Telcon: 2023 05 17
Peter Scheibel edited this page May 17, 2023
·
14 revisions
Held Wednesday May 17th, 9am PT (UTC -7:00)
- Peter Scheibel (host)
- Chris Green
- Jakov Petrina
- Jakov Smolic
- Mark Krentel
- Massimiliano Culpo
- Richarda Butler
- Tammy Dahlgren
- Todd Gamblin
- William Phan
- Brian Van Essen
This meeting is just Q&A: there are no pre-planned general Spack topics
- Wileam Phan asked about v0.20 coming out next week, Todd Gamblin provided highlights of the release.
- Chris Green on putting an issue with the concretizer in where he's trying to build successive environments, one after the other.
- A link dependency is being rebuilt, and the only difference is the CMake being built
- To compare them, you can
spack diff /hash1 /hash2
- Also possibly relevant: https://github.com/spack/spack/pull/33768 (you can include an env into another env)
- Also possibly relevant: https://github.com/spack/spack/pull/37438 (you can reconcretize with
unify:true
and without-f
and spack will not re-concretize already-concretized packages)
- Todd: mentioned new ability to reference buildcache specs by hash (e.g. in
spack install
,spack spec
)- Also relevant: https://github.com/spack/spack/pull/35975
- Brian: https://github.com/spack/spack/pull/37687 broke externals which define versions not listed in the package.py (see: https://github.com/spack/spack/pull/37733)
- Todd: you can see discussions on CI improvements at https://github.com/spack/spack/wiki/CI-Meeting-Notes
- in short: ongoing active effort to improve speed of CI
- Brian: updating
CachedCMakePackage
to handle common ROCm/CUDA options: https://github.com/spack/spack/pull/37592 - Chris: https://github.com/spack/spack/issues/37502
- If you make a faulty env, it cannot be removed
- Firstly: we should entirely avoid making an env if any part of it fails
- Secondly: when removing an env, do we need to remove it
- What about removing an env that is included by another env
- If you make a faulty env, it cannot be removed