Marshall Clow found some divide-by-zero warnings with UBSan in rand's binomial_distribution test. This eliminates the divide-by-zeros and describes in comments the numerical difficulties the test is having. Each of the problematic tests are exploring edge cases of the distribution.

llvm-svn: 177826

1 files changed, 65 insertions, 25 deletions

diff --git a/libcxx/test/numerics/rand/rand.dis/rand.dist.bern/rand.dist.bern.bin/eval.pass.cpp b/libcxx/test/numerics/rand/rand.dis/rand.dist.bern/rand.dist.bern.bin/eval.pass.cpp
index 36b9efd1732..3bfd2af6106 100644
--- a/libcxx/test/numerics/rand/rand.dis/rand.dist.bern/rand.dist.bern.bin/eval.pass.cpp
+++ b/libcxx/test/numerics/rand/rand.dis/rand.dist.bern/rand.dist.bern.bin/eval.pass.cpp
@@ -177,15 +177,23 @@ int main()
         }
         var /= u.size();
         double dev = std::sqrt(var);
-        skew /= u.size() * dev * var;
-        kurtosis /= u.size() * var * var;
-        kurtosis -= 3;
+        // In this case:
+        //   skew     computes to 0./0. == nan
+        //   kurtosis computes to 0./0. == nan
+        //   x_skew     == inf
+        //   x_kurtosis == inf
+        //   These tests are commented out because UBSan warns about division by 0
+//        skew /= u.size() * dev * var;
+//        kurtosis /= u.size() * var * var;
+//        kurtosis -= 3;
         double x_mean = d.t() * d.p();
         double x_var = x_mean*(1-d.p());
-        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
-        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
+//        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
+//        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
         assert(mean == x_mean);
         assert(var == x_var);
+//        assert(skew == x_skew);
+//        assert(kurtosis == x_kurtosis);
     }
     {
         typedef std::binomial_distribution<> D;
@@ -215,15 +223,23 @@ int main()
         }
         var /= u.size();
         double dev = std::sqrt(var);
-        skew /= u.size() * dev * var;
-        kurtosis /= u.size() * var * var;
-        kurtosis -= 3;
+        // In this case:
+        //   skew     computes to 0./0. == nan
+        //   kurtosis computes to 0./0. == nan
+        //   x_skew     == -inf
+        //   x_kurtosis == inf
+        //   These tests are commented out because UBSan warns about division by 0
+//        skew /= u.size() * dev * var;
+//        kurtosis /= u.size() * var * var;
+//        kurtosis -= 3;
         double x_mean = d.t() * d.p();
         double x_var = x_mean*(1-d.p());
-        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
-        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
+//        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
+//        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
         assert(mean == x_mean);
         assert(var == x_var);
+//        assert(skew == x_skew);
+//        assert(kurtosis == x_kurtosis);
     }
     {
         typedef std::binomial_distribution<> D;
@@ -333,15 +349,23 @@ int main()
         }
         var /= u.size();
         double dev = std::sqrt(var);
-        skew /= u.size() * dev * var;
-        kurtosis /= u.size() * var * var;
-        kurtosis -= 3;
+        // In this case:
+        //   skew     computes to 0./0. == nan
+        //   kurtosis computes to 0./0. == nan
+        //   x_skew     == inf
+        //   x_kurtosis == inf
+        //   These tests are commented out because UBSan warns about division by 0
+//        skew /= u.size() * dev * var;
+//        kurtosis /= u.size() * var * var;
+//        kurtosis -= 3;
         double x_mean = d.t() * d.p();
         double x_var = x_mean*(1-d.p());
-        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
-        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
+//        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
+//        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
         assert(mean == x_mean);
         assert(var == x_var);
+//        assert(skew == x_skew);
+//        assert(kurtosis == x_kurtosis);
     }
     {
         typedef std::binomial_distribution<> D;
@@ -371,15 +395,23 @@ int main()
         }
         var /= u.size();
         double dev = std::sqrt(var);
-        skew /= u.size() * dev * var;
-        kurtosis /= u.size() * var * var;
-        kurtosis -= 3;
+        // In this case:
+        //   skew     computes to 0./0. == nan
+        //   kurtosis computes to 0./0. == nan
+        //   x_skew     == inf
+        //   x_kurtosis == inf
+        //   These tests are commented out because UBSan warns about division by 0
+//        skew /= u.size() * dev * var;
+//        kurtosis /= u.size() * var * var;
+//        kurtosis -= 3;
         double x_mean = d.t() * d.p();
         double x_var = x_mean*(1-d.p());
-        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
-        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
+//        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
+//        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
         assert(mean == x_mean);
         assert(var == x_var);
+//        assert(skew == x_skew);
+//        assert(kurtosis == x_kurtosis);
     }
     {
         typedef std::binomial_distribution<> D;
@@ -409,14 +441,22 @@ int main()
         }
         var /= u.size();
         double dev = std::sqrt(var);
-        skew /= u.size() * dev * var;
-        kurtosis /= u.size() * var * var;
-        kurtosis -= 3;
+        // In this case:
+        //   skew     computes to 0./0. == nan
+        //   kurtosis computes to 0./0. == nan
+        //   x_skew     == -inf
+        //   x_kurtosis == inf
+        //   These tests are commented out because UBSan warns about division by 0
+//        skew /= u.size() * dev * var;
+//        kurtosis /= u.size() * var * var;
+//        kurtosis -= 3;
         double x_mean = d.t() * d.p();
         double x_var = x_mean*(1-d.p());
-        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
-        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
+//        double x_skew = (1-2*d.p()) / std::sqrt(x_var);
+//        double x_kurtosis = (1-6*d.p()*(1-d.p())) / x_var;
         assert(mean == x_mean);
         assert(var == x_var);
+//        assert(skew == x_skew);
+//        assert(kurtosis == x_kurtosis);
     }
 }