Re: Ada.Real_Time.Time_Last

[Niklas Holsti <niklas.holsti@tidorum.invalid>]

On 17-11-17 11:20 , Simon Wright wrote:
> Niklas Holsti <niklas.holsti@tidorum.invalid> writes:
>
>> On 17-11-15 16:28 , Simon Wright wrote:
>>> The only use case I can see for this is to allow us to write
>>>
>>>    delay until Ada.Real_Time.Time_Last;
>>>
>>> at the end of our (Ravenscar) main program.

    [snip]

>> (By the way, my Ravenscar programs always end in an eternal loop,
>> never in a delay until Time_Last.)
>
> Like
>
>    loop
>       null;
>    end loop;
>
> ?

No, the environment task becomes one of the working application tasks 
(why waste its stack space?), so more like

    loop
       Wait_For_Trigger_Or_Time;
       Do_Something_Useful;
    end loop;

> I'd have thought this would prevent the RTS from putting the processor
> to sleep, thus wasting energy.

In the only project I've done where the customer asked for sleep mode, 
and we implemented it (that was in C, not Ada) the customer found, 
during their system tests, that periodically entering and leaving sleep 
mode caused supply-voltage transients that coupled into the analog parts 
of the board and perturbed the instrument's ADC. So the SW was changed 
to use an idle loop instead of sleeping. YMMV.

>> (By the further way, it is a pity that Ada does not let Ada.Real_Time
>> provide the attribute functions Time'Min and Time'Max.)
>
> I suppose there'd have to be a way of indicating that a private type was
> scalar.

Long ago, I suggested (on c.l.a) an Ada extension by which a package 
could declare private types using the same classes of types that can be 
used for generic formal types. That would IMO not only be useful for 
programmers, it could formalise some of the informal text in the Ada RM. 
For example, RM A.4.8 defines the type Ada.Direct_IO.Count as

    type Count is range 0 .. implementation defined;

With the suggested extension, this could be written in legal (extended) 
Ada as

    type Count is private range <>;

That form still does not show that Count'First = 0, but with a little 
further extension one could perhaps leave only the upper bound 
unspecified, as in:

    type Count is private range 0 .. <>;

Perhaps it is time for me to revive this proposal...

However, in the case of Ada.Real_Time.Time, it may be too much to 
require that this type should be a scalar one, as it may need a 
combination of range and precision that exceeds that of the largest 
machine scalar.

-- 
Niklas Holsti
Tidorum Ltd
niklas holsti tidorum fi
       .      @       .