Fix casting negative double to unsigned int in GenerateGroupingPowder -- ornl-next (#39222)

Fix casting negative double to unsigned int in GenerateGroupingPowder (#39205)

* Fix casting negative double to unsigned int

A size_t is an unsigned int, but the way this method was casting
resulted in a negative value being cast to a large positive value. It
now uses an int64_t for the intermediate step instead.

* Add release note

* Results changed in this test

* Avoid int -> double conversion

Without this cast, the second component of the sum will be converted
from an int64_t to a double, resulting in a compiler warning and
potential loss of precision.

Co-authored-by: James Clarke <139879523+jclarkeSTFC@users.noreply.github.com>

Author: rboston628

Framework/DataHandling/src/GenerateGroupingPowder.cpp

@@ -79,8 +79,9 @@ class SphericalSectorLabelor : public Labelor {
       : Labelor(t_step, a_step, a_start), inv_tt_step(1.0 / tt_step), inv_aa_step(1.0 / aa_step),
         num_aa_step(int(std::ceil(360.0 / a_step))) {};
   size_t operator()(SpectrumInfo const &spectrumInfo, size_t i) override {
-    return static_cast<size_t>(spectrumInfo.twoTheta(i) * inv_tt_step) * num_aa_step +
-           static_cast<size_t>(std::floor((spectrumInfo.azimuthal(i) - aa_start) * inv_aa_step)) % num_aa_step + 1;
+    return static_cast<size_t>(
+        static_cast<int64_t>(spectrumInfo.twoTheta(i) * inv_tt_step * num_aa_step) +
+        static_cast<int64_t>(std::floor((spectrumInfo.azimuthal(i) - aa_start) * inv_aa_step)) % num_aa_step + 1);
   };
 };
 

Framework/DataHandling/test/GenerateGroupingPowder2Test.h

@@ -465,7 +465,7 @@ class GenerateGroupingPowder2Test : public CxxTest::TestSuite {
     const bool numberByAngle{true};
     const bool splitSides{true};
     const std::vector<int> groups_exp{7, 8, 9, 10, 11, 12, 13, 14};
-    const std::vector<double> pixel_groups_exp{11, 11, 11, 9, 9, 9, 9, 9, 9, 10, 10, 10, 12, 12, 12, 12, 12, 12};
+    const std::vector<double> pixel_groups_exp{12, 12, 12, 10, 10, 10, 9, 9, 9, 8, 8, 10, 10, 10, 12, 11, 11, 13};
     run_SNAPliteTest("PowderGrouping_no_azimuth_angle_numbering", ang1, ang2, numberByAngle, splitSides, groups_exp,
                      pixel_groups_exp);
   }

Framework/DataHandling/test/GenerateGroupingPowderTest.h

@@ -470,7 +470,7 @@ class GenerateGroupingPowderTest : public CxxTest::TestSuite {
     const bool numberByAngle{true};
     const bool splitSides{true};
     const std::vector<int> groups_exp{7, 8, 9, 10, 11, 12, 13, 14};
-    const std::vector<double> pixel_groups_exp{11, 11, 11, 9, 9, 9, 9, 9, 9, 10, 10, 10, 12, 12, 12, 12, 12, 12};
+    const std::vector<double> pixel_groups_exp{12, 12, 12, 10, 10, 10, 9, 9, 9, 8, 8, 10, 10, 10, 12, 11, 11, 13};
     run_SNAPliteTest("PowderGrouping_no_azimuth_angle_numbering", ang1, ang2, numberByAngle, splitSides, groups_exp,
                      pixel_groups_exp);
   }

docs/source/release/v6.13.0/Framework/Algorithms/Bugfixes/39205.rst

@@ -0,0 +1 @@
+- Fixed bug when calculating pixel groups in :ref:`GenerateGroupingPowder <algm-GenerateGroupingPowder>` where under certain conditions the number of pixels in a group was not calculated correctly. This could lead to incorrect grouping of pixels in the output workspace. The issue was caused by an incorrect handling of the pixel indices when calculating the groups. The fix ensures that the pixel indices are handled correctly, resulting in accurate grouping of pixels in the output workspace.