Stefan Behnel, 18.05.2010 06:59:
> Lisandro Dalcin, 17.05.2010 18:55:
>> On 17 May 2010 13:19, Robert Bradshaw wrote:
>>> On May 16, 2010, at 3:47 PM, Lisandro Dalcin wrote:
>>>> petsc4py would also qualify, but the dependency on numpy and core
>>>> PETSc could make it more cumbersome.
>>>
>>> Could either of these be installed in, and run from, your home
>>> directory? That would make things much easier (no root/admin
>>> privileges required).
>>
>> Of course, they can be installed at any place. However, the idea is to
>> also test against multiple Python versions, right? Then numpy should
>> also be available across Python versions ...
>
> Currently, we just pass tgz archives of the installed build results between
> build jobs. Adding a NumPy install step to that shouldn't be hard. It's
> used in a lot of Cython code, so it certainly qualifies as a standard
> dependency.

I added an intermediate step into the build pipelines of Python 2.4 through 
Py2-trunk that takes the freshly built CPython installation and installs 
"distribute", "pip" and "numpy" into it and then repackages it as tar.gz, 
so that the subsequent Cython build/test jobs can use it. Any other 
packages that should go in?

The problem I'm currently facing with this is that the NumPy tests now 
crash, e.g. here:

https://sage.math.washington.edu:8091/hudson/view/cython-devel/job/cython-devel-tests-py2-trunk-c/270/console

Could someone with NumPy experience take a look?

The corresponding NumPy 1.4.1 build log is here:

https://sage.math.washington.edu:8091/hudson/job/py2-trunk-ext-svn/2/consoleFull

It obviously lacks all external libraries, could that be a problem?

I'll remove the NumPy install for now, so that we get the tests back to 
normal. The installation is done in the script

        /scratch/hudson/Scripts/install_ext_packages.sh

Stefan
_______________________________________________
Cython-dev mailing list
[email protected]
http://codespeak.net/mailman/listinfo/cython-dev

Reply via email to