"Fatal error: sys/cdefs.h: No such file" when using anaconda gcc on amd64

[Juan Luis Cano]

Hello all,

I'm trying to provide a conda recipe for several FEniCS packages, and in
particular I wanted to use anaconda gcc to avoid incompatibilities
across Linux distributions[1]. However, while testing instant[2] in my
Ubuntu 14.04 machine, I get errors of this kind:

/home/.../_test/lib/gcc/x86_64-unknown-linux-gnu/4.8.2/include-fixed/features.h:338:25:
Fatal error: sys/cdefs.h: No such file or directory

# include <sys/cdefs.h>
^

compilation terminated.

I've read that this error has to do with 32/64bit, but I don't
understand what's happening since I am on 64bit and it used to work with
system gcc and gfortran. The package uses SWIG to compile extension
modules, so I need some help to debug this since I don't know if it's a
problem on SWIG, gcc or some specific Anaconda issue. I have copied some
relevant information in gist[3].

Best,

Juan Luis Cano

[1]: https://github.com/Juanlu001/fenics-recipes/issues/31
[2]: https://github.com/Juanlu001/fenics-recipes/tree/master/instant
[3]: https://gist.github.com/Juanlu001/02d678ab116d9eeba149

[Graham Markall]

I recall seeing a discussion of similar issues in this PR: https://github.com/conda/conda-recipes/pull/279#issuecomment-98254389

That PR has a long discussion about addressing some of the issues with gcc in anaconda and making it as portable as possible, which seems to be quite tricky - perhaps one of the suggestions in that thread can be used as a workaround?

--
Anaconda Community Support Group Brought to you by Continuum Analytics
--- You received this message because you are subscribed to the Google Groups "Anaconda - Public" group.
To unsubscribe from this group and stop receiving emails from it, send an email to anaconda+u...@continuum.io.
To post to this group, send email to anac...@continuum.io.
Visit this group at http://groups.google.com/a/continuum.io/group/anaconda/.

[juan...@gmail.com]

Thanks for the link Graham! Yes, I would need to see how can I fit that into the package I'm building but definitely explains the failures. I'll see what I can do.

Regards,

Juan Luis