A bug in Gnat, or in my code?

A bug in Gnat, or in my code?

[Björn Persson]

This program, when compiled with Gnat 3.3.2 and run on Fedora Core 1, 
raises Storage_Error on line 7 (because of a segmentation fault):

procedure Error is                        --  1
                                           --  2
    procedure Proc (NullS : String) is     --  3
       type SA is access String;           --  4
       Point : SA := new String (1 .. 0);  --  5
    begin                                  --  6
       Point.all := NullS;                 --  7
    end Proc;                              --  8
                                           --  9
    S : String := "s";                     -- 10
begin                                     -- 11
    Proc(S(3 .. 1));                       -- 12
end Error;                                -- 13

If I change 3 to 2 on line 12 there is no error. As far as I can see, 
(3..1) should be as legal as (2..1).

Am I right that this is a bug in Gnat?

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Preben Randhol]

Bj�rn Persson <spam-away@nowhere.nil> wrote on 11/03/2004 (08:22) :
> This program, when compiled with Gnat 3.3.2 and run on Fedora Core 1, 
> raises Storage_Error on line 7 (because of a segmentation fault):
> 
> procedure Error is                        --  1
>                                           --  2
>    procedure Proc (NullS : String) is     --  3
>       type SA is access String;           --  4
>       Point : SA := new String (1 .. 0);  --  5
>    begin                                  --  6
>       Point.all := NullS;                 --  7
>    end Proc;                              --  8
>                                           --  9
>    S : String := "s";                     -- 10
> begin                                     -- 11
>    Proc(S(3 .. 1));                       -- 12
> end Error;                                -- 13
> 
> If I change 3 to 2 on line 12 there is no error. As far as I can see, 
> (3..1) should be as legal as (2..1).
> 
> Am I right that this is a bug in Gnat?

I don't get this behaviour in gnat 3.15p so it is probably a gnat 3.3.2
bug. That said what are you trying to do with the code? It looks wierd
to me.

-- 
Preben Randhol -------------- http://www.pvv.org/~randhol/Ada95 --
                 �For me, Ada95 puts back the joy in programming.�

Re: A bug in Gnat, or in my code?

[Georg Bauhaus]

Preben Randhol <randhol+abuse@pvv.org> wrote:
: 
: I don't get this behaviour in gnat 3.15p so it is probably a gnat 3.3.2
: bug.

Fine with GCC 3.1 (1252) on Mac OS X too.

Re: A bug in Gnat, or in my code?

[Björn Persson]

Georg Bauhaus wrote:

> Preben Randhol <randhol+abuse@pvv.org> wrote:
> : 
> : I don't get this behaviour in gnat 3.15p so it is probably a gnat 3.3.2
> : bug.
> 
> Fine with GCC 3.1 (1252) on Mac OS X too.

Okay, seems like it might be a regression then.

I guess I should try a 3.4 snapshot before I file a bug report.

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Björn Persson]

Preben Randhol wrote:

> That said what are you trying to do with the code? It looks wierd
> to me.

Of course that code is useless. That's just a stripped test case. The 
bug occurs in To_Unbounded_String, called from this procedure:

       procedure klyv(argument : in string;
                      namn : out unbounded_string;
                      text : out unbounded_string) is
          use Ada.Strings;
          use ada.strings.fixed;
          pos : natural;
       begin
          pos := index(argument, "=");
          if pos = 0 then
             pos := argument'last + 1;
          end if;
          namn := to_unbounded_string(
                     trim(argument(argument'first .. pos - 1), both));
          text := to_unbounded_string(
                     argument(pos + 1 .. argument'last));
       end klyv;

It splits a string of the format "name=value" into two. If there is no 
equal sign, the whole string goes into Namn, and Text should become an 
empty string. In that case To_Unbounded_String is called with 
Argument(Argument'Last+2..Argument'Last), which triggers the bug.

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Preben Randhol]

In comp.lang.ada, you wrote:
> Preben Randhol wrote:
>
>> That said what are you trying to do with the code? It looks wierd
>> to me.
>
> Of course that code is useless. That's just a stripped test case. The 
> bug occurs in To_Unbounded_String, called from this procedure:
>
>        procedure klyv(argument : in string;
>                       namn : out unbounded_string;
>                       text : out unbounded_string) is
>           use Ada.Strings;
>           use ada.strings.fixed;
>           pos : natural;
>        begin
>           pos := index(argument, "=");
>           if pos = 0 then
>              pos := argument'last + 1;
>           end if;
>           namn := to_unbounded_string(
>                      trim(argument(argument'first .. pos - 1), both));
>           text := to_unbounded_string(
>                      argument(pos + 1 .. argument'last));
>        end klyv;
>
> It splits a string of the format "name=value" into two. If there is no 
> equal sign, the whole string goes into Namn, and Text should become an 
> empty string. In that case To_Unbounded_String is called with 
> Argument(Argument'Last+2..Argument'Last), which triggers the bug.

I see. I think you have a bug in that you forgot to trim the text.
Anyhow I made a test of your program and ran it on gnat 3.3.3 from
Debian and it works there. No crash. So maybe it is only the packaging
of RedHat? A program shouldn't segmentation fault and this is typically
an indication of a packaging/installation error.


   % gnatmake -f -gnatv klyv_test.adb 
   gcc-3.3 -c -gnatv klyv_test.adb

   GNAT 3.3.3 (Debian)
   Copyright 1992-2002 Free Software Foundation, Inc.

   Compiling: klyv_test.adb (source file time stamp: 2004-03-11 14:46:48)
    46 lines: No errors
   gnatbind -x klyv_test.ali
   gnatlink klyv_test.ali

   % ./klyv_test 
   Ada 95 = morsomt => Ada 95&morsomt
   Ada 95 morsomt => Ada 95 morsomt&

Here is my program:

klyv_test.adb
-------------------------------------
with Ada.Text_IO; use Ada.Text_IO;
with Ada.Strings;
with Ada.Strings.Fixed;
with Ada.Strings.Unbounded;  use Ada.Strings.Unbounded; 

procedure klyv_test is

   procedure klyv(argument : in string;
                  namn : out unbounded_string;
                  text : out unbounded_string) is
      use Ada.Strings;
      use ada.strings.fixed;
      pos : natural;
   begin
      pos := index(argument, "=");
      if pos = 0 then
         pos := argument'last + 1;
      end if;
      namn := to_unbounded_string(
                 trim(argument(argument'first .. pos - 1), both));
      text := to_unbounded_string(
                 trim(argument(pos + 1 .. argument'last), both));
   end klyv;

   Hoggestabbe1 : constant String := "Ada 95 = morsomt";
   Hoggestabbe2 : constant String := "Ada 95 morsomt";
   Tekst_foer : Unbounded_String;
   Tekst_etter : Unbounded_String;
begin

   klyv (Hoggestabbe1, Tekst_foer, Tekst_etter);

   Put_Line (Hoggestabbe1 & " => " & 
             To_String (Tekst_foer) & "&" &
             To_String (Tekst_etter));

   klyv (Hoggestabbe2, Tekst_foer, Tekst_etter);

   Put_Line (Hoggestabbe2 & " => " & 
             To_String (Tekst_foer) & "&" &
             To_String (Tekst_etter));


end klyv_test;
-------------------------------------

-- 
Rox-Filer; *the* file manager => http://rox.sf.net/

Re: A bug in Gnat, or in my code?

[Preben Randhol]

In comp.lang.ada, you wrote:
> Preben Randhol wrote:
>
>> That said what are you trying to do with the code? It looks wierd
>> to me.
>
> Of course that code is useless. That's just a stripped test case. The 
> bug occurs in To_Unbounded_String, called from this procedure:
>
>        procedure klyv(argument : in string;
>                       namn : out unbounded_string;
>                       text : out unbounded_string) is
>           use Ada.Strings;
>           use ada.strings.fixed;
>           pos : natural;
>        begin
>           pos := index(argument, "=");
>           if pos = 0 then
>              pos := argument'last + 1;
>           end if;
>           namn := to_unbounded_string(
>                      trim(argument(argument'first .. pos - 1), both));
>           text := to_unbounded_string(
>                      argument(pos + 1 .. argument'last));
>        end klyv;
>
> It splits a string of the format "name=value" into two. If there is no 
> equal sign, the whole string goes into Namn, and Text should become an 
> empty string. In that case To_Unbounded_String is called with 
> Argument(Argument'Last+2..Argument'Last), which triggers the bug.

I see. I think you have a bug in that you forgot to trim the text.
Anyhow I made a test of your program and ran it on gnat 3.3.3 from
Debian and it works there. No crash. So maybe it is only the packaging
of RedHat? A program shouldn't segmentation fault and this is typically
an indication of a packaging/installation error.


   % gnatmake -f -gnatv klyv_test.adb 
   gcc-3.3 -c -gnatv klyv_test.adb

   GNAT 3.3.3 (Debian)
   Copyright 1992-2002 Free Software Foundation, Inc.

   Compiling: klyv_test.adb (source file time stamp: 2004-03-11 14:46:48)
    46 lines: No errors
   gnatbind -x klyv_test.ali
   gnatlink klyv_test.ali

   % ./klyv_test 
   Ada 95 = morsomt => Ada 95&morsomt
   Ada 95 morsomt => Ada 95 morsomt&

Here is my program:

klyv_test.adb
-------------------------------------
with Ada.Text_IO; use Ada.Text_IO;
with Ada.Strings;
with Ada.Strings.Fixed;
with Ada.Strings.Unbounded;  use Ada.Strings.Unbounded; 

procedure klyv_test is

   procedure klyv(argument : in string;
                  namn : out unbounded_string;
                  text : out unbounded_string) is
      use Ada.Strings;
      use ada.strings.fixed;
      pos : natural;
   begin
      pos := index(argument, "=");
      if pos = 0 then
         pos := argument'last + 1;
      end if;
      namn := to_unbounded_string(
                 trim(argument(argument'first .. pos - 1), both));
      text := to_unbounded_string(
                 trim(argument(pos + 1 .. argument'last), both));
   end klyv;

   Hoggestabbe1 : constant String := "Ada 95 = morsomt";
   Hoggestabbe2 : constant String := "Ada 95 morsomt";
   Tekst_foer : Unbounded_String;
   Tekst_etter : Unbounded_String;
begin

   klyv (Hoggestabbe1, Tekst_foer, Tekst_etter);

   Put_Line (Hoggestabbe1 & " => " & 
             To_String (Tekst_foer) & "&" &
             To_String (Tekst_etter));

   klyv (Hoggestabbe2, Tekst_foer, Tekst_etter);

   Put_Line (Hoggestabbe2 & " => " & 
             To_String (Tekst_foer) & "&" &
             To_String (Tekst_etter));


end klyv_test;
-------------------------------------

Preben
-- 
Preben Randhol -------------- http://www.pvv.org/~randhol/Ada95 --
                 �For me, Ada95 puts back the joy in programming.�

Re: A bug in Gnat, or in my code?

[Björn Persson]

Preben Randhol wrote:

> I see. I think you have a bug in that you forgot to trim the text.

No, that's intentional. There might be cases where the value must be 
used as it is, with possible spaces. In fact, I just removed the Trim on 
the name. I improved another part of the program and then the trimming 
became unnecessary.

> Anyhow I made a test of your program and ran it on gnat 3.3.3 from
> Debian and it works there. No crash. So maybe it is only the packaging
> of RedHat? A program shouldn't segmentation fault and this is typically
> an indication of a packaging/installation error.

The Trim you inserted hides the bug, because Trim returns a more 
conventional null string that doesn't cause problems in 
To_Unbounded_String. After I removed it, your program crashes for me.

Did you try my test program on that version of Gnat?

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Preben Randhol]

On 2004-03-11, Bj�rn Persson <spam-away@nowhere.nil> wrote:
> Preben Randhol wrote:
>
>> I see. I think you have a bug in that you forgot to trim the text.
>
> No, that's intentional. There might be cases where the value must be 
> used as it is, with possible spaces. In fact, I just removed the Trim on 
> the name. I improved another part of the program and then the trimming 
> became unnecessary.
>
>> Anyhow I made a test of your program and ran it on gnat 3.3.3 from
>> Debian and it works there. No crash. So maybe it is only the packaging
>> of RedHat? A program shouldn't segmentation fault and this is typically
>> an indication of a packaging/installation error.
>
> The Trim you inserted hides the bug, because Trim returns a more 
> conventional null string that doesn't cause problems in 
> To_Unbounded_String. After I removed it, your program crashes for me.
>
> Did you try my test program on that version of Gnat?
>

% apt-get install gnat-3.3

(I just love the package system in Debian. One command and it is
installed :-) )

Now to the case. I removed both trims and now I get:


   % gnatmake -gnatv -f klyv_test
   gcc-3.3 -c -gnatv klyv_test.adb

   GNAT 3.3.3 (Debian)
   Copyright 1992-2002 Free Software Foundation, Inc.

   Compiling: klyv_test.adb (source file time stamp: 2004-03-11 17:12:00)
    45 lines: No errors
   gnatbind -x klyv_test.ali
   gnatlink klyv_test.ali
   k-083152(randhol:~)% ./klyv_test 
   Ada 95 = morsomt => Ada 95 & morsomt

   raised STORAGE_ERROR : stack overflow (or erroneous memory access)

Do you get a STORAGE_ERROR or do you get a segmentation fault?

-- 
Rox-Filer; *the* file manager => http://rox.sf.net/

Re: A bug in Gnat, or in my code?

[Björn Persson]

Preben Randhol wrote:

> % apt-get install gnat-3.3
> 
> (I just love the package system in Debian. One command and it is
> installed :-) )

Yes, that's nice. Yum does the same in Fedora, (And there's a version of 
Apt for Fedora too.)

> Now to the case. I removed both trims and now I get:
> 
> 
>    % gnatmake -gnatv -f klyv_test
>    gcc-3.3 -c -gnatv klyv_test.adb
> 
>    GNAT 3.3.3 (Debian)
>    Copyright 1992-2002 Free Software Foundation, Inc.
> 
>    Compiling: klyv_test.adb (source file time stamp: 2004-03-11 17:12:00)
>     45 lines: No errors
>    gnatbind -x klyv_test.ali
>    gnatlink klyv_test.ali
>    k-083152(randhol:~)% ./klyv_test 
>    Ada 95 = morsomt => Ada 95 & morsomt
> 
>    raised STORAGE_ERROR : stack overflow (or erroneous memory access)

Okay, then I won't report it to Redhat.

> Do you get a STORAGE_ERROR or do you get a segmentation fault?

You could say I get both. :-) Running from a shell I get the same 
message as you:

    [bjorn@groucho Parafras]$ ./error

    raised STORAGE_ERROR : stack overflow (or erroneous memory access)
    [bjorn@groucho Parafras]$ ./klyv_test
    Ada 95 = morsomt => Ada 95 & morsomt

    raised STORAGE_ERROR : stack overflow (or erroneous memory access)

Running in GDB (inside GPS) I first get "Program received signal 
SIGSEGV, Segmentation fault." If I try to continue from there, 
Storage_Error is raised and the program terminates. So the SIGSEGV 
signal is translated to a Storage_Error.

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Björn Persson]

There is now a bug report in the GCC Bugzilla, if anyone is interested:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14543

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Per Sandberg]

It is fixed in the latest ACT version of GNAT
/Per

Bj�rn Persson wrote:
> There is now a bug report in the GCC Bugzilla, if anyone is interested:
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14543
> 

Re: A bug in Gnat, or in my code?

[Manuel Collado]

Bj�rn Persson wrote:
> There is now a bug report in the GCC Bugzilla, if anyone is interested:
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14543

I've reproduced the bug with the DJGPP port of gcc 3.3.3

F:\PRUEBAS\ADA>gcc --version
gcc.exe (GCC) 3.3.3
Copyright (C) 2003 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

F:\PRUEBAS\ADA>gnatmake null_slice_error
gcc -c null_slice_error.adb
null_slice_error.adb:1:11: warning: file name does not match unit name, 
should be "nuslierr.adb"
gnatbind -x null_slice_error.ali
gnatlink null_slice_error.ali

F:\PRUEBAS\ADA>F:\PRUEBAS\ADA\null_slice_error.exe
Exiting due to signal SIGSEGV
General Protection Fault at eip=000018fa
...

-- 
To reply by e-mail, please remove the extra dot
in the given address:  m.collado -> mcollado

Re: A bug in Gnat, or in my code?

[Florian Villoing]

Bj�rn Persson <spam-away@nowhere.nil> wrote in message news:<3h_3c.85772$dP1.242518@newsc.telia.net>...
> Georg Bauhaus wrote:
> 
> > Preben Randhol <randhol+abuse@pvv.org> wrote:
> > : 
> > : I don't get this behaviour in gnat 3.15p so it is probably a gnat 3.3
>  .2
> > : bug.
> > 
> > Fine with GCC 3.1 (1252) on Mac OS X too.
> 
> Okay, seems like it might be a regression then.
> 
> I guess I should try a 3.4 snapshot before I file a bug report.

It seems to be OK on FreeBSD 5.2 with gcc-3.4 compiled from sources :

villoing:~/test$ gcc -v
Reading specs from /usr/home/villoing/install/gcc/bin/../lib/gcc/i386-unknown-freebsd5.2/3.4.0/specs
Configured with: ../gcc/configure --prefix=/home/villoing/install/gcc
--disable-shared --with-libiconv-prefix=/usr/local/lib
--disable-werror --enable-languages=c,ada
Thread model: posix
gcc version 3.4.0 20040305 (prerelease)



villoing:~/test$ gnatmake -f -gnatv error.adb
gcc -c -gnatv error.adb

GNAT 3.4.0 20040305 (prerelease) Copyright 1992-2004 Free Software
Foundation, Inc.

Compiling: error.adb (source file time stamp: 2004-03-12 16:11:18)
 15 lines: No errors
gnatbind -x error.ali
gnatlink error.ali



villoing:~/test$ ./error
villoing:~/test


Florian

Re: A bug in Gnat, or in my code?

[Björn Persson]

Manuel Collado wrote:

> I've reproduced the bug with the DJGPP port of gcc 3.3.3

Forwarded to GCC-Bugzilla. Further discussion probably fits best there.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14543

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Björn Persson]

Florian Villoing wrote:

> It seems to be OK on FreeBSD 5.2 with gcc-3.4 compiled from sources :

Forwarded to GCC-Bugzilla. Further discussion probably fits best there.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14543

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu

Re: A bug in Gnat, or in my code?

[Björn Persson]

Per Sandberg wrote:

> It is fixed in the latest ACT version of GNAT

Do you know of an earlier ACT release where the bug is present? 
Otherwise it may be that it only occurs in the GCC series.

-- 
Björn Persson

jor ers @sv ge.
b n_p son eri nu