[tzs]

> Pointers to free'd memory are akin to uninitialized pointers, so free(p) followed by if (p==q) is an instrument of arson

What's the reason for this?

[jcranmer]

Given the following code:

    void *p = malloc(N);
    do_random_stuff(p);
    void *q = malloc(N);

With this rule, the compiler can conclude that p and q cannot alias, even if it doesn't have body of do_random_stuff. Without it, it would first have to prove that p is never freed before calling q, which is basically impossible (moving the body of intervening code into a different file, for example, would do the trick).

[tedunangst]

900 years ago there was a CPU which stored pointers in special registers and trapped if you loaded a pointer with an invalid segment. And so loading the pointer into a register to compare it would crash.

[coliveira]

Using a freed pointer is incorrect behavior, a bug in shorter terms. If you do anything with a freed pointer (other than assigning new memory), you're inviting all kinds of bugs (independent of what the compiler might be doing with your code).

[xigoi]

Obviously dereferencing a freed pointer is incorrect behavior, but what harm is there in using its numerical value?

[Dylan16807]

I can't tell you exactly why but it's consistent with just about everything else involving p being undefined, and the result of the comparison would be useless anyway.

[tzs]

Why would the comparison be useless?

I can imagine situations where a pointer q might sometimes be a copy of pointer p and sometimes might point to something else, and the code wants to free q if and only if it is not a copy of p (because p has been free'd earlier).

[Dylan16807]

Because a new object can have the same address as p, so comparing to p isn't enough to tell you if you have a copy of p or a live pointer to something else.