You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
During the work on #715, I discovered that our conda-forge CI tests intend to use the cache action, but never do so successfully. The whole issue is described in detail in actions/cache/issues/1177.
After finding a way to fix the conda-forge CI tests, I reset the branch to throw out all unwanted commits, including all attempts of fixing the cache path. This also prevents scope creep.
One further source of scope creep might be this: while looking into mamba as a replacement for conda, I found that micromamba in particular comes with a setup action that has caching integrated. Pending @khaeru's opinion, we might switch to mamba/micromamba entirely rather than just using the mamba solver in conda. In that case, the paths here don't need fixing; instead, we can replace the whole cache action with the one integrated in the setup-micromamba action.
If we do want to keep conda and need to fix the paths, it is worth noting that our own setup-conda action provides an output called cache-path, which might in theory allow us to automatically use the correct path for caching our conda installation. In practice, we currently run the cache action before setup-conda, so we don't receive the path information in time. Currently, that output is not used anywhere in the workflow, so it might not work out of the box.
The text was updated successfully, but these errors were encountered:
During the work on #715, I discovered that our conda-forge CI tests intend to use the
cache
action, but never do so successfully. The whole issue is described in detail in actions/cache/issues/1177.After finding a way to fix the conda-forge CI tests, I reset the branch to throw out all unwanted commits, including all attempts of fixing the cache path. This also prevents scope creep.
One further source of scope creep might be this: while looking into mamba as a replacement for conda, I found that micromamba in particular comes with a setup action that has caching integrated. Pending @khaeru's opinion, we might switch to mamba/micromamba entirely rather than just using the mamba solver in conda. In that case, the paths here don't need fixing; instead, we can replace the whole
cache
action with the one integrated in thesetup-micromamba
action.If we do want to keep conda and need to fix the paths, it is worth noting that our own
setup-conda
action provides an output calledcache-path
, which might in theory allow us to automatically use the correct path for caching our conda installation. In practice, we currently run thecache
action beforesetup-conda
, so we don't receive the path information in time. Currently, that output is not used anywhere in the workflow, so it might not work out of the box.The text was updated successfully, but these errors were encountered: