Closed sierra-moxon closed 7 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
d45970a
) 62.09% compared to head (0579243
) 62.09%. Report is 1 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
I agree it's a little bit pushy w/re to upgrade if this action fails. I did use the methodology downstream in biolink-model
to find a breaking change in linkml 1.6.4
that wasn't caught by the existing linkml tests, nor was it caught by the poetry.lock version of the biolink-model repo (as that was still at linkml 1.6.3
).
The "pushy" part was that in order to get my biolink-model
PR to pass this dependency check, I had to update my poetry.lock file in biolink to use linkml 1.6.4
and then change the biolink-model
code generation to follow the new pattern introduced in linkml 1.6.4
. I'm glad I caught this and I put in a linkml PR to support the old way for awhile so other models downstream don't inadvertently run into this, but it was a bit hard to decide to tie an upgrade of linkml
to an unrelated PR so that the PR would pass its tests.
add a check dependencies action that removes the poetry.lock file and reinstalls the linkml-runtime environment (getting the latest compatible packages in the pyproject.toml specification) and then runs the tests as a way of testing the code in a PR on the latest package dependencies
partially resolves #1749.
After discussion during developer days hackathon, we want to limit our downstream build dependencies (e.g. not build out bmt, etc from LinkML packages directly), but makes sense to take this first step towards "latest possible" dependency checking.