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
Previously, we had some weird bugs with the CI pipeline sometimes
failing (#400), but not always reproducible. Namely, sometimes the CI
failed, due to `mypy` not finding the `sdk` types when running from the
`compliance_tool` environment. Since currently, we are at a point where
it is impossible to reproduce the failing CI (at least for me), I
decided to clean up the job definitions a little bit and make some
things more explicit.
Namely, instead of calling scripts like `pip` or `mypy` from their PATH,
we now explicitly call them via `python -m pip` and `python -m mypy`.
This theoretically ensures, that it always uses the script we just
installed with the dependencies and not something the VM already had
in its path via `actions/setup-python@v5`.
This should ensure that a script like `mypy` actually has all the
necessary dependencies installed.
Secondly, we had a `pip install -e ../sdk[dev]`, therefore installing
the development dependencies of the `sdk` in the `compliance_tool` CI
check. This is technically incorrect, since we use the `sdk` as external
dependency and therefore shouldn't depend on the development
dependencies. I therefore removed this.
Lastly, the `sdk-readme-codeblocks` check uses `bash` syntax. In theory,
the Ubuntu environment should use `bash` by default, but now it is made
explicit.
Fixes#400 (hopefully)
0 commit comments