dev meeting 2025 10 01

New Link

https://meet.google.com/bmh-mwbw-aqo

Agenda

• Cmdliner 2.0 (@shonfeder, @ElectreAAS)
• Action hashing via @gridbugs new hasher interface? (@Leonidas-from-XIV)
• Clarifications around RPC situation in non-watch scenarios (@ElectreAAS)
• Branch protections for CI (@shonfeder, @Leonidas-from-XIV)
• Someone in Tarides' should probably have maintain perms (@shonfeder)

Meeting notes

• Cmdliner 2.0 (@shonfeder, @ElectreAAS)
  • https://github.com/ocaml/dune/issues/12515
  • Any considerations?
    • Ali: most/any modifications in our vendored cmdliner
    • Tiny derisking subtask, whether it needs any additional work beyond the update
    • @ElectreAAS will do the initial derisking effort
    • If complex work, let's triage to determine priority
• Action hashing via @gridbugs new hasher interface? (@Leonidas-from-XIV)
  • Rudi: worth investigating
  • Upsides?
    • code hi
    • @Leonidas-from-XIV: potential perf improvement
    • @rgrinberg: for null builds only
    • @rgrinberg: needs to change reused actions, like running a file etc.
  • Issue to track: TODO
• Clarifications around RPC situation in non-watch scenarios (@ElectreAAS)
  • Struggling with getting test server to work in watch mode
  • What should happen when normal dune build is running (not using watch mode) and second dune command is run
    • Conclusion: Generally should not be allowed?
  • Some RPC commands can be handled when run not in watch mode on JS internal dune
    • Ask it for e.g. progress and sets of errors
  • go_with_rpc function that doesn't actually start server?
  • @rgrinberg: at some point we wanted to run RPC server without watch mode in some limited situations, so made it lazy
    • but we then realized that this laziness is not useful.
  • Conclusion:
    • if we start the RPC server in all conditions then shut it down, then we don't need the lazy starting
    • but we still need "ensure ready"
• Branch protections for CI (@shonfeder, @Leonidas-from-XIV)
  • @rgrinberg: Recent breakage is because of test dependency being updated
  • @Leonidas-from-XIV: reducing breakage on main is a motive, but allowing merge when ready is more important
  • relation to merge queue?
    • This may be the problem that was an issue when testing merge queues last time
  • Conclusion: There are no objections, and we will proceed
  • Which CI should be required?
    • Let's start with enabling all of them, and we can selectively disable ones that prove flaky
  • Followup: whomever in Tarides is granted merge perms can configure this
• Someone in Tarides' should probably have maintain perms (@shonfeder)
  • Whoever will be backup release coordinator in Tarides should get the maintenance perms

Attendees: @rgrinberg, @Leonidas-from-XIV, @shonfeder, @art-w, @ElectreAAS, @Alizter, @Sudha247