Hi Robert,
A few weeks ago, I fixed a problem in TF1 (version 3.02) when
storing a copy of the fitted function.
Your test works with this version.
I do not see any way to circumvent the problem in 3.01.
Rene Brun
On Thu, 27 Sep 2001, Robert Feuerbach wrote:
>
> Hi,
>
> I'm using ROOT version 3.01/06 on a LinuxRH7.1 machine, and am having
> problems using TH1::Fit when using a user-defined function. We were able
> to create a short program that has the same problems:
>
> /*---------------------------------------------------------*/
> #include <TF1.h>
> #include <TMath.h>
> #include <TH1.h>
> // #include <TROOT.h> /* needed for ROOT 2.25 */
>
> Double_t fitf(Double_t *x, Double_t *par) {
> Double_t arg=0;
> Double_t result=0.;
> arg = (x[0]-par[1])/par[2];
> result = par[0]*TMath::Exp(-0.5*arg*arg)+par[3];
> return result;
> }
>
> void main(int argc, char *argv[]) {
>
> // TROOT root("myroot","stuff"); /* needed for ROOT 2.25 */
> TF1 *myfunc = new TF1("myfunc",fitf,-2.,2.,4);
> //TF1 *myfunc = new TF1("myfunc","gaus",-2.,2.); /* doesn't hang up */
> myfunc->SetParameters(1000.,0.,.4,10.);
>
> TH1F *hist = new TH1F("hist1","test",100,-4.,4.);
> hist->FillRandom("myfunc",1000);
>
> myfunc->SetParameters(200.,-1.,.5,100.);
> hist->Fit("myfunc","R0");
> // hist->Fit("myfunc","R0N"); /* doesn't seg-fault/hang up */
>
> }
> /*---------------------------------------------------------*/
>
> After the actual fit is done (and the results printed to the screen), the
> program just hangs, and never exits from the hist->Fit line. If we tell
> ROOT NOT to store the function (with the 'N' option), it exits just
> fine. This works if we use a pre-defined function (such as "gaus") or
> run it under ROOT 2.25.
>
> Would you please look into this?
>
> Thank you,
> Rob Feuerbach
>
> --
> /***************************************************
> * Robert Feuerbach feuerbac@ernest.phys.cmu.edu *
> * CMU Physics Department (412) 268-2772 *
> * Pittsburgh, PA 15213 FAX: (412) 681-0648 *
> ***************************************************/
>
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:00 MET