Problem with C function pointers

Problem with C function pointers

[Andrew Skiba]

Hello.
I'm writing a thin interface library to DirectFB. This library uses
structures with pointers to functions, emulating C++ vtables in C. The
addresses of those functions are set inside DirectFB, so those are not
valid Ada addresses. So, when I try to call such function (or assign
it to an access function variable), I get

raised CONSTRAINT_ERROR : dfb.adb:171 access check failed

The question is how do I eliminate the check? I'm using GNAT and the
declarations of relevant objects are:

type SetCooperativeLevel_func is access function
(thiz:IDirectFBCore_ptr; level:CooperativeLevel) return DFBResult;
pragma Convention (C, SetCooperativeLevel_func); -- this does not help
type IDirectFBCore is record
  ...
  SetCooperativeLevel:SetCooperativeLevel_func; -- this field
    -- points to some C function inside DirectFB library
  ...
end record;
pragma Convention (C, IDirectFBCore); -- this does not help, too :-(
thiz:IDirectFBCore;
f:SetCooperativeLevel_func;

f:=thiz.SetCooperativeLevel; -- here exception is raised
thiz.SetCooperativeLevel (...); -- or here exception is raised

The situation remains the same if I make Unchecked_Conversion, still I
have to call a function some time, and then the exception will be
raised anyway.

Any suggests will be appreciated.
Regards,
Andrew Skiba

Re: Problem with C function pointers

[Stephen Leake]

skibochka@yahoo.com (Andrew Skiba) writes:

> The question is how do I eliminate the check? 

You don't want to eliminate the check! but you do want to solve the
problem. 

> type SetCooperativeLevel_func is access function
> (thiz:IDirectFBCore_ptr; level:CooperativeLevel) return DFBResult;
> pragma Convention (C, SetCooperativeLevel_func); -- this does not help
> type IDirectFBCore is record
>   ...
>   SetCooperativeLevel:SetCooperativeLevel_func; -- this field
>     -- points to some C function inside DirectFB library
>   ...
> end record;
> pragma Convention (C, IDirectFBCore); -- this does not help, too :-(
> thiz:IDirectFBCore;
> f:SetCooperativeLevel_func;
> 
> f:=thiz.SetCooperativeLevel; -- here exception is raised

Hmm. At this point "thiz" has the value "null". What do you expect?

I suspect you left out some code, that sets "thiz". please post a
small compilable example, if possible.

Have you tried running under gdb, and examining "thiz"?

Just out of curiosity, what does DirectFB do?
-- 
-- Stephe

Re: Problem with C function pointers

[Andrew Skiba]

> Hmm. At this point "thiz" has the value "null". What do you expect?
Well, you guessed! :-)

> I suspect you left out some code, that sets "thiz". please post a
> small compilable example, if possible.
Obviously, the code I left out passed the pointer to "thiz" to a C
function, which filled the rest of the structure, including the
pointers to the functions. But I did an error in passing the pointer
to "thiz", so "thiz" remained "null".

> Have you tried running under gdb, and examining "thiz"?
No, I printed it instead :-) It was "null", as you said.

> Just out of curiosity, what does DirectFB do?
This is a substitution for Windows DirectX on Linux. The FB stands for
Frame Buffer, but DirectFB takes care not only of the screen (frame
buffer), but also sound, keybord, mouse, as you might expect from
DirectX. An interesting example of it is "links" web browser, which
was originally text based, but with DirectFB you can see the graphics
on linux console. Good thing for embedded and mini-distributions,
which have no X.

Thanks again,
Andrew Skiba.

P/S still waiting for someone to tell me "hey, this library was ported
to Ada!"

Re: Problem with C function pointers

[Stephen Leake]

skibochka@yahoo.com (Andrew Skiba) writes:

> > I suspect you left out some code, that sets "thiz". please post a
> > small compilable example, if possible.
> Obviously, 

Well, I have learned from many years of posting here never to assume
the "obvious", but instead to ask for precise details. Often the
problem is that the poster is assuming something they shouldn't (which
was the case here :).

> > Have you tried running under gdb, and examining "thiz"?
> No, I printed it instead :-) It was "null", as you said.

Learn gdb; it's much better than Put_Line.

> > Just out of curiosity, what does DirectFB do?
> This is a substitution for Windows DirectX on Linux. The FB stands for
> Frame Buffer, but DirectFB takes care not only of the screen (frame
> buffer), but also sound, keybord, mouse, as you might expect from
> DirectX. An interesting example of it is "links" web browser, which
> was originally text based, but with DirectFB you can see the graphics
> on linux console. Good thing for embedded and mini-distributions,
> which have no X.

Hmm. So it _doesn't_ run under X Windows? then it is _not_ a
"substitution" for DirectX, since that runs under the equivalent of X
for Windows (which doesn't really have a separate name).

How big is it, compared to X Windows?

Now we have a third target for GtkAda :).

-- 
-- Stephe

Re: Problem with C function pointers

[Andrew Skiba]

After I wrote all answers, I saw the whole message became off-topic.
I'm sorry for that, and I'm ready to continue the duscussion in any
other place, if there are further questions.

> Well, I have learned from many years of posting here never to assume
> the "obvious", but instead to ask for precise details. Often the
> problem is that the poster is assuming something they shouldn't (which
> was the case here :).
I would never have such problem with C++, I mean, I would immediately
know what the problem is, but with Ada I tried to find a black cat in
a dark room, when it wasn't there :)

> > > Have you tried running under gdb, and examining "thiz"?
> > No, I printed it instead :-) It was "null", as you said.
> 
> Learn gdb; it's much better than Put_Line.
I use it for a very long time, may be you heard, that it works for C,
too :) The problem is, DirectFB has issues with debugging on the same
monitor, because it OBVIOUSLY (:-) has exclusive control of the video
card. Usually they are debugging via ssh, or a com-port. For simple
cases put_line isn't that bad.

> Hmm. So it _doesn't_ run under X Windows?
Did I say that? Of course, it runs under X Windows. Try tuxrunner game
for an example what it is. And, there is http://directfb.org/ to find
other cute projects.

> How big is it, compared to X Windows?
Very small. That's the point, after all :-) It does not have to do
much, linux kernel has device /dev/fb (if you configure it right), so
the library only makes memcopy to it. Well, it's oversimplified, but
the principle is right, I hope.

> Now we have a third target for GtkAda :).
Sorry, I did not understand that joke. If you meant to run GTK on
DirectFB, that's done already, as well as mplayer, SDL, FreeVO and so
on. I need it for a project, similar to FreeVO.

Regards,
Andrew Skiba.

Re: Problem with C function pointers

[Stephen Leake]

skibochka@yahoo.com (Andrew Skiba) writes:

> > > > Have you tried running under gdb, and examining "thiz"?
> > > No, I printed it instead :-) It was "null", as you said.
> > 
> > Learn gdb; it's much better than Put_Line.

> I use it for a very long time, may be you heard, that it works for C,
> too :) The problem is, DirectFB has issues with debugging on the same
> monitor, because it OBVIOUSLY (:-) has exclusive control of the video
> card. Usually they are debugging via ssh, or a com-port. For simple
> cases put_line isn't that bad.

Ok, that makes sense. But it might be worth setting up a serial line
terminal so you can run gdb for DirectFB programs.

> > Hmm. So it _doesn't_ run under X Windows? 

> Did I say that? Of course, it runs under X Windows. Try tuxrunner
> game for an example what it is. And, there is http://directfb.org/
> to find other cute projects.

Well now I'm confused. If DirectXB runs under X Windows, why can't
you run gdb in one window, with DirectXB in another?

> > Now we have a third target for GtkAda :).

> Sorry, I did not understand that joke. If you meant to run GTK on
> DirectFB, that's done already, as well as mplayer, SDL, FreeVO and so
> on. I need it for a project, similar to FreeVO.

Ok. Have you tried GtkAda? I'm not sure it's a given that if Gtk
works, GtkAda works. That would be nice to know.

-- 
-- Stephe

Re: Problem with C function pointers

[Ad Buijsen]

Stephen Leake wrote:
> skibochka@yahoo.com (Andrew Skiba) writes:
> 
>
> 
>>Did I say that? Of course, it runs under X Windows. Try tuxrunner
>>game for an example what it is. And, there is http://directfb.org/
>>to find other cute projects.
> 
> 
> Well now I'm confused. If DirectXB runs under X Windows, why can't
> you run gdb in one window, with DirectXB in another?
> 


Ah, semantic confusion...  DirectFB controls the hardware directly, and
there is a special XServer that runs on (top of) DirectFB, so , yes, 
DirectFB can run under X.

  Ad