pjm3@ns2.CC.Lehigh.EDU (PAUL JASON MARTINO) writes:

>I am getting some strange results running the following code on Borland 4.5.
>
>void f( void * );
>void f( const void * );
>
>void main()

main() should return an `int'.

>    {
>    char *c;
>
>    f( c );          // does not issue an error
>    f( (char *) 0 ); // error: ambiguous
>    }
>
>I can find nothing in the standard doc that should make the f(c) call be
>non-ambiguous. Borland 3.1 flags f(c) as an error. Our product PC-Lint for
>C/C++ flags both of them as an error. I was curious is anyone knows who is
>correct.

Neither PC-Lint nor Borland 3.1 is correct, IMHO - both statements
should be unambiguous, and should invoked `f(void *)'.
But Borland's behaviour seems particularly strange, since there
shouldn't be any difference between `c' and `(char *)0' in this case.

I base my conclusion on the Sept working draft.
The conversion sequences involved in this case are

 S1: (char *) -> (void *)

and
 S2: (char *) -> (void *) -> (const void *)

Now 13.2.3.2/3 states:

 Standard conversion sequence S1 is a better conversion sequence
 than standard conversion sequence S2 if
  - S1 is a proper subsequence of S2, or ...

Hence S1 is a better conversion sequence than S2, and so there is
no ambiguity - the code should invoke `f(void *)'.

P.S. Followups redirected to comp.std.c++.

--
Fergus Henderson - fjh@munta.cs.mu.oz.au