Gnat error msg meaning?

Gnat error msg meaning?

[tmoran]

Perhaps I haven't recovered from the eggnog yet, but why does Gnat give
an error message, what does it mean, and what should I do?

unique_str.ads:11:23: Only static constraints allowed for parent
discriminants in the partial view

relevant source:

package root is
  type String_Counts is range 0 .. 50_000;

  type String_Character_Counts is range 0 .. 100_000_000;

  type Store_Type(Max_Strings : String_Counts;
                  Max_Characters : String_Character_Counts;
                  Task_Safe : Boolean)
    is new Ada.Finalization.Controlled with private;
  ...

package unique_str is
  type Store_Type(Max_Strings : root.String_Counts;
                  Max_Characters : root.String_Character_Counts;
                  Task_Safe : Boolean;
                  Comparison : Comparison_Map_Ptr)
           is new root.Store_Type(Max_Strings, Max_Characters, Task_Safe)
--                    ^ error location
           with private;

Re: Gnat error msg meaning?

[tmoran]

> the language is requiring you to specify actual values
> for the discriminants (Max_Strings, Max_Characters, Task_Safe)
  I don't think so.  I give them actual values when declaring an object
of that type.
  I'm thinking now it might be a Gnat bug since two other compilers seem
to have no complaints.  Anybody seen anything similar?

Re: Gnat error msg meaning?

[Simon Wright]

tmoran@acm.org writes:

> package unique_str is
>   type Store_Type(Max_Strings : root.String_Counts;
>                   Max_Characters : root.String_Character_Counts;
>                   Task_Safe : Boolean;
>                   Comparison : Comparison_Map_Ptr)
>            is new root.Store_Type(Max_Strings, Max_Characters, Task_Safe)

             is new root.Store_Type

>            with private;

compiles OK with GCC 3.2.

Re: Gnat error msg meaning?

[tmoran]

> >            is new root.Store_Type(Max_Strings, Max_Characters, Task_Safe)
>
>              is new root.Store_Type
>
> >            with private;
>
> compiles OK with GCC 3.2.
  Interesting.  Dropping mention of the parent's constraints until the
full, private, declaration, does seem to solve the problem with Gnat.  One
of the other compilers doesn't seem to care either way, but the third one
points out RM 3.7(13) "The parent subtype shall be constrained;".   Who's
right?
  In the meantime, sounds like I need two different versions of this
spec for two different compilers. :(