access-to-classwide has non-static size?

access-to-classwide has non-static size?

[Simon Wright]

I'm creating a PO which has a member which is a bounded vector of
pointer-to-classwide,

   type Event_Base is abstract tagged limited private;
   type Event_P is access all Event_Base'Class;

...

   package Event_Queues is new Ada.Containers.Bounded_Vectors
     (Index_Type   => Positive,
      Element_Type => Event_P);

...

   protected type Dispatchable_Events is
      ...
   private
      ...
      The_Events : Event_Queues.Vector (Capacity => Capacity);

and when compiling with the Ravenscar profile (GNAT GPL 2014, GCC 4.9.1,
GCC 5.0.0 20150208) the compiler says

   warning: component has non-static size
   warning: creation of protected object of type "Dispatchable_Events"
    will violate restriction No_Implicit_Heap_Allocations

Is this (the non-static size part!) right?

I tried declaring Event_P "with Size => 64"; no difference.

I think, studying the output of -gnatR, a pointer-to-classwide is a thin
pointer, i.e. 4 bytes on arm-eabi).

Re: access-to-classwide has non-static size?

[J-P. Rosen]

Le 27/03/2015 09:26, Simon Wright a écrit :
>    protected type Dispatchable_Events is
>       ...
>    private
>       ...
>       The_Events : Event_Queues.Vector (Capacity => Capacity);
> 
> and when compiling with the Ravenscar profile (GNAT GPL 2014, GCC 4.9.1,
> GCC 5.0.0 20150208) the compiler says
> 
>    warning: component has non-static size
>    warning: creation of protected object of type "Dispatchable_Events"
>     will violate restriction No_Implicit_Heap_Allocations
> 
> Is this (the non-static size part!) right?

It's not the component of the Vector (Event_P) which is non static,
it's the component of the PO (i.e. The_Events).

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr

Re: access-to-classwide has non-static size?

[Simon Wright]

Simon Wright <simon@pushface.org> writes:

>    type Event_Base is abstract tagged limited private;
>    type Event_P is access all Event_Base'Class;
>
> ...
>
>    package Event_Queues is new Ada.Containers.Bounded_Vectors
>      (Index_Type   => Positive,
>       Element_Type => Event_P);
>
> ...
>
>    protected type Dispatchable_Events is
>       ...
>    private
>       ...
>       The_Events : Event_Queues.Vector (Capacity => Capacity);
>
> and when compiling with the Ravenscar profile (GNAT GPL 2014, GCC 4.9.1,
> GCC 5.0.0 20150208) the compiler says
>
>    warning: component has non-static size
>    warning: creation of protected object of type "Dispatchable_Events"
>     will violate restriction No_Implicit_Heap_Allocations
>
> Is this (the non-static size part!) right?

If I declare an Event_Queue.Vector at library level rather than in a PO
I don't get the warnings. So this is a GNAT problem, maybe?

Re: access-to-classwide has non-static size?

[Simon Wright]

"J-P. Rosen" <rosen@adalog.fr> writes:

> Le 27/03/2015 09:26, Simon Wright a écrit :
>>    protected type Dispatchable_Events is
>>       ...
>>    private
>>       ...
>>       The_Events : Event_Queues.Vector (Capacity => Capacity);
>> 
>> and when compiling with the Ravenscar profile (GNAT GPL 2014, GCC 4.9.1,
>> GCC 5.0.0 20150208) the compiler says
>> 
>>    warning: component has non-static size
>>    warning: creation of protected object of type "Dispatchable_Events"
>>     will violate restriction No_Implicit_Heap_Allocations
>> 
>> Is this (the non-static size part!) right?
>
> It's not the component of the Vector (Event_P) which is non static,
> it's the component of the PO (i.e. The_Events).

Thanks, J-P. I need to do the allocation explicitly, then. I was hoping
to be able to do it at elaboration.

Re: access-to-classwide has non-static size?

[Simon Wright]

Simon Wright <simon@pushface.org> writes:

> If I declare an Event_Queue.Vector at library level rather than in a PO
> I don't get the warnings. So this is a GNAT problem, maybe?

No, it's a Simon problem :-)

Re: access-to-classwide has non-static size?

[J-P. Rosen]

Le 27/03/2015 09:53, Simon Wright a écrit :
>> It's not the component of the Vector (Event_P) which is non static,
>> > it's the component of the PO (i.e. The_Events).
> Thanks, J-P. I need to do the allocation explicitly, then. I was hoping
> to be able to do it at elaboration.

You didn't show where Capacity was declared. If it's a variable, it's
clearly non static. Did you try it with a constant?

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr

Re: access-to-classwide has non-static size?

[Simon Wright]

"J-P. Rosen" <rosen@adalog.fr> writes:

> Le 27/03/2015 09:53, Simon Wright a écrit :
>>> It's not the component of the Vector (Event_P) which is non static,
>>> > it's the component of the PO (i.e. The_Events).
>> Thanks, J-P. I need to do the allocation explicitly, then. I was hoping
>> to be able to do it at elaboration.
>
> You didn't show where Capacity was declared. If it's a variable, it's
> clearly non static. Did you try it with a constant?

It's a constraint to the PO (sorry I left that out).

I've worked out how to do it at elaboration now:

   The_Instance_Events : access Event_Queues.Vector :=
     new Event_Queues.Vector (Capacity => Capacity);

(compiled but not yet executed - more challenges still to overcome!)

Re: access-to-classwide has non-static size?

[J-P. Rosen]

Le 27/03/2015 15:18, Simon Wright a écrit :
> I've worked out how to do it at elaboration now:
> 
>    The_Instance_Events : access Event_Queues.Vector :=
>      new Event_Queues.Vector (Capacity => Capacity);

So, you are avoiding No_Implicit_Heap_Allocation by having an explicit
heap allocation... ;-). Generally, you have No_Implicit_Heap_Allocation
if you want to be certain to have no dynamic allocation at all.

BTW, you started this thread explaining that you were aiming for the
Ravenscar profile, but No_Implicit_Heap_Allocation is not part of Ravenscar.

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr

Re: access-to-classwide has non-static size?

[Egil H H]

On Friday, March 27, 2015 at 3:58:42 PM UTC+1, J-P. Rosen wrote:
> but No_Implicit_Heap_Allocation is not part of Ravenscar.
>

According to the RM, it is:
http://www.adaic.org/resources/add_content/standards/12rm/html/RM-D-13.html

Re: access-to-classwide has non-static size?

[sbelmont700]

On Friday, March 27, 2015 at 10:18:51 AM UTC-4, Simon Wright wrote:
> 
> I've worked out how to do it at elaboration now:
> 
>    The_Instance_Events : access Event_Queues.Vector :=
>      new Event_Queues.Vector (Capacity => Capacity);
> 
> (compiled but not yet executed - more challenges still to overcome!)

Don't forget the 2012 'bounded' vector, which I presume would be acceptable for your situation ("Bounded vector objects should be implemented without implicit pointers or dynamic allocation.")

-sb

Re: access-to-classwide has non-static size?

[J-P. Rosen]

Le 27/03/2015 16:08, Egil H H a écrit :
> On Friday, March 27, 2015 at 3:58:42 PM UTC+1, J-P. Rosen wrote:
>> but No_Implicit_Heap_Allocation is not part of Ravenscar.
>>
> 
> According to the RM, it is:
> http://www.adaic.org/resources/add_content/standards/12rm/html/RM-D-13.html
> 
Oops, I missed that one. Strange, since Ravenscar is supposed to address
only concurrency issues, and this one is not related to concurrency.

-- 
J-P. Rosen
Adalog
2 rue du Docteur Lombard, 92441 Issy-les-Moulineaux CEDEX
Tel: +33 1 45 29 21 52, Fax: +33 1 45 29 25 00
http://www.adalog.fr

Re: access-to-classwide has non-static size?

[Simon Wright]

sbelmont700@gmail.com writes:

> On Friday, March 27, 2015 at 10:18:51 AM UTC-4, Simon Wright wrote:
>> 
>> I've worked out how to do it at elaboration now:
>> 
>>    The_Instance_Events : access Event_Queues.Vector :=
>>      new Event_Queues.Vector (Capacity => Capacity);
>> 
>> (compiled but not yet executed - more challenges still to overcome!)
>
> Don't forget the 2012 'bounded' vector, which I presume would be
> acceptable for your situation ("Bounded vector objects should be
> implemented without implicit pointers or dynamic allocation.")

It is the 2012 bounded vector.

There are issues;

   with Ada.Containers.Bounded_Vectors;
   package V is new Ada.Containers.Bounded_Vectors
     (Positive, Some_Access_Type);

is treated by GNAT GPL 2014 and FSF GCC 4.9.1 as if it violated
No_Implicit_Heap_Allocation (and in fact the generated code does indeed
call _gnat_malloc()). GCC 5.0.0 doesn't do this.

Still having similar problems with a big object with POs and tasks. You
would think that if the compiler was going to generate a .ali which says
the unit violates a restriction then it would also generate a warning
message, so I guess this is another example of the same bug as above -
or similar, anyway.