ci: save rust-cache only on main/staging pushes by ilblackdragon · Pull Request #2609 · nearai/ironclaw

ilblackdragon · 2026-04-18T04:24:55Z

Summary

Adds save-if gating to every Swatinem/rust-cache step in code_style.yml and test.yml so the cache is only written when github.ref is refs/heads/main or refs/heads/staging. PR jobs still restore — they just skip the upload.
Adds push triggers on main and staging to code_style.yml so those pushes actually exercise each matrix leg and refresh the cache that PRs will restore from. Without this, the save-if guard would mean the cache is never refreshed.
Gates no-panics on pull_request events (it diffs against github.event.pull_request.base.sha, which is empty on pushes) and teaches the code-style roll-up to accept "skipped" for it on push.

Motivation

Right now on #2471's latest CI run, each Clippy matrix leg shows:

Step	Duration
`cargo clippy`	~5m 36s
Post rust-cache (save)	in progress, already > 5m

The actions cache save step is compressing and uploading a multi-GB `target/` directory on every PR job, even though PR jobs have unique cache keys and no other PR will read them. Only pushes to long-lived branches should refresh the cache.

Expected impact: ~5 min saved per Clippy / Tests matrix leg on PRs. No change to cache hit rate for subsequent PRs — main / staging pushes refresh the same keys PRs restore from.

Test plan

This PR itself should show each Clippy matrix leg finishing roughly when `cargo clippy` finishes — no trailing cache-save wall clock
After merge, a push to `staging` should run the code_style workflow and save each cache key
Subsequent PRs should get warm cache restores on the main/staging-saved keys

gemini-code-assist · 2026-04-18T04:25:01Z

Note

Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported.

The three clippy matrix entries (all-features, default, libsql-only) each had a distinct cache key, so each leg maintained its own multi-GB target/ in the GitHub Actions cache. GHA caps repo cache at 10 GB and evicts LRU, so these three near-duplicates crowd out other useful entries and forced each leg to re-warm after eviction. Switch to shared-key: clippy for the Linux matrix and shared-key: clippy-windows for the Windows matrix. Whichever leg finishes (and saves) first wins the slot; the other two restore from that cache on the next run. Because --all-features builds a superset of the artifacts needed by default / libsql-only, a shared cache is incrementally useful for every leg even when the saver wasn't all-features. This is independent of and complementary to #2609 (save-if gating); together they reduce total cache churn per clippy job to near-zero wall clock on PR runs after the first main/staging push refreshes the slot.

Copilot

Pull request overview

Adjusts GitHub Actions caching behavior to avoid spending minutes uploading Rust target/ caches on PR runs, while still keeping caches warm via long-lived branches.

Changes:

Add save-if to all Swatinem/rust-cache steps in test.yml and code_style.yml to skip cache uploads except on main/staging.
Add push triggers for code_style.yml on main and staging to refresh caches outside PRs.
Gate no-panics to pull_request only and update the roll-up job to treat it as allowed-to-be-skipped on pushes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
.github/workflows/test.yml	Adds `save-if` to rust-cache steps so PR runs restore without uploading caches.
.github/workflows/code_style.yml	Adds push triggers + `save-if` on rust-cache; gates `no-panics` to PR-only and updates roll-up logic to allow skipped.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-18T04:28:16Z

@@ -101,11 +109,14 @@ jobs:
    - uses: Swatinem/rust-cache@e18b497796c12c097a38f9edb9d0641fb99eee32 # v2
      with:
        key: clippy-windows-${{ matrix.name }}


In clippy-windows, the rust-cache entry keyed as clippy-windows-${{ matrix.name }} will likely never get written after this change: the job only runs on pull_request (because github.base_ref is empty on push), and PR runs now have save-if false because github.ref is refs/pull/.... This means restores for this job will always be cold. Consider either (a) populating this cache on a push-triggered workflow/job, or (b) reusing an existing Windows cache key that is saved on pushes (e.g., align with windows-${{ matrix.name }} from test.yml) so clippy-windows can restore something that actually exists.

Suggested change

key: clippy-windows-${{ matrix.name }}

key: windows-${{ matrix.name }}

Good catch — fixed. clippy-windows now uses key: windows-${{ matrix.name }} (matches test.yml's windows-build), so it restores from a cache that test.yml's push-to-main runs actually populate. Kept the save-if on the clippy-windows step for defense but in practice it won't trigger (the job only runs on PRs).

Copilot · 2026-04-18T04:28:16Z

      - uses: Swatinem/rust-cache@e18b497796c12c097a38f9edb9d0641fb99eee32 # v2
        with:
          key: ${{ matrix.name }}
+          save-if: ${{ github.ref == 'refs/heads/main' || github.ref == 'refs/heads/staging' }}


save-if currently checks only github.ref. When this workflow is invoked via workflow_call (e.g., from scheduled/dispatch workflows) github.ref can still be refs/heads/main, which would allow cache uploads even though the event isn’t a push. If the intent is truly “save only on main/staging pushes”, gate additionally on github.event_name == 'push' (or otherwise ensure non-push callers don’t upload).

Right — when staging-ci.yml calls test.yml via workflow_call from its hourly schedule, github.ref resolves to refs/heads/main and the save-if would fire. Updated all save-if guards in test.yml (and code_style.yml for consistency) to also check github.event_name == 'push' so only real pushes save.

* ci: share rust-cache across clippy matrix legs The three clippy matrix entries (all-features, default, libsql-only) each had a distinct cache key, so each leg maintained its own multi-GB target/ in the GitHub Actions cache. GHA caps repo cache at 10 GB and evicts LRU, so these three near-duplicates crowd out other useful entries and forced each leg to re-warm after eviction. Switch to shared-key: clippy for the Linux matrix and shared-key: clippy-windows for the Windows matrix. Whichever leg finishes (and saves) first wins the slot; the other two restore from that cache on the next run. Because --all-features builds a superset of the artifacts needed by default / libsql-only, a shared cache is incrementally useful for every leg even when the saver wasn't all-features. This is independent of and complementary to #2609 (save-if gating); together they reduce total cache churn per clippy job to near-zero wall clock on PR runs after the first main/staging push refreshes the slot. * ci(clippy): run only all-features on push Adds if: github.event_name == 'pull_request' || matrix.name == 'all-features' to the Linux clippy matrix so push events (main/staging) run only the all-features leg. This pins the saver for the new shared-key slot: --all-features builds a superset of the artifacts needed by default and libsql-only, so when those PR legs cache-restore they always start from the richest possible baseline rather than whichever leg happened to win a three-way race. PRs still run all three legs, so lint coverage is unchanged on the path that matters (before merge). Push events are only exercised after a PR has already passed, so the redundant two legs were just warming a cache anyway.

Swatinem/rust-cache was saving the multi-GB target/ directory at the end of every CI job — including PR jobs that will never be read from again by a different PR. The post-cache step was adding 5+ minutes of wall clock to each Clippy and Tests matrix leg, on top of the actual compile time. This change: - Adds save-if gating to every rust-cache step in code_style.yml and test.yml so the cache is only written when github.ref is refs/heads/main or refs/heads/staging. PR jobs still restore, they just skip the upload. - Adds push triggers on main and staging to code_style.yml so those pushes actually exercise each matrix leg and refresh the cache entry that PRs will restore from next. Without this, the cache would never be refreshed after the one-time manual warm-up. - Gates no-panics on pull_request events since it diffs against github.event.pull_request.base.sha, which is empty on push events. Expected impact on PR wall clock: ~5 minutes saved per Clippy / Tests matrix leg, no change to cache hit rate for subsequent PRs.

Combines five reductions targeting the same root cause: clippy and cargo check were running the same compile work multiple times across matrix legs and platforms. - Drop `windows-build` from test.yml. Its `cargo check` matrix is fully subsumed by `clippy-windows`, which already type-checks `--all --benches --tests --examples` on the same configs. - Collapse PR-blocking clippy to `--all-features` only (Linux + Windows). Lint findings are almost never feature-gated, and the remaining configs (`default`, `libsql-only`) now run as `clippy-extra` / `clippy-windows-extra` jobs gated to `push` events on long-lived branches, where they verify feature combinations without slowing PRs. - Share rust-cache slot between clippy and tests via `shared-key: build` (Linux) and `shared-key: build-windows`. Clippy and `cargo test` produce overlapping artifacts for the same feature set, so a single slot serves both jobs and avoids duplicating multi-GB target/ caches. - Replace `cargo install cargo-component --locked || true` with a composite action backed by `taiki-e/install-action`. Drops a multi-minute source build to seconds and stops swallowing failures. Applied to test.yml, coverage.yml, and release.yml. - Add `paths-ignore` filters for top-level docs, LICENSE files, PNG assets, issue/PR templates, and `docs/**` (mintlify site). Conservative scope: any `.md` file actually `include_str!()`'d by the binary (workspace seeds, engine prompts, channel READMEs) remains unfiltered. Includes `save-if` guards on every rust-cache step so PR jobs are restore-only — complementary to #2609 and #2610. Gates `no-panics` to PR events (it diffs against the PR base SHA, unavailable on push) and teaches the `code-style` roll-up to accept `skipped` for the push-only / PR-only conditional jobs.

* ci: share rust-cache across clippy matrix legs The three clippy matrix entries (all-features, default, libsql-only) each had a distinct cache key, so each leg maintained its own multi-GB target/ in the GitHub Actions cache. GHA caps repo cache at 10 GB and evicts LRU, so these three near-duplicates crowd out other useful entries and forced each leg to re-warm after eviction. Switch to shared-key: clippy for the Linux matrix and shared-key: clippy-windows for the Windows matrix. Whichever leg finishes (and saves) first wins the slot; the other two restore from that cache on the next run. Because --all-features builds a superset of the artifacts needed by default / libsql-only, a shared cache is incrementally useful for every leg even when the saver wasn't all-features. This is independent of and complementary to nearai#2609 (save-if gating); together they reduce total cache churn per clippy job to near-zero wall clock on PR runs after the first main/staging push refreshes the slot. * ci(clippy): run only all-features on push Adds if: github.event_name == 'pull_request' || matrix.name == 'all-features' to the Linux clippy matrix so push events (main/staging) run only the all-features leg. This pins the saver for the new shared-key slot: --all-features builds a superset of the artifacts needed by default and libsql-only, so when those PR legs cache-restore they always start from the richest possible baseline rather than whichever leg happened to win a three-way race. PRs still run all three legs, so lint coverage is unchanged on the path that matters (before merge). Push events are only exercised after a PR has already passed, so the redundant two legs were just warming a cache anyway.

Swatinem/rust-cache was saving the multi-GB target/ directory at the end of every CI job — including PR jobs that will never be read from again by a different PR. The post-cache step was adding 5+ minutes of wall clock to each Clippy and Tests matrix leg, on top of the actual compile time. This change: - Adds save-if gating to every rust-cache step in code_style.yml and test.yml so the cache is only written when github.ref is refs/heads/main or refs/heads/staging. PR jobs still restore, they just skip the upload. - Adds push triggers on main and staging to code_style.yml so those pushes actually exercise each matrix leg and refresh the cache entry that PRs will restore from next. Without this, the cache would never be refreshed after the one-time manual warm-up. - Gates no-panics on pull_request events since it diffs against github.event.pull_request.base.sha, which is empty on push events. Expected impact on PR wall clock: ~5 minutes saved per Clippy / Tests matrix leg, no change to cache hit rate for subsequent PRs.

Copilot AI review requested due to automatic review settings April 18, 2026 04:24

github-actions Bot added scope: ci CI/CD workflows size: S 10-49 changed lines risk: medium Business logic, config, or moderate-risk modules contributor: core 20+ merged PRs labels Apr 18, 2026

Copilot started reviewing on behalf of ilblackdragon April 18, 2026 04:25 View session

ilblackdragon mentioned this pull request Apr 18, 2026

ci: share rust-cache across clippy matrix legs #2610

Merged

4 tasks

Copilot AI reviewed Apr 18, 2026

View reviewed changes

ilblackdragon force-pushed the ci/clippy-cache-save-if branch from ac1e66b to f15656c Compare April 18, 2026 05:05

ilblackdragon merged commit b93098d into staging Apr 18, 2026
7 checks passed

ilblackdragon deleted the ci/clippy-cache-save-if branch April 18, 2026 05:07

ilblackdragon mentioned this pull request Apr 18, 2026

ci: drop duplicate matrix legs and consolidate caches #2614

Closed

4 tasks

github-actions Bot mentioned this pull request Apr 18, 2026

chore: promote staging to staging-promote/158f5f7d-24597161022 (2026-04-18 05:31 UTC) #2615

Merged

ilblackdragon mentioned this pull request Apr 18, 2026

ci: speed up CI feedback loop #2566

Merged

6 tasks

henrypark133 mentioned this pull request Apr 21, 2026

chore: release #2606

Merged

coderabbitai Bot mentioned this pull request Jun 4, 2026

Man pages: render from the CLI, ship on every channel kjanat/runner#37

Merged

Navigation Menu

ci: save rust-cache only on main/staging pushes - #2609

ci: save rust-cache only on main/staging pushes#2609
ilblackdragon merged 1 commit into
stagingfrom
ci/clippy-cache-save-if

ilblackdragon commented Apr 18, 2026

Uh oh!

gemini-code-assist Bot commented Apr 18, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

ilblackdragon Apr 18, 2026

Uh oh!

Copilot AI Apr 18, 2026

Uh oh!

ilblackdragon Apr 18, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	key: clippy-windows-${{ matrix.name }}
	key: windows-${{ matrix.name }}

Conversation

ilblackdragon commented Apr 18, 2026

Summary

Motivation

Test plan

Uh oh!

gemini-code-assist Bot commented Apr 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

ilblackdragon Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

ilblackdragon Apr 18, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants