Re: Joy of this, Joy of that

On 11/27/24 00:55, Louis Krupp wrote:

On 11/26/2024 2:22 PM, Lawrence D'Oliveiro wrote:

On Tue, 26 Nov 2024 08:32:32 +0000, Richard Kettlewell wrote:
>

The examples show that freeing the same pointer twice can be
detected reliably.

They do no such thing.

free(p);
printf("%s(): returned from first free() call\n", __func__);
>
free(p);
printf("%s(): returned from second free() call\n", __func__);
>
Is that or is that not freeing the same pointer twice?

 In the example in the man page, if the two free() calls are back to back with no intervening malloc() calls, the double free() is detected. I tried adding a malloc(1000) call in the middle:
      free(p);
     printf("%s(): returned from first free() call\n", __func__);
      malloc(1000);
     printf("%s(): returned from second malloc(1000) call\n", __func__);
      free(p);
     printf("%s(): returned from second free() call\n", __func__);
 and the program ran to completion with no apparent problem.
 

Yes, possibly this malloc reassigned the same address as the unseen original malloc. In general, without fat pointers, it is hard to see how free could detect that memory had been reassigned.
It is interesting to know that unoptimised compilations try to spot double free, but only a little interesting. I remember we ran a tool called Purify to spot this type of error, amongst other errors. Presumably that injected additional code to check reliably.