Physics::Elasticity::Kinematics::w: change intermediate value to 'auto'

This fixes a bug uncovered by our recent base::Tensor() cleanup: The
intermediate tensor type used for the gradient in the w() function
erroneously casts the result to a Tensor<1, dim, double> which for
example for automatic differentiation then removes the underyling AD
type:

/srv/temp/testsuite-d8F1wK0J/dealii/include/deal.II/base/tensor.h:1316:13: error: cannot convert 'const value_type' (aka 'const Sacado::Fad::Exp::GeneralFad<Sacado::Fad::Exp::DynamicStorage<double>>') to 'value_type' (aka 'double') without a conversion operator
  : values{{value_type(initializer[indices])...}}
            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Passes all kinematics/physics_functions tests in the testsuite

Fixes #16517

diff --git a/include/deal.II/physics/elasticity/kinematics.h b/include/deal.II/physics/elasticity/kinematics.h
index 666c3e09146920881e43a98f3591b5bc641ab08d..a39d9c2a5323d601aeb59595fd176714c956e919 100644 (file)
--- a/include/deal.II/physics/elasticity/kinematics.h
+++ b/include/deal.II/physics/elasticity/kinematics.h
@@ -388,7 +388,7 @@ Physics::Elasticity::Kinematics::w(const Tensor<2, dim, Number> &F,
 {
   // This could be implemented as w = l-d, but that would mean computing "l"
   // a second time.
-  const Tensor<2, dim> grad_v = l(F, dF_dt);
+  const auto grad_v = l(F, dF_dt);
   return internal::NumberType<Number>::value(0.5) *
          (grad_v - transpose(grad_v));
 }