Skip to main content

conda-forge core meeting 2024-08-07

Add new agenda items under the Your __new__() agenda items heading

Attendees​

NameInitialsGitHub IDAffiliation
Sylvain CorlaySC@SylvainCorlayQuantStack
Thorsten BeierTB@derthorstenQuantStack
Filipe FernandesFF@ocefpafconda-forge
Klaus ZimmermannKZ@zklausQuansight
Axel ObermeierHV@h-vetinari
Jaime Rodríguez-GuerraJRG@jaimergpQuansight/cf
Dasha GurovaDGdashagurovaAnaconda/conda
John KirkhamJKjakirkhamNVIDIA/cf

X people total

Standing items​

  • [ ]

From previous meeting(s)​

  • [ ]

Active votes​

  • [ ]

Your __new__() agenda items​

  • (HV) Add emscripten as a target (RFC)
    • JRG: Ok with proceeding as long as nothing breaks and not too much burden is imposed on infra/core team.
    • TB: emscripten-forge "cheap" version of conda-forge. Mostly focused on Python packages, some of them are complicated (scipy, numpy).
    • IF: Concerns with changing ABIs with each compiler release, which will require a rebuild that is costly at the conda-forge scale.
    • ED: What would alleviate that concern?
      • IF: Stable ABI, shared builds
      • WV align emscripten-forge with conda-forge. import recipes from CF.
      • SC: WASI doesn't bring everything we need from emscripten.
      • FF: CF is a "do-ocracy". if someone wants to take on the work, great, go for it. if you want to do it as an experimental thing, then go do it. will only know details once we start doing it. need to explain the difference between this and pyodide. need to teach people how to do this. we are user-focused. we are building something of value for our users.
    • From the chat:
      • ED: So regarding emscripten, was the decision to “try it out and report back on complications”?
      • FF: I'm +1 to that. We will only know all the details once we try.
      • IF: Yeah, try out things and report back before we start building things
      • Clarification at the end: update emscripten feedstock, start with some simple libraries (e.g. zlib)
  • (HV) Drop PyPy (RFC)
    • IF: If we know it's going to be sunset eventually, we shouldn't wait and just drop it.
    • WV: Removing the migrator and eventually let it fade out? CFEP vote?
    • IF: No need for CFEP. Maybe just have a discussion in the migrator removal PR.
    • HV: Thinking of a softer approach, like pausing it for a few months and then removing it.
    • JRG: Adding an announcement about the intent to drop and leaving opportunities for folks that do depend on PyPy to know that is happening.
  • (HV) Compiler updates:
    • I'd like to finish the upgrade to GCC 13 / Clang 17; currently blocked on an issue with gfortran_impl_osx -- help welcome
    • When LLVM 19 goes GA in September, I'd like to migrate our Fortran compilers on windows soon after; we've been waiting for this for a long time, and flang 19 seems ready (tested various feedstocks with the RCs). We also have the functionality in place that Isuru asked for to restrict the migration to feedstocks that actually build for windows.
    • KZ: Docs / hints to help people move to flang?
    • HV: Most things should be taken care of by the activation scripts. The number of feedstocks is under 100ish so it's not a large amount.
    • KZ: Concerned about esmf, which custom build system, but well
  • (WV) v2 recipe work & cf-scripts work
  • (WV) should we use enums in conda-smithy? :)
  • (SC) Heads up: mamba 2.0 rc: https://medium.com/@QuantStack/introducing-mamba-2-0-0e8d5c6d1d0c
  • FF: Adding SC, WV, JRG to the financial team for OpenCollective approvals.
  • WV: Looking into how to leverage sigstore and containers for conda packaging, TUF and a proper signing story.

Pushed to next meeting​

  • [ ]

CFEPs​

  • [ ]