[Insight-developers] test_svd failing on windows

[Luis Ibanez]

Jim,

You are rigth, the test is crashing at run time.
Here is the callstack from the debugger:

_NMSG_WRITE(int 10) line 221
abort() line 44 + 7 bytes
vnl_matrix<std::complex<float> >::assert_finite() line 1064 + 8 bytes
vnl_svd<std::complex<float> >::vnl_svd<std::complex<float> >(const 
vnl_matrix<std::complex<float> > & {...}, double 0.00000000000000) line 86
test_svd_recomposition(const char * 0x0058ccc0, double 
1.0000000000000e-005, std::complex<float> * 0x00000000) line 169 + 16 bytes
test_svd() line 193 + 22 bytes
test_svd(int 0, int 0) line 197 + 21 bytes
main(int 1, char * * 0x00222508) line 100 + 19 bytes
mainCRTStartup() line 338 + 17 bytes
KERNEL32! 77e7eb69()


Luis

======================================================================================

Miller, James V (Research) wrote:

>Yes,
>
>moxel2.crd is building Release. The test just ran for my RelWithDebInfo
>build using .NET.  
>
>It looks (from the Dashboard) like the test is really crashing.
>
>I am looking through the changes I made at the moment.  Most of the 
>changes should only have affected Borland (specific versions of 
>vcl header files for Borland etc.)  The svd code may include a 
>header file that it did not before (vcl_cmath) but otherwise the
>svd code should be unchanged.
>
>Jim
>
>