TaJetClustering crashes with DataNotAvailableException occasionally

[tmadlener]

We are currently running the standard miniDST workflow which occasionally crashes in the TaJetClustering module if the input collection is not present. We get the following error message:

Vertex collection (PrimaryVertex) is empty !
Vertex collection (PrimaryVertex) is empty !
***********************************************
A runtime error occured - (uncaught exception):
       lcio::DataNotAvailableException: LCEventImpl::getCollection: collection not in event:PFOsminuse
Marlin will have to be terminated, sorry.
***********************************************

The first two lines here originate from the IsolatedLeptonTagging processor:

MarlinReco/Analysis/IsolatedLeptonTagging/src/IsolatedLeptonTaggingProcessor.cc

Line 236 in 442d5cd

std::cerr << "Vertex collection (" << _colPVtx << ") is empty !" << std::endl;

This in turn also means that the IsolatedLeptonTagging processor does not produce an output collection. However, the TaJetClustering expects this collection to be present, and doesn't properly handle the case when it isn't:

MarlinReco/Analysis/TauFinder/src/TaJetClustering.cc

Lines 186 to 187 in 442d5cd

	LCCollection* col = evt->getCollection( _pfoCollectionName );
	if( !col ){ /*cout << "Failed to obtain PFO collection!" << endl;*/ return; }

[tmadlener]

Looking into the whole chain that is run by the miniDST workflow it might be easier to simply fill an empty collection in the IsolatedLeptonTagging, in order to not break downstream consumers who expect a collection to be there, resp. do not handle a missing collection gracefully.

The cleaner solution would obviously be to make all processors handle missing collections appropriately. E.g. if a missing collection could simply be interpreted as no or empty input.