[SciPy-user] Core dump and import errors during scipy.test()

James Turner jturner@gemini....
Thu Jul 31 17:06:50 CDT 2008


> #0  0xb61359c0 in dqawfe_ (f=0xb612a6e8 <quad_function>, a=0xbfff94a0,
>     omega=0xbfff94c0, integr=0xbfff94a8, epsabs=0xbfff94d8, 
> limlst=0xbfff94b4,
>     limit=0xbfff94b8, maxp1=0xbfff94bc, result=0xbfff94d0, 
> abserr=0xbfff94c8,
>     neval=0xbfff9484, ier=0xbfff9488, lst=0xbfff948c, alist=0x88024a8,
>     blist=0x8802640, rlist=0x8805698, elist=0x8805830, iord=0x8719b70,
>     nnlog=0x87566b8, chebmo=0x8802f80) at 
> scipy/integrate/quadpack/dqawfe.f:267
> 267        10 l = dabs(omega)

I had a look at the file scipy/integrate/quadpack/dqawfe.f and
realized that whilst "dabs()" is a double-precision built-in
function, it is being assigned to an integer, "l" on line 267.
I wondered if the core dump is just due to their different
lengths, so I changed "l" to be declared as double precision and
now that crash seems to have gone away, but I'm getting another
one from the same _vq.so library:

Loaded symbols for 
/astro/iraf/i686/gempylocal/lib/python2.5/site-packages/scipy/cluster/_vq.so
#0  gesdd_ (min_lwork_=0xbfff6af0, max_lwork_=0xbfff6af0, prefix=0x1, m=0x5,
     n=0xbfff6b20, compute_uv_=0x0, __g77_length_prefix=-1249493092)
     at scipy/linalg/src/calc_lwork.f:39
39      MNTHR = INT( MINMN*11.0D0 / 6.0D0 )

The last lines from test(verbosity=2) are:
Check the behavior when the inputs and outputs are multidimensional. ... ok
Make sure that appropriate exceptions are raised when invalid valuesFloating 
exception (core dumped)

(...seems like they weren't?)

Are these kinds of test crashes a common occurence? It seems like
they should be if they're declaration errors? Or does no-one else
build the official public SciPy from source?

I still have to look into my other problem with the imports, but
it sounds like that one is already fixed in the latest SVN anyway.

Thanks!

James.



More information about the SciPy-user mailing list