conda-forge core meeting 2023-06-14
Add new agenda items under the Your __new__() agenda items
heading
Attendees
Name | Initials | GitHub ID | Affiliation |
---|---|---|---|
matthew r becker | mrb | beckermr | cf |
Filipe Fernandes | FF | ocefpaf | conda-forge |
John Kirkham | JK | jakirkham | NVIDIA/cf |
Cheng H. Lee | CHL | chenghlee | Anaconda/cf |
Dave Clements | DPC | tnabtaf | Anaconda |
Jaime Rodríguez-Guerra | JRG | jaimergp | Quansight/cf |
Ken Odegard | KO | kenodegard | Anaconda |
James Loy | JML | jmlphd | Anaconda |
Mark Anderson | MAA | markan | Anaconda |
17 people total
Standing items
- [ ]
From previous meeting(s)
- (HV) Windows Fortran migration (see below)
Active votes
- [ ]
Your new() agenda items
- (JK) Summarize recent Linux build outage issue
- https://github.com/conda-forge/status/issues/144
- all linux runners were down
- package on defaults (Anaconda
main
) uploaded to wrong subdir - broke libsolv in repodata fetch - so early in jobs that it broke everything
- discussed with anaconda, they fixed it
- we removed defaults everywhere
- how to improve:
- process on anaconda side
- conda-index fix to reject bad packages
- JK:
- schema and validation tool for repodata so we can check things
- more tests!
- JRG:
- relax conditions in libsolv that made it barf on this package
- maybe apply patch to our feedstock of libsolv
- CHL
- maybe worth raising how to improve in next week's conda community meeting; e.g., Anaconda customers have long-tail of backwards compatibilty needs (
free
channel still gets used)
- maybe worth raising how to improve in next week's conda community meeting; e.g., Anaconda customers have long-tail of backwards compatibilty needs (
- Wolf:
- we can ask if we can roll back this change - maybe change to warning
- the change: https://github.com/openSUSE/libsolv/blob/2fc0cb31a2010d70f3f9197189730400edbd6ef4/ext/repo_conda.c#L315-L319
- (JK) CUDA 12
- cuda 12 migrator is going!
- 30 feedstocks are done
- changes to commit messaging by Daniel
- (JK) GLIBC 2.28
- docker image PR working - needs to be merged after addressing comments
- interesting discussion about earthly versus Docker for containers.
- MB Was thinking of making changes to smithy, but will change docker images to alma8 instead
- (MB) Big lift remaining is to rebuild all the CDTs
- (JK) Windows + ARM
- Some discussion with MS about CI runners.
- Jaime will be added to thread
- (JK) Drop some older CUDAs
- Probably time to drop CUDA 10.2 at least
- Docker Support Policy
- https://gitlab.com/nvidia/container-images/cuda/blob/master/doc/support-policy.md
- TL;DR some older images are planned for removal soon
- Have explored this before
- What should the new minimum be?
- 11.0+ (lower version of 11?)
- Do we want to keep or drop 11.1 here?
- 11.2+
- CUDA Enhanced Compatibility
- 11.0+ (lower version of 11?)
- (MvN) Pinged Keith about CUDA 10.2 and he is ok dropping
- everyone says we should drop 10.2 (JRG to check with openmm and google colab)
- JRG: Colab is on CUDA 11.8 now! 🎉
- (HV) Boost unification
- (HV) Windows Fortran migration
- Open since 2 years; fortran-on-win-migratability will be necessary for scipy soon to support meson builds
- Migrator work that Isuru wanted has been done - can we start?
- Side benefit: Would enable openblas+openmp on windows
- (JRG) SDG reports: opt-in CI, conda/schemas
- (WV) ceps for new conda-build YAML schema
- (DPC) PackagingCon 2023 CFP and registration are open
- (DPC) Conda newsletter is coming
- Anything in particular from conda-forge to highlight?
- Maybe Asmit's work?
- CUDA 12
- Conda-Forge NumFOCUS
- Look for issues in conda-forge/blog for ideas.
Pushed to next meeting
- [ ]
CFEPs
- [ ]