[SciPy-dev] newcore and gcc4/gfortran (only 3 tests fail now)

[Pearu Peterson]

On Mon, 21 Nov 2005, Arnd Baecker wrote:

> On Mon, 21 Nov 2005, Pearu Peterson wrote:
>
>> On Mon, 21 Nov 2005, Arnd Baecker wrote:
>>
>>> ((Doing the build with
>>>    python setup.py config_fc --fcompiler=gnu95 build
>>>  fails - as expected ))
>>
>> Hmm, why would that be expected?
>
> Well, somehow I remember that gfortran was not considered
> to be ready for prime time - can't confirm this as scipy.org
> seems to be down at the moment. In addition, I might
> be running a version which is not young enough ...

I hear that gfortran is causing problems in OSX systems, here, on debian 
linux box, it behaves better.

>> What errors do you get? Here
>>
>>    python setup.py config_fc --fcompiler=gnu95 build
>>
>> works fine (I just get a segfault in one of cephes tests when building
>> with gfortran).
>
> OK, I should have been more precise: the build works fine.
> But already on import I get:
>
> Python 2.4.1 (#1, Sep 12 2005, 23:33:18)
> [GCC 4.0.2 20050901 (prerelease) (SUSE Linux)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
>>>> import scipy
> Importing test to scipy
> Importing base to scipy
> Importing basic to scipy
> Failed to import basic
> /home/abaecker/BUILDS2/Build_87_gcc4/inst_scipy_newcore/lib64/python2.4/site-packages/scipy/lib/blas/fblas.so:
> undefined symbol: e_wsfe
<snip>

I suspect that your blas/lapack libraries are built with g77 3.x. Try 
adding -lg2c to linkage command or upgrade gfortran.

>
> Running scipy.test(10,10) gives
>
> check_bdtrik (scipy.special.basic.test_basic.test_cephes)
> Segmentation fault
> (the same one you get?)

Yep. I get it also on a 32-bit system.

Pearu