FastJetContrib tag 1.033
In this branch the Externals/FastJetContrib/CMakeLists.txt files has been changed to retrieve the fjcontrib-1.033 tarball (previously it was retrieving 1.030), and the MD5 hash has also been updated. The 1.033 tag of FastJetContrib contains fixes to some memory leaks in the EnergyCorrelator package that were present in 1.030 and become very obvious when running large samples on the grid. Here is the EnergyCorrelator revision log: https://fastjet.hepforge.org/trac/log/contrib/contribs/EnergyCorrelator/trunk.
During compilation there were messages about a previously applied patch, so the line in the FastJetContrib CMakeLists.txt related to the patch was also removed. The patch file and directory are still present but can easily be removed.
The changes are very minor (3 lines in the CMakeLists.txt file), but the details of how this was tested are below in case something more stringent is required.
Testing was performed with this package:
atlasexternals is a submodule of that package, with the package filters set to keep only FastJet and FastJetContrib so that only these externals are recompiled locally (compilation of FastJetContrib fails if FastJet is not also compiled locally, even though there are no changes to Externals/FastJet).
After checking out the submodule (the 1.0 branch of atlas/externals), the jkinghor/atlasexternals fork was added as a remote and this branch (1.0_fjcontrib-1.033) was merged in to get the CMakeLists.txt changes.
The package is designed to do jet reconstruction using the JetDualUseConfig PyROOT setup authored by Pierre-Antoine Delsart. After compiling (including the local recompilation of FastJet and FastJetContrib!), the package is used to reconstruct jets and run a jetmodifier which retrieves some of the EnergyCorrelatorGeneralizedFunctions from fjcontrib/EnergyCorrelator, in particular the ones related to the N2 jet substructure moment.
Original testing was done with AnalysisBase 21.2.13 (as is the current setup in the branch on gitlab), but I recently re-tested with the head of atlas/athena 21.2 and AnalysisBase 21.2.23. Grid jobs with this branch were successful, so I assume (without running valgrind or something similar) that the memory leaks are indeed fixed!
Please let me know if any more testing is required for this MR!