Re: Trying to import a C function

["Warren W. Gay VE3WWG" <ve3wwg@cogeco.ca>]

Steve wrote:
> "Preben Randhol" <randhol+valid_for_reply_from_news@pvv.org> wrote in
> message
> news:slrnbqk7aq.q3.randhol+valid_for_reply_from_news@kiuk0156.chembio.ntnu.no...
> 
>>Hi
> 
> [snip]
> 
>>Now what is the equivalent of unsigned long long int in Ada?
> 
> In GNAT on x86:
> 
>   unsigned long long   =  Interfaces.Unsigned_64
> 
> You could define your own using a modular type definition, but isn't that
> the reason the Interfaces module is there in the first place?
> 
> Steve
> (The Duck)

I find that it is generally easier, and more portable, to handle
macros and size differences in a C wrapper function. This
particularly applies to Linux, since so much changes there still.

If you enforce a parameter size of 64 bits between your C wrapper
and the Ada code, you can always be assured that you won't get
burned on the size differences.

Furthermore, if you are targeting multiple platforms/UNIces, then
the structure definition, its padding and alignment will also
give you trouble (and in fact this can potentially be the case
with different Linux releases/distros). For this reason, it is
sometimes better to have the C wrapper take the structure
elements and pass them back to Ada in safe sized arguments
individually (or perhaps stable structure definitions that
guarantee sizes and alignments).

If you use the C wrapper approach, you have a better chance of
only visiting that code once. If you choose to avoid the
wrapper, you'll likely be looking at the interface again,
with the next release or platform. ;-)

The only other alternative is to run a process that generates
the interface (or interface parameters) prior to compiling. I
believe that FLORIST does something along this line (but
it has been a while since I have looked at that side of
things).

-- 
Warren W. Gay VE3WWG
http://home.cogeco.ca/~ve3wwg