Re: [Bug-gsl] Question on version 2.3 test outputs

[Patrick Alken]

Hi, thanks for the report. These arent serious failures so you can use
the library safely. I will adjust the tests so hopefully they will pass
for you in the future

On 12/12/2016 12:55 AM, Weber wrote:
> Hi,
>
> I just was trying out the new GSL 2.3 release and am seeing some test
> failures. I've included the output from these tests below. Here are
> the specifics:
>
> Hardware: Desktop PC, Intel Core i7 (6-cores), 24GB RAM.
>
> OS: Windows 7 64-bit
>
> GSL Builds: Cygwin (32-bit), native Windows (32-bit and 64-bit).
>
> These are local builds -- not binaries downloaded from the internet.
> Which test fails depends on whether the build is 32 or 64-bit bits --
> the same test fails with both the 32-bit Cygwin and 32-bit native
> Windows builds.
>
> The 64-bit build fails the multifit test, and the 32-bit builds fail
> the multifit_nlinear test. All of the other tests (forty-something of
> them) are passing. I have given no command line options to any of the
> tests -- don't know if that's required or not.
>
> I also have similar working builds of GSL versions 1.8, 2.0, 2.1 and
> 2.2 -- no test failures with them.
>
> Let me know if that's not enough information. I have working Debug
> builds for both 32 and 64-bit native Windows so I can set breakpoints
> and all that stuff if it would help for me to do any looking.
>
> -- aweatherguy --
>
> ------------------------------------------------------------------------
>
>
> ======== GSL 2.3 x64 Build (Win7 native) ==============
>
> Starting test-multifit...
> FAIL: shaw gcv: n=20 p=20 lambda (1.2492178093196711e-005 observed vs
> 1.2492178094905104e-005 expected) [21908]
> GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
>
> ======= GSL 2.3 x86 Build  (Win7 native) ==============
>
> Starting test-multifit_nlinear...
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/box3d
> did not converge, status=exceeded max number of iterations [39959]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/weighted/box3d
> did not converge, status=exceeded max number of iterations [39963]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/box3d
> did not converge, status=exceeded max number of iterations [40457]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/weighted/box3d
> did not converge, status=exceeded max number of iterations [40461]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/box3d
> did not converge, status=exceeded max number of iterations [42066]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/weighted/box3d
> did not converge, status=exceeded max number of iterations [42070]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/box3d
> did not converge, status=exceeded max number of iterations [42564]
> FAIL:
> trust-region/levenberg-marquardt+accel/scale=more/solver=svd/fdfvv/weighted/box3d
> did not converge, status=exceeded max number of iterations [42568]
> GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
> trs = levenberg-marquardt
> solver = cholesky
> solver = qr
> solver = svd
> trs = levenberg-marquardt+accel
> solver = cholesky
> solver = qr
> solver = svd
> trs = dogleg
> solver = cholesky
> solver = qr
> solver = svd
> trs = double-dogleg
> solver = cholesky
> solver = qr
> solver = svd
> trs = 2D-subspace
> solver = cholesky
> solver = qr
> solver = svd
>