Skip to content

Conversation

jumerckx
Copy link
Collaborator

@jumerckx jumerckx commented Aug 5, 2025

I'm not yet sure this does anything. But if this would work, we could split the tests in many more separate jobs.

@jumerckx jumerckx force-pushed the jm/ci branch 2 times, most recently from eca8191 to b5b55c6 Compare August 5, 2025 14:02
Comment on lines 36 to 39
version: ["1.10", "1.11"]
os: [ubuntu-24.04, ubuntu-22.04-arm, macOS-latest]
runtime: ["PJRT", "IFRT"]
assertions: [false, true]
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For what is worth, I personally like expanding lists in the one-per-line format, like in

- "1.10"
- "1.11"
# - 'nightly'
os:
- ubuntu-24.04
# `ubuntu-22.04-arm` is considered more stable than `ubuntu-24.04-arm`:
# <https://github.com/orgs/community/discussions/148648#discussioncomment-12099554>.
- ubuntu-22.04-arm
# Disable `macOS-13` until
# <https://github.com/EnzymeAD/Reactant.jl/issues/867> is resolved.
# - macOS-13
- macOS-latest
test_group:
- core
- neural_networks
- integration
runtime:
- "PJRT"
- "IFRT"
assertions:
- false
which makes it easy to add remove individual items without messing up with git-blame.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, I'll take this into account.
I'm still mainly messing around with things for now.

Do you think it makes sense to have a cache warmup step with cache keys independent of jobs and testgroup?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If it works, it can make sense.

@jumerckx
Copy link
Collaborator Author

jumerckx commented Aug 5, 2025

packages are still being precompiled in the test step. I'm looking for a culprit, this seems suspicious:

┌ Warning: Could not use exact versions of packages in manifest, re-resolving
└ @ Pkg.Operations /opt/hostedtoolcache/julia/1.10.10/aarch64/share/julia/stdlib/v1.10/Pkg/src/Operations.jl:1829

@giordano
Copy link
Member

giordano commented Aug 5, 2025

You may want to drastically reduce the matrix (ideally to a single element) while you investigate how to make this work, otherwise you'll waste loads of resources and everybody's time.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants