On Mon, Feb 4, 2013 at 9:39 PM, Nathaniel Smith <njs@pobox.com> wrote:
On Mon, Feb 4, 2013 at 12:36 PM, Matthew Brett <matthew.brett@gmail.com> wrote:
> Hi,
>
> On Mon, Feb 4, 2013 at 12:27 PM, Ondřej Čertík <ondrej.certik@gmail.com> wrote:
>> On Sun, Feb 3, 2013 at 2:57 AM, David Cournapeau <cournape@gmail.com> wrote:
>>> On Sun, Feb 3, 2013 at 12:28 AM,  <josef.pktd@gmail.com> wrote:
>>>> On Sat, Feb 2, 2013 at 6:14 PM, Matthew Brett <matthew.brett@gmail.com> wrote:
>>>>> Hi,
>>>>>
>>>>> I see there is no Windows 64 bit installer for the 1.7 rc1.
>>>>
>>>> related:
>>>> Is there any chance to get newer mingw or mingw-w64 support "soonish"?
>>>
>>> The problem has no solution until we can restrict support to windows 7
>>> and above. Otherwise, any acceptable solution would require user to be
>>> an admin.
>>
>> The installer is built with this VM/scripts:
>>
>> https://github.com/certik/numpy-vendor
>>
>> currently the VM itself is 32 bit. I think that might be upgraded to 64bit,
>> and maybe it's possible to use 64 bit Wine:
>>
>> http://wiki.winehq.org/Wine64
>>
>> but then we would need to figure out how to use Mingw with 64 bits.
>>
>> I would be very happy to accept patches to the above repository.
>>
>> Alternatively, if the actual Windows 64bit machine would have to be used,
>> is there any way to automate the process? Would you compile it from command line
>> (cmd.exe), just like I do in Wine? I would much prefer if we can figure out
>> how to do this in Wine, so that the process can be automated and other people
>> can easily reproduce it.
>
> I wonder whether getting ming64 to work on 64 bit Wine is too hard to
> get working before the release?  I often can't get 32-bit Wine
> working, and we've apparently got problems with mingw64 on native
> windows.
>
> As a short term fix, how about an Amazon image with the Windows 64 bit
> compilers on it?

-1 on providing an "official" solution which will require admin rights for the produced installer and not work for scipy.

>
> Or can Christophe Gohlke help us out here?

As a temporary measure it might make sense to just deem the cgolke
builds official and upload them to the usual places -- or at least, it
seems like it might make sense to me, but it depends on what
Christophe thinks :-).

I'm +0 on that. MSVC + MKL is currently the only real option for 64-bit binaries, so providing Christoph's installers as "official" could make sense. On the other hand, Christoph has a matching set of other packages on his site, so users will be better off being redirected there than just grabbing only a numpy installer from SF and not finding a scipy one there.

Ralf