Following on from ATR-21284, I discarded as un-workable the solutions to the following-unwanted-edged from HLTPassRaw problem which tried to resolve this via adding extra metadata.
The only remaining solutions it to add extra event data structure into the nav graph to prevent the following of unwanted-edges.
This MR adds extra Filter nodes into HLTNav_Summary which make sure that each exploration out from the HLTPassRaw node for a given chain only jumps into the graph at a node which definitely passed the event for that chain, not simply passed the hypo for that chain.
This is one half of the updates needed to do offline feature matching (cc @jburr)
In addition
- Remove mentions of rerun from L1Decoder (cc @tbold)
- Rework the manually created nav graph used to test feature retrieval. Update it to the new(ish) paradigm of ComboHypos and single InputMaker output collections. Include the new Filter objects added by this MR.