🚀 go-pugleaf

In article <DH220v.9r@thinkage.on.ca>, dat@thinkage.on.ca (David Adrien Tanguay) writes:
>How is the following concept for strict conformance? (with appropriate context)
>
>struct s_t {
>	int this, that;
>	/* this structure has a variable length tail */
>};
>
>#define tail_of(p) ((char*) ((p)+1))
>
>void f(size_t tail_size) {
>	struct s_t *sp = malloc(sizeof(*sp) + tail_size);
>	char *the_tail_is = tail_of(sp);
>	...
>}
>
>If tail_size < sizeof(struct s_t), can tail_of() fail because of the +1?
>I.e., assuming pointers carry bounds, can the implicit cast from malloc to sp
>re-boundize the pointer value to an integral number of struct s_t's? Then the
>cast to char* in tail_of would produce a pointer with the truncated bounds so
>you still couldn't access the entire malloc'd object that way.

Correct. But all you need to do is change the definition of tail_of
as follows:

#define tail_of(p) ((char*)(void*)(p) + sizeof(*(p)))

This way, when we convert p to a void*, we get back the exact same pointer
returned by malloc. This pointer can be used to access an array of any object
type. So we cast it to char*, and then skip over all the chars that make up
the struct s_t definition. The resulting pointer value points into the middle
of the mallocced char array, right at the beginning of the tail. This is
strictly conforming.


--

Chris Volpe				Phone: (518) 387-7766 (Dial Comm 8*833
GE Corporate R&D			Fax:   (518) 387-6560
PO Box 8, Schenectady, NY 12301		Email: volpecr@crd.ge.com