RE: 'const' with double indirection

[Andrew Haley <aph at gcc dot gnu dot org>]

Jacob van der Woude writes:
 > Thanks for the answer, Andrew.
 > 
 > But if we simple use your quoted part of the standard on the following
 > snippet:
 > 
 > void g1(const int * const List) {
 > }   /* g1 */
 > 
 > void f1(int * const List) {
 >     g1(List);
 > }   /* f1 */
 > 
 > void g2(const int * List) {
 > }   /* g2 */
 > 
 > void f2(int * List) {
 >     g2(List);
 > }   /* f2 */
 > 
 > there also should be a warning with each call of g.

Not necessarily.  We deliberately don't warn when passing a pointer to
a const-qualified version of a compatible type, presumably because it
would generate a ton of false positives.  However, in your first
example this wasn't a const-qualified version of a compatible type,
but a pointer to an incompatible type.

 > But there is no warning.
 > And there is no need for the warning (in both cases).
 > You can't fiddle with the arguments to direct or indirect modify the
 > original constants.

That's right.

 > With double (or more) level of indirection there is, except if everything is
 > 'const'.
 > So what is de difference with single or double indirection in respect to the
 > (not) generated warnings?
 > 
 > If you want me to discuss this on a C mailing list fiest, can you give me
 > the name of one?
 > 
 > With what (tool?) did you generate they explanations of declarions?

cdecl.

Andrew.