(j3.2006) elem_len argument to CFI_allocate

[Bader, Reinhold]

[...]
> 
> Questions:
> 
> 1) Have I missed an important case here?
> 
> 2) Did we intend to support Case 1?

If yes, there also needs to be a provision in CFI_deallocate to reset
the elem_len suitably, otherwise the distinguishability gets lost after
the first allocation; I'd also suggest having a macro CFI_DEFERRED_LENGTH
with a negative value which makes code using this feature more readable. 

> 
> 3) If no to 1 and 2, do we need the elem_len argument to CFI_allocate?

No. 

Regards
Reinhold
> 
> 
> Cheers,
> Bill
>