Numpy in Jupyter GCC-4.6 error

[PeterSprague]

I have Jupyterhub running in a jail with Python 3.5.

$ pip3 install numpy

installs with no errors

$ python35
>> import numpy
>> [aka no error]

In notebook cell:

import numpy as np

Gives error:

ImportError: /lib/libgcc_s.so.1: version GCC_4.6.0 required by /usr/local/lib/gcc48/libgfortran.so.3 not found

Ports installed include lapack, openblas, gcc, gcc46, cblas, pkgconfig

$ pip3 install scipy

Throws all kinds of errors related back to gcc 4.6 fortran

Any ideas how to fix this? Haven't found anything that works so far.

 

[PeterSprague]

Brute force solution:

$ portmaster gcc5
$ ln -s /usr/local/lib/gcc5 /usr/local/lib/gcc
$ ln -s /usr/local/bin/gcc5 /usr/local/lib/gcc

For this jail:
in /etc/make.conf

.if !empty(.CURDIR:M/usr/ports/*) && exists(/usr/local/bin/gcc5)
CC=gcc49
CXX=g++49
CPP=cpp49
.endif

Edit /etc/libmap.conf:

libgcc_s.so.1   gcc5/libgcc_s.so.1
libgomp.so.1    gcc5/libgomp.so.1
libobjc.so.3    gcc5/libobjc.so.4
libssp.so.0     gcc5/libssp.so.0
libstdc++.so.6  gcc5/libstdc++.so.6

See: http://unix.stackexchange.com/questions/130639/how-do-i-install-the-latest-gcc-on-freebsd

 

[soulcatcher]

Your solution is a bit overkill.
I had the same problem and the fix was to run jupyter notebook under the correct LD_LIBRARY_PATH:

export LD_LIBRARY_PATH=/usr/local/lib/gcc49
jupyter notebook

For the extended explanation see http://www.mianchen.com/running-my-own-jupyter-notebook-server-in-a-freebsd-jail/