[Numpy-discussion] ANN: Numpy 1.8.0 beta 1 release

[Christoph Gohlke]

On 9/3/2013 4:45 PM, Charles R Harris wrote:
>
>
>
> On Tue, Sep 3, 2013 at 5:40 PM, Christoph Gohlke <cgohlke at uci.edu
> <mailto:cgohlke at uci.edu>> wrote:
>
>     On 9/3/2013 2:51 PM, Charles R Harris wrote:
>     >
>     >
>     >
>     > On Tue, Sep 3, 2013 at 3:23 PM, Christoph Gohlke <cgohlke at uci.edu <mailto:cgohlke at uci.edu>
>     > <mailto:cgohlke at uci.edu <mailto:cgohlke at uci.edu>>> wrote:
>     >
>     >     On 9/1/2013 9:54 AM, Charles R Harris wrote:
>     >
>     >         Hi all,
>     >
>     >         I'm happy to announce the first beta release of Numpy 1.8.0.
>     >         Please try
>     >         this beta and report any issues on the numpy-dev mailing list.
>     >
>     >         Source tarballs and release notes can be found at
>     >https://sourceforge.net/__projects/numpy/files/NumPy/1.__8.0b1/
>     >         <https://sourceforge.net/projects/numpy/files/NumPy/1.8.0b1/>.
>     >         The Windows
>     >         and OS X installers will follow when the infrastructure issues
>     >         are dealt
>     >         with.
>     >
>     >         Chuck
>     >
>     >
>     >     Hello,
>     >
>     >     I tried numpy-1.8.0.dev-86a6e6c with msvc9 and MKL 11.1 on
>     >     win-amd64-py2.7. It builds OK but there are 23 test errors and 6
>     >     failures (attached).
>     >
>     >     Some 3rd party packages (e.g. scipy, numexpr, pytables, bottleneck,
>     >     pandas and matplotlib) that were built against numpy-MKL 1.7 fail
>     >     tests when used with numpy-MKL 1.8. Other packages test OK (e.g.
>     >     skimage, sklearn, statsmodels, mahotas, pygame). See
>     >     <http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7-__numpy-1.8.0.dev-86a6e6c/
>     <http://www.lfd.uci.edu/%7E__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7-__numpy-1.8.0.dev-86a6e6c/>
>     >     <http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7-numpy-1.8.0.dev-86a6e6c/>>
>     >     compared to
>     >     <http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7/
>     <http://www.lfd.uci.edu/%7E__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7/>
>     >     <http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7/>>.
>     >
>     >
>     >     I have not looked in more detail or at other Python versions yet.
>     >
>     >
>     > Thanks Christoph,
>     >
>     > Looks like some work to do. I wonder what is different between windows
>     > and linux here?
>     >
>     > Chuck
>     >
>
>     Looks like the fundamental PyArray_PyIntAsIntp function is broken on 64
>     bit Windows. 64 bit PyLong values are intermediately stored in a 32 bit
>     C long variable. But maybe I am missing something...
>     <https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L729>
>     <https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L767>
>
> My, that does look suspicious. That function is new in 1.8 I believe.
> Looks like it needs fixing whatever else it fixes.
>
> Chuck
>

In fact, using a npy_longlong instead of npy_long fixes all numpy test 
errors and failures. But it probably foils the recent optimizations.

Christoph