I understand Arch Linux isn't supported per the README, so this is just FYI. Maybe the ubuntu docker install is more concerning? I tried that as a backup, and it didn't work either.
I'm actively working on fixing this -- it's totally possible the issue is only my own, so if I find a solution, I'll make a PR.
TLDR:
I suspect either the boost version or setup.py for preventing installation on some systems. I think better tracking of required package versions could get around this.
Recall the Ubuntu Linux install instructions:
sudo apt-get install -y libboost-all-dev libgsl0-dev ccache
sudo apt-get install -y python-matplotlib # If you don't already have it
sudo apt-get install -y python-scipy # If you don't already have it
pip install venture-0.4.2.tar.gz
The Arch Linux equivalent is the following:
sudo pacman -S boost gsl ccache
pip2 install matplotlib scipy
I indeed have these packages:
$pacman -Q | egrep 'boost|gsl'
boost 1.67.0-4
boost-libs 1.67.0-4
gsl 2.5-1
$pip2 list | egrep 'matplotlib|scipy'
matplotlib 2.2.2
scipy 1.1.0
$ python2 --version
Python 2.7.15
Same thing with ubuntu docker image:
# apt list | egrep 'boost-all|libgsl'
libboost-all-dev/bionic,now 1.65.1.0ubuntu1 amd64 [installed]
libgsl-dev/bionic,now 2.4+dfsg-6 amd64 [installed]
# pip list | egrep 'matplotlib|scipy'
matplotlib (2.2.2)
scipy (1.1.0)
# python --version
Python 2.7.15rc1
They're higher versions than specified in conda/meta.yaml (which I'm referring to because there is no requirements.txt -- which is what pip would use), but this doesn't seem to be causing the problem (I'll try all downgraded versions next).
Here's the ubuntu case with the latest release:
# pip install v0.5.1.1.tar.gz
Processing ./v0.5.1.1.tar.gz
Complete output from command python setup.py egg_info:
fatal: not a git repository (or any of the parent directories): .git
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/tmp/pip-ZBvNtO-build/setup.py", line 85, in <module>
pkg_version, full_version = get_version()
File "/tmp/pip-ZBvNtO-build/setup.py", line 57, in get_version
'git', 'describe', '--dirty', '--long', '--match', tag, '--tags',
File "/usr/lib/python2.7/subprocess.py", line 223, in check_output
raise CalledProcessError(retcode, cmd, output=output)
subprocess.CalledProcessError: Command '['git', 'describe', '--dirty', '--long', '--match', 'release-0.5.1a20170816', '--tags']' returned non-zero exit status 128
Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-ZBvNtO-build/
I've tried running this from inside the Venture repo, but this doesn't work.
I also tried the 4.2 release, which gives a multi-page template error, which I also get in the 5.1 case if I call python setup.py install manually. It's a huge error, but I think the relevant part is this:
backend/new_cxx/src/pyutils.cxx:23:10: fatal error: boost/python/numeric.hpp: No such file or directory
#include <boost/python/numeric.hpp>
^~~~~~~~~~~~~~~~~~~~~~~~~~
However, this is already installed, so far as I can tell:
# apt-get install libboost-python-dev
libboost-python-dev is already the newest version (1.65.1.0ubuntu1).
Looks like others have run into this. I tried changing numeric.hpp to numpy.hpp, but maybe a downgrade of boost will work? I will let you know...
Anyway, all of that documentation is to say two important things: I suspect either the boost version or setup.py for preventing installation on some systems. I think better tracking of required package versions could get around this.
I understand Arch Linux isn't supported per the README, so this is just FYI. Maybe the ubuntu docker install is more concerning? I tried that as a backup, and it didn't work either.
I'm actively working on fixing this -- it's totally possible the issue is only my own, so if I find a solution, I'll make a PR.
TLDR:
I suspect either the boost version or setup.py for preventing installation on some systems. I think better tracking of required package versions could get around this.
Recall the Ubuntu Linux install instructions:
The Arch Linux equivalent is the following:
I indeed have these packages:
Same thing with ubuntu docker image:
They're higher versions than specified in conda/meta.yaml (which I'm referring to because there is no requirements.txt -- which is what pip would use), but this doesn't seem to be causing the problem (I'll try all downgraded versions next).
Here's the ubuntu case with the latest release:
I've tried running this from inside the Venture repo, but this doesn't work.
I also tried the 4.2 release, which gives a multi-page template error, which I also get in the 5.1 case if I call
python setup.py installmanually. It's a huge error, but I think the relevant part is this:However, this is already installed, so far as I can tell:
Looks like others have run into this. I tried changing numeric.hpp to numpy.hpp, but maybe a downgrade of boost will work? I will let you know...
Anyway, all of that documentation is to say two important things: I suspect either the boost version or setup.py for preventing installation on some systems. I think better tracking of required package versions could get around this.