NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/lang/c/17040

Text File | 1992-11-23 | 2.9 KB | 83 lines

Path: sparky!uunet!zaphod.mps.ohio-state.edu!usc!cs.utexas.edu!rutgers!concert!sas!mozart.unx.sas.com!sasghm
From: [e-mail] (Gary Merrill)
Newsgroups: comp.lang.c
Subject: Re: Is this ANSI?
Message-ID: <[e-mail]>
Date: 23 Nov 92 16:54:28 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail] (Noter of Newsworthy Events)
Organization: SAS Institute Inc.
Lines: 69
Originator: [e-mail]
Nntp-Posting-Host: theseus.unx.sas.com


In article <[e-mail]>, [e-mail] (Andrew Koenig) writes:
|> In article <[e-mail]> [e-mail] (Gary Merrill) writes:
|>
|> > I think "does not allow that technique" is a little strong.  What you
|> > *can* do is
|>
|>     struct FM
|>     {
|>        short data_len;
|>        char  data[1];
|>         };
|>
|>     struct FM *fmptr;
|>
|>     fmptr = malloc( sizeof(struct FM) + ??? );
|>
|> > where the ??? is the number of bytes of your data minus 1.  Some may
|> > regard this as "impure", but it is not forbidden by the standard and
|> > is eminently portable.
|>
|> Oh, really?  Can you give me concrete evidence that the C standard allows it?
|>
|> More specifically, after executing the code above, can you give me
|> evidence that referrint to fmptr->data[i] is allowed when i>1?
|> --
|>                 --Andrew Koenig
|>                   [e-mail]

Seems to follow quite directly from 3.3.2.1.  I don't see any
problem here at all.

One may want to argue that any reference beyond the end of the
array is not ensured of being a valid pointer (3.3.6), and I
concede the obvious fact that if a struct FM were *defined*,
then such a reference is "undefined".  However, in the example
given, the space is allocated via a malloc() call and hence
must conform to the requirements for such allocations (4.10.3).

I guess the question then reduces to:  Could an ANSI-conforming
compiler allocate memory in such a way that the given example
would not work?  I may have to admit that a compiler *could*
do this -- but it would appear to take a special effort of
perversity to accomplish it.  The memory allocated will be contiguous.
Then &fmptr->data[0] will point into this contiguous memory.
Now, how to arrange it so that &fmptr->data[2] does *not* point
into this contiguous memory ...  I suppose it can be done.

This seems especially perverse since one could easily address
the desired memory by a pointer initialized to &data[0].  Thus,
for example,

    char *p;

    p = &fmptr->data[0];

and

    p[1024]

(or whatever) is perfectly okay.  But data[1024] isn't.  It
isn't as though the memory isn't there, or that we can't
reference it.  Makes you wonder.


--
Gary H. Merrill  [Principal Systems Developer, C Compiler Development]
SAS Institute Inc. / SAS Campus Dr. / Cary, NC  27513 / [phone]
[e-mail] ... !mcnc!sas!sasghm