Hi,
You can work around the problem by disabling the Cint optimization:
.O 0
.x henry.C
Cheers,
Philippe.
> Axel, Philippe,
>
> Could you follow this problem that seems to appear only on 32 bits machines?
>
> Rene
>
> ----- Original Message -----
> From Stephan Zimmer <zimmerst_at_googlemail.com>
> Date Sat, 27 Feb 2010 18:43:45 +0100
> To Rene Brun <Rene.Brun_at_cern.ch>
> Cc <roottalk_at_root.cern.ch>
> Subject Re: [ROOT] Single call of procedure works while recurring
> call in loop returns error.
> Hi Rene,
> that's very unfortunate... so far I've tested the macro on different
> platforms with different root-versions; all with the same result:
> >>>>>>>>>>>>>>>>>>>>
> root [0] .X henry.C
> initial chi2/Ndf: 528.202
> Chi2/Ndf 528.202
> Chi2/Ndf 528.202
> Error: Can't call vector<double,allocator<double>
> >::vector<double,allocator<dou
> ble> >((class vector<double>***&)37128248) in current scope
> C:\Users\zimmer\Desk
> top\ex6\henry.C(15)
> Possible candidates are...
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(void);
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(vector<double,allocator<double>
> >::size_type n,co
> nst double& value=double());
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(const vector<double,allocator<double>
> >& x);
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(vector<double,allocator<double>
> >::const_iterator
> first,vector<double,allocator<double> >::const_iterator last);
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(void);
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(vector<double,allocator<double>
> >::size_type n,co
> nst double& value=double());
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(const vector<double,allocator<double>
> >& x);
> public: vector<double,allocator<double> >
> vector<double,allocator<double> >::vec
> tor<double,allocator<double> >(vector<double,allocator<double>
> >::const_iterator
> first,vector<double,allocator<double> >::const_iterator last);
> *** Interpreter error recovered ***
> >>>>>>>>>>>>>>>>>>>>>>>>>
> tested architectures and root versions:
> Linux, Ubuntu 9.10: 5.18, 5.26
> Linux, SLC 4: 5.22
> Linux SLC4 on lxplus: 5.21.04
> Windows 7, VC9+: 5.26
>
> Any guess what could cause this behavior?
> Any help is greatly appreciated.
>
> Cheers, Stephan
>
> On 2/27/2010 5:21 AM, Rene Brun wrote:
> > Hi Stephan,
> >
> > I cannot reproduce the problem. Executing your script, I get:
> > root [0] .L henry.C
> > root [1] henry()
> > initial chi2/Ndf: 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > Chi2/Ndf 528.202
> > root [2] .q
> >
> > I tried too with ACLIC and get the same result. To use it with
> ACLIC, I
> > had to make 2 small changes in the includes, see attachment.
> >
> > Rene
> >
> > Stephan Zimmer wrote:
> >> Oh sorry for that.
> >> The script to call is henry.C and the dat-file henry.dat (see
> >> attachment) I just wonder whether I did something stupid in my code.
> >> Thank you very much.
> >> The same behavior happens under 5.18/5.22/5.26
> >> Cheers, Stephan
> >>
> >> On 2/27/2010 5:01 AM, Rene Brun wrote:
> >>
> >>> Could you provide a complete script and the data file to execute
> it?
> >>>
> >>> Rene
> >>>
> >>> Stephan Zimmer wrote:
> >>>
> >>>> Hi ROOTers,
> >>>> I wonder if someone of you might understand why the following code
> >>>> crashes...
> >>>>
> >>>> double CotTheta(const double x_i, const double alpha[2]){
> >>>> // returns the estimate y(x_i) based on the theoretical distribution
> >>>> return alpha[0]*x_i - alpha[1]*pow(x_i,-1);
> >>>> }
> >>>>
> >>>> double CalculateChi2(const vector<double> x, const
> vector<double> y,
> >>>> const vect\
> >>>> or<double> yError, const double alpha[2]){
> >>>> // returns the chi2 value for a given set of parameters
> >>>> double sumchi2=0;
> >>>> for (unsigned int i = 0; i<x.size(); i++){
> >>>>
> >>>> sumchi2=sumchi2+(pow(yError[i],-2)*pow(y[i]-CotTheta(x[i],alpha),2));
> >>>> }
> >>>> return sumchi2;
> >>>> }
> >>>>
> >>>> void littleHenry(){
> >>>>
> >>>> TVirtualFitter::SetDefaultFitter("Minuit");
> >>>>
> >>>> // readout henry.dat
> >>>> vector<double> xVec;
> >>>> vector<double> xEVec;
> >>>> vector<double> yVec;
> >>>> vector<double> yEVec;
> >>>> double x(0),y(0),xE(0), yE(0);
> >>>> double output[3];
> >>>> FILE *data;
> >>>> char datafname[]="henry.dat";
> >>>>
> >>>> data = fopen(datafname,"r");
> >>>> while (fscanf(data, "%lf %lf %lf %lf",&x,&xE,&y,&yE) != EOF) {
> >>>> xVec.push_back(x);
> >>>> yVec.push_back(y);
> >>>> xEVec.push_back(xE);
> >>>> yEVec.push_back(yE);
> >>>> }
> >>>> double alpha[2];
> >>>> double chi2;
> >>>> alpha[0]=1e-4;
> >>>> alpha[1]=1e2;
> >>>> CalculateChi2(xVec,yVec,yEVec,alpha);
> >>>> --> that works...
> >>>>
> >>>>
> >>>> However, once I want to execute the routine CalculateChi2 more than
> >>>> twice, by putting it inside a loop I get a weird error message:
> >>>> for (int i = 0; i<5; i++){
> >>>> chi2=CalculateChi2(xVec,yVec,yEVec,alpha);
> >>>> cout << "Chi2/Ndf " << chi2/xVec.size() << endl;
> >>>> }
> >>>> ----> OUTPUT <-------
> >>>> Processing littleHenry.C...
> >>>> Chi2/Ndf 528.202
> >>>> Chi2/Ndf 528.202
> >>>> Error: Can't call vector<double,allocator<double>
> >>>>
> >>>>
> >>>>> ::vector<double,allocator<double> >((class
> >>>>>
> >>>> vector<double>***&)148530456) in current scope littleHenry.C:34:
> >>>> Possible candidates are...
> >>>> (in vector<double,allocator<double> >)
> >>>> vector.dll -1:-1 0 public:
> vector<double,allocator<double> >
> >>>> vector<double,allocator<double> >::vector<double,allocator<double>
> >>>>
> >>>>> (void);
> >>>>>
> >>>> vector.dll -1:-1 0 public:
> vector<double,allocator<double> >
> >>>> vector<double,allocator<double> >::vector<double,allocator<double>
> >>>>
> >>>>
> >>>>> (vector<double,allocator<double> >::size_type n,const double&
> >>>>>
> >>>> value=double());
> >>>> vector.dll -1:-1 0 public:
> vector<double,allocator<double> >
> >>>> vector<double,allocator<double> >::vector<double,allocator<double>
> >>>>
> >>>>
> >>>>> (const vector<double,allocator<double> >& x);
> >>>>>
> >>>> vector.dll -1:-1 0 public:
> vector<double,allocator<double> >
> >>>> vector<double,allocator<double> >::vector<double,allocator<double>
> >>>>
> >>>>
> >>>>> (vector<double,allocator<double> >::const_iterator
> >>>>>
> >>>> first,vector<double,allocator<double> >::const_iterator last);
> >>>> *** Interpreter error recovered ***
> >>>>
> >>>> I wonder if I somehow run out of the boundaries of the vector
> but that
> >>>> should not happen as they are not touched and treated as constants...
> >>>> Any ideas?
> >>>>
> >>>> Cheers,
> >>>> Stephan
> >>>>
> >>>>
> >
>
>
Received on Sun Feb 28 2010 - 15:43:45 CET
This archive was generated by hypermail 2.2.0 : Thu Mar 04 2010 - 05:50:03 CET