<div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote">On Tue, Sep 3, 2013 at 5:45 PM, Charles R Harris <span dir="ltr"><<a href="mailto:charlesr.harris@gmail.com" target="_blank">charlesr.harris@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote"><div><div class="h5">On Tue, Sep 3, 2013 at 5:40 PM, Christoph Gohlke <span dir="ltr"><<a href="mailto:cgohlke@uci.edu" target="_blank">cgohlke@uci.edu</a>></span> wrote:<br>

<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>On 9/3/2013 2:51 PM, Charles R Harris wrote:<br>
><br>
><br>
><br>
> On Tue, Sep 3, 2013 at 3:23 PM, Christoph Gohlke <<a href="mailto:cgohlke@uci.edu" target="_blank">cgohlke@uci.edu</a><br>
</div><div>> <mailto:<a href="mailto:cgohlke@uci.edu" target="_blank">cgohlke@uci.edu</a>>> wrote:<br>
><br>
>     On 9/1/2013 9:54 AM, Charles R Harris wrote:<br>
><br>
>         Hi all,<br>
><br>
>         I'm happy to announce the first beta release of Numpy 1.8.0.<br>
>         Please try<br>
>         this beta and report any issues on the numpy-dev mailing list.<br>
><br>
>         Source tarballs and release notes can be found at<br>
</div>>         <a href="https://sourceforge.net/__projects/numpy/files/NumPy/1.__8.0b1/" target="_blank">https://sourceforge.net/__projects/numpy/files/NumPy/1.__8.0b1/</a><br>
>         <<a href="https://sourceforge.net/projects/numpy/files/NumPy/1.8.0b1/" target="_blank">https://sourceforge.net/projects/numpy/files/NumPy/1.8.0b1/</a>>.<br>
<div>>         The Windows<br>
>         and OS X installers will follow when the infrastructure issues<br>
>         are dealt<br>
>         with.<br>
><br>
>         Chuck<br>
><br>
><br>
>     Hello,<br>
><br>
>     I tried numpy-1.8.0.dev-86a6e6c with msvc9 and MKL 11.1 on<br>
>     win-amd64-py2.7. It builds OK but there are 23 test errors and 6<br>
>     failures (attached).<br>
><br>
>     Some 3rd party packages (e.g. scipy, numexpr, pytables, bottleneck,<br>
>     pandas and matplotlib) that were built against numpy-MKL 1.7 fail<br>
>     tests when used with numpy-MKL 1.8. Other packages test OK (e.g.<br>
>     skimage, sklearn, statsmodels, mahotas, pygame). See<br>
</div>>     <<a href="http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7-__numpy-1.8.0.dev-86a6e6c/" target="_blank">http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7-__numpy-1.8.0.dev-86a6e6c/</a><br>


>     <<a href="http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7-numpy-1.8.0.dev-86a6e6c/" target="_blank">http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7-numpy-1.8.0.dev-86a6e6c/</a>>><br>


>     compared to<br>
>     <<a href="http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7/" target="_blank">http://www.lfd.uci.edu/~__gohlke/pythonlibs/tests/__20130902-win-amd64-py2.7/</a><br>
>     <<a href="http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7/" target="_blank">http://www.lfd.uci.edu/%7Egohlke/pythonlibs/tests/20130902-win-amd64-py2.7/</a>>>.<br>
<div>><br>
><br>
>     I have not looked in more detail or at other Python versions yet.<br>
><br>
><br>
> Thanks Christoph,<br>
><br>
> Looks like some work to do. I wonder what is different between windows<br>
> and linux here?<br>
><br>
> Chuck<br>
><br>
<br>
</div>Looks like the fundamental PyArray_PyIntAsIntp function is broken on 64<br>
bit Windows. 64 bit PyLong values are intermediately stored in a 32 bit<br>
C long variable. But maybe I am missing something...<br>
<<a href="https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L729" target="_blank">https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L729</a>><br>


<<a href="https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L767" target="_blank">https://github.com/numpy/numpy/blob/maintenance/1.8.x/numpy/core/src/multiarray/conversion_utils.c#L767</a>><br>


<br></blockquote></div></div><div>My, that does look suspicious. That function is new in 1.8 I believe. Looks like it needs fixing whatever else it fixes.<br><br></div></div></div></div></blockquote><div><br></div><div>BTW, do the tests pass with a 32 build?<br>
<br></div><div>Chuck <br></div></div></div></div>