Closed xylar closed 3 years ago
@jhkennedy, in general I have questions about these dependencies. Some of them seems to be dependencies of the dependencies. They don't need to be listed here unless cime
itself requires them.
cc @rljacob and @mkstratos.
A local build of the package was successful. (Ignore the failing CI, it's completely unrelated to this package.)
Should the build file be updated with this too with new build, version, and Python versions? https://github.com/E3SM-Project/e3sm-unified/blob/54c9c54f0cba36a96249e05e3d6b2db9f4d9f035/recipes/cime-env/build.bash#L6-L8
@jhkennedy, thanks for taking a look. I removed all of the dependencies I think are redundant. In addition to your list, I figured keeping scipy
likely makes sense. Let me know what you think.
@mkstratos, good call. I removed the very outdated build.bash
entirely. That's a confusing script to have because usually that name is reserved for a build step that is part of building the conda package, not for building the conda package itself. Also, the package is now noarch-python, meaning only one version is produced (so no need for separate builds for different python versions or OSes).
conda build -c conda-forge recipe/cime-env
should do the trick. I don't think we need a script for that. It's a very standard build.
@mkstratos, I'm going to go ahead and upload the package and deploy tomorrow morning. I'll leave this PR open for now in case I run into trouble with the deployment script. Feel free to comment if you notice anything wrong.
Uploaded: https://anaconda.org/e3sm/cime-env/files
@mkstratos, version 1.1.0 has been deployed on Cori, Compy, Anvil/Chrysalis and Cooley. It will load with the load_latest_cime_env.sh
script on each machine. Let me know how it goes so I can merge this PR if all is in order.
@mkstratos, version 1.1.0 has been deployed on Cori, Compy, Anvil/Chrysalis and Cooley. It will load with the
load_latest_cime_env.sh
script on each machine. Let me know how it goes so I can merge this PR if all is in order.
The new environment is working for me (EVV runs and generates pages correctly) on Cori, which is as far as my access goes.
@mkstratos, okay. I'll assume that others will let me know in the future of there are issues on other E3SM machines. Thanks for testing.
This merge also
hdf5
andlibnetcdf
to be consistent with conda-forge pinsnetcdf4
to the latest releasepython >= 3.7
json_tricks
(which is alivvkit
dependency, not a CIME or E3SM one).