[SciPy-Dev] memory corruption when running scipy 0.8 test suite

[Pauli Virtanen]

Wed, 14 Jul 2010 03:47:07 -0700, Christoph Gohlke wrote:
[clip]
> Finally, the "Real-valued Bessel domains", test_twodiags
> (test_linsolve.TestLinsolve), and
> test_linsolve.TestSplu.test_spilu_smoketest errors and failures
> disappear when I run the scipy tests with the -O flag, which tells
> Python to optimize generated bytecode:
> 
> python -O -c"import scipy;scipy.test()" -> 4 failures
> 
> python -c "import scipy;scipy.test()" -> 1 error, 6 failures

It also tells Python to remove all of the "assert" statements, which 
effectively removes some of the tests.

Recommendation for using "assert" is maybe not the best of ideas in Nose, 
and we should probably avoid that...

The airye stuff comes from Amos, and the nan failures either indicate 
issues with compilation of the Fortran codes, or (less likely) in the 
wrappers.

[clip]
> In the process of deleting *.pyo files and reinstalling other MKL
> dependent packages, the mysterious 64-bit ndimage/dsyevr failure also
> vanished.

So SuperLU still fails with MKL, but MKL causes no other additional 
failures?

Could you re-paste the errors+failures that you see in a "clean" 
installation?

Thanks,
Pauli