Re: getting started with Spark; generic queue package

[Stephen Leake <stephen_leake@stephe-leake.org>]

On Monday, July 8, 2019 at 8:48:37 AM UTC-7, Anh Vo wrote:
> On Friday, July 5, 2019 at 5:51:29 PM UTC-7, Stephen Leake wrote:
> > I'm finally getting around to trying Spark, mostly because it finally has some support for access types. For now, I'm using GNAT Community 2018; I'll upgrade to 2019 soon.
> > 
> > To start, I'm trying to add contracts to an existing bounded queue package:
> > 
> > generic
> >    type Item_Type is private;
> > package SAL.Gen_Bounded_Definite_Queues
> >   with Spark_Mode
> > is
> >    pragma Pure;
> > 
> >    type Queue_Type (Size : Positive) is tagged private;
> >    --  Size is maximum number of items in the queue.
> >    --  Tagged to allow Object.Method syntax.
> > 
> >    function Get_Overflow_Handling (Queue : in Queue_Type) return Overflow_Action_Type;
> >    procedure Set_Overflow_Handling (Queue : in out Queue_Type; Handling : in Overflow_Action_Type);
> >    --  See Add for meaning of Overflow_Handling. Default is Error.
> > 
> >    procedure Clear (Queue : in out Queue_Type)
> >    with Post => Queue.Count = 0;
> >    --  Empty Queue of all items.
> > 
> >    function Count (Queue : in Queue_Type) return Natural;
> >    --  Returns count of items in the Queue
> > 
> >    function Is_Empty (Queue : in Queue_Type) return Boolean
> >    with Post => (if Is_Empty'Result then Queue.Count = 0);
> >    --  Returns true if no items are in Queue.
> > 
> >    function Is_Full (Queue : in Queue_Type) return Boolean
> >    with Post => (if Is_Full'Result then Queue.Count = Queue.Size);
> >    --  Returns true if Queue is full.
> > 
> >    function Remove (Queue : in out Queue_Type) return Item_Type with
> >      Pre => Queue.Count > 0,
> >      Post => Queue.Count = Queue.Count'Old - 1;
> >    --  Remove head item from Queue, return it.
> >    --
> >    --  Raises Container_Empty if Is_Empty.
> > 
> >    --  other stuff deleted
> > private
> > 
> >    type Item_Array_Type is array (Positive range <>) of Item_Type;
> > 
> >    type Queue_Type (Size : Positive) is tagged record
> >       Overflow_Handling : Overflow_Action_Type := Error;
> > 
> >       Head  : Natural := 0;
> >       Tail  : Natural := 0;
> >       Count : Natural := 0;
> >       Data  : Item_Array_Type (1 .. Size);
> >       --  Add at Tail + 1, remove at Head. Count is current count;
> >       --  easier to keep track of that than to compute Is_Empty for
> >       --  each Add and Remove.
> >    end record;
> > 
> > end SAL.Gen_Bounded_Definite_Queues;
> > 
> > 
> > with SAL.Gen_Bounded_Definite_Queues;
> > package Prove_Integer_Queues is new SAL.Gen_Bounded_Definite_Queues (Integer);
> > 
> > 
> > Running gnatprove on this gives info messages like:
> > 
> > prove_integer_queues.ads:2:01: info: in instantiation at sal-gen_bounded_definite_queues.ads:36
> > prove_integer_queues.ads:2:01: info: implicit function contract not available for proof ("Count" might not return)
> > 
> > Is there any way to turn those off? I don't see such an option in 'gnatprove -help', nor in the Spark 2014 user guide (not I have read every page yet; I did go thru the tutorial).
> > 
> > Worse, in the gnatprove.out file, it says:
> > 
> >   Prove_Integer_Queues.Add at sal-gen_bounded_definite_queues.ads:66, instantiated at prove_integer_queues.ads:2 skipped
> > 
> > and so on, for every subprogram.
> > 
> > Is there a way to find out why it was skipped? And what do I need to do to fix this?
> > 
> > The body file has 'with Spark_Mode' on the package.
> > 
> > In general, is it useful to use Spark on generic packages? Section 4.3 in the Spark user guide says:
> > 
> >    Note that a generic package instance is considered to be declared at its   
> >    instantiation point.  For example, a generic package cannot be both 
> >    marked SPARK_Mode and instantiated in a subprogram body.
> > 
> > But I think that only applies if I run gnatprove on that instantiation; if I only run it on Prove_Integer_Queues, it should be ok. On the other hand, does that mean other instantiations of Gen_Bounded_Definite_Queues are "proved" in any sense?
> > 
> > Most of my packages rely on Ada.Finalization, which is not currently supported in Spark. Is there any plan to add that?
> > 
> > -- Stephe
> 
> There is a dedicated mailing list for SPARK. The mailing address is spark2014-discuss@lists.adacore.com. You will get a quick reply from the experts in this list.

ok, I'll post there.

> By the way, I tried to download your SAL. So, I can check it out with GNAT Community 2019. But, the server was not available/failed. 

I assume that's from http://stephe-leake.org/ada/sal.html; apparently I got halfway thru my most recent release process and stopped. Fixed now.

--  Stephe