NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/std/cplus/1763

Internet Message Format | 1992-12-12 | 1.8 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!uwm.edu!linac!att!att!allegra!alice!ark
From: [e-mail] (Andrew Koenig)
Newsgroups: comp.std.c++
Subject: Re: Zero-length structures and pointer comparisons
Message-ID: <[e-mail]>
Date: 12 Dec 92 19:56:40 GMT
Article-I.D.: alice.24398
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Reply-To: [e-mail] ()
Organization: AT&T Bell Laboratories, Murray Hill NJ
Lines: 38

In article <[e-mail]> [e-mail] (John MAX Skaller) writes:

> >It shouldn't need to do that.  At most one trampoline is ever necessary
> >for a given nested function during its lifetime, namely the one that
> >binds the lexically surrounding context that was current when the
> >block containing the definition of that function was entered.

>     Are you saying that if I have a recursive function with a nested
> function in it, and store the address of the nested function in scope
> at the time of entering the n'th recursion into an array in slot n,
> then executing the j'th slot will access the n'th activation record
> and not the j'th one? (I.e. all the functions in the array
> access the most recent activation record?)

No, that's just the opposite of that I said.

I said `the context that WAS current.'

Perhaps an example will make it clearer:

    void (*p)(void);
    void f(int n)
    {
        void g() { printf("%d\n", n); }
        if (n == 3)
            p = &g;
        if (n == 5)
            (*p)();
        else f(n+1);
    }

I am claiming that calling f(0) should print 3, not 5, because the
address stored in p should correspond to the stack frame in which
n is 3.  Indeed, that is just what gcc does.  It is also what every
other language I know does that supports such functions at all.
--
                --Andrew Koenig
                  [e-mail]