Seg fault for marker 24

Hello

I have a program compiled against ROOT 5.10 and from time to time, after I produce many histograms and print them in ~100 png files, I get segmetation fault in asim_straight_ellips() if I use marker type 24.
For marker type 30 the behaviour is even more strange - a lot of messages about invalid free()
But there is no seg fault if I use ps files instead of png
Are there any limitations for png file format ?

Cheers, Sasha

Could you send the shortest possible script reproducing this problem?

Rene

I do not see any problem with last root version with the marker types 24 and 30. See the attached png file.


the program is
alxr.usatlas.bnl.gov/lxr/source/ … onTool.cxx
and it is running inside Athena, so maybe there is memory corruption somewere else … who knows
As I said, i got crash only after I created many many png files, there was no problem with few files

The histogram which causes problem is filled like this:

TF1 * fit_gaus = new TF1("g","gaus");
for (ch=0; ch!=48; ++ch) {
   double Sigma=0.0, ErrS=0.0;
    if (hist1[ch]->GetEntries()>0) {
      hist1[ch]->Fit("g","N");
      Sigma = fit_gaus->GetParameter(2);
      ErrS = fit_gaus->GetParError(2);
   }
   final_hist1->SetBinContent(ch+1, Sigma);
   final_hist1->SetBinError(ch+1, ErrS);
}

Sigma is close to zero (0.2…0.5). If I add 1.0 to it, there is no crash
If I keep Sigma unchanged, but don’t print first 5 plots (i.e. don’t do canvas->Print() for them) crash happens 5 plots later

histograms are in
/castor/cern.ch/user/t/tilebeam/commissioning/tilemon_61111_Ped.0.root

*sigma in Tile/RawChannel/Summary folder are final histograms
and amp in Tile/RawChannel/LB are used in fit