NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/lang/c/20146

Internet Message Format | 1993-01-23 | 3.4 KB

Xref: sparky comp.lang.c:20146 comp.std.c:3406
Path: sparky!uunet!charon.amdahl.com!pacbell.com!decwrl!sdd.hp.com!zaphod.mps.ohio-state.edu!howland.reston.ans.net!bogus.sura.net!udel!gatech!hubcap!ncrcae!ncrhub2!ncrgw2!psinntp!heimdall!thor!scjones
From: [e-mail] (Larry Jones)
Newsgroups: comp.lang.c,comp.std.c
Subject: Re: offsetof in case labels
Message-ID: <[e-mail]>
Date: 24 Jan 93 18:27:54 GMT
References: <[e-mail]>
Sender: [e-mail]
Followup-To: comp.std.c
Lines: 56

In article <[e-mail]>, [e-mail] (Robert Swan) writes:
> We have a problem on various machines using the offsetof macro from
> stddef.h in case labels.  These machines all reject the expression as
> being non-constant.

Those compilers are not ANSI/ISO conforming.  The ANSI/ISO C Standard
clearly requires offsetof to expand to an integral constant expression
which, as you noted, is exactly what is required for a case label.
(Since this is a standards question, I've also posted this to comp.std.c
and redirected follow-ups there.)

> A representative of one of the machine vendors (IBM), has glibly stated
> that the compiler is doing the right thing by rejecting it.

He is mistaken.  By an amazing coincidence, X3J11 has received a request
for interpretation from someone at IBM on exactly this point.  This RFI
points out that it is not possible to write an offsetof macro that
expands to strictly conforming code (which is precisely why the macro
exists -- if you could do it yourself, the macro would be redundant) and
arguing that this makes any program that uses offsetof not strictly
conforming and thus an implementation is free to reject it.  This is, of
course, nonsense -- an otherwise strictly conforming program that uses
offsetof in a strictly conforming fashion is still strictly conforming
reguardless of the expansion of the macro.  The RFI, suspecting this to
be the committee's response, goes on to point out that this *requires*
an implementation to implement an extension of some sort in order to
implement offsetof and argues that this is inappropriate.  While this is
the most common method of implementing offsetof (generally by extending
the implementation's definition of constant expression, which the
standard allows, or by adding a new keyword), it is not, strictly
speaking, necessary.  All that is required is an omniscient preprocessor
that can expand offsetof to the correct integer constant -- not a
difficult task if the preprocessor is an integral part of the compiler
with access to the symbol table.

(These are my recollections of the RFI and the committee's discussions
-- there is no guarantee that they match anyone else's recollections or
the actual facts.  Those who care should obtain the collected Requests
for Interpretation and Responses which ANSI should be publishing as a
Technical Information Bulletin real soon now.)

> However, I'm pretty much at sea with the ANSI standard - we don't have a
> copy, and the excerpts I have seen of it suggest that it is written in
> sufficiently obscure legalese to cloud the superficial interpretation I
> might put on it.

On the contrary, it's actually quite readable for a language standard; a
bit more difficult than K&R Appendix A, perhaps, but not significantly
so.
----
Larry Jones, SDRC, 2000 Eastman Dr., Milford, OH  45150-2789  513-576-2070
[e-mail]
Pitiful.  Just pitiful. -- Calvin