`simulate.py` scans return only 1 event

[c-dilks]

Environment: (where does this bug occur, have you tried other environments)

• Which branch (often main for latest released): main

Steps to reproduce: (give a step by step account of how to trigger the bug)

1. simulate.py -t 6 -k 5 -n 30 (eta scan, used in CI)

Likely other tests (set with -t option) fail as well. simulate.py generates a Geant4 macro file then runs it with npsim; the tests that now (likely) fail have the following pattern:

/control/verbose 2
/run/initialize
/gps/verbose 2
/gps/particle pi+
/gps/number 1
/gps/position 0 0 0 cm

# eta scan, with varied azimuth for each eta bin

# eta=1.6  theta=22.828754832963863 deg
/gps/direction 0.2679262666315343 -0.2806111036063496 0.9216685544064713
/gps/ene/mono 4.862377599175745 GeV
/run/beamOn 1
/gps/direction 0.35670337364908533 -0.15260989170314782 0.9216685544064713
/gps/ene/mono 4.862377599175745 GeV
/run/beamOn 1
/gps/direction 0.355244674481362 -0.15597530917076796 0.9216685544064713
/gps/ene/mono 4.862377599175745 GeV
/run/beamOn 1
/gps/direction -0.3269047091825161 -0.2089506806224379 0.9216685544064713
/gps/ene/mono 4.862377599175745 GeV
/run/beamOn 1
/gps/direction 0.1315721232832751 0.3649874685424366 0.9216685544064713
/gps/ene/mono 4.862377599175745 GeV
/run/beamOn 1

...

Expected Result: (what do you expect when you execute the steps above)

An output file with many events, and stdout from npsim includes (irrelevant lines omitted):

ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 1 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 2 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 3 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 4 Begin event action. Access event related information.
...

Actual Result: (what do you get when you execute the steps above)

An output file with 1 event, and the corresponding stdout is repeated for event 0:

ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
ParticleHandler  INFO  +++ Event 0 Begin event action. Access event related information.
...

See example CI log

[c-dilks]

@wdconinc did something change in Geant4 or DD4hep such that the pattern of multiple /run/beamOn 1 calls (with varying gun directions, for example) now results in only a single event? Previously it ran as many events as /run/beamOn 1 calls.

[c-dilks]

@chchatte92 because of this issue, we don't really have CI working here on drich-dev; this would need to be fixed if you still want to move simulate.py to reconstruction_benchmarks, or you'll have to make your own npsim commands etc. there, or generate HEPMC files (however you prefer to run benchmarks is up to you).

EDIT well, CI does work here, this issue just means the eta and momentum scans all only process a single event.

[wdconinc]

I don't know specifically what could explain this. Geant4 formally starts a new G4Run with each /run/beamOn 1 call, and it might be that dd4hep now overwrite the previous run in the edm4hep file instead of adding events to a single run. Just a thought.

[c-dilks]

Could it be the change to PODIO frames?

That's one difference I see after comparing recent CI output to old (working) CI output:

• new logs
• old logs

All of the eta_scan jobs have this problem.

Also I was a bit wrong about the lines Event # Begin event action... incrementing, but I'm confident those linked old logs were from a workflow run in a working state, since they are from the PR merge commit that added these particular jobs.

[wdconinc]

The change would indeed likely have come with the update to the EDM4hep output plugin to write podio frames.