Re: ANN: Ada 2005 Math Extensions, 20100810 release

[Simon Wright <simon@pushface.org>]

I was thinking about the next release, and contemplating

   --  Obtain the eigenvalues of a non-symmetric real matrix.
   function Eigenvalues (A : Real_Matrix) return Complex_Vector;

It's easy enough to implement this in
Ada.Numerics.Generic_Complex_Arrays.Extensions, but it seems slightly
more natural to put it into Ada.Numerics.Generic_Real_Arrays.Extensions.

However, I'm puzzled about how to get Complex_Vector in there, because
it's declared in Generic_Complex_Arrays (and of course you'd want to use
the same instantiation throughout, not a local one).

Starting from the declaration of Generic_Complex_Arrays, I arrived at

   with Ada.Numerics.Generic_Complex_Arrays;
   with Ada.Numerics.Generic_Complex_Types;

   generic
      with package Complex_Types is new Ada.Numerics.Generic_Complex_Types
        (Real);
      with package Complex_Arrays is new Ada.Numerics.Generic_Complex_Arrays
        (Generic_Real_Arrays, Complex_Types);            -- <<<<<<<<<<<<<<<<<<
      use Complex_Arrays;
   package Ada.Numerics.Generic_Real_Arrays.Extensions is

      function Eigenvalues (A : Real_Matrix) return Complex_Vector;

   end Ada.Numerics.Generic_Real_Arrays.Extensions;

but GNAT complains about Generic_Real_Arrays where I've indicated,
saying it expects a package instance to instantiate a formal.

I'd been hoping that -- in the context of a child of
Generic_Real_Arrays -- the name would mean 'the current instantiation',
but clearly not (GNAT GPL 2010, GCC 4.5.0).

Any thoughts?