[Ada] Fix crash on instantiation in separate package body

[Eric Botcazou]

This is an old regression present on all active branches for a quite unusual 
setup where a generic package is declared and instantiated in a subunit, and
which comes from a thinko in the Has_Body.Find_Body routine of Sem_Elab.

Tested on x86-64/Linux, applied on all the active branches.


2026-03-23  Eric Botcazou  <ebotca...@adacore.com>

        PR ada/124607
        * sem_elab.adb (Has_Body.Find_Body): Fix thinko.


2026-03-23  Eric Botcazou  <ebotca...@adacore.com>

        * gnat.dg/generic_inst20.ads, gnat.dg/generic_inst20.adb: New test.
        * gnat.dg/generic_inst20-sub.adb: New helper.

-- 
Eric Botcazou
diff --git a/gcc/ada/sem_elab.adb b/gcc/ada/sem_elab.adb
index daabcd6a2ee..5e3823c3649 100644
--- a/gcc/ada/sem_elab.adb
+++ b/gcc/ada/sem_elab.adb
@@ -9890,7 +9890,7 @@ package body Sem_Elab is
                --  The corresponding subunit was previously loaded
 
                if Present (Lib_Unit) then
-                  return Lib_Unit;
+                  return Proper_Body (Unit (Lib_Unit));
 
                --  Otherwise attempt to load the corresponding subunit
 

package Generic_Inst20 is
   pragma Elaborate_Body;
   package Sub is
      function F return Boolean;
   end Sub;
end Generic_Inst20;

-- { dg-do compile }

package body Generic_Inst20 is
   package body Sub is separate;
end Generic_Inst20;

-- { dg-excess-errors "cannot generate code" }

separate (Generic_Inst20)

package body Sub is
   function F return Boolean is
   begin
      return True;
   end F;
   generic
   package G is
   end G;
   package I is new G;
end Sub;