[SciPy-dev] What instructions should I give scikits.bvp_solver users to get gfortran?

[David Cournapeau]

On Mon, Dec 7, 2009 at 11:07 PM,  <josef.pktd at gmail.com> wrote:
> On Mon, Dec 7, 2009 at 4:56 AM, Pauli Virtanen <pav+sp at iki.fi> wrote:
>> Sun, 06 Dec 2009 19:52:49 -0800, John Salvatier wrote:
>>
>>> Hello,
>>>
>>> I have recently mostly completed by boundary problem solver scikit (
>>> http://pypi.python.org/pypi/scikits.bvp_solver) and someone mentioned to
>>> me that it would be useful to provide instructions for getting gfortran
>>> in my documentation (http://packages.python.org/scikits.bvp_solver/).
>>> What instructions should I give my readers? For mac users, I suspect the
>>> best instructions are to tell them to use the Scipy Superpack, but I
>>> don't know what I should tell Windows users.
>>
>> Windows users are well off with Python(x,y), it comes with GFortran.
>> I believe also the newer (newest?) Mingw's come with GFortran, but I'm
>> not sure right now.
>
> The new release of MingW, that came out last summer, comes with
> gfortran.
>
> Do you know what the compatibility is between g77 and gfortran?
> Can extension for numpy be build with gfortran when numpy and
> lapack have been built with g77? Will we run into similar
> problems on Windows as the switch between g95 and gfortran
> on linux?

Yes, it will most likely be even worse, given that our situation is
already quite messy on windows with mingw / MSVS stuff,

David