Compile error about matscipy HOT 4 CLOSED

Oops, this is related to how different compiler handle C++11. James had looked at this in detail and probably knows the answer.

James, should we think about disabling the ring statistics part on some platforms?

Lars

from matscipy.

I've locally resolved this with Christoph but we still need a general solution. It seems C++11 is sufficiently well supported, the only problem is getting Python distutils to pass different flags when compiling C and C++ code, as passing -std-c++0 with C input files causes errors with some compilers. Setting the CXXFLAGS environment variable should work, but doesn't because of this bug:
http://bugs.python.org/issue1222585

from matscipy.

Should now be fixed. I had to delve deeper into numpy.distutils than I would like, but I've found a way to pass different options to C and C++ compilers which is working on our Travis-CI build slaves with g++ and clang++ and on Linux and OS X. There are now dictionaries of C and C++ options at the top of setup.py that can be customised on a per-compiler basis if this issue recurs.

Christoph, it would be great if you could revert the local changes made to setup.py (with git checkout setup.py) then pull again and check it now compiles out of the box for you with anaconda Python. I will then close the issue.

from matscipy.

It builds and installs. If I run into any issues in usage, I'll open another report. Many thanks!

from matscipy.