Groups | Search | Server Info | Keyboard shortcuts | Login | Register

Groups > comp.lang.c.moderated > #457

Re: Casting the first member of a struct (and back)

Show all headers | View raw

On 07/08/2013 07:06, stephan beal wrote:
> Hi, Gurus,
>
> i'm trying to clarify my understanding of how exactly the casting rules work for the first member of a C struct. For example...
>
> struct Base {
>    int foo;
>    ...
> };
>
> struct Subclass {
>    Base base;
>    ...
> };
>
> (struct typedefs elided)
>
> Subclass sub = { {...}, ... };
> Base * b = (Base*)⊂

In plain terms, take the memory of sub and treat it as a base. Looks 
fine to me because of the prohibition of initial padding in any object.
>
> It is my understanding (please correct me if i'm wrong) that that is
> completely kosher in C because C guarantees (IIUC) that the offset of
> the first member in the struct is 0.
>
> What i'm not clear on is if the reverse of that cast is legal:
>
> Subclass * original = (Subclass*)b;

Whiist this should work in the current context, it is dangerous as it 
relies on the memory addresses by a b* actually containing a subclass 
object. Hence I would expect that to be generally classed as undefined 
behaviour.

>
> (insofar as the coder knows that the conversion is legal, of course,
> e.g. by checking a type-id indicator stored in the base type.)

You are confusing 'legal' with 'will work as expected'. The 
implementation takes the programmer's word that a cast is valid. That is 
the nature of casts in C. They are effectively an unchecked way of 
changing the type of an area of memory.
>
> My intuition says, "if the offset is 0, the reverse cast must be just as legal because the address is the same,"
> but i am experienced enough to know that intuition is only a part of the story when it comes
? to details like this :/. Additionally, i seem to remember reading that 
C++ does _not_ guaranty
> that the addresses are the same for such casts.
Well C++ makes the same guarantees as C for Plain Old Data types (i.e 
structs that conform to the requirements of C) However, the general 
rules in C++ have to accommodate inheritance (possibly multiple). As a 
result C++ programmers avoid C-style casts because they are blunt tools 
in the context of C++.

I guess one of the resident C specialists will quote you chapter and verse.

Francis
-- 
comp.lang.c.moderated - moderation address: clcm@plethora.net -- you must
have an appropriate newsgroups line in your header for your mail to be seen,
or the newsgroup name in square brackets in the subject line.  Sorry.

Back to comp.lang.c.moderated | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

Casting the first member of a struct (and back) stephan beal <sgbeal@googlemail.com> - 2013-08-07 01:06 -0500
  Re: Casting the first member of a struct (and back) Francis Glassborow <francis.glassborow@btinternet.com> - 2013-08-07 13:51 -0500
    Re: Casting the first member of a struct (and back) Keith Thompson <kst-u@mib.org> - 2013-09-02 04:07 -0500
  Re: Casting the first member of a struct (and back) James Kuyper <jameskuyper@verizon.net> - 2013-08-07 13:51 -0500
  Re: Casting the first member of a struct (and back) Jasen Betts <jasen@xnet.co.nz> - 2013-08-07 13:51 -0500

csiph-web