Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

recipe_miles_* broken due to newest akima package versions #2695

Closed
sloosvel opened this issue Jun 28, 2022 · 5 comments
Closed

recipe_miles_* broken due to newest akima package versions #2695

sloosvel opened this issue Jun 28, 2022 · 5 comments
Labels

Comments

@sloosvel
Copy link
Contributor

All recipes using the latest version of the R package akima (v0.6-3.4) get stuck loading the library. I checked the releases for the package (https://github.com/cran/akima/tags) and there have been two new releases since the release of ESMValTool v2.5:

  • 0.6-3.4 (released in 04/22) : This one gets stuck loading the package
Loading required package: mapproj
Loading required package: akima

  • 0.6-3.3 (released in 04/22): This one also gets stuck
  • 0.6-2.3 (released in 11/21): This one works and finishes the diagnostic. I assume it's the same version that is using ESMValTool v2.5

Setting the R requirements in r_requirements.yml for the package like below seems to solve the issue:

akima:
  origin: cran
  version: 0.6-2.3
@zklaus
Copy link

zklaus commented Jun 28, 2022

Note that "in real life" this package should come from conda-forge as r-akima. If we don't manage to get things working with the latest version, we can pin there as <0.6_3.3.

@valeriupredoi
Copy link
Contributor

so they made a release a few days after another release - surely a bugfix release but not a great one by the looks of it. This is numerical stuff so I'd be inclined to pin to <=2.3 since all these numerical libs are prone to introduce differences in results with changing versions - what you guys think?

@zklaus
Copy link

zklaus commented Jun 28, 2022

General ramblings about numerical stuff aside, I agree that pinning to <=0.6_2.3 is reasonable here. There are substantial changes in the newer versions with no changelog or description of what was changed and why.

@sloosvel
Copy link
Contributor Author

Do all packages need to be delted from the R_requirements file then? because for instance the tools one cannot be found in conda forge if I move it to the environment file.

bouweandela added a commit that referenced this issue Jul 21, 2022
Closes #2606
Closes #2299
Workaround for #2695
Closes ESMValGroup/ESMValCore#1364?

Versioning

    Use setuptools-scm to give a more accurate version number. The version number is built up as follows:
    {major}.{minor}.{patch}.dev{distance}+g{commit}+d{date}.
    Examples:
        For released versions, it will remain unchanged, e.g. 2.5.0
        If there are additional commits since the release, the patch version number will be increased with one and the number of commits since the release and the current commit will be used to generate a more detailed description. For example, if you are on v2.5.0 of the tool and there have been 72 commits in your current branch since that release and the latest commit has the (hexadecimal) number a424c89 (you can find the number of all commits in the current branch by using git log), then the version number will look like this: 2.5.1.dev72+ga424c891d.
        If there are uncommitted changes, the letter d and the current date will be added to the version number, e.g. 2.5.1.dev72+ga424c891d.d20220719

CircleCI

    Use appropriate resources
    Upload compressed artifacts for fast runs
    Upload test results in a format that CircleCI understands, enabling test summaries and statistics
    Only run documentation test nightly, readthedocs build is now configured to fail on warnings

Readthedocs

    Use mamba build
    Fail on warnings

Docker

    Use mamba container as a basis

Python development installation

    Install all development dependencies from conda-forge

R development installation

    Install all development dependencies from conda-forge

Stickler CI

    Remove the configuration file because we are not using this service anymore

Updates to dependencies

    Pin cf-units to a version that is compatible with the tests for the v2.6 release (Test failure due to change in cf-units ESMValCore#1655)
    Pin r-akima (recipe_miles_* broken due to newest akima package versions #2695)
    Pin xesfm to 0.3.0 (Old xesmf version dependency creates problems #2729)
    Remove xesmf dependency numba (Dependency refinement sparse, numba pangeo-data/xESMF#158)
@valeriupredoi
Copy link
Contributor

@sloosvel this pin now blocks updating support to Python 3.12 - do you think we can relax it? See #3501

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants