[SciPy-User] recipe for installation on Mac OS X?

[Ralf Gommers]

On Thu, Feb 24, 2011 at 7:59 PM, Ben Willmore <bdeb at willmore.eu> wrote:
> Hi,
>
> I'm trying to install scipy on Mac OS X 10.6 and have not succeeded
> after a fair bit of effort. I wonder if anyone can provide some
> guidance.
>
> I have tried:
>
> 1. Apple's Python 2.6 and the SciPy superpack from
> <http://stronginference.com/scipy-superpack/>
> -- scipy.test('1','10') fails with errors relating to ARPACK, such as [1].
>
Those ARPACK tests are supposed to fail currently on 64-bit, so if
that's the only issue you're fine. The superpack author just should
have marked them as knownfail.


> 2. python.org Python 2.7, gfortran and the compilation instructions
> from <http://www.scipy.org/Installing_SciPy/Mac_OS_X>
> -- scipy.test('1','10') fails; I didn't store the error.
>
Can't help there without more details.

> 3. python.org Python 2.7, and corresponding prebuilt SciPy/NumPy
> binaries from <http://www.scipy.org/Download>
> -- scipy.test('1','10') fails with errors relating to Weave, e.g. [2].
>
That should just work, and I've never seen that error before. Is it
possible you mixed the 32-bit and 64-bit versions of
Python/NumPy/SciPy?

> 4. plus a number of less well defined approaches, none of which has
> resulted in a SciPy that passes tests.
>
> Can anyone tell me what I've done wrong with the above, or provide a
> different strategy for successful installation of a recent SciPy on
> Mac OS X. I am non-naive about compilers etc and willing to put in a
> bit of effort to get this working satisfactorily.
>
Nothing too much you did wrong as far as I can tell. I'd suggest going
with the 2.6 superpacks and python from python.org, or building from
source. In the latter case, just give us a build log, compiler
versions and test errors and we'll go from there.

Cheers,
Ralf

P.S. Macports & friends also (often) works, but it'll be a lot harder
to help you if something goes wrong there.