Fix invalid memory access in ClassifyPhotonElectronAlg

The lines

photons.erase( std::remove_if( photons.begin(), photons.end(),
                                   [&]( const auto& hypo ) { return CaloMomentum( hypo ).pt() < m_eTcut; } ),
                   photons.end() );
    electrons.erase( std::remove_if( electrons.begin(), electrons.end(),
                                     [&]( const auto& hypo ) { return CaloMomentum( hypo ).pt() < m_electrEtCut; } ),
                     electrons.end() );

produced invalid memory access in very rare events when corrected calo hypos have pT lower than the same pT cut applied before correction. It seems KeyedContainer does not support remove_if? (cc @sponce) (worth to say that erasing elements happens very rarely, something like 1:1000 events) This was spotted when running (not yet merged) new test from Moore!386 (merged) (one has to run on more than 1000 events to see a crash).

Another small fix is changing m_eTcut -> m_photonEtCut in two places (perhaps a typo? cc @cmarinbe).

Thanks @sstahl for spotting the problem with erasing element in the very first commit.

Should go with Brunel!964 (merged) for references updates

added Calo label

added 1 commit

• b59543d3 - fixed erasing elements from KeyedContainer

Compare with previous version

added 1 commit

• 2c90a896 - Fixed formatting

Compare with previous version

changed the description

approved this merge request

unmarked as a Work In Progress

As discussed in private mails, let's merge this to get rid of the segfaults for now. As a second step, I'd like to remove the KeyedContainers used here. But it will be in a separate merge request.

assigned to @rjhunter

Could you put a bit of information in the description about where this seg fault occurs please? Was it seen in a nightly build or ci-test at some point? I'm happy to test it in the meantime.

/ci-test --merge

changed the description

resolved all threads

added lhcb-gaudi-head label

• [2020-02-15 00:11] Validation started with lhcb-gaudi-head#2533
• [2020-02-16 00:07] Validation started with lhcb-gaudi-head#2534
• [2020-02-17 00:10] Validation started with lhcb-gaudi-head#2535
• [2020-02-18 00:10] Validation started with lhcb-gaudi-head#2536
• [2020-02-19 00:10] Validation started with lhcb-gaudi-head#2537
• [2020-02-20 00:11] Validation started with lhcb-gaudi-head#2538
• [2020-02-23 00:07] Validation started with lhcb-gaudi-head#2539

mentioned in issue Moore#136 (closed)

assigned to @peilian and unassigned @rjhunter

This MR is far behind the master and there are some further discussions. It also seems that it triggered the failure in nightly test: https://lhcb-nightlies.web.cern.ch/logs/tests/nightly/lhcb-gaudi-head/2536/x86_64-centos7-gcc9-opt/Moore/ Should we unlable the lhcb-gaudi-head and wait for further updates?

added 83 commits

• 2c90a896...282b0731 - 82 commits from branch master
• b6ef3ddc - Merge remote-tracking branch 'origin/master' into fix_ClassifyPhotonElectronAlg

Compare with previous version

added 1 commit

• e9916fe2 - Apply suggestion to CaloFuture/CaloFutureReco/src/ClassifyPhotonElectronAlg.cpp

Compare with previous version