Delphes problems: 3 questions for the authors...

[attilio]

Hello,
I've been playing with Delphes a lot in the past couple of weeks and I have 3 problems... one is a minor one but not the others. I'm using the master branch from GitHub and macosx. Let's start with the first one:

All tracks related quantities are filled in file modules/ParticlePropagator.cc but the curvature. I guess a command like

particle->C = 1./r;

is missing around L341.

The second one is a lot more weird. Here is how to reproduce it:
1) Compile Delphes with Pythia (I'm using pythia8307)
2) Uncomment the track branch in cards/delphes_card_CMS_PileUp.tcl
3) Produce few events with PU (I've used the MB file provided in delphes) using the card in 2) and examples/Pythia8/configNoLHE.cmnd
4) Dump the Generated Particles branch and the Track branch
5) Look at the generated particle associated to each track (when it exists). The first few lines are ok (I mean the track is a reasonable match to the gen particle, looking at Pt,Eta, Phi) but after few tenths of tracks the comparison is screwed and there is no relation at all among the track and the associated particle
6) If you look carefully the generated particle belongs to a previous event. Weird... isn't it?

Example entry n.4
track n.17:
Trk(pt,eta,phi) 5.56361 1.82801 2.42623 and the associated
Gen(pt,eta,phi) 5.01931 1.82801 2.42611
This is ok... but same entry and different track:
track n.150: 
Trk(pt,eta,phi) 1.62629 0.898563 -0.837461 and the associated
Gen(pt,eta,phi) 0.106294 -4.18368 2.71955
This Gen particle belongs to the Gen event n.2

7) My impression is that the PileUp Merger has some problem and break the track-particle relation (this problem appears only when you produce events with PU)

Last but not least... this is not a delphes problem but is related to the external package TrackCovariance. It's a very useful package and I'm using it a lot but it looks to me it a has a serious memory leak... I've tried to locate it but I'm not really an expert so I failed. What I see is that the memory used by root start to increase as soon as you start to call the VertexFit class and after 1000 events it reach a couple of GB and eventually crash if you run on a lot of events. You can see it simply running the macro examples/ExamplePVtxFind.C.

That's all folks. I like a lot using delphes and I hope some one could give me some hint on 2) and 3).

Best

Attilio