Installation problem with Numpy 1.3 on Windows AMD64
Hi! I just upgraded to Python 2.6.2 (from 2.5) on Windows AMD64 in order to use Numpy 1.3 for AMD64 and got the following error: - pythonw.exe has stopped working Numpy was installed both per-machine and per-user but the error persists. Python 2.6.2 works without Numpy. Any ideas? Dinesh
On Sat, May 30, 2009 at 8:02 AM, Dinesh B Vadhia <dineshbvadhia@hotmail.com> wrote:
Hi! I just upgraded to Python 2.6.2 (from 2.5) on Windows AMD64 in order to use Numpy 1.3 for AMD64 and got the following error:
- pythonw.exe has stopped working
Numpy was installed both per-machine and per-user but the error persists. Python 2.6.2 works without Numpy.
The 64 bits build is experimental on windows - there is one fundamental issue which I have not been able to nail down yet. It may be a numpy bug, a python bug, or a mingw-w64 bug, I am not sure yet. But it means that under some conditions, numpy crashes before import. If you need a working 64 bits numpy, the best bet is to build it by yourself using the MS compilers (the 64 bits compilers are available for free if you install the platform SDK 6.0a or later). But then you won't be able to build scipy on top of it unless you manage to use a 64 bits fortran compiler. David
re: "But it means that under some conditions, numpy crashes before import." It it helps the debugging, I have a standard Windows 64-bit configuration. Please let me know when this build is fixed ... Cheers Dinesh From: David Cournapeau Sent: Saturday, May 30, 2009 2:14 AM To: Discussion of Numerical Python Subject: Re: [Numpy-discussion] Installation problem with Numpy 1.3 onWindows AMD64 On Sat, May 30, 2009 at 8:02 AM, Dinesh B Vadhia <dineshbvadhia@hotmail.com> wrote:
Hi! I just upgraded to Python 2.6.2 (from 2.5) on Windows AMD64 in order to use Numpy 1.3 for AMD64 and got the following error:
- pythonw.exe has stopped working
Numpy was installed both per-machine and per-user but the error persists. Python 2.6.2 works without Numpy.
The 64 bits build is experimental on windows - there is one fundamental issue which I have not been able to nail down yet. It may be a numpy bug, a python bug, or a mingw-w64 bug, I am not sure yet. But it means that under some conditions, numpy crashes before import. If you need a working 64 bits numpy, the best bet is to build it by yourself using the MS compilers (the 64 bits compilers are available for free if you install the platform SDK 6.0a or later). But then you won't be able to build scipy on top of it unless you manage to use a 64 bits fortran compiler. David
participants (2)
-
David Cournapeau
-
Dinesh B Vadhia