ci(pip-release): drop the removed buildjet cache provider - #3125
Merged
Conversation
BuildJet is shutting down its GitHub Actions service, so Swatinem/rust-cache removed the provider in v2.9.2 (2026-08-06). The workflow floats on `@v2`, so it picked the removal up automatically and every `linux` job now logs "The `cache-provider` `buildjet` is not valid". The action swallows the error, so the jobs still pass — they just run with no caching at all and a red annotation. Rather than re-point at `cache-provider: github`, drop the step. It was not earning its keep: - rust-cache doesn't include the matrix in its key, and the workspace has a single Cargo.lock, so all 8 expansions of the `linux` job shared one key (`v0-rust-linux-Linux-x64-2e010167-37d8d8b3`) and one ~2.25 GB entry — polluted across 4 target triples and 2 crates. - Job durations are the same with the cache warm and with it gone entirely (12-19 min either way). - The repo cache already sits at ~9.7/10 GB. Moving this 2.25 GB blob onto the GitHub budget would LRU-evict the ci.yml workspace and e2e caches that the PR loop actually depends on — the exact thrash 93b9e5c fixed. Also removes delete-buildjet-cache.yml (a manual `buildjet/cache-delete` dispatch with nothing left to delete) and refreshes the nightly.yml comment that pointed at BuildJet as the escape hatch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Contributor
|
😎 Merged successfully - details. |
trunk-io
Bot
added
trunk-not-ready
trunk-queued
trunk-testing
trunk-tests-passed
and removed
trunk-not-ready
trunk-queued
trunk-testing
labels
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
BuildJet is shutting down its GitHub Actions service, so
Swatinem/rust-cacheremoved thebuildjetcache provider in v2.9.2 (released 2026-08-06).pip-release.ymlfloats on@v2, so it picked the removal up automatically and everylinuxjob now logs:The action swallows the error, so the jobs still pass — they just run with no caching and a red annotation (example).
Rather than re-point at
cache-provider: github, this drops the step, because it wasn't earning its keep:All 8 matrix expansions shared one cache key. rust-cache doesn't include the matrix in its key and the workspace has a single
Cargo.lock, so every expansion restored the samev0-rust-linux-Linux-x64-2e010167-37d8d8b3/ same ~2.25 GB blob — polluted across 4 target triples and 2 crates, populated by whichever expansion won the save race.Removing it costs nothing measurable. Durations with the cache warm (run of 2026-08-05) vs. with it gone entirely (today):
Moving it onto the GitHub budget would thrash. The repo cache is at ~9.7/10 GB across 23 entries; this 2.25 GB blob is ~23% of the cap and would LRU-evict the
e2e(1.69 GB) andworkspace(1.27 GB ×2) caches the PR loop depends on — the exact thrash 93b9e5c fixed. Keying it per-target so the 8 jobs stop colliding would need ~18 GB, which doesn't fit either.Reasoning is recorded inline so the step doesn't get re-added. Also removes
delete-buildjet-cache.yml(a manualbuildjet/cache-deletedispatch with nothing left to delete) and refreshes thenightly.ymlcomment that recommended BuildJet as the escape hatch for nightly's cache budget.Unrelated to the macOS wheel failure in that same run — that was a transient
DEPTH_ZERO_SELF_SIGNED_CERTwhile maturin-action downloaded maturin.