Skip to content
Pierre Roux edited this page Jun 11, 2025 · 1 revision

Attending: Yves Bertot, Alessandro Bruni, Cyril Cohen, Pierre Roux, Kazuhiko Sakaguchi, Théo Stotskopf, Enrico Tassi, Quentin Vermande

  • quick announce: nasty filename (absence of) conflict now fixed in master: https://github.com/rocq-prover/rocq/pull/20601 Nix and OPAM should now behave the same in case of duplicated file

  • Large scale documentation of mathcomp: phase 1 request for annotations

    • three projects: "translate" proof scripts from Lean to Rocq able to translate some proof terms (success rate currently around 30%)

    • Crrrocq: interaction between LLM and proof assistants

    • LM4Docq: generate documentation using LLMs

      • add docstrings to all element sof the mathcomp library
      • exploit them in natural language queries
      • need some work to audit the currently generated docstrings https://github.com/LLM4Rocq/LLM4Docq
      • where are the docstrings stored: currently in a separate database, ultimately we want them in the file
      • Théo probably attending Rocq'n'share
      • need to schedule something together: Rocq'n'share, sharing days? announce it on mathcomp dev zulip channel
  • Clarify the authorship of multinomials (before September if possible) https://github.com/math-comp/multinomials/issues/59

    Discuss it with Pierre-Yves

  • https://github.com/math-comp/math-comp/wiki still reads "Next Sharing day: 2025-05-28" did we decide something?

    Let's go for July 9th. Kazuhiko or Quentin will announce it.

  • Export all_boot in all_order,... ? (c.f. https://rocq-prover.zulipchat.com/#narrow/channel/237664-math-comp-users/topic/Error.20with.20syntax.20levels.20all_fingroup.20vs.20all_algebra/with/523275229 )

    Needs more discussion, for instance at Rocq'n'share

  • chair for next meeting (July 9th): Pierre

  • PR triage

    • https://github.com/math-comp/math-comp/pull/1438
      • new heuristic for pattern selection skipping implicit arguments
      • sometime breaks with phantom, workaround with multiple implicits
      • selecting wrong instance can be very bad performancewise (not a new problem but needs to be careful)
      • discuss design in Paris (end of June)
      • complete overlays during next sharing days (July 9th)

Clone this wiki locally