GitLab

Edward Moyse authored Sep 20, 2021 and Atlas Nightlybuild committed Sep 20, 2021

GeneratorObjectsTPCnv: Clear flow information when reading.

See merge request atlas/athena!46581

(cherry picked from commit 5c99a242)

a8bd69cf GeneratorObjectsTPCnv: Clear flow information when reading.

Fix fourvector initialization for hepmc3 case.

Scott Snyder authored Mar 17, 2021 and Edward Moyse committed Mar 17, 2021

Get the DataPool code working with hepmc3.

Some additional work is probably needed later to deal with the additional
dynamic allocations done by hepmc3.

Remove `atlas_depends_on_subdirs` and cleanup library dependencies.

Cleanup of the GeneratorObjectsTPCnv: removed wrong comments and moved copypasted code into a separate include

For HepMC2
```
const HepMC::GenParticlePtr == HepMC::GenParticle* const
HepMC::ConstGenParticlePtr == const HepMC::GenParticle*
```

Susumu Oda authored May 28, 2020 and Frank Winklmeier committed May 28, 2020

Remove unnecessary stuff from InDetEventAthenaPool_test.txt. Cosmetic changes on InDetSimDataCollectionCnv_p1_test.cxx.

`SG::CurrentEventStore::store()` is still used during P->T conversion as this has to work with
the pile-up digitization use-case, where multiple `StoreGateSvc` instances are used.

This commit updates the `HepMcParticleLink` class and the `HepMcParticleLinkCnv_p1`,
`HepMcParticleLinkCnv_p2`, `SiHitCollectionCnv_p3` and `TRT_HitCollectionCnv_p4` classes
so that the `HepMcParticleLink` unit tests pass.
It should also fix the problem seen in ATLASSIM-4539, which occurred when there were no
HepMcParticleLinks to the hard-scatter GenEvent.

This change updates the HepMcParticleLink unit tests, so that they cover all use-cases found
in current production jobs.

Library dependency fixes.

Update transient HepMcParticleLink to use unsigned int rather than unsigned short to store the eventIndex

This change should allow the HepMcParticleLink to accomodate all allowed values of `GenEvent::event_number`
i.e. up to `INT_MAX` = 21^31 -1 = 2147483647.

A new persistent class is not being added at this time, but existing workarounds allow event numbers larger
than max unsigned short to be correctly recovered for links pointing to the first event in the McEventCollection.
In the standard production workflow, pile-up background event numbers are reset to to numbers in the range
0 to 2^15 -1, which should be OK up to <mu> values of 840 with current settings.

A new persistent version of HepMcParticleLink will be introduced after the migration to HepMC3.

In the case that the standard production workflow is being used
then the first event in the `McEventCollection` is the only one
where the `eventIndex` could exceed the maximum value of unsigned
short. In this case we can work around the issue by using an
`m_mcEvtIndex` of zero as a special case, in which `m_mcEvtIndex`
should be interpreted as the position of the `GenEvent` in the
`McEventCollection` rather than the value of `GenEvent::event_number()`.

Similar changes are made to SiHitCollectionCnv_p3.cxx and TRT_HitCollectionCnv_p4.cxx.

Andrii Verbytskyi authored Apr 22, 2020 and Frank Winklmeier committed Apr 22, 2020

This reverts commit 84d28a23.

This commit extends `HepMcParticleLinkCnv_p1_test.cxx` and `HepMcParticleLinkCnv_p2_test.cxx`
to check links to `GenParticle` objects which are not part of the first `GenEvent` in the
`McEventCollection`. It also adds a check to ensure that links are correct when the position
of a `GenEvent` in the `McEventCollection` rather than its event number is used.

In the 21.0 branch an eventIndex of zero means that the transient `HepMcParticleLink` looks
at the first event in the `McEventCollection`, otherwise the value returned by
`HepMcParticleLink::eventIndex()` represents the GenEvent::event_number of a `GenEvent` in
the `McEventCollection` and this is used to find the appropriate `GenEvent` in the `McEventCollection`
when using the link.

NB There is a weakness here that if more than one `GenEvent` has the same event number all links will
point to the first `GenEvent` which matches. This will be fixed after the migration to `HepMC3`.

Also dropped references to separate pile-up `McEventCollections` as the use of these is not supported
in 21.0 releases.