Open
Description
- Sooner or later we would want to make a release 😉
- Since 3.6.0.0, release artifacts are generated in the gitlab mirror repo: https://gitlab.haskell.org/haskell/cabal, pushing a tag
- Last repo builds are failing for some archs, f.e. https://gitlab.haskell.org/haskell/cabal/-/pipelines/48561
- aarch64-darwin: https://gitlab.haskell.org/haskell/cabal/-/jobs/969257
- x86_64-freebsd13: https://gitlab.haskell.org/haskell/cabal/-/jobs/969254
- The config of those builds have diverged from the hls one, which builds the same archs
- Afaics the hls config was inspired in cabal one but got several patches thereafter
- For example in hls nix is not used anymore for aarch64-darwin-m1
- The gitlab builds does not run any test suite so we dont have the certainly they will work
- But start to run the test suite could open a can of worms, from legitimate tests fails to errors only visible running the tests
- So some sort of manual testing could complement automatic tests if they are not reliable
So i think before the release we should
- fix failing builds (it will need update ci config)
- refactor ci config to make it as simple as possible (removing nix config f.e.)
- run the test suite in as many archs as possible, without blocking the release
Activity