[SciPy-User] Numpy / Scipy build / test errors on Mac OS X Lion with Macports
Charles R Harris
Wed Nov 23 12:04:54 CST 2011
On Wed, Nov 23, 2011 at 10:54 AM, Christoph Deil <
> On Nov 23, 2011, at 3:26 PM, Robert Kern wrote:
> Would it be helpful if I file tickets with the build and test logs for
> numpy and scipy using the XCode and Macports compilers, or are these
> problems well-known and simply too hard to fix?
> I don't know what problems you are referring to, so yes, reporting
> them would help.
> I tried to install numpy 2.0.0.dev-7f302cc on
> Mac OS X Lion 10.7.2 (11C74)
> XCode 4.2.1 (4D502)
> using these compilers / python:
> $ which python; python
> Python 2.7.2 (default, Nov 23 2011, 11:40:08)
> [GCC 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2335.15.00)] on
> $ which gcc; gcc --version
> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 (Based on Apple Inc. build
> 5658) (LLVM build 2336.1.00)
> $ which gfortran; gfortran --version
> GNU Fortran (GCC) 4.4.6
> The build log contains many errors such as:
> _configtest.c:1:20: error: endian.h: No such file or directory
> _configtest.c:5: error: size of array 'test_array' is negative
> _configtest.c:7: error: 'SIZEOF_LONGDOUBLE' undeclared (first use in this
> _configtest.c:8: error: 'HAVE_DECL_SIGNBIT' undeclared (first use in this
> _configtest.c:7: error: 'Py_UNICODE_WIDE' undeclared (first use in this
> but it all starts with this warning:
> numpy/core/setup_common.py:86: MismatchCAPIWarning: API mismatch detected,
> the C API version numbers have to be updated. Current C api version is 6,
> with checksum eb54c77ff4149bab310324cd7c0cb176, but recorded checksum for C
> API version 6 in codegen_dir/cversions.txt is
> e61d5dc51fa1c6459328266e215d6987. If functions were added in the C API, you
> have to update C_API_VERSION in numpy/core/setup_common.py.
> I tried rm -r and get a fresh git clone as suggested here:
> Same result.
> Nevertheless I do get a numpy that works for the most part, running the
> full test
> gives only one RuntimeWarning related to the power function and one
> failure in polyfit:
> RuntimeWarning: invalid value encountered in power
> np.power(out, 0.5, out=out, casting='unsafe')
> FAIL: Tests polyfit
> Traceback (most recent call last):
> line 622, in test_polyfit
> assert_almost_equal(a, a_)
> line 155, in assert_almost_equal
> err_msg=err_msg, verbose=verbose)
> line 221, in assert_array_almost_equal
> header='Arrays are not almost equal')
> line 186, in assert_array_compare
> verbose=verbose, header=header)
> line 677, in assert_array_compare
> raise AssertionError(msg)
> Arrays are not almost equal
> (mismatch 100.0%)
> x: array([ 4.25134878, 1.14131297, 0.20519666, 0.01701 ])
> y: array([ 1.9345248 , 0.49711011, 0.10202554, 0.00928034])
Don't worry about this one, it comes from Travis changing the unmasked
version of polyfit which changes the singular values, etc.
What is the reason for the C API Mismatch?
> I also see it with clang and on a linux box so I'm not sure if it's a
> problem with my installation or with numpy HEAD.
> How can I resolve it?
> Is this the cause for the following compile errors?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SciPy-User