Fix various compiler warnings

These solve warnings when compiled with gcc 14 using a set of warnings enabled that we use in one of our projects:
  -Wall
  -Wnon-virtual-dtor
  -Wzero-as-null-pointer-constant
  -Wduplicated-branches
  -Wundef
  -Wvla
  -Wpointer-arith
  -Wextra
  -Wno-unused-parameter

Author: aroffringa

Diff for: include/xtensor/xassign.hpp

@@ -456,16 +456,21 @@ namespace xt
         constexpr bool simd_strided_assign = traits::simd_strided_assign();
         if (linear_assign)
         {
-            if (simd_linear_assign || traits::simd_linear_assign(de1, de2))
-            {
-                // Do not use linear_assigner<true> here since it will make the compiler
-                // instantiate this branch even if the runtime condition is false, resulting
-                // in compilation error for expressions that do not provide a SIMD interface.
-                // simd_assign is true if simd_linear_assign() or simd_linear_assign(de1, de2)
-                // is true.
-                linear_assigner<simd_assign>::run(de1, de2);
-            }
-            else
+            // Do not use linear_assigner<true> here since it will make the compiler
+            // instantiate this branch even if the runtime condition is false, resulting
+            // in compilation error for expressions that do not provide a SIMD interface.
+            // simd_assign is true if simd_linear_assign() or simd_linear_assign(de1, de2)
+            // is true.
+            if constexpr(simd_assign) {
+                if(simd_linear_assign || traits::simd_linear_assign(de1, de2))
+                {
+                    linear_assigner<true>::run(de1, de2);
+                }
+                else
+                {
+                    linear_assigner<false>::run(de1, de2);
+                }
+            } else
             {
                 linear_assigner<false>::run(de1, de2);
             }

Diff for: include/xtensor/xfixed.hpp

@@ -153,7 +153,7 @@ namespace xt
                 (L == layout_type::row_major) || (L == layout_type::column_major),
                 "Layout not supported for fixed array"
             );
-#if (_MSC_VER >= 1910)
+#if (defined(_MSC_VER_) && _MSC_VER >= 1910)
             using temp_type = std::index_sequence<X...>;
             return R({workaround::get_computed_strides<L, I, temp_type>(shape[I] == 1)...});
 #else

Diff for: include/xtensor/xmath.hpp

@@ -1091,7 +1091,7 @@ namespace xt
 #define XTENSOR_GCC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
 
 // Workaround for MSVC 2015 & GCC 4.9
-#if (defined(_MSC_VER) && _MSC_VER < 1910) || (defined(__GNUC__) && GCC_VERSION < 49999)
+#if (defined(_MSC_VER) && _MSC_VER < 1910) || (defined(__GNUC__) && XTENSOR_GCC_VERSION < 49999)
 #define XTENSOR_DISABLE_LAMBDA_FCT
 #endif
 

Diff for: include/xtensor/xsort.hpp

@@ -1242,7 +1242,6 @@ namespace xt
     template <layout_type L = XTENSOR_DEFAULT_TRAVERSAL, class E>
     inline auto argmin(const xexpression<E>& e)
     {
-        using value_type = typename E::value_type;
         auto&& ed = eval(e.derived_cast());
         auto begin = ed.template begin<L>();
         auto end = ed.template end<L>();
@@ -1272,7 +1271,6 @@ namespace xt
     template <layout_type L = XTENSOR_DEFAULT_TRAVERSAL, class E>
     inline auto argmax(const xexpression<E>& e)
     {
-        using value_type = typename E::value_type;
         auto&& ed = eval(e.derived_cast());
         auto begin = ed.template begin<L>();
         auto end = ed.template end<L>();

Diff for: include/xtensor/xutils.hpp

@@ -30,7 +30,7 @@
 
 #include "xtensor_config.hpp"
 
-#if (_MSC_VER >= 1910)
+#if (defined(_MSC_VER) && _MSC_VER >= 1910)
 #define NOEXCEPT(T)
 #else
 #define NOEXCEPT(T) noexcept(T)