Closed alexeagle closed 4 months ago
I'm starting on this here: https://github.com/bazel-contrib/.github/pull/13
Yes, we can absolutely use the action, though I wonder if rulesets want caches enabled at all - it seems to me that a cacheless setup is preferred, at least that's what we do in rules_ruby.
What are the pros and cons of using this for rulesets? How do you feel they are different from other projects?
The main reason I'm not using the caches in rulesets is that I want to make sure the toolchains can be installed from scratch. This way I could catch regressions in repository rules that for example download and compile Ruby.
It's useful just for the bazelrc to set enable_bzlmod, and I think the repository cache as well since it contains content-addressed bytes and only accessed when the user supplies an integrity hash. So, I think the answer is yes :)
As part of moving this repo to bazel-contrib, I'm curious if we should consider changing https://github.com/bazel-contrib/rules-template/blob/main/.github/workflows/ci.yaml to use this. I wrote that reusable action that it references but it's not sophisticated about the caching.