Defect with formal package parameters

Defect with formal package parameters

[Eric Hughes]

GNAT 2008 (just installed) fails to compile the following example,
which was cut down from existing code to isolate the problem.  This is
the error message.
> foo.ada-spec:16:34: actual for "Operation" in actual instance does not match formal

Changing the syntax to eliminate "others =>", which is what the
commented-out line does, allows the example to compile without error.
Unfortunately, that's not a viable workaround, because the real code
requires named parameters.

At present I am assuming this is a defect with GNAT 2008 rather than
with the Ada 2005 language definition.  I had the identical problem
with GNAT 2007.  I had stopped development back then, a few months
ago, in the apparently-misplaced hope that a new compiler version
would fix this.

I should point out, for those without significant experience in
generic programming, that this defect completely blocks a huge class
of generic layering techniques that are primarily of interest to
library authors.  Code that does not admit significant reconfiguration
does not require this technique.

Eric
=======================================================
package Foo is
   pragma Elaborate_Body( Foo ) ;

   generic
      with procedure Operation is <> ;
   package Signature is end ;

   procedure Operation_Actual is null ;
   package Impl is new Signature( Operation => Operation_Actual ) ;

   generic
      with package S is new Signature( others => <> ) ;
      -- with package S is new Signature( <> ) ;
   package Module is end ;

   package M is new Module( S => Impl ) ;
end ;
=======================================================

Re: Defect with formal package parameters

[Simon Wright]

Eric Hughes <eric.eh9@gmail.com> writes:

> At present I am assuming this is a defect with GNAT 2008 rather than
> with the Ada 2005 language definition.  I had the identical problem
> with GNAT 2007.

You don't mention the Bugzilla (or other) reference you reported this
under?

Re: Defect with formal package parameters

[Georg Bauhaus]

Eric Hughes wrote:
> GNAT 2008 (just installed) fails to compile the following example,
...
> =======================================================
> package Foo is
>    pragma Elaborate_Body( Foo ) ;
> 
>    generic
>       with procedure Operation is <> ;
>    package Signature is end ;
> 
>    procedure Operation_Actual is null ;
>    package Impl is new Signature( Operation => Operation_Actual ) ;
> 
>    generic
>       with package S is new Signature( others => <> ) ;
>       -- with package S is new Signature( <> ) ;
>    package Module is end ;
> 
>    package M is new Module( S => Impl ) ;
> end ;
> =======================================================

Another workaround:

package Foo is
   pragma Elaborate_Body( Foo ) ;

   generic
      with procedure Operation is <> ;
   package Signature is end ;

   procedure Operation is null ;
   package Impl is new Signature ;

   generic
      with package S is new Signature( others => <> ) ;
   package Module is end ;

   package M is new Module( S => Impl ) ;
end ;

Re: Defect with formal package parameters

[Eric Hughes]

Eric Hughes wrote:
>    procedure Operation_Actual is null ;
>    package Impl is new Signature( Operation => Operation_Actual ) ;

On Jun 19, 4:48 pm, Georg Bauhaus <rm.tsoh.plus-
bug.bauh...@maps.futureapps.de> wrote:
> Another workaround:
>
>    procedure Operation is null ;
>    package Impl is new Signature ;

The whole point is that I was stacking these package instantiations
two layers deep, and this would cut out the second layer.

Eric

Re: Defect with formal package parameters

[Eric Hughes]

On Jun 19, 1:49 pm, Simon Wright <simon.j.wri...@mac.com> wrote:
> You don't mention the Bugzilla (or other) reference you reported this
> under?

Please feel free to do so.

Eric