Re: [Pnet-developers] Use of ILNativeUInt vs. unsigned long in pointer casting and manipulation

[Peter Colson]

On 24/09/2004, at 11:36 AM, Rhys Weatherley wrote:

> On Friday 24 September 2004 11:40 am, Peter Colson wrote:
>
> > OK. I just want to understand how the resultant pointer can be used 
> > as a
> > real 'type' pointer in an environment that uses 8 byte pointers, if
> > it's sourced from a 32 bit integral?
>
> "ILType *" is not a pointer.  It is a pointer-sized union between 
> 32-bit
> integers, class pointers, value type pointers, and complex types.  The
> low-order bits indicate which arm of the union is being used.  
> Essentially,
> it is the following structure, packed into one word:
>
> struct
> {
>     int what_is_this : 2;
>     union
>     {
>         ILInt32 element_type;
>         ILClass *class_ptr;
>         ILClass *value_type_ptr;
>         ILComplexType *complex_ptr;
>     };
> };
>
> It's a space-saving hack.  One which is too late to change.  If your 
> platform
> cannot bit-pack into low-order bits, then it is time to get a new 
> platform.
> It is that simple.

I think I mentioned in a previous post that the later release 
environment
we're now in allows for 8 byte pointers that can be bit manipulated. 
That is
no longer a problem. I understand how the low-order bits are being used
to hold type information.

But your ILType* union above does resolve some elements to pointers
(ILClass*, ILComplexType*) once the 'what_is_this' element is stripped.

Do these 'pointers' then get used as real pointers that point to 
structures
defining the class or complex type? I would guess they do?

But if the ILType* is sourced from a function like ILMetaUncompressData
that returns a 4 byte unsigned long (which is one example I've seen) do
we then have a problem on a platform using 8 byte pointers when (if) we
need to use the ILClass* or ILComplexType* as a 'real' pointer?

(Or are these union elements not really used as pointers either?)

Regards,
Peter Colson.