Re: Prototypes and parameters

[Barbara Morris <bmorris@xxxxxxxxxx>]

Tom Liotta wrote:
>
> Barbara:
>
> Can you clarify one thing? What do you mean by "which would not work" when 
>you speak about a pointer for a varying length field? Is that to mean such an 
>address cannot be passed? or that the called procedure could not correctly 
>return a value without messing things up in the calling procedure? Would that 
>still be strictly true even if both caller and callee used an agreed length, 
>e.g., the caller set the length appropriately before the call? I've never 
>looked at what would happen for a varying length field in this case; this 
>makes me curious.

Tom, as long as you -always- pass the address of a varying length field,
and the called procedure expects a pointer to a varying length field,
everything would work fine.

But if you sometimes pass the address of a fixed-length field, and
sometimes a varying-length field, the called procedure wouldn't know
what to expect.  Unless you add an extra parameter to indicate the type.

If you pass a pointer, and decide you're only going to allow the address
of one type to be passed, you'll have to enforce this through
documentation and knuckle-rapping.  The compiler won't help you.  But if
you're only going to allow one type to be passed, just prototype it as
that type and let the compiler help you.

Barbara