[ROOT] A loop with 2 different fits = only one pass through loop

From: Dr. John Krane (jkrane@netzero.com)
Date: Thu Aug 19 2004 - 21:36:55 MEST


I'm running a stand-alone program with ROOT libraries (v4.00/03) 
compilied into it so I can use Minuit.  Because I am doing complicated 
fits, I can't just fit a histogram, I instead do:

  TVirtualFitter *minuit = TVirtualFitter::Fitter(0, 150);
  minuit->SetFCN(dummy_func);   // wrapper for my full-blown fit routine
...etc. and all is well.  I have the command


inside a loop that adds a new data point each time through.  Again, all 
this is working just fine, but today I decided I would like to look at 
the slope of the residuals of this fit.  So I figured I would use TH1D 
for the residuals and the method Fit("pol1").  (Actually, I made my own 
TF1 so I can get the slope parameter.)  So into the loop I put:

      TF1 p1("p1","[0]+x*[1]",0.0,100.0);

...and this works as well...once.  When I get back up to the 
minuit->ExecuteCommand("MINI") part, I get a terrible crash:

 *** Break *** segmentation violation
 Generating stack trace...
 0x40fbe949 in TMinuit::Eval(int, double*, double&, double*, int) + 0x2b 
from /bin/root/lib/libMinuit.so
 0x40fc2dfc in TMinuit::mnderi() + 0x47c from /bin/root/lib/libMinuit.so
 0x40fcb8f7 in TMinuit::mnmigr() + 0x20b from /bin/root/lib/libMinuit.so
 0x40fc45d6 in TMinuit::mnexcm(char const*, double*, int, int&) + 0x686 
from /bin/root/lib/libMinuit.so
 0x40fba057 in TFitter::ExecuteCommand(char const*, double*, int) + 0x31 
from /bin/root/lib/libMinuit.so
 0x080598fc in main + 0x22fa from simulate_signals
 0x420156a4 in __libc_start_main + 0xe4 from /lib/tls/libc.so.6
 0x0804ead9 in TFile::TFile[in-charge](char const*, char const*, char 
const*, int) + 0x31 from simulate_signals

Now, it seems from reading through the roottalk archives that Greg 
Landsberg had this problem back in March 2003, running in ROOT with a 
macro.  His solution was to save the pointer gMinuit before his second 
fit and then restore the pointer afterward.  Because I'm running 
stand-alone, I don't have a gMinuit, I only have my pointer to the 
TVirtualFitter.  The roottalk archives indicate that Rene has fixed 
things, but I'm still stuck.  I'm using version 4.00/03, which dates 
from 10 May 2004, so I'm not sure if this is the same problem Greg had 
after all.

Could somebody please advise me on how to get out of this little hole 
I'm stuck in?

    - John

This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:09 MET