A note on product-convolution for generalized subexponential distributions