Closed mmagnuski closed 9 years ago
Thanks Mikolaj! Good eye. I’m now hoping you’re spot on. (It’s a simple answer.)
The Travis script is something I hacked together, just trying for that 'green bar’ as fast as possible. :) I’ll try conda environs (post-SFN).
On Oct 15, 2015, at 4:05 AM, Mikolaj Magnuski notifications@github.com wrote:
I took a look at the errors that Travis is spitting out. I was wondering if they might be related to the fact that you are using both virtualenv and conda (or is it that travis always uses virtualenv?). For example virtualenv usues python 2.7.9, but the miniconda version has python 2.7.10 - so there are actually two python versions on the machine. There is also some noise in the installation process - numpy (version 1.10) is first installed via conda but then pip install -r requirements.txt uninstalls it and installs the required version 1.9. Maybe using conda environments is worth trying out? For example:
conda create -n testenv --yes pip python=2.7 conda update conda --yes source activate testenv conda install --yes --file requirements.txt You can also take a look at the travis script that seaborn is using https://github.com/mwaskom/seaborn/blob/master/.travis.yml I would be happy to help with this, but I am using a Windows machine
— Reply to this email directly or view it on GitHub https://github.com/voytekresearch/pacpy/issues/19.
Looks like it was was conda/pip install nonsense dance I created in the first .yml draft. Now resolved. Thanks @mmagnuski!
:+1:
I took a look at the errors that Travis is spitting out. I was wondering if they might be related to the fact that you are using both
virtualenv
andconda
(or is it that travis always usesvirtualenv
?). For example virtualenv usues python 2.7.9, but the miniconda version has python 2.7.10 - so there are actually two python versions on the machine. There is also some noise in the installation process - numpy (version 1.10) is first installed viaconda
but thenpip install -r requirements.txt
uninstalls it and installs the required version 1.9. Maybe using conda environments is worth trying out? For example:You can also take a look at the travis script that seaborn is using I would be happy to help with this, but I am using a Windows machine :cry: