search

drphilmarshall / OM10

Tools for working with the Oguri & Marshall (2010) mock catalog of strong gravitational lenses
MIT License
8 stars 19 forks source link

Travis failiing because requirements.txt is out of order? #41

Closed drphilmarshall closed 7 years ago

drphilmarshall commented 7 years ago

Infuriatingly.

mbaumer commented 7 years ago

The Travis stack trace says that scikit-learn can't find scipy during its install. It seems like issues like this with the Travis default python, pip, and scipy are known: https://gist.github.com/dan-blanchard/7045057 Their suggested solution (including template .travis.yml file) is to download and install miniconda every time Travis runs, and use that as your working python install. I remember this method being suggested at AHW as well--any thoughts on using this to avoid this problem?

drphilmarshall commented 7 years ago

Thanks Mike! I've resorted to conda in other projects for other reasons, perhaps I should just switch to that every time. ill try it out.

On Fri, Nov 4, 2016 at 2:06 PM, Mike Baumer notifications@github.com wrote:

The Travis stack trace says that scikit-learn can't find scipy during its install. It seems like issues like this with the Travis default python, pip, and scipy are known: https://gist.github.com/dan-blanchard/7045057 Their suggested solution (including template .travis.yml file) is to download and install miniconda every time Travis runs, and use that as your working python install. I remember this method being suggested at AHW as well--any thoughts on using this to avoid this problem?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/drphilmarshall/OM10/issues/41#issuecomment-258547396, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY9whYzmx5I0-FQKhaVfFIsaRrqQmhks5q655BgaJpZM4KKjco .

mbaumer commented 7 years ago

I understand your "last resort" thinking--it does seem a bit silly to have to download and install an entire python distribution every time you run your tests--when (in theory) the travis system already has its own python in place. But from Travis' point of view, perhaps they see it as trading more computational overhead each time a test set runs for less human overhead in maintaining the diverse python configurations users demand. Just a thought to make us feel less bad about using it as a default strategy!

drphilmarshall commented 7 years ago

:-) Travis is now happy, as of PR #44.