loaded from file...
Info in <TGeoManager::SetTopVolume>: Top volume is cave. Master volume is cave
Info in <TGeoNavigator::BuildCache>: --- Maximum geometry depth set to 100
Info in <TGeoManager::Voxelize>: Voxelizing...
Info in <TGeoManager::CountLevels>: max level = 5, max placements = 14153
Info in <TGeoManager::CloseGeometry>: 45747 nodes/ 698 volume UID's in FAIR geometry
Info in <TGeoManager::CloseGeometry>: ----------------modeler ready----------------
Info in <TGeoNodeMatrix::CheckOverlaps>: Checking overlaps for cave and daughters within 0.1
Check overlaps:      [==========] 45747 [100.00 %]           0:16
Info in <TGeoNodeMatrix::CheckOverlaps>: Number of illegal overlaps/extrusions : 2

root [2] FAIRGeom->PrintOverlaps()
=== Overlaps for FAIRGeom ===
 = Overlap ov00000: cave/Cavern_1 overlapping cave/SHiPMagnet_1/magyoke_1 ovlp=94
 = Overlap ov00001: cave/Cavern_1 overlapping cave/SHiPMagnet_1/MCoil3_1 ovlp=69
root [3]

I assume this is not the desired behaviour of the switch to the web graphics?

[ferdymercury]

pinging @linev

[linev]

Yes, geometry checker is integrated in GL-based geompainter library.
Probably one can move TGeoChecker class in separate library,
but for now --web=off is only solution to use it.

[olantwin]

Yes, geometry checker is integrated in GL-based geompainter library. Probably one can move TGeoChecker class in separate library, but for now --web=off is only solution to use it.

Thanks for this additional information.

Is there a comparable workaround for PyROOT?

[linev]

One can set WebGui.Display: off in .rootrc file.
Or call ROOT.gROOT.SetWebDisplay("off") in the beginning of python script

[ferdymercury]

One can also use a rootlogon.py file or, maybe in the future, just an env variable, see #18770

[olantwin]

Thanks, while .rootrc and rootlogon.py might be useful for users, ROOT.gROOT.SetWebDisplay("off") is probably the way to go in our framework, as I can't control the user's environment or logon scripts.

Thanks to all for helping to find the workarounds!

To fix the underlying issue, the geometry checker would need to be made independent of the GL geometry visualisation?