Description
Part of the reason we had the unexpected Windows errors last night was because we hadn't actually been testing on Windows, since I only test locally on Linux. I don't plan on having a Windows test setup that I use regularly, so we need to setup CI to help out here. Here are proposed changes to CI:
If the current branch is "staging" (and maybe if the target branch of a PR is "staging"), then the following changes to the CI pipeline would apply:
-
An action would retrieve built artifacts from any repos upstream from it (is there a way to do this via github actions? probably) and place the wheels in a directory
-
Would need to modify pip arguments to prefer the files in the directory -- we do this via robotpy-build build-dep I think.
-
Undecided: if a pipeline completes successfully, should it automatically trigger a staging branch build for all of its dependents?
If we're able to do this successfully, then no more running CI on main, only on PRs/tags/staging. Rationale is that they just use up github runner allocations.