gsl_ran_discrete_preproc "bug"

Hi,

I ran over a bug in gsl_ran_dicrete_preproc that manifests only on
certain computers. I've attached info and test case backtrack of two
computers where bug manifests. One of them is Versalogics Cobra
(http://www.versalogic.com/products/ds.asp?productid=164#Specs) -
Pioneer 3-DX onboard computer, and other is Toshiba Satelite Pro U300;
you can see that from the dmidecode.txt. Versalogic's Cobra runs
Debian lenny (gsl version 1.11+dfsg-1) and Toshiba Ubuntu 8.10 (gsl
version 1.11-2).

The problem is with bigs and smalls stack (randist/discrete.c).
Although the same code is used when calculating the size of stacks and
when inserting (that's when the problem manifests) in them it seems it
works differently (FPU unit?). I've solved this temporarily with
allocating stacks of fixed size (Kevents).

--

-- 
Srećko Jurić-Kavelj
+385913444837

Sanctus Eusebius Hieronymus: "Parce mihi, Domine, quia dalmata sum!"

_______________________________________________
Bug-gsl mailing list
Bug-gsl <at> gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gsl

Re: gsl_ran_discrete_preproc "bug"

At Mon, 4 May 2009 15:52:49 +0200,
Srećko Jurić-Kavelj wrote:
> The problem is with bigs and smalls stack (randist/discrete.c).
> Although the same code is used when calculating the size of stacks and
> when inserting (that's when the problem manifests) in them it seems it
> works differently (FPU unit?). I've solved this temporarily with
> allocating stacks of fixed size (Kevents).

Thanks for the bug report. As you say, it does appear to be an FPU
problem causing the test to give a different result in the two cases.

I'll modify the code to use a single test.

--

-- 
Brian Gough
(GSL Maintainer)

Support freedom by joining the FSF!
http://www.fsf.org/associate/support_freedom/join_fsf?referrer=37

[bug #26502] abort in gsl_ran_discrete_preproc

URL:
  <http://savannah.gnu.org/bugs/?26502>

                 Summary: abort in gsl_ran_discrete_preproc
                 Project: GNU Scientific Library
            Submitted by: bjg
            Submitted on: Fri 08 May 2009 11:55:52 AM BST
                Category: Runtime error
                Severity: 3 - Normal
        Operating System: 
                  Status: Confirmed
             Assigned to: None
             Open/Closed: Open
                 Release: 1.12
         Discussion Lock: Any

    _______________________________________________________

Details:

Hi,

I ran over a bug in gsl_ran_dicrete_preproc that manifests only on
certain computers. I've attached info and test case backtrack of two
computers where bug manifests. One of them is Versalogics Cobra
(http://www.versalogic.com/products/ds.asp?productid=164#Specs) -
Pioneer 3-DX onboard computer, and other is Toshiba Satelite Pro U300;
you can see that from the dmidecode.txt. Versalogic's Cobra runs
Debian lenny (gsl version 1.11+dfsg-1) and Toshiba Ubuntu 8.10 (gsl

[bug #26503] blas drotm: error in vector size checking

URL:
  <http://savannah.gnu.org/bugs/?26503>

                 Summary: blas drotm: error in vector size checking
                 Project: GNU Scientific Library
            Submitted by: bjg
            Submitted on: Fri 08 May 2009 11:59:35 AM BST
                Category: Runtime error
                Severity: 3 - Normal
        Operating System: 
                  Status: Confirmed
             Assigned to: bjg
             Open/Closed: Open
                 Release: 1.12
         Discussion Lock: Any

    _______________________________________________________

Details:

From: Marco Maggi <mrc.mgg <at> gmail.com>
To: bug-gsl <at> gnu.org
Subject: [Bug-gsl] blas drotm: error in vector size checking
Date: Thu, 30 Apr 2009 11:31:34 +0200

In GSL version 1.12:

int
gsl_blas_drotm (gsl_vector * X, gsl_vector * Y, const double P[])

[bug #26503] blas drotm: error in vector size checking

Update of bug #26503 (project gsl):

                  Status:               Confirmed => Fixed                  
             Open/Closed:                    Open => Closed                 

    _______________________________________________________

Follow-up Comment #1:

fixed by 07806b6858c8ebf866bc1fad024f637d1e2df505

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?26503>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

Re: problem in cubic solver

At Tue, 28 Apr 2009 17:11:31 +0200,
Lorenzo Moneta wrote:
>   This patch fails the current test for polynomial in gsl, however in  
> my opinion, this is acceptable, because the test condition is too  
> strict. With a slight change of the test coefficient, you will have a  
> similar failure also with the current version.

I've applied the patch and rewritten the angular calculation in terms
of R2 / Q3 so that it is based on the same variables as the condition
(R2 < Q3)--this should avoid the argument of acos being out of range.
Thanks for reporting the problem.

--

-- 
Brian Gough

Re: GSL Build on HPUX-11 Problems

At Tue, 21 Apr 2009 16:18:40 -0400,
Warren Gray wrote:
> Perhaps the ./configure tests can be expanded to configure this in
> an improved way?

Although it may not solve your specific problem here, I have checked
in some improvements to the tests for inline.

--

-- 
Brian Gough

Re: gsl_ran_discrete_preproc "bug"

At Mon, 4 May 2009 15:52:49 +0200,
Srećko Jurić-Kavelj wrote:
> The problem is with bigs and smalls stack (randist/discrete.c).
> Although the same code is used when calculating the size of stacks and
> when inserting (that's when the problem manifests) in them it seems it
> works differently (FPU unit?). I've solved this temporarily with
> allocating stacks of fixed size (Kevents).

Could you try the attached patch and see if it fixes the problem (I
think I see how the probelm occurs but I wasn't able to reproduce it
locally on x86 or x86_64).  Thanks.

--

-- 
Brian Gough
(GSL Maintainer)

Support freedom by joining the FSF!
http://www.fsf.org/associate/support_freedom/join_fsf?referrer=37

_______________________________________________
Bug-gsl mailing list
Bug-gsl <at> gnu.org
http://lists.gnu.org/mailman/listinfo/bug-gsl

Re: gsl_ran_discrete_preproc "bug"

Hi,

I've just tested the patch on Versalogic's Cobra (Pioneer 3-DX onboard
comp.) and it works.

As for the sampling results in the test case, they seem to be sound,
but I haven't tested thoroughly.

$ ./randist_test
0 52
1 265
2 152
3 20
4 63
5 145
6 25
7 142
8 94
9 221

Thanks.

2009/5/14 Brian Gough <bjg <at> gnu.org>:
> At Mon, 4 May 2009 15:52:49 +0200,
> Srećko Jurić-Kavelj wrote:
>> The problem is with bigs and smalls stack (randist/discrete.c).
>> Although the same code is used when calculating the size of stacks and
>> when inserting (that's when the problem manifests) in them it seems it
>> works differently (FPU unit?). I've solved this temporarily with
>> allocating stacks of fixed size (Kevents).

Re: gsl_ran_discrete_preproc "bug"

At Fri, 15 May 2009 10:39:55 +0200,
Srećko Jurić-Kavelj wrote:
> I've just tested the patch on Versalogic's Cobra (Pioneer 3-DX onboard
> comp.) and it works.
> 
> As for the sampling results in the test case, they seem to be sound,
> but I haven't tested thoroughly.

Thank you. I've committed that to the repository for the next release.

--

-- 
Brian Gough
(GSL Maintainer)

Support freedom by joining the FSF 
http://www.fsf.org/associate/support_freedom/join_fsf?referrer=37