Segmentation Fault in TRef

Hello,

I’m getting a segmentation fault in TRef that I don’t quite understand…

TRef::operator= (this=0x1753820, obj=<optimized out>) at root-6.06.08/core/base/src/TRef.cxx:301
301        SetUniqueID(uid);
(gdb) info locals
uid = 1

I used to have the same code working before I made some changes. I now have an abstract base class Event and a derived class SimEvent.

class Event: public TObject {
   // ...
   ClassDef(Event, 1);
}

class SimEvent: public Event{
   // ...
   TClonesArray* ecalSimHits;
   TClonesArray* trackerSimHits;
   TClonesArray* simParticles;

   ClassDef(SimEvent, 1);
}

The SimEvent class is basically just a container for a set of TClonesArray.

For instance, one of my output classes has a single TRef to another class object like …

class SimTrackerHit: public TObject {
    // ....
    TRef simParticle{nullptr};
}

It seems to work fine to add the objects to the event and persist it so long as the TRef is not set. But when I try to set a TRef on one of the output objects in my test, it immediately crashes.

Can anyone suggest how I might go about debugging this to see why TRef is crashing here when the unique ID is being set? Is a unique ID of 1 valid in this case? (In case it matters, I’m testing the output by adding 3 objects to the event.)

All I can think of is that ROOT is for some reason seeing the base Event class instead of SimEvent so the unique ID is not valid, but I don’t have enough understanding of the internals to say why that would be.

Thanks.

–Jeremy

Okay, it seems to be working with…

myRef.setObject(anObject);

Instead of

myRef = anObject;

Not sure why though, as it used to work before using the overloaded assignment operator.