Hi, Rene:
Thank you a lot for your help. But I still can not understand why one addition call
of the fit function before the for loop can "solve" this problem. It seems for me
that the problem is not only because the second call of the TF1 constructor.
Huaizhang
Rene Brun wrote:
> hi Masa,
> This is not a CINT bug. It is simply a bookeeping problem between
> the ROOT classes TF1, TMethodCall and the CINt class G__CallFunc.
> The TF1 constructor when called with a CINT function creates a TMethodCall
> object.
> In turn the TMethodCall constructor (via the Init function) creates a
> G__CallFunc
> object. So far, so good. However when a TF1 object is deleted, the TMethodCall
> destructor deletes the G__CallFunc object. At the next call to TF1 with the
> same interpreted function, CINT returns correctly that the function is unknown.
> I could remove the call to the G__CallFunc destructor in TMethodCall destructor.
> However, this will generate some memory leakage.
> The destructor of G__CallFunc being empty, in principle no damage should happen
> when invoking this destructor several times. However, the problem is somewhere
> with a non reentrant Init or SetFunc function.
>
> Rene Brun
>
> Masaharu Goto wrote:
> >
> > Hello Huaizhang,
> >
> > Thank you for reporting a bug.
> >
> > I'd like to look into this. For doing so, will you give me complete
> > information about your example. In this case, I do not know how 'hist'
> > is created. I'd appreicaite if you send me small and complete example.
> >
> > Masaharu Goto
> >
> > >
> > >I wrote the following program.
> > >
> > >Double_t RatioFunc(Double_t *t, Double_t *par) {
> > > Double_t R;
> > > R=par[0]*cos(2*3.14159265*0.0002291*(1-par[1]*1.0e-6)*t[0]+par[2])+par[3];
> > > return R;
> > >}
> > >
> > >TF1* RatioFit(TH1F* h, Float_t t_start, Float_t t_end) {
> > >
> > > TF1 *RatioFit=new TF1("RatioFit", RatioFunc, 0, 700000, 4);
> > >
> > > RatioFit->SetParNames("A", "R", "'f#", "offset");
> > > RatioFit->SetRange(t_start, t_end);
> > > RatioFit->SetParameters(0.4, 120, 0, 0);
> > >
> > > h->Fit("RatioFit", "R");
> > > return RatioFit;
> > >}
> > >
> > >Then I tried to do
> > >
> > >int i;
> > >TF1* fit;
> > >for(i=0; i<5; i++) fit=RatioFit(hist, t, 400000.);
> > >
> > >After the first fit, which is o.k., it complain "Function:RatioFit cannot be
> > >compiled".
> > >However, the following program works well.
> > >
> > >int i;
> > >TF* fit;
> > >fit=RatioFit(h, t, 40000);
> > >for(i=1; i<5; i++) fit=RatioFit(hist, t, 40000);
> > >
> > >I wonder why those two programs make difference.
> > >
> > >Huaizhang Deng
--
Huaizhang Deng
516-344-5731
P. O. Box 426
Upton, NY 11973
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:27 MET