Assigning to dereferenced null access variable

Assigning to dereferenced null access variable

[Vincent Marciante]

In the following code, I expected a constraint_error to occur
after the line containing "???" but it did not occur.  Is that 
correct behavior?

Also, after that line, I expected the code in the two blocks 
to behave the same but that was not the case.  The following is the output that I get from OS/2 GNAT 3.12

abcd_Sequence.The_Characters.all = "abcd"
Null_Characters(1..0) = Null_Characters
Null_Sequence.The_Characters = null.
Null_Sequence.The_Characters(1..0) = Null_Characters
Null_Sequence.The_Characters(1..0) := Null_Characters(1..0) succeeded ?
Null_Sequence.The_Characters(1..0) := Null_Characters caused Constraint_Error ?


with Ada.Text_IO;  Use Ada.Text_IO;

procedure Null_Sequence_Assignment is

    type Characters is array (Positive range <>) of Character;
    type Reference is access Characters;
    type Character_Sequence is
        record
            The_Length : Natural := 0;
            The_Characters : Reference;
        end record;

    Null_Characters : Characters(1..0);
    abcd_Characters : Characters := "abcd";

    Null_Sequence : Character_Sequence;
    abcd_Sequence : Character_Sequence := (4, new Characters'(1=>'a',2=>'b',3=>'c',4=>'d'));

begin

    abcd_Characters(1..0) := Null_Characters(1..0);
    abcd_Characters(1..0) := Null_Characters;
    Null_Characters       := abcd_Characters(1..0);


    abcd_Sequence.The_Characters(1..0) :=  
        Null_Characters;

    abcd_Sequence.The_Characters(1..Null_Sequence.The_Length) :=
        Null_Sequence.The_Characters(1..Null_Sequence.The_Length);


    abcd_Sequence.The_Characters(1..0) :=
        Null_Sequence.The_Characters(1..0);

    abcd_Sequence.The_Characters(1..0) :=
        Null_Characters;

    put_line("abcd_Sequence.The_Characters.all = """
            & String(abcd_Sequence.The_Characters.all) & '"');  

--- ???

    Null_Sequence.The_Characters(1..0) :=   

      -- Why no Constraint_Error here do to
      -- dereferencing of null access value?
      -- Is it due to allowed optimization?

        abcd_Sequence.The_Characters(1..0);                                      
        


    if Null_Characters(1..0) = Null_Characters then
        Put_Line("Null_Characters(1..0) = Null_Characters");
    else
        Put_Line("Null_Characters(1..0) /= Null_Characters");
    end if;


    if Null_Sequence.The_Characters = null then
        Put_Line("Null_Sequence.The_Characters = null.");
    else
        Put_Line("Null_Sequence.The_Characters /= null.");
    end if;

    if Null_Sequence.The_Characters(1..0) = Null_Characters then
        Put_Line("Null_Sequence.The_Characters(1..0) = Null_Characters");
    else
        Put_Line("Null_Sequence.The_Characters(1..0) /= Null_Characters");
    end if;


    begin
        Null_Sequence.The_Characters(1..0) :=
            Null_Characters(1..0);

            Put_Line("Null_Sequence.The_Characters(1..0) := "
                    &"Null_Characters(1..0) succeeded ?");
            -- I thought that dereferencing The_Characters
            -- would have caused Constraint_Error to occur.
    exception
        when Constraint_error =>
            Put_Line("Null_Sequence.The_Characters(1..0) := "
                    &"Null_Characters(1..0) caused Constraint_Error");
    end;  

    
    begin
        Null_Sequence.The_Characters(1..0) := 
            Null_Characters; 

            Put_Line("Null_Sequence.The_Characters(1..0) := "
                    &"Null_Characters succeeded");
    exception
        when Constraint_error =>
            Put_Line("Null_Sequence.The_Characters(1..0) := "
                    &"Null_Characters caused Constraint_Error ?");
            -- I thought, because Null_Characters(1..0) = Null_Characters,
            -- that the assignment would have succeeded, as it did
            -- in the previous block.
    end;  

end;

Simpler Question (was Assigning to dereferenced null access variable)

[Vincent Marciante]

--Okay, I cut down the question to:
--
--Why don't both blocks in the following code behave similarly?
--I get the following output using OS/2 GNAT 3.12p
--
--The Null_Characters(1..0) assignment succeeded
--The Null_Characters assignment caused Constraint_Error
--
--If the above indicates a compiler defect then what is the
--correct behavior?

with Ada.Text_IO; use Ada.Text_IO;
procedure Test_Null_Assignments is

    type Characters is array (Positive range <>) of Character;
    Null_Characters : Characters(1..0);

    type Reference is access Characters;
    type Character_Sequence is
        record
            The_Characters : Reference;
        end record;

    Null_Sequence : Character_Sequence;
begin
    begin
        Null_Sequence.The_Characters(1..0) := Null_Characters(1..0);  
        Put_Line("The Null_Characters(1..0) assignment succeeded");
        --OS/2 GNAT 3.12 succeeds
    exception
        when Constraint_Error => 
        Put_Line("The Null_Characters(1..0) assignment caused
Constraint_Error");
    end;

    begin
        Null_Sequence.The_Characters(1..0) := Null_Characters;  
        Put_Line("The Null_Characters assignment succeeded");
    exception
        when Constraint_Error => 
        Put_Line("The Null_Characters assignment caused
Constraint_Error");
        --OS/2 GNAT 3.12 raises Constraint_Error
    end;
end;

Re: Simpler Question (was Assigning to dereferenced null access variable)

[Vincent Marciante]

Jean-Pierre Rosen wrote:
> 
> "Vincent Marciante" <marciant@li.net> a �crit dans le message news:3A0847ED.49AF@li.net...
> >[...]
> >     begin
> >         Null_Sequence.The_Characters(1..0) := Null_Characters;
> >         Put_Line("The Null_Characters assignment succeeded");
> >     exception
> >         when Constraint_Error =>
> >         Put_Line("The Null_Characters assignment caused
> > Constraint_Error");
> >         --OS/2 GNAT 3.12 raises Constraint_Error
> >     end;
> > end;
> Null_Sequence.The_Characters is a null pointer, there is no array to point to. This is what causes Constraint_Error, nothing to do
> with the assignment.
> 
> --
> ---------------------------------------------------------
>            J-P. Rosen (Rosen.Adalog@wanadoo.fr)
> Visit Adalog's web site at http://pro.wanadoo.fr/adalog

I thought that null_sequence.the_characters being a null pointer
should always cause constraint_error.   I will assume that you 
mean that the fact that no constraint_error ocurred in the first 
block (repeated below) indicates a compiler defect, so, I will send
a report to ACT.   Thanks.

    begin
        Null_Sequence.The_Characters(1..0) := Null_Characters(1..0);  
        Put_Line("The Null_Characters(1..0) assignment succeeded");
        --OS/2 GNAT 3.12 succeeds
    exception
        when Constraint_Error => 
        Put_Line("The Null_Characters(1..0) assignment caused Constraint_Error");
    end;

Re: Simpler Question (was Assigning to dereferenced null access variable)

[Jean-Pierre Rosen]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 766 bytes --]


"Vincent Marciante" <marciant@li.net> a �crit dans le message news: 3A0847ED.49AF@li.net...
>[...]
>     begin
>         Null_Sequence.The_Characters(1..0) := Null_Characters;
>         Put_Line("The Null_Characters assignment succeeded");
>     exception
>         when Constraint_Error =>
>         Put_Line("The Null_Characters assignment caused
> Constraint_Error");
>         --OS/2 GNAT 3.12 raises Constraint_Error
>     end;
> end;
Null_Sequence.The_Characters is a null pointer, there is no array to point to. This is what causes Constraint_Error, nothing to do
with the assignment.

--
---------------------------------------------------------
           J-P. Rosen (Rosen.Adalog@wanadoo.fr)
Visit Adalog's web site at http://pro.wanadoo.fr/adalog

GNAT for OS/2 (was Re: Simpler Question (was Assigning ...))

[Vincent Marciante]

Robert Dewar wrote:
> 
> No need, you are using a pretty old version of the compiler
> (3.12)

If glladly use a more recent public GNAT version if there was 
one for OS/2 :)   Might ACT make one available when GNAT gets 
fully integrated into GCC 3.0?   We OS/2ers are accustomed (and 
greatful) for getting only every other public version ;)

Re: Simpler Question (was Assigning to dereferenced null access variable)

[Vincent Marciante]

Robert Dewar wrote:

> ... it is hard to imagine a real program being
> affected by this :-) 

Actually, similar code was part of the unbounded string 
components that are part of the original Booch Components!
Executing that code when compiled with another compiler 
caused a constraint_error but not with GNAT.

P.S.

A corrected version of the string components is being prepared.

Re: Simpler Question (was Assigning to dereferenced null access variable)

[Robert Dewar]

In article <3A089948.5979@li.net>,
  Vincent Marciante <marciant@li.net> wrote:
> so, I will send a report to ACT

No need, you are using a pretty old version of the compiler
(3.12) and this problem of not checking for a constraint error
in the case of a null pointer pointing to nothing was fixed
a while ago (needless to say this was found in some peculiar
test program, it is hard to imagine a real program being
affected by this :-) Interestingly the ACATS/ACVC tests
did not test this particular case.


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: Simpler Question (was Assigning to dereferenced null access variable)

[Robert Dewar]

In article <3A0A24DF.38D9@li.net>,
  Vincent Marciante <marciant@li.net> wrote:
> Robert Dewar wrote:
>
> > ... it is hard to imagine a real program being
> > affected by this :-)
>
> Actually, similar code was part of the unbounded string
> components that are part of the original Booch Components!
> Executing that code when compiled with another compiler
> caused a constraint_error but not with GNAT.

Yes, but what I am saying is that it is hard to believe that
any code *depends* on getting a constraint_error for assigning
nothing in such a case :-)


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: GNAT for OS/2 (was Re: Simpler Question (was Assigning ...))

[tjerick]

On Thu, 9 Nov 2000 03:46:49, Vincent Marciante <marciant@li.net> 
wrote:

> Robert Dewar wrote:
> > 
> > No need, you are using a pretty old version of the compiler
> > (3.12)
> 
> If glladly use a more recent public GNAT version if there was 
> one for OS/2 :)   Might ACT make one available when GNAT gets 
> fully integrated into GCC 3.0?   We OS/2ers are accustomed (and 
> greatful) for getting only every other public version ;)

Yes, I'd glady accept a new release of the OS/2 version on GNAT 3.1x. 
Please?

Tim Erickson