Telcon: 2021 12 08
Peter Scheibel edited this page Dec 8, 2021
·
14 revisions
This is the next planned Spack meeting
- Peter Scheibel
- Tammy Dahlgren
- Mark Krentel
- Massimiliano Culpo
- Asher Mancinelli
-
Tammy: if a user comments on a PR, sometimes we assign them to it
- Not all these folks have merge privileges
- Do we clarify to them that we can merge it
- If a package lists a maintainer, we may want to wait
- Could a bot alert us if the maintainer doesn't review it for one week? (If the bot is event-driven, then maybe this would be hard, although there are bot actions that close threads with no activity, so we could have the bot ping us instead of closing)
- If the maintainer has merge privileges, we could automatically assign them
- Examples of auto-closer (which we could repurpose for alerts after no activity):
- Also, we can assign whoever we want to a PR and leave a comment that they can tell us if they don't have time to review it
-
Asher: failure for large CI job (this is for a local CI)
- Core issue: generate job for Spack CI takes too many resource
- Artifacts for child pipelines are not created
- Example: https://gitlab.pnnl.gov/exasgd/frameworks/exago/-/tree/gitlab-spack-ci-debug
- Check if using parent of https://github.com/spack/spack/pull/26264 helps
-
Peter: Can we get rid of csh support?
- I always run
bash
first thing when I start a shell on Mac OS - Does anywhere require CSH support and have no option for bash support?
- We didn't discuss this here, but based on some follow-up in slack there are folks who specifically want to use csh
- I always run
- Mark: libelf/elfutils
- Related to
hpctoolkit
, which has started transitively linking to both libelf and elfutils when building with+rocm
- Related to
- Multiple providers of same virtual package
- Case 1: we want blas from X and lapack from Y
- Case 2: we actually want several instances of the same virtual (e.g. to run different instances of MPI for the same root)
- Possibly revisit: https://github.com/spack/spack/discussions/24966
- Concerning improvements to
spack develop
- Concerning improvements to
- Harmen (not sure if I can join): there are packages like libblastrampoline / libmpitrampoline which provide
a blas / lapack / mpi interface to link to, and forwards calls to an actual blas / lapack / mpi provider lib.
E.g. julia uses it to link to a blas interface for its binary deps, allowing the user to switch blas provider
at runtime, avoiding abi issues. The problem is Spack only allows one provider per dag, but these packages
both provides the virtual and depends on the virtual. How do we deal with this "composition" type of pattern?
I was thinking: maybe we can relax unique provider per dag to unique provider per subgraph connected through
link type deps? Then libblastrampoline can provide blas, and depend on blas as a run-type dependency.
- Mark Krentel: https://github.com/spack/spack/issues/27157