Skip to content

dev meeting 20220303

Nathan Rebours edited this page Mar 4, 2022 · 1 revision

Present at the meeting:

  • Marek Kubica (@Leonidas-from-XIV)
  • Nathan Rebours (@NathanReb)

Current plan

  • Nathan keeps working on the tezos opam-monorepo build
  • Nathan attempts to fix the vendor variable bug in lockfiles
  • Marek finishes the work on hybrid mode
  • Marek starts working on opam-overlays CI

Meeting notes

Marek focused on improving the hybrid mode solver, it's now ready for review. He had to use an unsatisfying workaround to re-inject the contraints into the second solver run. Nathan might have a solution for this, the constraints can be re-injected in the solver context in a cleaner way.

Nathan was off for a week but he got back to working on tezos since. It's making good progress. A few dune ports needed to be fixed, Etienne helped quite a bit with hacl-star where we now have a simpler port but which works much better. The tezos team confirmed that in the short-term they will be able to drop the dependency on bls12-381-legacy. This is good news but it won't happen overnight so Nathan uploaded a repackaged version to dune-universe which fixes the conflicting packages issue. There is also a problem with ometrics which both vendors the merlin repository (a patched version of it) and depends on it through opam, which causes build failure when used in an opam monorepo project. Nathan reached out to the maintainers suggesting a variety of possible fixes. Finally Nathan started openning PRs to the tezos repository so that it is in a state compatible with opam-monorepo. The first contribution is to slightly relax = constraints on packages that need a dune port. The next step is to open a PR upgrading some package which had their latest version ported to dune upstream.

We discussed the vendor variable issue again, Nathan talked to Raja about the various solutions we had in mind so we have a clearer view of what would work or not. We talked about the ?vendor syntax which could do the job quite well!

We also talked about the CI for opam-overlays. While the overall solution on how we will keep dune ports up to speed with upstream repository still needs to be better specified, it is clear that having a CI that can check whether a dune ports builds correctly on opam-overlays is beneficial and will be required no matter how we proceed.