It looks like conda changed the default versions it looks for today, because conda install numpy scipy
does not currently work:
$conda create numpy scipy -n test-build-issue3 --dry-run
Error: Unsatisfiable package specifications
Hint: the following combinations of packages create a conflict with the
remaining packages:
- numpy 1.7*
- scipy
If I give specific versions, I can set up a working environment locally.
conda create numpy=1.7.1 scipy=0.13.0 pandas=0.13.0 matplotlib=1.3 PIL -n test-build --dry-run
But on Travis, while conda reports having installed and linked pandas, the test suite raises an ImportError:
$ nosetests --nologcapture -a '!slow'
E
(...)
ImportError: No module named pandas
One example of a complete failed build is here.
The test suite was passing fine yesterday. I have lots of experience with Travis but less experience running and debugging conda on it. Any suggestions?
Thanks for the folks over at the anaconda mailing list, this is resolved.
Today, ContinuumIO updated conda, but not miniconda, and this was the cause of my woes. To protect oneself from this situation, add this line to before-install:
- conda update --yes conda
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With