[ATR-19997] Propagate Filter output collections in RDOtoRDOTrigger. Tidy up feature access.
Working towards fixing the crash reported here https://its.cern.ch/jira/browse/ATR-19997?focusedCommentId=2778617&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-2778617
Compliance fixes to the InputMakers
- We should not be re-attaching existing
"muons"link. It can be fetched from the parent (
"seed") Decision object. cc @sshaw
- We should not be re-attaching existing jet link as
"feature". It can be fetched from the parent (
"seed") Decision object. Also - only HypoAlgs should attach links with the name
"feature". (If this turns out to be really needed, then the name of the link should change.)
- We should use
"view"to minimise hard-coded strings.
TrigBjetBtagHypoAlgMT.cxx to fail online validation, the presence of a
"feature" link in the InputMaker's output tricked the validation into thinking that
TrigBjetBtagHypoAlgMT.cxx was acting like a combi-hypo
- Add dummy
TrigBjetBtagHypoAlgMT.cxx, until a real feature is ready. cc @cvarni
I found that Filter navigation collections were not being propagated to the output file.
collectFilters were being given a sequencer which was one level too high, but
collectHypos had a slightly different logic and was able to cope. Hence some output was generated and I did not spot before.
- Collect all hypos and filters from
- Add extra protection in
TrigCompositeUtils::getRejectedDecisionNodesto not follow invalid
seedlinks. Prints warning that navigation information is incomplete in the file.
This then caused Tau monitoring to crash as we started to find tau features - but return invalid element links. This I believe is understood and will be fixed by @carquin as discussed on mattermost (it's the
Py:TriggerConfig WARNING The producer of the TAUCaloRoIs not in the menu, it's outputs won't ever make it out of the HLT issue)
- Make invalid online taus print a WARNING and then continue