Re: using offsetof with g++

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Matt Austern <austern@apple.com> writes:

| On Nov 6, 2003, at 11:47 AM, Gabriel Dos Reis wrote:
| 
| > I'm sure Matt remembers that when having a dinner in Kona with Bjarne
| > and others, I brought this issue but could not give any convinving
| > arguments as answers to the following basic question
| >
| >    Why do you want to take the offsetof a non-POD?
| >
| > (I know Matt gave at least two reasons but I can't recall; I just
| > remember mine was not convincing enough).
| 
| The basic question can be phrased as: Why do you want to use
| offsets instead of pointers to members?  Here are three possible
| answers:
|   1. You want to have non-C++ code that manipulates C++ objects.
|        You can pass the non-C++ code a void* and an integer offset,
|        but you can't very well pass it a complicated structure like a
|        pointer-to-member.
|   2. You want to do things that pointers-to-members don't support,
|        like arithmetic on offsets to iterate through data members.
|   3. You have legacy code that takes offsets, that you wrote when
|        you had a compiler that failed to enforce the offsetof
|        restrictions.  Now you've got a stricter compiler.  You could
|        rewrite your code in terms of pointers to member, but you
|        don't want to.
| 
| I don't think #3 is very convincing---it's a case for a compiler switch
| that relaxes restrictions, but it's not a case for changing the language
| standard.  #2 may or may not be convincing; it might serve better as
| a case for making pointers to members more powerful.  #1 strikes
| me as the best argument, since there is real need for interoperability
| between C++ and other languages.
| 
| I don't remember which of these reasons I used when I was talking
| to you and Bjarne last week; I don't think I was very coherent that
| night.
 
I don't think it was you not being coherent: the place was very noisy.

I seem to recall I heard at least #2 and probably #1.  I was wondering
whether allowing arithmetic on pointer-to-members would cover #2 and
possibily  #1.  I guess this is something we may have to look at more
seriously next meeting. 

-- Gaby