Re: Documentation of memcpy and undefined behavior in memset

[Rich Felker <dalias@libc.org>]

On Thu, Jul 06, 2017 at 08:02:12PM +0300, Alexander Monakov wrote:
> On Thu, 6 Jul 2017, Rich Felker wrote:
> > FWIW, I think GCC may do aggressive optimization based on the
> > assumption that memcpy implies the pointer points to an object (of
> > size at least 1)
> 
> The compiler can deduce that the pointer is non-null (and that's
> fine), but otherwise I don't see what possible optimizations could
> take place. Did you have something specific in mind?

It could presumably move loads from after a branch to before. E.g.

	memcpy(q,p,0);
	if (whatever) {
		y=*p;
		...
	}
	/* y not used after here */

to:

	memcpy(q,p,0);
	y=*p;
	if (whatever) {
		...
	}
	/* y not used after here */

If p points to one past the end of an object that ends on a page
boundary, this transformation could introduce a crash.

Rich