Skip to content

Use Minuit2 in histogram fit doing in LArCellDeadOTxCorr

Walter Lampl requested to merge wlampl/athena:DeadOTXMinuit2 into master

ROOT's TH1F::Fit(...) method by default uses minuit that is known be not thread-safe. This is most likely the reason for the crash observed reported as ATLASRECTS-5968.

The only way I found to convince ROOT to use minuit2 w/o relying on another global static variable was to copy a few lines from TH1.cxx to here. This way the ROOT::Math::MinimizerOption is kept on the local stack. Not nice but works.

The fit results are not exactly identical, there are generally difference two digits after the decimal point. For a correction like this one, this is irrelevant.

Merge request reports