Make sure there is a TGeoNavigator per thread before calling Contains

Required to make FTHitEfficiencyMonitor work when multi-threaded, as currently the geometry navigation is not configured for multi-threading.

Requires LHCb!4652 (merged) where the number of threads is set.

I provided a detailed description of the relevant functions in LHCb!4652 (merged), and then on slide 8 (titled "Usage") of this very old presentation describing the parallelisation of the ROOT Geometry library (the only relevant documentation I could find) it shows the required change in this MR https://indico.cern.ch/event/193979/contributions/357987/attachments/280434/392148/ParallelizingROOTgeometry.pdf

I think this is the correct ROOT way to do it, but in future we can probably find a better way to do it.

---

Validated by

• Core Software
• RTA
• Simulation

Core/src/DeIOV.cpp

@@ -27,6 +27,9 @@
 #include <sstream>
 #include <vector>
 
+#include "TGeoManager.h"
+#include "TGeoNavigator.h"
+
 using lhcb::geometrytools::toTransform3D;
 
 /// Initialization of sub-classes
@@ -115,6 +118,13 @@ void LHCb::Detector::detail::DeIOVObject::print( int indent, int flg ) const {
 bool LHCb::Detector::detail::DeIOVObject::isInside( const ROOT::Math::XYZPoint& globalPoint ) const {
   auto     localPoint = toLocal( globalPoint );
   Double_t p[3]{toDD4hepUnits( localPoint.x() ), toDD4hepUnits( localPoint.y() ), toDD4hepUnits( localPoint.z() )};
+  // Geometry navigation requires thread local navigators
+  // manager.SetMaxThreads(n_threads) should be set once the geometry is locked in the initialisation of DD4hepSvc
+  // then you need to check there is a TGeoNavigator for the thread
+  auto&          manager = dd4hep::Detector::getInstance().manager();
+  TGeoNavigator* nav     = manager.GetCurrentNavigator();
+  if ( !nav ) nav = manager.AddNavigator();
+  assert( nav != nullptr );
   return detector().solid()->Contains( p );
 }