Cross compile problem: #include <Python.h> and -lpython2.7

Curtis Olson

I have successfully installed python on my omega2+ and run python code there. I have successfully built the cross compiler tools and can compile C/C++ code for my omega2+.

However, I have some C++ code that calls the python C++ API. It needs to #include <Python.h> when compiling and needs to link to -lpython2.7. The cross compiler finds my host system Python.h which ultimately leads to compile #error "Unknown word size". Also, libpython2.7.so doesn't exist in my cross compile tool chain.

Has anyone gone down this path before that can offer a recommendation on how to setup my cross compiler tool chain to handle this?

(If anyone is curious, the code I'm attempting to compile is a UAV autopilot ... I'm evaluating the feasibility of flying the omega2+ on board my aircraft.) Project code (fully open-source) is here for anyone that wants to browse it: https://github.com/AuraUAS/aura-core

Thanks in advance,

Curt.

Curtis Olson

Ok, I'll reply to myself here because I think I found a solution.

Download python-2.7.13 source code and extract it.
Run configure with all the options needed by the LEDE cross compile tools
Run make and the compile ran to successful completion.
make install failed miserably because of some pc host vs. omega stuff that was beyond me to figure out.
manually copied the Include tree from python src to ...source/staging_dir/toolchain-mipsel_24kc_gcc-5.4.0_musl/include/python2.7/ and manually copied the pyconfig.h file as well.
manually copied the libpython2.7.a file to ...source/staging_dir/toolchain-mipsel_24kc_gcc-5.4.0_musl/lib/
fiddled with the build system in my own software because autoconf couldn't run python-config and get the right answer.
compiled my code successfully.

All that's left is it test run in on the actual omega2 which I will hopefully be able to do this weekend.