Re: Another Ada/ARM Cortex M4F tutorial up, with some interesting results

["G.B." <bauhaus@futureapps.invalid>]

On 22.09.14 07:51, Mike Silva wrote:
> Sure wish I had this 20-30 years ago!
>
> http://www.embeddedrelated.com/showarticle/625.php
>

The source uses GNAT's own 'Enum_Rep, in part for use during
compile time checks.  I wonder if the effect can be achieved
in portable Ada, at very little cost.

The following example code is set up to generate static
failures. Is there a way to reliably get rid of the entities
declared for testing only, as, I guess, they can potentially
make it to object code?

package Plain_Ada is

    type Word is mod 2**16;

    type Lights is new Word;
    Red    : constant Lights :=  2#0001#;
    Green  : constant Lights :=  2#0010#;
    Blue   : constant Lights :=  2#0101#;  -- deliberate
    Yellow : constant Lights :=  2#0100#;

    Flashy : constant Lights := Red
      or Green
      or Blue
      or Yellow;

private
    package Compile_Time_Checks is

       type Yes is new Boolean range True .. True;

       Expected_Bits : constant Yes :=
         Yes (Flashy = 2#0111#
                and
              (Blue or 1) /= 2#0000#);

       No_Overlap : constant := (case 1 is
                                    when Red or Green => 1,
                                    when Red or Blue => 1,
                                    when Red or Yellow => 1,
                                    when Green or Blue => 1,
                                    when Green or Yellow => 1,
                                    when Blue or Yellow => 1,
                                    when others => 1);
    end Compile_Time_Checks;
end Plain_Ada;

One way not have traces of the tests in code is, I suppose,
to put the tests into some pramga that establishes Boolean
context, such as pragma Assert.