Re: Ada and Motif libraries

[Simon Wright <simon@pogner.demon.co.uk>]

kilgallen@eisner.decus.org (Larry Kilgallen) writes:

> In article <x7vk9b01ot8.fsf@pogner.demon.co.uk>, Simon Wright <simon@pogner.demon.co.uk> writes:
> > Unless I missed an update (entirely possible), there's a lot of work
> > to do to convert the binding machinery to allow big-endian processors
> > (specifically, SPARC).
> 
> I thought the method DEC posted on Gatekeeper was supposed to work
> with big-endian machines as well, because it read through the C header
> files and built a C program to emit the Ada declarations.

It does indeed build a C program as you say. But the problem I fixed
(solution not available for posting (if anyone's interested) unless I
get permission) revolved around the positions of bit fields in
records. I don't think that DEC had any interest in big-endian
machines.

For example,

    type Event_Mask_Type is
        record
            Key_Press           : Boolean;

needs with it

        for Event_Mask_Type use 
        record
            Key_Press               at 0 range 0 .. 0;

on a little-endian machine and

        for Event_Mask_Type use 
        record
            Key_Press               at 0 range 31 .. 31;

on a big-endian one. I have to say I can't see why Ada compilers force
this grief on us, we don't compute 'Position differently on different
machines. But then I came late to big-endian representations so they
have always seemed slightly crazy to me!

And you can't use

        for Event_Mask_Type'Bit_Order use System.Low_Order_First;

(on GNAT anyway, which takes advantage of RM13.5.3(7).)