Re: [GD-General] Offset a void pointer

[Antoine C. <a.c...@gm...>]

On Thu, 14 Oct 2004 15:11:04 +0100, Richard Fabian <ra...@gm...> wrote:
> if you are referring to this:
> 
> http://gcc.gnu.org/gcc-3.4/changes.html
> from the section starting:
> "The cast-as-lvalue extension has been removed for C++ and deprecated
> for C and Objective-C. In particular, code like this: "
> 
> then that is just that GCC has caught up with MS compilers in
> dissallowing the implicit lvalue return of standard C style casts

I don't think so. I think it's more that they're catching up with iso
c++ 99 standard, as most compilers, including MS ones, do. In fact,
that change happened at the same time they rewrote their c++ parser
from scratch to attain better standard conformance.
They refer is as "cast-to-lvalue *extension*", which imo imply that
it's not part of the standard.

> 
> Still appears to be perfectly legal to cast to an explicit lvalue type...

Not if you consider that "cast-as-lvalue", as they put it, might mean
"using a cast expression as if it was a lvalue". I don't see any
exception being made here depending on what you're casting to.

> 
> Please correct me if i am wrong on this... At least in my head it
> certainly seems to make no sense to take out casting to a modifiable
> type...

I don't know, it doesn't surprise me they forbid it. It seems to me
that about the only thing it could be used for is to specify the size
of data pointed by a void pointer, as you shown before. What would it
mean in other cases ? What would it do ?

> 
> Regarding Ola Olsson: static_cast is a little bit safer, as the types
> Cavey Gerard was referring to were reasonably compatible. AFAIK
> reinterpret_cast is used to allow you to do more "dodgy" casting, e.g.
> between a function pointer and a float, something which static_cast
> does not do...

reinterpret_cast is indeed the most permissive C++ cast, but iirc it's
usage is not guaranteed to be portable (for reasons like endianess)