test fail on Mac OS 10.9.3/python 3.4.0: ImportError: dlopen(...): =mage found. #4737

Closed
SashaOv opened this Issue May 23, 2014 · 11 comments

Comments

Projects
None yet
4 participants

SashaOv commented May 23, 2014

I have installed numpy using pip but full tests fail. Here is the full log: https://gist.github.com/SashaOv/0403c4ef38af00ecaf0d

Contributor

matthew-brett commented May 23, 2014

How did you install Python? MacPython?

Did numpy compile, or did you get the pypi wheel ?

SashaOv commented May 23, 2014

It was python 3.4.0 installed with PyEnv.

The numpy was compiled with gcc:

> gcc -v
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 5.1 (clang-503.0.40) (based on LLVM 3.4svn)
Target: x86_64-apple-darwin13.2.0
Thread model: posix
Contributor

matthew-brett commented May 23, 2014

I don't know pyenv ; does it compile a new version of Python?

What do you get for python -c "import distutils.util; print(distutils.util.get_platform())"

I guess that you did pip install numpy and pip downloaded the numpy source and compiled it?

SashaOv commented May 23, 2014

Please see the linked Gist https://gist.github.com/SashaOv/0403c4ef38af00ecaf0d for the full log.

Pyenv is a shell wrapper and doesn't modify the Python code. It is using the stock version of python, in this case 3.4.0

Contributor

matthew-brett commented May 23, 2014

The gist gives the log of the test run but not the installation.

I don't know what 'stock python' is. Your version string is:

Python version 3.4.0 (default, May 20 2014, 20:42:24) [GCC 4.2.1 Compatible Apple LLVM 5.1 (clang-503.0.40)]

That matches your gcc version information exactly, implying to me that it was compiled on your machine, maybe via homebrew. The MacPython python.org binary has this version information:

Python 3.4.0 (v3.4.0:04f714765c13, Mar 15 2014, 23:02:41) 
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin

so I am pretty sure you don't have that.

Owner

pv commented May 24, 2014

The failures are from f2py tests, so it's also possible the Fortran compiler is not compatible.

SashaOv commented May 24, 2014

Using the latest build of gfortran:

    > gfortran -v
Using built-in specs.
Target: i686-apple-darwin8
Configured with: /Builds/unix/gcc/gcc-4.2/configure --prefix=/usr/local --mandir=/share/man --program-transform-name=/^[cg][^.-]*$/s/$/-4.2/ --build=i686-apple-darwin8 --host=i686-apple-darwin8 --target=i686-apple-darwin8 --enable-languages=fortran
Thread model: posix
gcc version 4.2.3

Let me know if you need more info. I am using the miniconda build with Python 3.3.0.

Contributor

matthew-brett commented May 24, 2014

Is there a conda support list you could try?

Your gfortran targets i686, I guess that may be a problem, if your Python is 64-bit.

SashaOv commented May 25, 2014

Yep this must be the problem: python is 64 bit while gfortran is run in i686 mode. Is it possible to pass -arch x86_64 to gfortran during the build?

Contributor

matthew-brett commented May 26, 2014

It sounds like this is a conda configuration issue - could you please take this up on some conda support forum? Maybe https://groups.google.com/a/continuum.io/forum/#!forum/conda

Thanks.

Contributor

juliantaylor commented Aug 31, 2014

seems like an issue numpy can't do much about, so closing. Please reopen if there is something we can do.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment