A strange bug in rf708_bphysics.C

[Zhihong_Shen]

Dear experts,
I found a trange bug in line 169 of rf708_bphysics.C.
For the RooBDecay, if we change the definition of bcpg to

RooBDecay bcpg("bcpg", "bcpg", dt, tau, DG, RooConst(1), fsinh, fcos, fsin, RooConst(0.), truthModel, RooBDecay::DoubleSided);

where dm is changed to constant 0.
The following bug will be obtained:

input_line_61:2:89: error: array subscript is not an integer
Double_t TFormula____id9800548585748267179(Double_t *x){ return TMath::Exp(-TMath::Abs(x[x[2]])/x[1])*TMath::Cos(x[x[2]]*x[2]) ; }
                                                                                        ^~~~~
input_line_61:2:115: error: array subscript is not an integer
Double_t TFormula____id9800548585748267179(Double_t *x){ return TMath::Exp(-TMath::Abs(x[x[2]])/x[1])*TMath::Cos(x[x[2]]*x[2]) ; }
                                                                                                                  ^~~~~
input_line_62:2:89: error: array subscript is not an integer
Double_t TFormula____id9800548585748267179(Double_t *x){ return TMath::Exp(-TMath::Abs(x[x[2]])/x[1])*TMath::Cos(x[x[2]]*x[2]) ; }
                                                                                        ^~~~~
input_line_62:2:115: error: array subscript is not an integer
Double_t TFormula____id9800548585748267179(Double_t *x){ return TMath::Exp(-TMath::Abs(x[x[2]])/x[1])*TMath::Cos(x[x[2]]*x[2]) ; }
                                                                                                                  ^~~~~
Error in <prepareMethod>: Can't compile function TFormula____id9800548585748267179 prototype with arguments Double_t*
Error in <TFormula::InputFormulaIntoCling>: Error compiling formula expression in Cling
Error in <TFormula::ProcessFormula>: Formula "exp(-abs(x[x[2]])/x[1])*cos(x[x[2]]*x[2])" is invalid !
[#0] FATAL:InputArguments -- RooFormula 'exp(-abs(@0)/@1)*cos(@0*@2)_dt_tau_0' did not compile or is invalid.
Input:
	exp(-abs(@0)/@1)*cos(@0*@2)
Passed over to TFormula:
	exp(-abs(x[x[2]])/x[1])*cos(x[x[2]]*x[2])
terminate called after throwing an instance of 'std::runtime_error'
  what():  RooFormula 'exp(-abs(@0)/@1)*cos(@0*@2)_dt_tau_0' did not compile or is invalid.
Input:
	exp(-abs(@0)/@1)*cos(@0*@2)
Passed over to TFormula:
	exp(-abs(x[x[2]])/x[1])*cos(x[x[2]]*x[2])

However, when we change the definition of dm to

RooRealVar dm1("dm1", "delta m(B0)", 0);

The bug disappears.
This is very strange.
Could anyone give me some advide?

My Root version is 6.24.00.
The same bug also appear in ROOT6.22

Thanks inadvance!

Best Wishes,
Zhihong

[jonas]

Hi @Zhihong_Shen,

thanks for reporting this problem!

Indeed, it’s a bug in the RooFormula that behaves wrongly if one of the variable has just an integer as a name, like RooConst(0.0) whose name is simply “0”. That’s why it works when you give your zero constant a different name, like “dm1” which you tried later where you didn’t hit the bug. So as you noticed, the workaround is easy

I have created a PR to fix the bug:

[RF] Correctly treat variables with integer names in RooFormula

It will be fixed in the next patch release.

Thanks again for reporting and have a nice evening!
Jonas

[Zhihong_Shen]

@jonas Thanks very much !