[v638][RF] Ensure RooHistPdf values are positive in RooHistPdf::doEval()

roofit/roofitcore/src/RooHistPdf.cxx

@@ -192,6 +192,9 @@ void RooHistPdf::doEval(RooFit::EvalContext &ctx) const
 
    auto xVals = ctx.at(_pdfObsList[0]);
    _dataHist->weights(output.data(), xVals, _intOrder, true, _cdfBoundaries);
+   for (auto &ret : output) {
+      ret = std::max(ret, 0.0);
+   }
 }
 
 

roofit/roofitcore/test/testRooHistPdf.cxx

@@ -2,33 +2,80 @@
 // Authors: Jonas Rembser, CERN 03/2023
 
 #include <RooDataHist.h>
+#include <RooFitResult.h>
+#include <RooHelpers.h>
 #include <RooHistPdf.h>
 #include <RooLinearVar.h>
 #include <RooRealIntegral.h>
 #include <RooRealVar.h>
+#include <RooWorkspace.h>
 
 #include <gtest/gtest.h>
 
 // Verify that RooFit correctly uses analytic integration when having a
 // RooLinearVar as the observable of a RooHistPdf.
 TEST(RooHistPdf, AnalyticIntWithRooLinearVar)
 {
-    RooRealVar x{"x", "x", 0, -10, 10};
-    x.setBins(10);
+   RooRealVar x{"x", "x", 0, -10, 10};
+   x.setBins(10);
 
-    RooDataHist dataHist("dataHist", "dataHist", x);
-    for(int i = 0; i < x.numBins(); ++i) {
-       dataHist.set(i, 10.0, 0.0);
-    }
+   RooDataHist dataHist("dataHist", "dataHist", x);
+   for (int i = 0; i < x.numBins(); ++i) {
+      dataHist.set(i, 10.0, 0.0);
+   }
 
-    RooRealVar shift{"shift", "shift", 2.0, -10, 10};
-    RooRealVar slope{"slope", "slope", 1.0};
-    RooLinearVar xShifted{"x_shifted", "x_shifted", x, slope, shift};
+   RooRealVar shift{"shift", "shift", 2.0, -10, 10};
+   RooRealVar slope{"slope", "slope", 1.0};
+   RooLinearVar xShifted{"x_shifted", "x_shifted", x, slope, shift};
 
-    RooHistPdf pdf{"pdf", "pdf", xShifted, x, dataHist};
+   RooHistPdf pdf{"pdf", "pdf", xShifted, x, dataHist};
 
-    RooRealIntegral integ{"integ", "integ", pdf, x};
+   RooRealIntegral integ{"integ", "integ", pdf, x};
 
-    EXPECT_DOUBLE_EQ(integ.getVal(), 90.);
-    EXPECT_EQ(integ.anaIntVars().size(), 1);
+   EXPECT_DOUBLE_EQ(integ.getVal(), 90.);
+   EXPECT_EQ(integ.anaIntVars().size(), 1);
+}
+
+// Regression test for https://github.com/root-project/root/issues/21159, which
+// uncovered that the values were not clipped to be positive when evaluating a
+// RooHistPdf with the new vectorizing evaluation backend.
+TEST(RooHistPdf, EnsurePositiveValuesInFFTConvPdf)
+{
+   RooHelpers::LocalChangeMsgLevel changeMsgLvl(RooFit::WARNING);
+
+   using namespace RooFit;
+
+   RooWorkspace ws{"ws"};
+
+   // Observable
+   ws.factory("mass[5500,6100]");
+
+   // Signal Gaussian
+   ws.factory("mean[5800,5795,5805]");
+   ws.factory("sigma[8.1,1,30]");
+   ws.factory("Gaussian::sigpdf(mass, mean, sigma)");
+
+   // Argus background
+   ws.factory("argus_par[-55]");
+   ws.factory("argus_m0[5665]"); // 5800 - 135
+   ws.factory("ArgusBG::argus(mass, argus_m0, argus_par)");
+
+   // Resolution Gaussian (mean fixed to 0)
+   ws.factory("Gaussian::resolution(mass, 0, sigma)");
+
+   // Convolution
+   ws.factory("FFTConvPdf::conv(mass, argus, resolution)");
+
+   // Combined model (signal + background)
+   ws.factory("SUM::model(0.3*sigpdf, conv)");
+
+   // Retrieve objects
+   auto &mass = *ws.var("mass");
+   auto &model = *ws.pdf("model");
+
+   std::unique_ptr<RooDataSet> data{model.generate(mass, 1000)};
+
+   std::unique_ptr<RooFitResult> fit_result{model.fitTo(*data, Save(true), PrintLevel(-1))};
+
+   EXPECT_EQ(fit_result->status(), 0) << "The fit should succeed with status 0";
 }