Re: test release 1.11.90

["Brian Gladman" <brg at gladman dot plus dot com>]

----- Original Message ----- 
From: "Brian Gough" <bjg@gnu.org>
To: <gsl-discuss@sourceware.org>
Sent: Tuesday, December 09, 2008 10:57 PM
Subject: test release 1.11.90


> There is a test release for gsl-1.12 at
>
>  ftp://alpha.gnu.org/gnu/gsl/gsl-1.11.90.tar.gz
>  ftp://alpha.gnu.org/gnu/gsl/gsl-1.11.90.tar.gz.sig
>
> Please try it out and report any problems here.

The results with Microsoft Visual Studi 2008 Professional running on Vista 
Ultimate x64 with both the Microsoft and Intel compilers (v11.0) are as 
follows:

1. linalg\test.c is not C89 compliant as it contains declarations in the 
function body of vector_alloc():

gsl_vector *
vector_alloc (size_t n)
{
 size_t p[5] = {3, 5, 7, 11, 13};
 static size_t k = 0;

 size_t stride = p[k];
 k = (k + 1) % 5;

 gsl_block * b = gsl_block_alloc (n * stride); 
<<<<<<<<<<<
 gsl_vector * v = gsl_vector_alloc_from_block (b, 0, n, stride); 
<<<<<<<<<<<
 v->owner = 1;
 return v;
}

2. The Microsoft Compiler -- with one exception, all tests pass for both 
static and DLL libraries for both 32 and 64 bit compilation.   The exception 
is that the file fft.c fails to compile correctly in both 32 and 64 bit mode 
when optimisation is turned on. This bug has been reported to Microsoft.

3 The Intel Comipler (v11) --  with one exception, all tests pass for both 
static and DLL libraries for both 32 and 64 bit compilation. The exception 
is that testode reports errors in the 64-bit build:

FAIL: bsimp linear step(0) (0.57999999999999996 observed vs 
0.78000000000000003 expected) [49]
FAIL: bsimp exponential step(0) (14.879731724872837 observed vs 
14.881219772446466 expected) [50]
FAIL: bsimp cosine-sine step(0) (0.36235775447667362 observed vs 
0.36142553436718422 expected) [51]
FAIL: bsimp cosine-sine step(1) (0.93203908596722629 observed vs 
0.93240097764180585 expected) [52]
FAIL: bsimp classic_stiff step(1) (0 observed vs -9.9895000171712489e-005 
expected) [54]
FAIL: bsimp evolve_apply reached maxiter [155]
FAIL: bsimp linear[0,4] evolve(0) (1 observed vs 9 expected) [156]
FAIL: bsimp evolve_apply reached maxiter [157]
FAIL: bsimp exp[0,2] evolve(0) (1 observed vs 7.3890560989306504 expected) 
[158]
FAIL: bsimp sine[0,2] evolve(0) (-1.#IND observed vs -0.41614683654714241 
expected) [159]
FAIL: bsimp sine[0,2] evolve(1) (-1.#IND observed vs 0.90929742682568171 
expected) [160]
FAIL: bsimp sine[0,2] w/errors evolve(0) (-1.#IND observed 
vs -0.41614683654714241 expected) [161]
FAIL: bsimp sine[0,2] w/errors evolve(1) (-1.#IND observed vs 
0.90929742682568171 expected) [162]
FAIL: bsimp evolve_apply reached maxiter [163]
FAIL: bsimp stiff[0,1] evolve(0) (1 observed vs 0.73575888234288467 
expected) [164]
FAIL: bsimp stiff[0,1] evolve(1) (0 observed vs -0.36787944117144233 
expected) [165]
FAIL: bsimp evolve_apply reached maxiter [166]
FAIL: bsimp stiff[0,5] evolve(0) (1 observed vs 0.013475893998170934 
expected) [167]

It also hangs before completing.  I have not yet had the time to look into 
this failure in any detail.

This report requires minor changes to some GSL files to adjust to 
differences on Windows.

   best regards and my thanks to all GSL contributors.

         Brian Gladman