Re: [Bug-gsl] multilarge_nlinear test failures for Solaris x86/SPARC 32/64-bit with libgsl 2.2.1 and 2.3

[Patrick Alken]

Hello,

  Your proposed fix looks fine to me. I've updated the git repository.

Patrick

On 01/10/2017 02:46 PM, Rich Burridge wrote:
> Hi,
>
> I'm updating the version of libgsl in Solaris and am seeing test failures
> for the multilarge_nlinear tests for both 32-bit and 64-bit on both x86
> and SPARC platforms, for both libgsl version 2.2.1 and 2.3.
>
> Here's the x86 / 32-bit failures:
>
> ...
>
> FAIL: test
> ==========
>
> trs = levenberg-marquardt
> trs = levenberg-marquardt+accel
> trs = dogleg
> trs = double-dogleg
> trs = steihaug-toint
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05 
> expected)
> [8929]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected) [8930]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/weighted/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05
> expected) [8933]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/weighted/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected)
> [8934]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05 
> expected)
> [9437]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected) [9438]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/weighted/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05
> expected) [9441]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=levenberg/weighted/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected)
> [9442]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/powell_badly_scaled 
>
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05 
> expected)
> [9945]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/powell_badly_scaled 
>
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected) [9946]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/weighted/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05
> expected) [9949]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/weighted/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected)
> [9950]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/powell_badly_scaled 
>
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05 
> expected)
> [10453]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/powell_badly_scaled 
>
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected) 
> [10454]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/weighted/powell_badly_scaled 
> i=0 (1.09815933187144788e-05 observed vs 1.09815932969975976e-05
> expected) [10457]
> FAIL:
> trust-region/steihaug-toint/solver=cholesky/scale=more/weighted/powell_badly_scaled 
> i=1 (9.10614672185895024 observed vs 9.10614673986700218 expected)
> [10458]
> FAIL test (exit status: 1)
>
> The others are similar.
>
>
> If I adjust the tolerance in the test_powell3.c test program with the
> following change:
>
> --- gsl-2.2.1/multilarge_nlinear/test_powell3.c.orig    2017-01-10 
> 10:33:08.905770897 +0000
> +++ gsl-2.2.1/multilarge_nlinear/test_powell3.c 2017-01-10 
> 10:47:34.330655639 +0000
> @@ -2,7 +2,7 @@
>  #define powell3_P         2
>
>  static double powell3_x0[powell3_P] = { 0.0, 1.0 };
> -static double powell3_epsrel = 1.0e-10;
> +static double powell3_epsrel = 1.0e-8;
>
>  static double powell3_J[powell3_N * powell3_P];
>
>
> then the tests successfully pass.
>
> I can easily patch our version of libgsl for this, but if this is an
> acceptable change, it would be nice if you could make a similar 
> adjustment
> to the master sources.
>
> Thanks.