GNAT Ada.Numerics.Aux question

GNAT Ada.Numerics.Aux question

[Happy Segfault]

Hi all!  

I'm trying out GtkAda 2.2.0 on my Red Hat 9 system using gcc 3.2.2,
and the testgtk OpenGL demo crashes with a constraint error.  I think
I've tracked down the problem to a call to Ada.Numerics.Aux.Sin, which
is a thin binding to the libm sin function.  What could be going wrong
here?  Why isn't Sin (X) getting the expected value in the program
below?  Oddly, Sqrt (X) seems to work fine.

--
-- test_sin.adb
--
with Ada.Text_IO;
with Ada.Numerics.Aux;
 
procedure test_sin is
  X : Ada.Numerics.Aux.Double;
begin
  X := Ada.Numerics.Aux.Double (2.0);
  X := Ada.Numerics.Aux.Sqrt (X);
  Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
  -- output is ' 1.41421356237309505E+00' which is expected
  X := Ada.Numerics.Aux.Double (30.0 * 3.14159 / 180.0);
  X := Ada.Numerics.Aux.Sin (X);
  -- use the pragma Import (C, Sin, "sin"); from Ada.Numerics.Aux
  Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
  -- output is '0.E+17' but 0.5 is the expected result
  Ada.Text_IO.Put_Line (Float'Image (Float (X)));
  -- raises constraint error
end test_sin;

Thanks!  

John

Re: GNAT Ada.Numerics.Aux question

[Anders Wirzenius]

"Happy Segfault" <happysegfault@yahoo.com> wrote in message
news:ed66a020.0307281917.333e3587@posting.google.com...
> Hi all!
>
> I'm trying out GtkAda 2.2.0 on my Red Hat 9 system using gcc 3.2.2,
> and the testgtk OpenGL demo crashes with a constraint error.  I
think
> I've tracked down the problem to a call to Ada.Numerics.Aux.Sin,
which
> is a thin binding to the libm sin function.  What could be going
wrong
> here?  Why isn't Sin (X) getting the expected value in the program
> below?  Oddly, Sqrt (X) seems to work fine.
>
> --
> -- test_sin.adb
> --
> with Ada.Text_IO;
> with Ada.Numerics.Aux;
>
> procedure test_sin is
>   X : Ada.Numerics.Aux.Double;
> begin
>   X := Ada.Numerics.Aux.Double (2.0);
>   X := Ada.Numerics.Aux.Sqrt (X);
>   Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
>   -- output is ' 1.41421356237309505E+00' which is expected
>   X := Ada.Numerics.Aux.Double (30.0 * 3.14159 / 180.0);
>   X := Ada.Numerics.Aux.Sin (X);
>   -- use the pragma Import (C, Sin, "sin"); from Ada.Numerics.Aux
>   Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
>   -- output is '0.E+17' but 0.5 is the expected result
>   Ada.Text_IO.Put_Line (Float'Image (Float (X)));
>   -- raises constraint error
> end test_sin;
>
> Thanks!
>
> John

Works on Windows XP, Gnat 3.15p:

---------- Run the executable ----------
 1.41421356237309505E+00
 4.99999616987255739E-01
 5.00000E-01
Normal Termination
Output completed (0 sec consumed).

Anders

Re: GNAT Ada.Numerics.Aux question

[arvids lemchens]

Hello John,

happysegfault@yahoo.com am 28.07.03 um 20:17 in comp.lang.ada:
> I'm trying out GtkAda 2.2.0 on my Red Hat 9 system using gcc 3.2.2,

Compilerrelated Problem?

> -- test_sin.adb

ich@pc3:~$ gm test_sin.adb

GNATMAKE 3.14p (20010503) Copyright 1995-2001 Free Software Foundation,
Inc.
  "test_sin.ali" being checked ...
  -> "test_sin.adb" time stamp mismatch
gnatgcc -c -gnatE -gnatf -gnato -gnatwa test_sin.adb
test_sin.adb:2:18: warning: "ada.numerics.aux" is an internal GNAT unit
test_sin.adb:2:18: warning: use of this unit is non-portable and
version-dependent
End of compilation
gnatbind -x test_sin.ali
gnatlink test_sin.ali
ich@pc3:~$ ./test_sin
1.41421356237309505E+00
4.99999616987255739E-01
5.00000E-01
ich@pc3:~$

Wondering if Ada.Numerics.Generic_Elementary_Functions wouldn't be a
better choice than ada.numerics.aux?

And if not, why?

MvfG,

Arvids

Re: GNAT Ada.Numerics.Aux question

[Preben Randhol]

arvids lemchens wrote:
> Hello John,
> 
> happysegfault@yahoo.com am 28.07.03 um 20:17 in comp.lang.ada:
>> I'm trying out GtkAda 2.2.0 on my Red Hat 9 system using gcc 3.2.2,
> 
> Compilerrelated Problem?

Yes gnat 3.2 has problems with sin and cos (at least it had then I tried
it). Use Gnat 3.15p instead until the gcc 3.x becomes stable.

You'll find it here => http://libre.act-europe.fr/GNAT/
-- 
Ada95 is good for you.
http://www.crystalcode.com/codemage/MainMenu/Coding/Ada/IntroducingAda.php

Re: GNAT Ada.Numerics.Aux question

[Preben Randhol]

Happy Segfault wrote:
> Why isn't Sin (X) getting the expected value in the program
> below?

Because 3.2 has a bug with regards to sin and cos. Use Gnat 3.15p until
gcc 3.x becomes stable.

You'll find Gnat 3.15p here: http://libre.act-europe.fr/GNAT/

-- 
Ada95 is good for you.
http://www.crystalcode.com/codemage/MainMenu/Coding/Ada/IntroducingAda.php

Re: GNAT Ada.Numerics.Aux question

[Happy Segfault]

Preben Randhol <randhol+abuse@pvv.org> wrote in message news:<slrnbiccgs.1pf.randhol+abuse@kiuk0152.chembio.ntnu.no>...
> Happy Segfault wrote:
> > Why isn't Sin (X) getting the expected value in the program
> > below?
> 
> Because 3.2 has a bug with regards to sin and cos. Use Gnat 3.15p until
> gcc 3.x becomes stable.
> 
> You'll find Gnat 3.15p here: http://libre.act-europe.fr/GNAT/

Thanks for your response!  I was able to find a bug in GCC's bugzilla
that looks related (although it's not Ada.Numerics.Aux):

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11391

Bad news is that it's confirmed in 3.3.  

John

Re: GNAT Ada.Numerics.Aux question

[Preben Randhol]

Happy Segfault wrote:
> Bad news is that it's confirmed in 3.3.  

Use Gnat 3.15p instead. I guess gcc 3.x will be stable around gcc 4.0

-- 
Ada95 is good for you.
http://www.crystalcode.com/codemage/MainMenu/Coding/Ada/IntroducingAda.php

Re: GNAT Ada.Numerics.Aux question

[Gerald]

happysegfault@yahoo.com (Happy Segfault) wrote in message news:<ed66a020.0307281917.333e3587@posting.google.com>...
> Hi all!  
> 
> I'm trying out GtkAda 2.2.0 on my Red Hat 9 system using gcc 3.2.2,
> and the testgtk OpenGL demo crashes with a constraint error.  I think
> I've tracked down the problem to a call to Ada.Numerics.Aux.Sin, which
> is a thin binding to the libm sin function.  What could be going wrong
> here?  Why isn't Sin (X) getting the expected value in the program
> below?  Oddly, Sqrt (X) seems to work fine.
> 
> --
> -- test_sin.adb
> --
> with Ada.Text_IO;
> with Ada.Numerics.Aux;
>  
> procedure test_sin is
>   X : Ada.Numerics.Aux.Double;
> begin
>   X := Ada.Numerics.Aux.Double (2.0);
>   X := Ada.Numerics.Aux.Sqrt (X);
>   Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
>   -- output is ' 1.41421356237309505E+00' which is expected
>   X := Ada.Numerics.Aux.Double (30.0 * 3.14159 / 180.0);
>   X := Ada.Numerics.Aux.Sin (X);
>   -- use the pragma Import (C, Sin, "sin"); from Ada.Numerics.Aux
>   Ada.Text_IO.Put_Line (Ada.Numerics.Aux.Double'Image (X));
>   -- output is '0.E+17' but 0.5 is the expected result
>   Ada.Text_IO.Put_Line (Float'Image (Float (X)));
>   -- raises constraint error
> end test_sin;
> 
> Thanks!  
> 
> John

You may use gcc-gnat from redhat-rawhide.

Here is what I got on a RH 9.0 system with gcc-gnat from
redhat-rawhide:
gnatmake -O2 test_sin.adb
gcc -c -O2 test_sin.adb
test_sin.adb:2:18: warning: "ada.numerics.aux" is an internal GNAT
unit
test_sin.adb:2:18: warning: use of this unit is non-portable and
version-dependent
gnatbind -x test_sin.ali
gnatlink test_sin.ali

./test_sin 
 1.41421356237309505E+00
 4.99999616987255739E-01
 5.00000E-01

gnatmake -v

GNATMAKE 3.3 20030715 (Red Hat Linux 3.3-14) Copyright 1995-2002 Free
Software Foundation, Inc.

..

you have to install a new binutils package from rawhide too.

Best, Gerald