Mac OS X (El Capitan 10.11.6) STEME installation

Question

[themantalope]

Hi All,

I'm having some trouble getting STEME (v 1.9.1) to compile on my mac (OS X v10.11.6, MacBook Pro, mid-2015). I'm using gcc-4.2 (installed with homebrew), and I've installed the latest versions of boost, boost-python, seqan and fftw using homebrew. However, I keep on getting a compiler error when attempting to build the package. I have the environment variable ARCHFLAGS set to -arch i386 -arch x86_64. When I run the command python setup.py build after configuring, I am getting this error:

fatal error: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/lipo: can't open input file: /var/folders/4x/f3rlcx6s3ng53wr5nkrnlhb00000gp/T//ccOhxhL9.out (No such file or directory)

So it seems that lipo is trying to open a temporary file that it either doesn't have access to, or that it thinks exists, but then doesn't. Has anyone else had this issue with STEME on a mac? Has anyone been able to get STEME to successfully compile on a mac?

[johnbaronreid]

Hi, I'm the author of STEME. Thanks for your interest :) It has been a long time since I compiled it on a Mac. I'm pretty sure it should be possible as all the dependencies are available. Having said that, the error message doesn't mean anything to me. Did you try it without setting ARCHFLAGS? Also are you using gcc-4.2 because the help says so? I'm pretty sure a more recent gcc would be a better bet. I think you just need those architecture settings when compiling boost and fftw3, not when compiling STEME. I hope that helps, just let me know if not, John. P.S. Just to check you do know there is a web-server with most of the important functionality? http://sysbio.mrc-bsu.cam.ac.uk/STEME

[themantalope]

Yep, but I would like to use the glmnet feature for scan discriminiation as well as the spacing analysis. Thanks for the comments, I'll keep trying!

[themantalope]

I've now tried several later versions of the gcc compiler (4.5 - 6). The compilation gets a bit farther but still gets stuck. Here is the output for where the error is being thrown:

c++/pvalues/shifted_hirji.cpp:36:21: required from here /usr/local/include/boost/range/end.hpp:48:26: error: could not convert 'boost::detail::multi_array::const_multi_array_view<T, NumDims, TPtr>::end() const [with T = double; long unsigned int NumDims = 1ul; TPtr = const double*; boost::detail::multi_array::const_multi_array_view<T, NumDims, TPtr>::const_iterator = boost::detail::multi_array::array_iterator<double, const double*, mpl_::size_t<1ul>, const double&, boost::iterators::random_access_traversal_tag>]()' from 'boost::detail::multi_array::const_multi_array_view<double, 1ul, const double*>::const_iterator {aka boost::detail::multi_array::array_iterator<double, const double*, mpl_::size_t<1ul>, const double&, boost::iterators::random_access_traversal_tag>}' to 'boost::range_detail::extract_iterator<boost::detail::multi_array::const_multi_array_view<double, 1ul, const double*>, true>::type {aka boost::detail::multi_array::array_iterator<double, double*, mpl_::size_t<1ul>, double&, boost::iterators::random_access_traversal_tag>}' return c.end();

It seems that there is some kind of casting error. Is there a different version of boost that I should use when trying to compile?

[themantalope]

I am able to get the code dependent on boost to build, however I am now having problems with SeqAn dependent code building. Which version of SeqAn should I build with? Does version 2.2.0 build properly on other Linux systems?

[johnbaronreid]

Also I did find this similar error message to do with installing matplotlib on Mac: http://stackoverflow.com/questions/8117719/cant-install-matplotlib-on-os-x-with-pip Perhaps using a different python as suggested there might help. Can you check if you can install matplotlib? If not perhaps your package manager can help.