"Ada Gems": Constructor functions

"Ada Gems": Constructor functions

[Tomek Wałkuski]

Hi,

motivated by this article posted at adacore.com I have written some
code:

---------
t.ads
---------
with Ada.Strings.Unbounded;
use Ada.Strings.Unbounded;

package T is
    type Object (<>) is tagged limited private;

   function Create_Object (Name : in String) return Object;
   function Construct return Object;
   function Name (This : in Object) return String;

   type Object is tagged limited
      record
         Name : Unbounded_String;
      end record;
end T;


---------
t.adb
---------
package body T is
   function Create_Object (Name : in String) return Object is
      return (Name => To_Unbounded_String (Name));
   end Create_Object;

   function Construct return Object is
      return Create_Object (Name => "Object without a name!");
   end Construct;

   function Name (This : in Object) return String is
   begin
      return To_String (This.Name);
   end Name;

   type Object is tagged limited
      record
         Name : Unbounded_String;
      end record;
end T;


---------
test_t.adb
---------
with T;

procedure Test_T is
   Some_Object : constant T.Object := Construct;
begin
   null;
end Test_T;

------------------------------
... and tried to compile it.

(Gentoo Linux, gnat-gpl-4.1.3.2007)

Compiler said:
"
+================GNAT BUG DETECTED==================+
| GPL 2007 (20070402-41) (i686-pc-linux-gnu) Assert_Failure namet.adb:
687  |
(...)
compilation abandoned
gnatmake: "test_t.adb" compilation error
"

Is it REALLY a compiler bug? Or am I only dumb? :) Maybe I'm doing
something, somewhere wrong?

Re: "Ada Gems": Constructor functions

[Jeffrey R. Carter]

Tomek Wa�kuski wrote:
> 
> with Ada.Strings.Unbounded;
> use Ada.Strings.Unbounded;
> 
> package T is
>     type Object (<>) is tagged limited private;
> 
>    function Create_Object (Name : in String) return Object;
>    function Construct return Object;
>    function Name (This : in Object) return String;
> 
>    type Object is tagged limited
>       record
>          Name : Unbounded_String;
>       end record;
> end T;

T has a private type (Object) but no private part.

> package body T is
>    function Create_Object (Name : in String) return Object is
>       return (Name => To_Unbounded_String (Name));
>    end Create_Object;
> 
>    function Construct return Object is
>       return Create_Object (Name => "Object without a name!");
>    end Construct;
> 
>    function Name (This : in Object) return String is
>    begin
>       return To_String (This.Name);
>    end Name;
> 
>    type Object is tagged limited
>       record
>          Name : Unbounded_String;
>       end record;
> end T;

Type Object is repeated here.

> with T;
> 
> procedure Test_T is
>    Some_Object : constant T.Object := Construct;
> begin
>    null;
> end Test_T;

Construct is not directly visible here.

> Is it REALLY a compiler bug? Or am I only dumb? :) Maybe I'm doing
> something, somewhere wrong?

It's really a compiler error. A compiler should be able to handle the 
errors in your code without crashing.

FWIW, I got this with another version:

gnatmake -O2 -gnaton -fstack-check test_t.adb
gcc -c -O2 -gnaton -fstack-check test_t.adb
test_t.adb:4:39: "Construct" is not visible
test_t.adb:4:39: non-visible declaration at t.ads:8
t.ads:11:04: declaration of full view must appear in private part
gnatmake: "test_t.adb" compilation error

-- 
Jeff Carter
"Death awaits you all, with nasty, big, pointy teeth!"
Monty Python & the Holy Grail
20

Re: "Ada Gems": Constructor functions

[Robert A Duff]

Tomek Wa�kuski <tomek.walkuski@gmail.com> writes:

> Is it REALLY a compiler bug? Or am I only dumb? :) Maybe I'm doing
> something, somewhere wrong?

Well, there are several mistakes in the code you posted.  Jeff mentioned
some.  Also, there are a couple of missing "begin"s in function bodies.

But "GNAT BUG DETECTED" really means what it says -- if you make a
mistake, the compiler should tell you about it, rather than crashing
with the bug-box message.  So yes, this is a compiler bug, by
definition.

Are you sure the code you posted is identical to the code that caused
the bug box?

- Bob

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

The "working" code is:

t.ads:
with Ada.Strings.Unbounded;
use Ada.Strings.Unbounded;

package T is
   type Object (<>) is tagged limited private;

   function Create_Object (Name : in String) return Object;
   function Construct return Object;
   function Name (This : in Object) return String;

   private
   type Object is tagged limited
      record
         Name : Unbounded_String;
      end record;
end T;


t.adb:
package body T is
   function Create_Object (Name : in String) return Object is
   begin
      return (Name => To_Unbounded_String (Name));
   end Create_Object;

   function Construct return Object is
   begin
      return Create_Object (Name => "Object without a name!");
   end Construct;

   function Name (This : in Object) return String is
   begin
      return To_String (This.Name);
   end Name;
end T;


test_t.adb:
with T;

procedure Test_T is
   Some_Object : constant T.Object := Construct;
begin
   null;
end Test_T;



------
tomek@genbox [~/Desktop/test_t] $ gnatmake test_t.adb
gnatgcc -c test_t.adb
test_t.adb:4:39: "Construct" is not visible
test_t.adb:4:39: non-visible declaration at t.ads:8
gnatmake: "test_t.adb" compilation error

(and no bug report)

Why "Construct" is not visible?

Re: "Ada Gems": Constructor functions

[Robert A Duff]

Tomek Wa kuski <tomek.walkuski@gmail.com> writes:

>    Some_Object : constant T.Object := Construct;
...
> Why "Construct" is not visible?

You need to say T.Construct, just like you said T.Object.
Or else you need a use clause.

- Bob

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

On 6 Wrz, 15:55, Robert A Duff <bobd...@shell01.TheWorld.com> wrote:
> You need to say T.Construct, just like you said T.Object.
> Or else you need a use clause.

I tried this before:

tomek@genbox [~/Desktop/test_t] $ gnatmake test_t.adb
gnatgcc -c test_t.adb
gnatgcc -c t.adb
+===========================GNAT BUG
DETECTED==============================+
| GPL 2007 (20070402-41) (i686-pc-linux-gnu) Assert_Failure namet.adb:
687  |
| Error detected at t.adb:
4:7                                              |
| Please submit a bug report by email to
report@adacore.com.               |
| GAP members can alternatively use GNAT
Tracker:                          |
| http://www.adacore.com/ section 'send a
report'.                         |
| See gnatinfo.txt for full info on procedure for submitting
bugs.         |
| Use a subject line meaningful to you and us to track the
bug.            |
| Include the entire contents of this bug box in the
report.               |
| Include the exact gcc or gnatmake command that you
entered.              |
| Also include sources listed below in gnatchop
format                     |
| (concatenated together with no headers between
files).                   |
| Use plain ASCII or MIME
attachment.                                      |
+==========================================================================
+

Please include these source files with error report
Note that list may not be accurate in some cases,
so please double check that the problem can still
be reproduced with the set of files listed.

t.adb
t.ads

compilation abandoned
gnatmake: "t.adb" compilation error


----
Submit bug report?

Re: "Ada Gems": Constructor functions

[anon]

Yes, you did find a GNAT BUG!

To find the line numbers and errors try 

gnat1 t.adb -gnatl 2>t.lst
gnat1 test_t.adb -gnatl 2>test_t.lst

But to make a report to Adacore:
First compile using 

gnatmake test_t.adb 1>1 2>2

-- Note: File '2' should have the error info that Adacore is 
--                asking about.


Then send Adacore the following list of files 

attach file => t.ads 

attach file => t.adb

attach file => test_t.adb

attach file => 1

-- Note: Send File 1 only if it has any special info else skip
--

attach file => 2


I use Ada 95 to compile and did not receive the BUG ERROR 
only a few normal compiling errors.

Now, for a compilable corrected version in GNAT Ada 95.  
Did not use GNAT Ada 2005 compiler. 

changes to code:
--- removed  => denote the line is commented out
--- Added    => denote the line is added to source 
other comments added for info only


-----------
-- t.ads --
-----------
with Ada.Strings.Unbounded;
use Ada.Strings.Unbounded;

package T is
--    type Object (<>) is tagged limited private; --------- removed
                                   -------- because can not initial
                                   -------- a limited type in Ada 95

    type Object (<>) is private;  --------- Added

   function Create_Object (Name : in String) return Object;
   function Construct return Object;
   function Name (This : in Object) return String;

private ------------------------------ Added
--   type Object is tagged limited -------- removed
                                   -------- because can not initial
                                   -------- a limited type in Ada 95

   type Object is  ------------------------ Added
      record        
         Name : Unbounded_String;
      end record;
end T;

-----------
-- t.adb --
-----------
package body T is
  function Create_Object (Name : in String) return Object is
    begin  --------- Added
      return (Name => To_Unbounded_String (Name));
    end Create_Object;

  function Construct return Object is
    begin  --------- Added
      return Create_Object (Name => "Object without a name!");
    end Construct;

  function Name (This : in Object) return String is
    begin
      return To_String (This.Name);
    end Name;

---------------------------------------- removed 
--                                 ----- because it conflicts 
--                                 ----- with t.ads version
--   type Object is tagged limited
--      record
--         Name : Unbounded_String;
--      end record;
end T;

----------------
-- test_t.ads --
----------------
with T;
use  T; ------ Added 
        ------ could of added "T." to Construct instead

procedure Test_T is
   Some_Object : constant T.Object := Construct;
begin
   null;
end Test_T;




In <1189015827.384331.17490@k79g2000hse.googlegroups.com>,  =?iso-8859-2?q?Tomek_Wa=B3kuski?= <tomek.walkuski@gmail.com> writes:
>Hi,
>
>motivated by this article posted at adacore.com I have written some
>code:
>
>---------
>t.ads
>---------
>with Ada.Strings.Unbounded;
>use Ada.Strings.Unbounded;
>
>package T is
>    type Object (<>) is tagged limited private;
>
>   function Create_Object (Name : in String) return Object;
>   function Construct return Object;
>   function Name (This : in Object) return String;
>
>   type Object is tagged limited
>      record
>         Name : Unbounded_String;
>      end record;
>end T;
>
>
>---------
>t.adb
>---------
>package body T is
>   function Create_Object (Name : in String) return Object is
>      return (Name => To_Unbounded_String (Name));
>   end Create_Object;
>
>   function Construct return Object is
>      return Create_Object (Name => "Object without a name!");
>   end Construct;
>
>   function Name (This : in Object) return String is
>   begin
>      return To_String (This.Name);
>   end Name;
>
>   type Object is tagged limited
>      record
>         Name : Unbounded_String;
>      end record;
>end T;
>
>
>---------
>test_t.adb
>---------
>with T;
>
>procedure Test_T is
>   Some_Object : constant T.Object := Construct;
>begin
>   null;
>end Test_T;
>
>------------------------------
>.... and tried to compile it.
>
>(Gentoo Linux, gnat-gpl-4.1.3.2007)
>
>Compiler said:
>"
>+================GNAT BUG DETECTED==================+
>| GPL 2007 (20070402-41) (i686-pc-linux-gnu) Assert_Failure namet.adb:
>687  |
>(...)
>compilation abandoned
>gnatmake: "test_t.adb" compilation error
>"
>
>Is it REALLY a compiler bug? Or am I only dumb? :) Maybe I'm doing
>something, somewhere wrong?
>

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

Report sent.

So.. It is not possible to use Ada2005 standard features mentioned in
Ada Gems article with GNAT complier?

Re: "Ada Gems": Constructor functions

[Robert A Duff]

Tomek Wa kuski <tomek.walkuski@gmail.com> writes:

> Report sent.

Thanks.

> So.. It is not possible to use Ada2005 standard features mentioned in
> Ada Gems article with GNAT complier?

Apparently not with the version you are using.  :-(

It does work with the latest version of GNAT, though.

- Bob

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

Got reply:

"
Dear Tomek, the current version of the compiler reports:

 > gnatmake test_t.adb
 > gcc -c test_t.adb
 > test_t.adb:4:40: cannot initialize entities of limited type
 > gnatmake: "test_t.adb" compilation error

which seems expected.
"

Huh?

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

Also, this same code compiled on Windows with GNAT GPL 2007 throws:

> gnatmake test_t.adb
t.adb:4:07: "" is undefined
t.adb:4:07: actual for "From" must be a variable
gnatmake: "c:\documents and settings\tomek\pulpit\test_t\t.adb"
compilation error


There is GNAT bug on every Gentoo box I have got, on Windows I can't
compile it (with... this same compiler) and AdaCore can't reproduce
this bug...

Re: "Ada Gems": Constructor functions

[Georg Bauhaus]

Tomek Wa kuski wrote:
> Also, this same code compiled on Windows with GNAT GPL 2007 throws:
> 
>> gnatmake test_t.adb
> t.adb:4:07: "" is undefined
> t.adb:4:07: actual for "From" must be a variable
> gnatmake: "c:\documents and settings\tomek\pulpit\test_t\t.adb"
> compilation error

But no GNAT Bug Box this time?

 
> There is GNAT bug on every Gentoo box I have got, on Windows I can't
> compile it (with... this same compiler) and AdaCore can't reproduce
> this bug...
> 

Did they mention which compiler they have used when trying to
reproduce the bug box? 

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

On 11 Wrz, 09:51, Georg Bauhaus <bauhaus.rm.t...@maps.futureapps.de>
wrote:
> But no GNAT Bug Box this time?
>
No.

Someone wrote to me, that he managed to compile this code on Windows
XP with this same compiler...

Weird...

> Did they mention which compiler they have used when trying to
> reproduce the bug box?
>
They have said only "current version of the compiler reports".

My friend, also on Gentoo, can reproduce this bug as well (so I have
contacted with Gentoo devs, maybe it is Gentoo issue).

Tomek

Re: "Ada Gems": Constructor functions

[Anh Vo]

On Sep 11, 12:51 am, Georg Bauhaus
<bauhaus.rm.t...@maps.futureapps.de> wrote:
> Tomek Wa kuski wrote:
> > Also, this same code compiled on Windows with GNAT GPL 2007 throws:
>
> >> gnatmake test_t.adb
> > t.adb:4:07: "" is undefined
> > t.adb:4:07: actual for "From" must be a variable
> > gnatmake: "c:\documents and settings\tomek\pulpit\test_t\t.adb"
> > compilation error
>
> But no GNAT Bug Box this time?

I got the same result, no GNAT Bug Box, on Windows using GNAT-2007-
GPL. However, it worked fine on Red Hat with gcc-4.3-20070907
snapshot. One interesting observation, if the return statement is
qualified as shown below, it is compiled on Windows, too.

  return Object'(Name => To_Unbounded_String (Name));

AV

Re: "Ada Gems": Constructor functions

[Tomek Wa kuski]

On 11 Wrz, 23:33, Anh Vo <anhvofrc...@gmail.com> wrote:
> One interesting observation, if the return statement is
> qualified as shown below, it is compiled on Windows, too.
>
>   return Object'(Name => To_Unbounded_String (Name));
>
This compiles on Windows and also on Gentoo box with gnat-
gpl-4.1.3.2007 (which is in fact GNAT GPL 2007 from AdaCore).

This also works:

return New_Object : Object do
    New_Object.Name := To_Unbounded_String (Name);
end return;

Mystery solved? :) So... bug must persist in Ada Gems article :)

Tomek