Two MAYALIAS in cs_tests/recur8.c should be NOALIAS?

[taquangtrung]

Hi,

For the two MAYALIAS in this test case, I think they should be NOALIAS?

This is because c and &b are aliases, and b, z1, z2 are variables declared differently, hence &b, &z1, &z2 cannot be aliases?

Could you advise if my understanding is correct?

// cs_tests/recur8.c

#include "aliascheck.h"
int z1,z2;
void foo(int **p);
void bar(int **a){
	int *c, b;
	*a = &b;
	c = *a;
	MUSTALIAS(c,&b);
	MAYALIAS(c,&z1);   // it should be no-alias if strong updates are enabled
	MAYALIAS(c,&z2);
	foo(a);
}


void foo(int** p){
	p = malloc(10);
	*p = &z2;
	bar(p);
}

int main(){
	int **x, *y;
	x = &y;
	y = &z1;
	foo(x);
}

[yuleisui]

Ideally, yes. but again, this is a conservative case, since heap strong updates are not enabled by the underlying analysis.

[taquangtrung]

Hi again,

I'm working on a flow-sensitive analysis technique. For this program, it is non-terminating due to the recursive call of foo and bar, and my analysis tool doesn't terminate as well. Could you advise if this test case can be fixed? E.g., by adding a base case to the recursive call?

Thank you!

[yuleisui]

Just done by adding a termination condition: e622672