[test] Fix error in TAT.hpp test for old compiler.

For gcc 8, arithmetic operator over valarray cannot duduce these
scalar type automatically. So, this commit converts the scalartype to
the correct type manually.

tests/test_scalar.cpp

@@ -15,28 +15,28 @@ TEST(test_scalar, tensor_and_number) {
     }
     ASSERT_THAT(a.storage(), ElementsAreArray(std::begin(s_a), std::end(s_a)));
     auto b = a + 1.0;
-    std::valarray<std::complex<double>> s_b = s_a + 1.0;
+    std::valarray<std::complex<double>> s_b = s_a + std::complex<double>(1.0);
     ASSERT_THAT(b.storage(), ElementsAreArray(std::begin(s_b), std::end(s_b)));
     auto c = 1.0 + a;
-    std::valarray<std::complex<double>> s_c = 1.0 + s_a;
+    std::valarray<std::complex<double>> s_c = std::complex<double>(1.0) + s_a;
     ASSERT_THAT(c.storage(), ElementsAreArray(std::begin(s_c), std::end(s_c)));
     auto d = a - 1.0;
-    std::valarray<std::complex<double>> s_d = s_a - 1.0;
+    std::valarray<std::complex<double>> s_d = s_a - std::complex<double>(1.0);
     ASSERT_THAT(d.storage(), ElementsAreArray(std::begin(s_d), std::end(s_d)));
     auto e = 1.0 - a;
-    std::valarray<std::complex<double>> s_e = 1.0 - s_a;
+    std::valarray<std::complex<double>> s_e = std::complex<double>(1.0) - s_a;
     ASSERT_THAT(e.storage(), ElementsAreArray(std::begin(s_e), std::end(s_e)));
     auto f = a * 1.5;
-    std::valarray<std::complex<double>> s_f = s_a * 1.5;
+    std::valarray<std::complex<double>> s_f = s_a * std::complex<double>(1.5);
     ASSERT_THAT(f.storage(), ElementsAreArray(std::begin(s_f), std::end(s_f)));
     auto g = 1.5 * a;
-    std::valarray<std::complex<double>> s_g = 1.5 * s_a;
+    std::valarray<std::complex<double>> s_g = std::complex<double>(1.5) * s_a;
     ASSERT_THAT(g.storage(), ElementsAreArray(std::begin(s_g), std::end(s_g)));
     auto h = a / 1.5;
-    std::valarray<std::complex<double>> s_h = s_a / 1.5;
+    std::valarray<std::complex<double>> s_h = s_a / std::complex<double>(1.5);
     ASSERT_THAT(h.storage(), ElementsAreArray(std::begin(s_h), std::end(s_h)));
     auto i = 1.5 / (a + 1);
-    std::valarray<std::complex<double>> s_i = 1.5 / (s_a + 1);
+    std::valarray<std::complex<double>> s_i = std::complex<double>(1.5) / (s_a + std::complex<double>(1));
     ASSERT_THAT(i.storage(), ElementsAreArray(std::begin(s_i), std::end(s_i)));
 }
 
@@ -62,7 +62,7 @@ TEST(test_scalar, tensor_and_tensor) {
     std::valarray<double> s_e = s_a * s_b;
     ASSERT_THAT(e.storage(), ElementsAreArray(std::begin(s_e), std::end(s_e)));
     auto f = a / (b + 1);
-    std::valarray<double> s_f = s_a / (s_b + 1);
+    std::valarray<double> s_f = s_a / (s_b + double(1));
     ASSERT_THAT(f.storage(), ElementsAreArray(std::begin(s_f), std::end(s_f)));
 }
 
@@ -111,6 +111,6 @@ TEST(test_scalar, tensor_and_tensor_inplace) {
     s_a -= s_b;
     ASSERT_THAT(a.storage(), ElementsAreArray(std::begin(s_a), std::end(s_a)));
     a /= b + 1;
-    s_a /= s_b + 1;
+    s_a /= s_b + double(1);
     ASSERT_THAT(a.storage(), ElementsAreArray(std::begin(s_a), std::end(s_a)));
 }