RD lines crashing when no PVs

The following error occurs when running hlt2_pp_thor.qmt test (after about 15k events):

rd_detached_dimuon#2                  ERROR Empty vertex container passed to Sel::calculateBestVertex()
HLTControlFlowMgr                     FATAL Event failed in Node TwoBodyCombiner/rd_detached_dimuon#2

Quick look shows a lot of lines in RD don't have proper filters.

So to-do list:

• (always) apply proper (already pre-defined) filters, instead of privately using upfront_reconstruction at each line
• b_to_hh(h)gamma(...) lines can be merged in one file

added RD RTA hlt2 labels

mentioned in merge request !1279 (merged)

@wouter

Have you observed this crash today for the first time or already some time ago? In order to pinpoint which set of lines can cause this crash.

About b_to_hh(h)gamma(...) lines: they are currently being ported from LoKi to ThOr and their whole structure will be changed.

Few days ago. Easiest to pinpoint is to check which lines use directly upfront_reconstruction in their definition, this should be rd_prefilter or rd_prefilter_nopvs or something like that. I quickly replaced all these with the former and this solves it, as it requires pvs in the event (but which one exactly should be chosen carefully of course).

Thanks, that is exactly what I am planning to do. Give me a few minutes to open some hotfix.

Related hotfix: !1404 (merged)

Tested locally with !1404 (merged) and RD lines are now fine, but saw error with some BandQ lines (pinging @mengzhen ), so possibly all groups should check they are requiring PV.

Hi Miro, thank you for the information !

In bandq directory, both hlt2 and sprucing lines are using a prefilter containing a PV requirement, which is defined in a individual file.

Seems that the only difference between rd and bandq prefilter is that, bandq is using make_pvs and rd is using make_pvs_v2. Is the suggestion to use make_pvs_v2 to be consistent with the exact PVs we are using for line builders ?

If that's the case, this might be a shared issue by many WGs, for example b2oc and b2cc.

mentioned in merge request !1404 (merged)

mentioned in merge request !1225 (merged)

make_pvs_v2 is actually about to become obsolete, so please don't use, if you have a choice.

@wouter Good point, thanks. So a standard make_pvs should be used or even a different function?

use make_pvs ... of course, if it then doesn't run, keep using make_pvs_v2, and we'll migrate later.

marked the checklist item b_to_hh(h)gamma(...) lines can be merged in one file as completed

marked the checklist item (always) apply proper (already pre-defined) filters, instead of privately using upfront_reconstruction at each line as completed

closed