Re: Combining entry_call, accept_statment and terminate_statment

[tmoran@acm.org]

> task body TT is
> begin
>    loop
>       select
>          p.X;
>          -- dealing with new data
>       or accept Config;
>          -- modify the task
>       or terminate;
>       end select;
>    end loop;
> end TT;
   It seems to me there are three kinds of inputs TT has to deal with:
regular data, please-reconfig, and please-terminate.  From that point of
view, TT could just deal with one supplier of input, rather than the three
different suppliers in the current code:
  task body TT is
  begin
     loop
       p.X(What_To_Do, Possible_Data);
       case What_To_Do is
         when Regular_Data => ... -- dealing with new data
         when Reconfigure  => ... -- modify the task
         when Please_Terminate => exit;
       end case;
     end loop;
  end TT;

The protected object p then has to convey more heterogeneous information,
and may need to use Task_IDs if it needs to deliver different info to
different callers, but it will be a queueing, rather than polling, solution.