dev meeting 20230322

Present: @alizter @ejgallego @emillon @nojb @rgrinberg

• meeting notes
  • some teams (opam) are putting notes on wiki and happy with that
  • we used to do that, then switch to a google docs, then stopped
  • let's try to resume that effort
  • @emillon to take notes by default for now
• #7376 - cinaps confused by symlinks
• release process
  • should we release on main (as previously done) or off a branch (like ocaml/ocaml)
  • Rudi: this should not disrupt development of new features. it looked difficult for ocaml/ocaml
  • Ali: coq releases off a branch and backporting is difficult
  • Etienne: we can try as we did before (main) but by communicating the "freeze" (not the right word maybe) date more in advance
  • tentative date: 2023-04-19 (t+4 weeks) so that the pace corresponds to the last releases
• source fetching
  • concerns about vendoring parts of opam
  • this is a lot of code (could have been reimplemented) but will need this for integration with rest of opam (solver etc)
  • some code in the opam libs can be removed
  • discussion about solvers and why they're needed
  • can we improve on opam? (MVS, allowing several versions of a lib) - maybe we can do that but later
• doc effort
  • meta-pages have been split
  • now documenting where contents should be recycled
  • once done, can process document per document
  • new content is also written (ex: the ocaml ecosystem)
  • once we get the hang of it, this can be parallelized with interested people
  • README improved too
• performance of dune builds with coq (hott) https://github.com/HoTT/Coq-HoTT/pull/1687
• flake.nix - using opam-nix vs nixpkgs
• let= performance
  • Ali: We can write pretty things but how can we optimize them if they are slow? (can we ppx?)
  • do the pretty thing by default, if this shows up in perf reports switch to the optimized one
  • it's impossible to use ppx in dune