convolve_2d fails with image types using channel type float32_t

[harshitpant1]

Actual behavior

convolve_2d fails to convolve images of type with channel type float32_t.

Expected behavior

It should work, as it does for other image types.

C++ Minimal Working Example

#include <boost/gil.hpp>
#include <vector>
namespace gil = boost::gil;
int main()
{
    gil::gray32f_image_t img_in(25, 25);
    gil::gray32f_image_t img_out(img_in.dimensions());
    std::vector<float> vec{0,0,0,0,1,0,0,0,0};
    gil::detail::kernel_2d<float> kernel(vec.begin(), vec.size(), 1, 1);
    gil::detail::convolve_2d(view(img_in), kernel, view(img_out));
    return 0; 
}

Environment

• Compiler version: MS Visual C++ version 19.28.29333
• Version (Git ref or <boost/version.hpp>): 107500

[harshitpant1]

adding [0], on line 376, seems to work. Perhaps the compiler doesn't like doing two implicit conversions.

aux_total +=
      src_view(col_boundary, row_boundary)[0] *
      kernel.at(flip_ker_row, flip_ker_col);