cast to pointer to array - c vs. c++ (or gcc vs. g++) -

the following code compiles (and works expected) when compiled c code (gcc file suffix '.c') throws error when compiled c++ code (g++ of gcc file suffix '.cpp'):

int sum_square_matrix(int *matrix, int size) {     int (*m)[size] = (int(*)[])matrix;     ... } 

the following error thrown when compiled g++ (or gcc suffix .cpp):

c_vs_cpp.c:4:32: error: cannot convert ‘int (*)[]’ ‘int (*)[(((sizetype)(((ssizetype)size) + -1)) + 1)]’ in initialization      int (*m)[size] = (int(*)[])matrix; 

the motivation behind such code processing square matrix double dimension array when dimension not known in advance.

so questions are:

1. is there in c++ standard different c standard makes casting illegal?
2. any other suggested method acheiving same goal (other manually accessing elements in matrix using ugly arithmetics , service functions). preferably common way supported in both c/c++.

1. is there in c++ standard different c standard makes casting illegal?

yes. c++ standard doesn't allow variable length arrays †. c standard (since c99).

2. any other suggested method acheiving same goal (other manually accessing elements in matrix using ugly arithmetics , service functions).

it's pretty trivial wrap ugly arithmetics inside helper type. simple example:

template<class t> struct matrix_view {     t* matrix;     size_t size;      t* operator[](size_t i) {         return matrix + * size;     } };  template<class t> auto make_matrix_view(t* matrix, size_t size) {     return matrix_view<t>{matrix, size}; }  int sum_square_matrix(int *matrix, int size) {     auto m = make_matrix_view(matrix, size);     // ... } 

with more boilerplate can make matrix_view usable in standard algorithms , range based loops (only outer loop though; inner loop isn't quite trivial).

preferably common way supported in both c/c++.

perhaps design macro uses shown matrix_view in c++ , whatever cast method supported in c.

---

† gcc support vla language extension, why error message doesn't reason error. whatever reason, c++ implementation doesn't agree conversion. apparently if state size in cast:

int (*m)[size] = (int(*)[size])matrix; 

i don't think has c++ standard though, difference between standard c , gcc implementation of vla extension c++.