Re: [csmith-bugs] GCC Complaint: "init from incompatible pointer type"

[John Regehr <regehr@cs.utah.edu>]

Eric, is the Csmith code actually wrong or is it just giving an annoying 
warning?  In general it's fine to add qualifiers, just not to delete them. 
 For example Linux uses the idiom below which looks a lot like your 
example code but with one fewer levels of indirection.  The important part 
is that when you make an atomic_t, it does not need to be volatile.

John


typedef struct {
    int counter;
} atomic_t;

#define atomic_read(v)      (*(volatile int *)&(v)->counter)
#define atomic_set(v,i)     (((v)->counter) = (i))




On 05/23/2013 08:36 AM, Eric Eide wrote:
> Csmith (@ commit ac43da8, circa Mar 20 2013) sometimes generates code like
> this:
>
> 	void foo()
> 	{
> 	    int *p = 0;
> 	    volatile int **q = &p;
> 	}
>
> GCC 4.6.3 complains about this code, scolding:
>
> 	warning: initialization from incompatible pointer type [enabled by
> 	default]
>
> I noticed this because my predicate script for C-Reduce looks for "incompatible
> pointer type" warnings to weed out bad reduction steps.  When the original
> program leads to such warnings, of course, my reductions cannot get off the
> ground.
>
> If you want a Csmith command line that leads to such generated code, please let
> me know.
>
> Eric.