(j3.2006) TKR for C Interop

[Aleksandar Donev]

On Tuesday 27 February 2007 15:46, Dan Nagle wrote:

> The TR is about calling C with Fortran funny stuff.
> This is a walk through major changes to Fortran's interface scheme.
Agreed (if by this you mean IGNORE something or other).
The original issue I was talking about, which is about whether the same 
binding label can appear in more than one Fortran interface block, has 
everything to do with the TR. In particular, it provides an answer to a very 
fundamental question about the TR:

Should type and/or rank information be included in the descriptors, that is, 
when one creates a descriptor one must supply them, and one can get them out 
of a descriptor previously created? [The old proposal does this for rank 
only.]

If the answer is yes, then there better be some use for it. If one cannot use 
the same C procedure to handle arguments of different type and rank, then 
including TKR info in the API is useless, and adds burden to the programmer 
(since it complicates the descriptor API).

If we want to say "no", then we are making for rather poor TR design, IMO, but 
at least it is consistent. The proposed "let's also add type information", 
without a change to chapter 16, is totally inconsistent.

Best,
Aleks