Compilation halts on GNAT 3-13p, RedHat6.2 Linux

Compilation halts on GNAT 3-13p, RedHat6.2 Linux

[Frank]

Hi!

The following code causes my compilation to halt (possibly loop forever).
It's beyond doubt that there are errors in the code, but the compilator
doesn't output anything.
My GNAT/gtk is recompiled with s-stratt.ad? for GLADE.
Could there be someone that could try to compile this on a un-GLADE GNAT
3-13p installation,
and give me the result? Please also note which parameters you succeded with
.

I have tried:

gnatgcc -c -O2 pa_tennis_engine.adb
and without options at all.

Frank

----------------------------------------------------------------------------
------------------------------------
package PA_TENNIS_ENGINE is
  pragma Remote_Call_Interface;

  type TYPE_GET_RACKET is access procedure (P_POS : out
PA_POSITION.TYPE_POSITION);
  type TYPE_MOVE_MY_RACKET is access procedure (P_POS : in
PA_POSITION.TYPE_POSITION);

  procedure REGISTER  (P_REQUESTORS_MOVE_RACKET : out TYPE_MOVE_MY_RACKET;
                       P_REQUESTORS_GET_RACKET  : out TYPE_GET_RACKET;
                      P_OPPONENTS_GET_RACKET   : out TYPE_GET_RACKET);

end PA_TENNIS_ENGINE;



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

with TEXT_IO;
package body PA_TENNIS_ENGINE is

  procedure REGISTER (P_REQUESTORS_MOVE_RACKET : out TYPE_MOVE_MY_RACKET;
                      P_REQUESTORS_GET_RACKET  : out TYPE_GET_RACKET;
                      P_OPPONENTS_GET_RACKET   : out TYPE_GET_RACKET)
  is
  begin
      P_OPPONENTS_GET_RACKET := GET_L_RACKET'Access;
  end REGISTER;


end PA_TENNIS_ENGINE;

Re: Compilation halts on GNAT 3-13p, RedHat6.2 Linux

[Mark Johnson]

Frank wrote:

> Hi!
>
> The following code causes my compilation to halt (possibly loop forever).
> It's beyond doubt that there are errors in the code, but the compilator
> doesn't output anything.
> My GNAT/gtk is recompiled with s-stratt.ad? for GLADE.
> Could there be someone that could try to compile this on a un-GLADE GNAT
> 3-13p installation,
> and give me the result? Please also note which parameters you succeded with
> .

Hmm. Very odd. GNAT 3.15w goes into an infinite loop with the code you
provided.

You can compile the spec w/o any problems [other than the obvious errors].
  gcc -c -O2 pa_tennis_engine.ads

You can replace the assignment statement in the body w/ a null; & it compiles
[with errors].
  gcc -c -O2 pa_tennis_engine.adb
  pa_tennis_engine.ads:6:01: "PA_POSITION" is undefined (more references
follow)

If you generate a dummy procedure for Get_L_Racket such as...
    procedure Get_L_Racket(P_Pos : out PA_POSITION.Type_Position) is
    begin
       null;
    end;
you get the following error messages...
  gcc -c -O2 pa_tennis_engine.adb
  pa_tennis_engine.adb:20:45: prefix must statically denote a remote subprogram

  pa_tennis_engine.adb:20:45: not subtype conformant with declaration at
unknown location
  pa_tennis_engine.adb:20:45: functions can only match functions
  pa_tennis_engine.ads:6:01: "PA_POSITION" is undefined (more references
follow)

Once you did that & then comment it out, you still get the same set of error
messages.

Again, very odd - obviously you are exercising a piece of code that doesn't get
tested very often :-).

Needless to say - the code is broke. I was looking through the ARM E.2.3 & it
looks like you are violating one of those restrictions. The errors above may
give you a hint as to which one is being violated.
  --Mark
PS: The last one of these I noticed & forwarded to ACT, they had a fix quite
quickly. Of course, you need support to get the updated version in a timely
manner.

Re: Compilation halts on GNAT 3-13p, RedHat6.2 Linux

[Frank]

Hi!

Have you installed Glade, and recompiled with s-stratt.ads/s-stratt.adb when
trying this?

Thank you for help,
Frank

Re: Compilation halts on GNAT 3-13p, RedHat6.2 Linux

[Frank]

Hi!
Inspired by your hacking, I hacked some more:-)
Regarding the result:
>
> If you generate a dummy procedure for Get_L_Racket such as...
>     procedure Get_L_Racket(P_Pos : out PA_POSITION.Type_Position) is
>     begin
>        null;
>     end;
> you get the following error messages...
>   gcc -c -O2 pa_tennis_engine.adb
>   pa_tennis_engine.adb:20:45: prefix must statically denote a remote
subprogram
>
>   pa_tennis_engine.adb:20:45: not subtype conformant with declaration at
> unknown location
>   pa_tennis_engine.adb:20:45: functions can only match functions
>   pa_tennis_engine.ads:6:01: "PA_POSITION" is undefined (more references
> follow)

I managed to recreate this error also, when I declared the procedure in the
package body and _not_ declaring it in the package specification part.
If I declare it in the spec part, the error "prefix must statically denote a
remote subprogram" disappeared. It is in that way I have programmed it in my
original source code, this is just a lobotomized version of a program that I
am going to send to report@gnat.com to show them a easy example
of my problem.

>
> Once you did that & then comment it out, you still get the same set of
error
> messages.
>

I am not quite sure I understand this paragraph, are you saying that you
commented out a change in the source code, saved and the compilator still
reported the errors from the commented code???

Regarding:

>You can replace the assignment statement in the body w/ a null; & it
compiles
>[with errors].
>  gcc -c -O2 pa_tennis_engine.adb
>  pa_tennis_engine.ads:6:01: "PA_POSITION" is undefined (more references
>follow

I have recreated this situation, further:
My impression is that if I move the declaration of the dummy function
_above_ the procedure REGISTER the compilation-hang disappeares.
Or if there is a declaration of it in the spec part of the package the
problem disappeares.

Frank