Invalid code generated in the gradient when using operators

[gojakuch]

reproducer:

#include "clad/Differentiator/Differentiator.h"
#include <iostream>

namespace TN {
    template <typename T>
    struct Test2 {
        T operator()(T x) {
            return 2*x;
        }
        T operator[](T x) {
            return 4*x;
        }
    };
}

namespace clad {
namespace custom_derivatives {
namespace class_functions {
    template <typename T>
    void operator_subscript_pullback(::TN::Test2<T>* obj, T x, T d_u, ::TN::Test2<T>* d_obj, T* d_x) {
        (*d_x) += 4*d_u;
    }
}}}

double fn2(double x, double y) {
    TN::Test2<double> t;
    auto q = t[x];
    return q;
}

int main() {
  double dx, dy;
  auto df2 = clad::gradient(fn2);
  std::cout << "f2(3, 4)=" << fn2(3.0, 4.0) << '\n';
  dx = 0; dy = 0;
  df2.execute(3.0, 4.0, &dx, &dy);
  std::cout << "df/dx(3, 4)=" << dx << '\n';
  std::cout << "df/dy(3, 4)=" << dy << '\n';
}

generates code with the following line:

double q = operator[](x);

which is invalid and should be replaced with double q = t[x];

[vgvassilev]

double q = operator[](x); is missing double q = t.operator[](x);. However, I agree it should appear the way you suggested.