Storage_Error: mine's or gnat's fault?

Storage_Error: mine's or gnat's fault?

[Alejandro R. Mosteo]

I'm seeing this (sanitized) stack trace when using the 
*Ada.Containers.Holders* package (both 2015/2016 version):

--8<--
raised STORAGE_ERROR : stack overflow or erroneous memory access

Ada.Tags.Displace at s-stoele.adb:97
Rx.Producers.Holders.Definites.Reference at a-coinho.adb:379
Rx.Examples.Strings.Subscribe at rx-observable.adb:33
Rx.Examples.Strings."&" at rx-observable.adb:81
Rx.Examples.Basic at rx-examples-basic.adb:14
Main at b__rx-examples-basic.adb:345
--8<--

Since my code has not a single pointer, "new", unchecked operation or 
access I feel the responsibility might not be mine... Also the offending 
line is a bit of unfathomable (to me) Address arithmetic. I'd post a 
sample but the code is spread among several files right now. I could try 
to synthesize a self-contained example if it seems a legit bug in the 
sages' opinion.

In the meantime I'm going to try with an indefinite list as container.

Álex.

Re: Storage_Error: mine's or gnat's fault?

[Alejandro R. Mosteo]

On 03/06/16 23:01, Alejandro R. Mosteo wrote:
> I'm seeing this (sanitized) stack trace when using the
> *Ada.Containers.Holders* package (both 2015/2016 version):
>
> --8<--
> raised STORAGE_ERROR : stack overflow or erroneous memory access
>
> Ada.Tags.Displace at s-stoele.adb:97
> Rx.Producers.Holders.Definites.Reference at a-coinho.adb:379
> Rx.Examples.Strings.Subscribe at rx-observable.adb:33
> Rx.Examples.Strings."&" at rx-observable.adb:81
> Rx.Examples.Basic at rx-examples-basic.adb:14
> Main at b__rx-examples-basic.adb:345
> --8<--
>
> Since my code has not a single pointer, "new", unchecked operation or
> access I feel the responsibility might not be mine... Also the offending
> line is a bit of unfathomable (to me) Address arithmetic. I'd post a
> sample but the code is spread among several files right now. I could try
> to synthesize a self-contained example if it seems a legit bug in the
> sages' opinion.
>
> In the meantime I'm going to try with an indefinite list as container.

Using the Indefinite_Doubly_Linked_List the problem is equally suspect 
(this is code from a-cidlli.adb):

    function First_Element (Container : List) return Element_Type is
    begin
       if Checks and then Container.First = null then
          raise Constraint_Error with "list is empty";
       end if;

       return Container.First.Element.all; -- <-- RAISES
    end First_Element;

I'll try to produce a small trigger.

>
> Álex.

Re: Storage_Error: mine's or gnat's fault?

[Jeffrey R. Carter]

On 06/03/2016 02:43 PM, Alejandro R. Mosteo wrote:
> 
>    function First_Element (Container : List) return Element_Type is
>    begin
>       if Checks and then Container.First = null then
>          raise Constraint_Error with "list is empty";
>       end if;
> 
>       return Container.First.Element.all; -- <-- RAISES
>    end First_Element;

What is Checks? If it is/returns False when the list is empty, you'll try to
dereference null.

-- 
Jeff Carter
"I wave my private parts at your aunties."
Monty Python & the Holy Grail
13

Re: Storage_Error: mine's or gnat's fault?

[Jeffrey R. Carter]

On 06/03/2016 03:45 PM, Jeffrey R. Carter wrote:
> On 06/03/2016 02:43 PM, Alejandro R. Mosteo wrote:
>>
>>    function First_Element (Container : List) return Element_Type is
>>    begin
>>       if Checks and then Container.First = null then
>>          raise Constraint_Error with "list is empty";
>>       end if;
>>
>>       return Container.First.Element.all; -- <-- RAISES
>>    end First_Element;
> 
> What is Checks? If it is/returns False when the list is empty, you'll try to
> dereference null.

My mistake. That's the code for the container, not your code.

-- 
Jeff Carter
"I wave my private parts at your aunties."
Monty Python & the Holy Grail
13

Re: Storage_Error: mine's or gnat's fault?

[Simon Wright]

"Alejandro R. Mosteo" <alejandro@mosteo.com> writes:

> Using the Indefinite_Doubly_Linked_List the problem is equally suspect
> (this is code from a-cidlli.adb):
>
>    function First_Element (Container : List) return Element_Type is
>    begin
>       if Checks and then Container.First = null then
>          raise Constraint_Error with "list is empty";
>       end if;
>
>       return Container.First.Element.all; -- <-- RAISES
>    end First_Element;

For info, in Ada.Containers.Helpers.Generic_Implementation,
in a-conhel.ads,

      Checks : constant Boolean := Container_Checks'Enabled;

And, using Gautier's example, on OS X, neither FSF GCC 6.1.0 nor GNAT
GPL 2016 show the problem (a-cidlli.adb is the same in both compilers
bar the runtime exception, so they are clearly of a similar vintage).

Re: Storage_Error: mine's or gnat's fault?

[Alejandro R. Mosteo]

On 04/06/16 20:50, Simon Wright wrote:
> "Alejandro R. Mosteo" <alejandro@mosteo.com> writes:
>
>> Using the Indefinite_Doubly_Linked_List the problem is equally suspect
>> (this is code from a-cidlli.adb):
>>
>>     function First_Element (Container : List) return Element_Type is
>>     begin
>>        if Checks and then Container.First = null then
>>           raise Constraint_Error with "list is empty";
>>        end if;
>>
>>        return Container.First.Element.all; -- <-- RAISES
>>     end First_Element;
>
> For info, in Ada.Containers.Helpers.Generic_Implementation,
> in a-conhel.ads,
>
>        Checks : constant Boolean := Container_Checks'Enabled;
>
> And, using Gautier's example, on OS X, neither FSF GCC 6.1.0 nor GNAT
> GPL 2016 show the problem (a-cidlli.adb is the same in both compilers
> bar the runtime exception, so they are clearly of a similar vintage).

Thanks for the info, although I don't think I'm suppressing any checks, 
but I will double-check (ha).

Álex.