Calling C++ methods from Ada..

Calling C++ methods from Ada..

[Ganesh Ramasivan]

The following is a small test program in which the Ada function is
implemented in C++.

------------------------------
with System;

procedure CppTest is

procedure SET_TIME (hour: System.Address);
pragma import(CPP, SET_TIME, "Time::SetTime");

begin
  null;
end CppTest;

--------------------------------

I get the following error when I compile this code using the GHS Ada
compiler under Solaris 2.8. Can anybody please tell me what I am doing
wrong?

c8966@gatekeeper: cpp>build CppTest.bld
Compiling /home/c8966/ada/cpp/CppTest.ada because "CppTest.o" doesn't
exist
/home/c8966/ada/cpp/CppTest.ada: Error: line 83 col 10 LRM:B.1(26), No
compatible C++ type for System.Address_Target

Error: build failed
Build terminated


Thanks,

Ganesh

Re: Calling C++ methods from Ada..

[Luke A. Guest]

What compiler are you using?

Luke.

Re: Calling C++ methods from Ada..

[Jeff C r e e.m]

"Luke A. Guest" <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> wrote 
in message 
news:pan.2004.12.01.23.05.04.344825@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk...
> What compiler are you using?
>
> Luke.
>

It was not obvious but he said GHS (Green Hills) in the original message. 

Re: Calling C++ methods from Ada..

[Brian May]

>>>>> "Ganesh" == Ganesh Ramasivan <ganesh.ramasivan@gdcanada.com> writes:


    Ganesh> procedure CppTest is

    Ganesh> procedure SET_TIME (hour: System.Address); pragma
    Ganesh> import(CPP, SET_TIME, "Time::SetTime");

    Ganesh> begin null; end CppTest;

Is Time::SetTime a static member function? You didn't give the
definition in your E-Mail so I can't tell..

If it isn't, then it is a class method, and requires a hidden *this
parameter.

Furthermore, if it is a virtual function, then a static binding may
not be appropriate.

Check the GNAT reference manual for details on the following pragmas
(section 10.2 Interfacing to C++):

pragma CPP_Class
pragma CPP_Constructor
pragma CPP_Vtable

Note I haven't used any of these, don't understand (yet) how to use
them, and I don't know of any examples (do a web search).
-- 
Brian May <bam@snoopy.apana.org.au>

Re: Calling C++ methods from Ada..

[Brian May]

>>>>> "Luke" == Luke A Guest <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> writes:

    Luke> What compiler are you using?

Good point - I assumed he was using GNAT, so the solution I gave may
not be relevant if he is not using GNAT. I am not sure if the pragmas
I gave are GNAT specific or not.
-- 
Brian May <bam@snoopy.apana.org.au>

Re: Calling C++ methods from Ada..

[Luke A. Guest]

On Thu, 02 Dec 2004 22:12:03 +1100, Brian May wrote:

>>>>>> "Luke" == Luke A Guest <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> writes:
> 
>     Luke> What compiler are you using?
> 
> Good point - I assumed he was using GNAT, so the solution I gave may
> not be relevant if he is not using GNAT. I am not sure if the pragmas
> I gave are GNAT specific or not.

Yeah, I didn't see the reference to GHS, I must've just skimmed past ;-),
but the code he posted wouldn't work on GNAT anyway because you have to
provide GNAT the mangled C++ names, which is crap IMO.

Luke.

Re: Calling C++ methods from Ada..

[Adrien Plisson]

Luke A. Guest wrote:
> but the code he posted wouldn't work on GNAT anyway because you have to
> provide GNAT the mangled C++ names, which is crap IMO.

and compiler dependent... and even rarely documented...

-- 
rien

Re: Calling C++ methods from Ada..

[Martin Krischik]

Luke A. Guest wrote:

> On Thu, 02 Dec 2004 22:12:03 +1100, Brian May wrote:
> 
>>>>>>> "Luke" == Luke A Guest
>>>>>>> <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> writes:
>> 
>>     Luke> What compiler are you using?
>> 
>> Good point - I assumed he was using GNAT, so the solution I gave may
>> not be relevant if he is not using GNAT. I am not sure if the pragmas
>> I gave are GNAT specific or not.
> 
> Yeah, I didn't see the reference to GHS, I must've just skimmed past ;-),
> but the code he posted wouldn't work on GNAT anyway because you have to
> provide GNAT the mangled C++ names, which is crap IMO.

The original idea was a C++ parser which generated the bindings. With the
complexity of C++ the only way to go. However the parser never
happend :-( .

Martin
-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: Calling C++ methods from Ada..

[Luke A. Guest]

On Fri, 03 Dec 2004 09:59:22 +0100, Martin Krischik wrote:

> The original idea was a C++ parser which generated the bindings. With the
> complexity of C++ the only way to go. However the parser never
> happend :-( .

That's just overkill, the way to do it is like the example the OP posted.
Have the compiler take a pragma which determines which C++ compiler the
binding is from, this determines the name mangling method (as they're all
different), then when the Ada compiler comes across an import of a C++
method (with class scoping), it can generate the correct bindings.

Luke.

Re: Calling C++ methods from Ada..

[Luke A. Guest]

So, the example could be written like this:

pragma import(CPP, GCC3, SET_TIME, "void Time::SetTime(...)");

Although, I'm not sure why he would be passing in a pointer to a set time
function ;-)

Luke.

Re: Calling C++ methods from Ada..

[Luke A. Guest]

On Fri, 03 Dec 2004 09:59:22 +0100, Martin Krischik wrote:

> The original idea was a C++ parser which generated the bindings. With the
> complexity of C++ the only way to go. However the parser never
> happend :-( .

I wonder how much work it would be to add a switch to the GCC C++ compiler
to enable generation of GNAT bindings (when --languages has ada passed on
the configuration command line)?

Luke.

Re: Calling C++ methods from Ada..

[Brian May]

>>>>> "Luke" == Luke A Guest <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> writes:

    Luke> So, the example could be written like this: pragma
    Luke> import(CPP, GCC3, SET_TIME, "void Time::SetTime(...)");

    Luke> Although, I'm not sure why he would be passing in a pointer
    Luke> to a set time function ;-)

Of course you need at least one pointer - don't forget the hidden C++
*this parameter...

(I am still assuming the function in question is static).
-- 
Brian May <bam@snoopy.apana.org.au>

Re: Calling C++ methods from Ada..

[Georg Bauhaus]

Luke A. Guest <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> wrote:
: On Fri, 03 Dec 2004 09:59:22 +0100, Martin Krischik wrote:
: 
:> The original idea was a C++ parser which generated the bindings. With the
:> complexity of C++ the only way to go. However the parser never
:> happend :-( .
: 
: I wonder how much work it would be to add a switch to the GCC C++ compiler
: to enable generation of GNAT bindings (when --languages has ada passed on
: the configuration command line)?

Maybe there are some traces of this idea in GCC/GNAT for .NET/.GNU/.WHATEVER?
(I'm guessing this because much in .NET is about extensible types with
operations as part of a common language infrasructure. So if GCC is
enabled to build type information for MSIL (?), the information needed to
generate Ada bindings should be right there.)


-- Georg

Re: Calling C++ methods from Ada..

[David Botton]

See Martin Carlisles' A#. It's in there.

http://www.usafa.af.mil/dfcs/bios/mcc_html/a_sharp.html

David Botton


On 2004-12-04 08:22:07 -0500, Georg Bauhaus 
<sb463ba@l1-hrz.uni-duisburg.de> said:

> Luke A. Guest <laguest@n_o_p_o_r_k_a_n_d_h_a_m.abyss2.demon.co.uk> wrote:
> : On Fri, 03 Dec 2004 09:59:22 +0100, Martin Krischik wrote:
> : :> The original idea was a C++ parser which generated the bindings. With the
> :> complexity of C++ the only way to go. However the parser never
> :> happend :-( .
> : : I wonder how much work it would be to add a switch to the GCC C++ compiler
> : to enable generation of GNAT bindings (when --languages has ada passed on
> : the configuration command line)?
> 
> Maybe there are some traces of this idea in GCC/GNAT for .NET/.GNU/.WHATEVER?
> (I'm guessing this because much in .NET is about extensible types with
> operations as part of a common language infrasructure. So if GCC is
> enabled to build type information for MSIL (?), the information needed to
> generate Ada bindings should be right there.)
> 
> 
> -- Georg