In case it helps, I was able to build the graph-tool v2.20 conda package successfully using the recipe I put in this repo: I uploaded the package to the conda-forgery anaconda channel:

You can find the conda recipe in the recipe folder and commands required to build the package in the .gitlab-ci.yml file. The package is built using the same linux-anvil docker container that is used to build conda-forge packages. Most of the dependencies are available in the conda-forge channel, but you also need to add the conda-forgery channel (I should probably come up with a better name) in order to install pycairo.

Note that this builds graph-tool without graphics support. Graphis support requires too many dependencies that I had difficulty compiling from source.

On Tue, Feb 28, 2017 at 1:12 PM, P-M <> wrote:

I have updated cgal to 4.7 so that now works and after applying the
libreadline workaround it compiles and installs but I still get C++
signature mismatches.

I have ripped off Anaconda entirely, re-installed it, installed the latest
Boost in it using only one source to make sure all packages are compatible
(conda-forge). Running "./configure --prefix=$HOME/.local CXX="g++-5"
CXXFLAGS="-std=gnu++14 -Wno-unused-local-typedefs"
LDFLAGS="-L/home/pmj27/anaconda2/lib" --enable-openmp" seems to work and
completes succesfully, however, the configuration summary prints "Boost CPP
flags: -I/usr/include" which is not what I want and which I guess is causing
my later problems?
                             CONFIGURATION SUMMARY
Using python version:   2.7.13
Python interpreter:     /home/pmj27/anaconda2/bin/python2.7
Installation path:

C++ compiler (CXX):     g++-5
C++ compiler version:   5.4.0

Prefix:                 /home/pmj27/.local
Pkgconfigdir:           ${libdir}/pkgconfig

Python CPP flags:       -I/home/pmj27/anaconda2/include/python2.7
Python LD flags:        -L/home/pmj27/anaconda2/lib -lpython2.7
Boost CPP flags:        -I/usr/include
Boost LD flags:         -lboost_iostreams -lboost_python -lboost_regex
Numpy CPP flags:
Sparsehash CPP flags:
CGAL CPP flags:         -I/usr/include
CGAL LD flags:          -L/usr/lib -lCGAL -lCGAL_Core -lgmp
Expat CPP flags:        -I/usr/include
Expat LD flags:         -L/usr/lib -lexpat
Cairomm CPP flags:      -std=c++11 -I/usr/include/cairomm-1.0
-I/usr/lib/x86_64-linux-gnu/cairomm-1.0/include -I/usr/include/cairo
-I/usr/include/glib-2.0 -I/usr/lib/x86_64-linux-gnu/glib-2.0/include
-I/usr/include/pixman-1 -I/usr/include/freetype2 -I/usr/include/libpng12
-I/usr/include/sigc++-2.0 -I/usr/lib/x86_64-linux-gnu/sigc++-2.0/include
Cairomm LD flags:       -lcairomm-1.0 -lcairo -lsigc-2.0
OpenMP compiler flags:  -fopenmp
OpenMP LD flags:
Extra CPPFLAGS:         -I/home/pmj27/anaconda2/include
Extra CXXFLAGS:         -fopenmp -O3 -fvisibility=default
-fvisibility-inlines-hidden -Wno-deprecated -ftemplate-depth-250  -DNDEBUG
-Wall -Wextra -ftemplate-backtrace-limit=0 -std=gnu++14
Extra LDFLAGS:          -L/home/pmj27/anaconda2/lib

Using OpenMP:           yes
Using sparsehash:       yes
Using cairo:            yes

Is there any way of telling why it keeps defaulting back to using the boost
in /usr/include and a way of making sure it doesn't do so? The  config.log
is attached.

View this message in context:
Sent from the Main discussion list for the graph-tool project mailing list archive at
graph-tool mailing list