NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/lang/c/16390

Internet Message Format | 1992-11-12 | 3.3 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!agate!dog.ee.lbl.gov!horse.ee.lbl.gov!torek
From: [e-mail] (Chris Torek)
Newsgroups: comp.lang.c
Subject: Re: How does the Address-of Operator (&) really work?
Date: 12 Nov 1992 13:31:42 GMT
Organization: Lawrence Berkeley Laboratory, Berkeley
Lines: 69
Distribution: world
Message-ID: <[e-mail]>
References: <[e-mail]>
Reply-To: [e-mail] (Chris Torek)
NNTP-Posting-Host: 128.3.112.15

In article <[e-mail]> [e-mail]
(Andrew Choi) writes:
>  I was told that the address-of operator & (along with sizeof)
>is done at compilation [time]

This is not strictly correct.  Many systems will compute addresses
during compilation, rather than at runtime, but there is no such
guarantee in the language.  Furthermore, addresses of automatic
objects do not even exist at compile time:

    void pointless(void) {
        int i, *j = &i;
        ...
    }

The value `&i' is not a compile-time constant.

>therefore, it becomes legal to do the followings:
>
>  &(((struct S *) NULL)->field)

This also is false.  It usually works, and is how many (if not most)
implementations actually implement the ANSI `offsetof' macro, but it is
not guaranteed to work, or to return anything useful.  (This is *why*
ANSI *has* an offsetof macro.)

>However, what I don't understand is how it works under multiple
>files.  Considered the followings:
>
>  /* file "foo.c" */
>  char *a = "This is a test";
>
>  /* file "bar.c" */
>  char **b = &a;
>
>When the 2 files, "bar.c" and "foo.c" are separately compiled,
>what does the compiler put for "&a", since it does not know where
>the definition for "a" is?

The compiler must emit some sort of directive to the rest of the
compilation system---which may be called the `linker' on your machine
---telling it that it has to initialize `b' before calling main().
This initialization is typically performed at link time, although
again there is no such guarantee in the C language.

Remember that the ANSI standard says only that there is a *system* that
acts as specified by the standard.  It does not say how the system is
built, or how it is to do what it must.  It says only *what happens*.

The actual mechanics on traditional systems, at least for this case,
are that the `object files' are in a special format (e.g., a bit
stream, a sequence of records, a set of sections, or whatever) that
includes a `symbol table' and `relocation information'.  The relocation
information says things like:

    In file bar.o, at offset +104 in the read/write data section,
    insert the value of the global symbol `_a'.

The actual encoding varies wildly from one system to another.  Any
system that implements the ANSI C standard must provide *some*
mechanism by which a static pointer variable can be set to point to a
named static object plus an integer constant offset, or to a named
function (without an offset), but the standard (and this newsgroup)
will not say how *your* system does this.  Your system's particular
needs are unique to your system, and you should find them described
in documents (and newsgroups) also specific to your system.
--
In-Real-Life: Chris Torek, Lawrence Berkeley Lab CSE/EE ([phone])
Berkeley, CA        Domain:    [e-mail]