https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85800
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |WAITING
Last reconfirmed| |2018-05-16
CC| |rguenth at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, I can't reproduce it because the arrays happen to be not adjacent for me.
But if I change
if (c1 == c2)
// Always true if p and q have same integer representation.
arr3[i] = arr1[i];
else
arr3[i] = arr2[i];
to
if (c1 == c2)
// Always true if p and q have same integer representation.
arr3[i] = arr2[i];
else
arr3[i] = arr1[i];
I get 10 consistently. This is because then arr3 will indeed be A.
Changing main to
int main() {
struct { int B[4]; int A[4]; } a;
printf("%p %p\n", a.A, &a.B[4]);
store_10_to_p(a.A, &a.B[4]);
printf("%d\n", a.A[0]);
return 0;
}
also makes it work reliably for me.
I guess at -O3 you get store_10_to_p inlined. You can try
changing the function to
void __attribute__((noinline,noclone)) store_10_to_p(int *p, int *q) {
to see if that makes any difference.
As said, I don't see any bug here.
