Thread: Constant nothingness

Qualifying void as const doesn't seem to be valid at all, however GCC will accept "const void" as a legal type (though some compilers consider it an error). Has anyone ever encountered this in code, or can think of an interesting use for it?
Personally, I can't see any beyond maybe writing amusing/obsfucated code...

I dont really see any use of it. But what I'm suprised is that there are few compiler which does happly accept that. Which is kind a wired.

~ssharish

memcmp is prototyped this way.

Code:

 
  int memcmp(const void *s1, const void *s2, size_t n);

Void | C FAQs

There are 108 references to "const void" in the GNU C Manual, and most of them are in prototypes, eg. alphasort, fwrite, scandir, etc.

It's not pointless at all; many commands require a const, but not all of them impose a type; since void can be anything (as opposed to "nothing"), the parameter would be "const void".

Hmm this is interesting. Here is something which i came up with

Code:

#include <stdio.h>

const void * foo( void );

int main()
{
    printf("%s", (char *) foo() );
    
    getchar();
    return 0;
}



const void * foo( void )
{
      char *p = NULL;
      
      if( ( p = malloc( 10 ) ) != NULL )
      {
          strcpy(p, "hello");
          return p;
      }
}

~ssharish

@slinger, mk, ssharish:
To clarify, I'm speaking of "const void", not "const void *". The latter is extremely useful, pointing to a generic constant type, however the former doesn't make much sense since it is specifying a return type of nothingness which doesn't make sense to be considered "constant".

Consider the following two prototypes:

Code:

const void foo( int arg );
void bar( int arg );

They differ only in semantics. If a function returns "nothing", can it logically claim to return it as constant?

Yet GCC respects it as a legitimate type, consider:

Code:

const void foo( int arg );
void bar( int arg );
...
void (*fred)(int) = foo;

will generate a type error.

This is the paradox I'm getting at.

Originally Posted by @nthony

const void foo( int arg );

A prototype with a 'return' value of "const void" (as opposed to const void*) will produce a GCC warning ("function definition has qualified void return type"). If you actually try to use a return value from a void function, const or not you get an error.

That's not the purpose of the type. I think you are chasing a straw dog, and slowly at that It just ain't enough of anything to be a "paradox".

I get this on 4.1.2 on Debian, but 3.3.5 on Debian and 3.4.5 and 4.4.0 on MinGW are silent.

I'm not going after a solution or anything; just offering one of those "hmm, that's weird" moments.

>> I'm not going after a solution or anything; just offering one of those "hmm, that's weird" moments.

You think that's wierd? Well in C++ you can actually return void!

Code:

template < typename Type >
Type foo( void )
{
	return Type( );
}

int main( void )
{	
	foo< void >( );
}