reco::FormulaEvaluator segFaults if function string is not hard-coded/pre-defined

[kirschen]

Dear experts,

in preparation for a new JEC-release, the parametrization has been extended by a TMath::Exp-expression. However, the the reco::FormulaEvaluator segFaults with this extension.

After inspecting
https://github.com/cms-sw/cmssw/blob/master/CommonTools/Utils/src/FormulaEvaluator.cc#L562-L575
and following, it seems like any new function needs to be announced to the reco::FormulaEvaluator, i.e. if we are to introduce any new function, we should inject it to reco::FormulaEvaluator as early as possible. For now, we replaced "TMath::Exp" by "exp" in our parametrization to make it work.

Would it be possible to catch "undefined" parametrizations instead of segfaulting to make it obvious what goes wrong? Maybe just replace the return in
https://github.com/cms-sw/cmssw/blob/master/CommonTools/Utils/src/FormulaEvaluator.cc#L660
by an error?

Thanks,
Henning for the JERC-group

FYI: Previous "JetMET"-issues with reco::FormulaEvaluator e.g. #16716 (@Dr15Jones @slava77 )

[cmsbuild]

A new Issue was created by @kirschen .

@davidlange6, @Dr15Jones, @smuzaffar, @fabiocos can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[slava77]

@kirschen
I think that an error here is reasonable.
Would you like to submit a pull request for this?

[slava77]

assign analysis,reconstruction

[cmsbuild]

New categories assigned: reconstruction,analysis

@slava77,@perrotta,@monttj you have been requested to review this Pull request/Issue and eventually sign? Thanks

[slava77]

@kirschen
please clarify if you have a plan to make a PR with the change proposed by you when the issue was opened.
Thank you.

[kirschen]

Hi, sorry for the delay. Yes, I can take care of the small PR throwing an exception if none of the predefined functions are found in the next couple of days.

[kirschen]

Will link it here, then.

[slava77]

@kirschen
I wanted to check on the status of that small PR.
Please update.
Thank you.

[kirschen]

Issue is more generically resolved with this PR:
#22354

that gives a useful exception with the initial expression reported here
kirschen@lxplus067 src $ root jec_issue_Dec2017.C
root [0]
Processing jec_issue_Dec2017.C...
terminate called after throwing an instance of 'cms::Exception'
what(): An exception of category 'FormulaEvaluatorParseError' occurred.
Exception Message:
While parsing '2' could not parse beyond '[2](31./([6]+[7]100./3.(TMath::Max(0.,1.03091-0.051154pow(x,-0.154227))-TMath::Max(0.,1.03091-0.051154TMath::Power(208.,-0.154227)))+[8]0.021(-1.+1./(1.+'

[perrotta]

+1

• Solved with Properly handle parsing failure in FormulaEvaluator #22354, see reco::FormulaEvaluator segFaults if function string is not hard-coded/pre-defined #21851 (comment)

[perrotta]

@kirschen : you can close this issue

[kirschen]

Thanks, all!