Re: [PATCH]: Fix PR tree-optimization/21407

[Mark Mitchell <mark at codesourcery dot com>]

Daniel Berlin wrote:

> If i give you a pointer to a field, that would mean you could add or
> subtract bytes and legally get at any other member of the structure,
> containing structure, etc.

Yes, in my mind, that's the whole point of offsetof.

> > Do you think it's legal if the fields are in the other order:
>
>
> Yes, but only because of rules about initial members, etc.

It's certainly possible that I'm wrong.  Joseph is the person I trust 
most to know the C standard, and I'll be interested in his opinion.

But, casting pointers to aribtrary types is specifically allowed by the 
standard -- so long as you cast them back before you use them.  This 
program:

  struct S {
    int i;
    double d;
  };

  void f(int * p) {
    struct S* s = (struct S*) p;
    s->d = 3.0;
  }

  int main () {
    struct S s;
    f(&s.i);
  }

is definitely valid.  I don't think this has anything at all to do with 
common initial prefix rules; the only important thing is that "s" 
actually point to an object of type "struct S", which it does.  The 
initial prefix rules are about accessing a "struct Derived" using a 
"struct Base *", or vice versa; they give you an allowance to do that, 
even though you can't normally use a "struct Foo*" to access a "struct Bar".

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304