(j3.2006) [Interop-tr] definition of CFI_CDESC_T

[Robert Corbett]

On 04/24/12 02:16, N.M. Maclaren wrote:
> On Apr 24 2012, Malcolm Cohen wrote:
>>
>> (cc'd to interop-tr which is probably a better place for discussion).
>
> Now set as the Reply-to.
>
>> That does not look like "a type suitable for declaring a C descriptor" to me.
>>
>> Surely, given
>>   CFI_CDESC_T(2) x; then x must be a C descriptor, and therefore &x must be 
>> compatible with CFI_cdesc_t*, and x.base_addr et al should be usable; neither 
>> of those things appear to be true of this (that or my eyes have glazed over 
>> too much).
>>
>> It might be "a type suitable for declaring something [that might be] at least 
>> as big as a type suitable for declaring a C descriptor", but that is a horse 
>> of a decidedly different colour, no?
>
> Yes, but it is actually the horse that the TS is riding :-(  A long,
> long time ago, I suggested a macro that took the variable and declared
> it as CFI_cdesc_t *, pointing to a suitable area of memory.  However,
> that wasn't nice in other ways, and the decision was taken to merely
> allocate the memory - hence the example with a cast.
>
> The wording is wrong in C terms, but this is just another example of the
> dilemma of whether to use Fortran or C terminology.  But the sentence
> "A pointer to a variable declared using CFI_CDESC_T can be cast to
> CFI_cdesc_t *." makes it pretty clear how it is intended to be used.
> The example code delivers everything that is needed.
>
> It could be rephrased as something like:
>
>    "... and evaluates to a type of suitable size and alignment for
>    defining a variable to use as a C descriptor of that rank."

In the previous e-mail discussion of this issue, you pointed out that the
type should also be required to be an unqualified type.

Bob Corbett