offsetof "problem"

[Andrew Pollard <andrewp at andypo dot net>]

Hi All,

A checkin by "Jason Merrill"...

>date: 2002/02/01 19:31:20;  author: jason;  state: Exp;  lines: +26 -4
>        * typeck.c (build_component_ref): Always complain about offsetof
>        constructs on non-PODs.  Only make it an error for members of
>        virtual bases.

I realise that there must be a good reason behind this, but.....

How are you meant to define things that get the size of an arbitary
member, or the offset of a member.... eg

#define OFFSET_OF(type,field)                                                 \
            (reinterpret_cast <size_t>                                        \
                (reinterpret_cast <char *>                                    \
                    (&(reinterpret_cast <type *> (0))->field)))

#define SIZEOF_MEMBER(type,field)                                             \
            (sizeof ((reinterpret_cast <type *> (0))->field))

now whinge with a warning.... (gcc-3.1)

I realise that these were always a bit of dodgy code... but now all 
get out of gcc-3.1(20020329) are

> warning: invalid offsetof from non-POD type `struct A'; use 
>    pointer to member instead

Fair enough, but the types in question are basically POD types, but
they just have member functions and constructors/destructors.

I can understand that in the presence of inheritance, etc, that these
macros could be dodgy, but is there a 'proper' way of doing them?

Thanks,
 Andrew.
--
 Andrew Pollard, ASI/Brooks Automation  | home: andrew@andypo.net
670 Eskdale Road, Winnersh Triangle, UK | work: Andrew.Pollard@brooks.com
 Tel/Fax:+44 (0)118 9215603 / 9215660   | http://www.andypo.net