🚀 go-pugleaf

In article <45666e$amf@engnews2.Eng.Sun.COM> clamage@Eng.Sun.COM
(Steve Clamage) writes:

|> fjh@munta.cs.mu.OZ.AU (Fergus Henderson) writes:

|> >jaf3@motel6.cec.wustl.edu (Stephen W. Gevers) writes:

|> >>Andy J Robb <AJRobb@pavilion.co.uk> wrote:
|> >>:Ron Forrester <rjf@infograph.com> wrote:
|> >>:...
|> >>:It is madness if ANSI C++ allows implict casting of an integral value
|> >>:of a different size (zero) to a pointer but NOT ((void*)0).
|> >>
|> >>The reason that you cannot implicitly cast (void*) 0 to a null pointer
|> >>is because it would be a type hole (one that C currently allows):

|> >If that is the reason, then it is a bogus reason.
|> >Allowing `(void *)0' to be implicitly converted to a null
|> >pointer does not create any type holes.

|> This topic is discussed in D&E. Everywhere else in C++, the validity
|> of a cast is determined only by the types, not by the values. (With
|> one exception noted in a moment.) If we allow
|> 	T* p = (void*)0;
|> then why not
|> 	void* q = 0;
|> 	T* p = q;
|> Carrying this argument to its logical conclusion you should allow
|> 	extern void* q;
|> 	T* p = q
|> if q is null and not otherwise, requiring a runtime check. In fact,
|> you can get this effect using a dynamic_cast, which is always valid
|> for null pointers, and which is the cast that takes values into account.

|> Implicit casts are valid or not based only on the types, not the values.

|> I also want to address Forrester's comment about converting an integer
|> to a pointer. That is NOT the case. A literal zero in a pointer context
|> is not an integer, but a null pointer.

I think, actually, that it is an ``integral constant expression
evaluating to 0'', which will be treated as a null pointer constant in
certain contexts.

|> Notice the difference between
|> a constant zero in source text and an integer with value zero in
|> machine code.

But this is exactly what Fergus was saying.  There is already one
(very) special case; it wouldn't really take any more black magic to
make the constant expression ``(void*)0'' (or any expression which
evaluates to it) a special case, too.
-- 
James Kanze         Tel.: (+33) 88 14 49 00        email: kanze@gabi-soft.fr
GABI Software, Sarl., 8 rue des Francs-Bourgeois, F-67000 Strasbourg, France
Conseils en informatique industrielle --
                              -- Beratung in industrieller Datenverarbeitung