Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.c++ > #11011 > unrolled thread

CHello *p = 0; p->f(); /* what does spec say about this*/

Back to article view | Back to comp.lang.c++

Contents

  CHello *p = 0; p->f(); /* what does spec say about this*/ lovecreatesbeauty <lovecreatesbeauty@gmail.com> - 2011-10-20 07:24 -0700
    Re: CHello *p = 0; p->f(); /* what does spec say about this*/ Leigh Johnston <leigh@i42.co.uk> - 2011-10-20 16:12 +0100
    Re: CHello *p = 0; p->f(); /* what does spec say about this*/ "Bo Persson" <bop@gmb.dk> - 2011-10-20 17:45 +0200
      Re: CHello *p = 0; p->f(); /* what does spec say about this*/ lovecreatesbeauty <lovecreatesbeauty@gmail.com> - 2011-10-20 10:55 -0700
        Re: CHello *p = 0; p->f(); /* what does spec say about this*/ Victor Bazarov <v.bazarov@comcast.invalid> - 2011-10-20 14:09 -0400
        Re: CHello *p = 0; p->f(); /* what does spec say about this*/ Robert Wessel <robertwessel2@yahoo.com> - 2011-10-20 13:19 -0500

#11011 — CHello *p = 0; p->f(); /* what does spec say about this*/

class CHello{
public:
	void f(){};
};

int main()
{
	CHello *p = 0;
	p->f(); /* undefined behavior? what does spec say about this*/
	return 0;
}

[toc] | [next] | [standalone]

#11015

On 20/10/2011 15:24, lovecreatesbeauty wrote:
> class CHello{
> public:
> 	void f(){};
> };
>
> int main()
> {
> 	CHello *p = 0;
> 	p->f(); /* undefined behavior? what does spec say about this*/
> 	return 0;
> }

Yes it is undefined behaviour; you should not do it unless your name is 
Microsoft. :)

/Leigh

[toc] | [prev] | [next] | [standalone]

#11020

lovecreatesbeauty wrote:
> class CHello{
> public:
> void f(){};
> };
>
> int main()
> {
> CHello *p = 0;
> p->f(); /* undefined behavior? what does spec say about this*/
> return 0;
> }

You are not allowed to dereference a null pointer so, yes, this is 
undefined.


Bo Persson

[toc] | [prev] | [next] | [standalone]

#11025

On Oct 20, 11:45 pm, "Bo Persson" <b...@gmb.dk> wrote:
> lovecreatesbeauty wrote:
> > class CHello{
> > public:
> > void f(){};
> > };
>
> > int main()
> > {
> > CHello *p = 0;
> > p->f(); /* undefined behavior? what does spec say about this*/
> > return 0;
> > }
>
> You are not allowed to dereference a null pointer so, yes, this is
> undefined.

Thanks, But the book `C: A reference manual, 5th' also mentions that
stddef.h may define offsetof like this:

#define offsetof(type, member) ( (size_t) & ((type*)0) -> member )

[toc] | [prev] | [next] | [standalone]

#11026

On 10/20/2011 1:55 PM, lovecreatesbeauty wrote:
> On Oct 20, 11:45 pm, "Bo Persson"<b...@gmb.dk>  wrote:
>> lovecreatesbeauty wrote:
>>> class CHello{
>>> public:
>>> void f(){};
>>> };
>>
>>> int main()
>>> {
>>> CHello *p = 0;
>>> p->f(); /* undefined behavior? what does spec say about this*/
>>> return 0;
>>> }
>>
>> You are not allowed to dereference a null pointer so, yes, this is
>> undefined.
>
> Thanks, But the book `C: A reference manual, 5th' also mentions that
> stddef.h may define offsetof like this:
>
> #define offsetof(type, member) ( (size_t)&  ((type*)0) ->  member )

There is no provision in the Standard regarding the implementation of 
the C language 'offsetof' macro *except* that its 'type' argument is 
limited to the POD types.  That means that any other requirement of the 
Standard shall be followed when implementing that macro, which means 
that if it is implemented to dereference a null pointer, it's invalid.

As for the book, how is a book about implementation of some C language 
construct be relevant here?  This is a C++ newsgroup.

V
-- 
I do not respond to top-posted replies, please don't ask

[toc] | [prev] | [next] | [standalone]

#11027

On Thu, 20 Oct 2011 10:55:49 -0700 (PDT), lovecreatesbeauty
<lovecreatesbeauty@gmail.com> wrote:

>On Oct 20, 11:45 pm, "Bo Persson" <b...@gmb.dk> wrote:
>> lovecreatesbeauty wrote:
>> > class CHello{
>> > public:
>> > void f(){};
>> > };
>>
>> > int main()
>> > {
>> > CHello *p = 0;
>> > p->f(); /* undefined behavior? what does spec say about this*/
>> > return 0;
>> > }
>>
>> You are not allowed to dereference a null pointer so, yes, this is
>> undefined.
>
>Thanks, But the book `C: A reference manual, 5th' also mentions that
>stddef.h may define offsetof like this:
>
>#define offsetof(type, member) ( (size_t) & ((type*)0) -> member )


That's internal compiler magic for implementing offsetof.  It actually
invokes three* different kinds of undefined (by the standard) behavior
to produce a constant expression for the value desired.  With
sufficient knowledge of the compiler's internals, the *implementation*
can get away with that sort of thing.

Note that if the three undefined things are not supported by the
compiler in the required way, the compiler is still obligated to
provide offsetof, but will need to use different internal compiler
magic to do so.  The above trick is popular because many compilers,
particularly those supporting flat address spaces, actually Do The
Right Thing with that expression, making the implementation of
offsetof trivial.

Note that unlike your example, a compiler that supports those tricks
for implementing offsetof does not actually dereference the NULL
pointer. 


*It casts an integer zero to an object pointer, then performs
arithmetic on the result, and allows that in a constant expression.

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.c++

csiph-web