Tue, 14 Dec 2004 10:47:19 -0800
In investigating the subject issue, so far I haven't been able to get a
straight answer out of anybody. I retrieved several alleged examples of how
this is done from the Internet. Some of them don't work; others work, but
they don't do anything useful, the problem evidently being that Microsoft
Visual C++ is not properly linked to the Ada libraries and so can't
comprehend math functions, text_io, and so on. Such linking instructions as
it's been possible to scrape up off the bottom of the Internet or squeeze
out of vendors is either vague or obviously wrong. I've been told the
linking has to be done on the Ada side if Ada calls C++, but also if it's
the other way around. I've been told the exact opposite. I've been told
it's easy; I've been told it can't be done.
What we have here is a directory of Ada packages to be integrated with a
system in which the controlling code has been developed in Compaq Visual
FORTRAN under Microsoft Visual Studio 6.0 for Windows XP on an HP 7500 PC.
According to one alleged authority, the Ada packages cannot be invoked
directly from FORTRAN--C++ must intervene. Linking to C++, on the C++ side,
Ada95 object files generated on the ADA side by GNAT Adagide is not a
problem, provided the Ada code makes no reference to the Ada libraries,
directly or indirectly (such as containing "+" or "-", for example).
Attempting to link the Ada libraries to C++ on either the C++ or Ada side
doesn't seem to work, by which I mean I don't know how to do it; but I don't
know how to do it because the instructions for linking to Ada on the
Microsoft side are nonexistent, while the corresponding instructions for
linking to C++ on the Ada side seem to be incomplete.
My web search led me to one web site on which a frustrated user had posted
exactly the same question I have. In 1998. In the intervening six years,
there was one response from an expert: "I don't know."
I would appreciate any information I can get my hands on.