elab pragma placement

[Simon Belmont]

Where is the right spot to apply an elaboration pragma to a library level subprogram? GNAT (20210519-103) complains that in the below code C cannot with subprogram A.B because it's not Pure, but it certainly appears to be, and no amount of shuffling around the pragma in B seems to placate it. Playing with the declaration of T seems to change the behavior; is GNAT is just misreporting a different error?

Thanks

-sb


package A is
pragma Pure;
type T is limited null record;
end A;

procedure A.B (x : T) is
pragma Pure; -- is this the right spot?
begin
null;
end A.B;

with A.B; -- error: cannot depend on "B" (wrong categorization)
package C is
pragma Pure;
end C;

[Luke A. Guest]

On 21/01/2023 23:43, Simon Belmont wrote:
> Where is the right spot to apply an elaboration pragma to a library level subprogram? GNAT (20210519-103) complains that in the below code C cannot with subprogram A.B because it's not Pure, but it certainly appears to be,

To be pure it can contain only things that can be pre-elaborated, iirc,
so only types, no functions.

[AdaMagica]

A subprogram is a program unit and may be declared pure. I do not see that A.B performs any of 10.2.1(15ff).

I tried
procedure A.B (x : T) with Pure;
which is accepted together with its body and still got the compiler complaint.

Looks like a compiler problem.

[Rod Kay]

If you are on linux, then you could set the stack size with

$ ulimit -s unlimited
$ launch_my_app



Regards.

[Rod Kay]

Sorry, wrong thread :/ .