Only return a non empty cluster splitting probability container name, if it is either in the input file
or if a tool exist which would create it. This is a workaround to fix the configuration of combined
muon reconstruction in the trigger.

Allow the PixelClusterOnTrackTool to renounce the input cluster splitting probability map.
This is used to suppress implicit circular dependency warnings when the PixelClusterOnTrackTool
is used indirectly by the ambiguity resolver where the ambiguity resolver produces the cluster
splitting probability map which is read by the PixelClusterOnTrackTool.

Originally the cluster splitting probabilities were stored in a reserved word of the
corresponding PixelClusters. This however required to update a collection already stored
in StoreGate i.e. was not thread-safe.

Now the splitting probabilities are stored in separate maps, which are copied and updated
when passed from algorithm to algorithm. The algorithms record the map at the very beginning
to allow their tools (PixelClusterOnTrackTool) to access the map via read handles rather than
passing the map down the call chain explicitly.

If NnClusterizationFactory.useTTrainedNetworks is set, then we should
also clear NnClusterizationFactory.NnCollectionJSONReadKey,
to prevent having a spurious conditions dependency.

- remove the pixel covariance correction,
- remove dRmap based identification of collimated tracks,
- remove creation of dRmap.

All of the above is unused and was not properly migrated to athenaMT.

Robin Newhouse authored 4 years ago and Frank Winklmeier committed 4 years ago

Other changes include:
snake_case -> camelCase
stl math functions
Removing NN configuration from new tracking cuts
A temporary override of config filepath for testing purposes. Will be removed before final merge.

The RAWtoESD step of q431 was broken intentionally by the previous
commit because it removes a bunch of configurations options from the
hole search tool. This commit amends this by modifying several bits of
configuration where an ID hole search tool is created.

The TRT_SegmentsToTrack requires a PRDtoTrackMap to compute shared TRT hits. The
map is produced when creating the alternative TRT standalone tracks. Thus, the
has to be produced first.

Add configurable flag ofr CaloClusterEt for ROiSeed in NewTrackingCuts and changing default cutlevel to 19

Also, the SolenoidParametrization is turned into conditions data.

If the flag is set the TRT error scaling will use mu from either
the offline or online luminosity conditions data.

Further fixes for python 3.

Edson Carquin Lopez authored 5 years ago and Adam Edward Barton committed 5 years ago

TRT alignment code: move to conditions handle, migrate client TRT_DetElementsRoadMaker, remove callback and create new conditions alg.

Removing Cache

Adding TRT_DetElementsRoadCondAlg_xk to the sequence within the new tracking configuration

Remove alignment callback from TRT_DetElementsRoadMaker_xk

Including TRT_DetElementsRoadCondAlg into InDetTrigConfigRecNewTracking.py and cleaning unused variables, commented out lines and lines included for debugging

Updating/adding copyright statements.

Use the average interactions per crossing from the luminosity conditions
data for the TRT drift circle mu corrections. This works around the problem
that the average interactions per crossing is currently not set in the xAOD
event info.

The mu correction is current disabled in the configuration to avoid output
changes.

Anthony Morley authored 5 years ago and Walter Lampl committed 5 years ago

Clean up logic in GSF extrapolator  --  still needs some work but this is a little less verbose and confusing