ci(test): committed gap-suite runner + informational smoke-parity gate#5222
ci(test): committed gap-suite runner + informational smoke-parity gate#5222TheHypnoo wants to merge 5 commits into
Conversation
Add scripts/run_gap_tests.sh — a committed runner for the gap suite (test-files/test_gap_*.ts, 235 tests), replacing the out-of-repo /tmp/run_gap_tests.sh that CLAUDE.md pointed at. It is a thin wrapper over run_parity_tests.sh --filter test_gap_, so it reuses the one canonical normalizer, skip-list, and output cap (seed of the single-normalizer work), and gates on "no NEW failures vs known_failures.json" rather than run_parity_tests.sh's loose <80%-aggregate exit. Wire it into a new smoke-parity CI job. The gap suite is the highest-signal- per-second test Perry has and had no committed runner and no PR gate — a single-feature regression could merge green. The job is INFORMATIONAL for now (continue-on-error): the first runs surface which gap tests fail on the Linux image under node 26 so they can be triaged into known_failures.json; once curated + green, a follow-up drops continue-on-error and branch protection makes it required. Uses node 26 (the node-suite baseline oracle) and only allow-listed actions (no sccache), so it is not blocked by the org action allow-list. Also fix the stale CLAUDE.md parity-status line (28 -> 235 tests, /tmp -> scripts/run_gap_tests.sh). No source changes; no existing job altered.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughAdds ChangesGap test smoke gate
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/test.yml:
- Around line 284-295: Replace all mutable version tags with full commit SHAs
for the four GitHub Actions used in the smoke-parity job in the test.yml
workflow file. Update actions/checkout from `@v6` to its commit SHA,
dtolnay/rust-toolchain from `@stable` to its commit SHA, Swatinem/rust-cache from
`@v2` to its commit SHA, and actions/setup-node from `@v6` to its commit SHA. Each
action should be specified in the format owner/repo@<full-commit-sha> to ensure
reproducibility and prevent supply-chain drift.
In `@scripts/run_gap_tests.sh`:
- Around line 50-60: The script uses fixed temporary filenames like
/tmp/gap_all_fails.txt, /tmp/gap_known.txt, and /tmp/gap_new.txt which can
conflict when multiple script instances run concurrently. Replace all three
fixed temp filenames with dynamically generated unique names using mktemp or by
incorporating a process-specific identifier (such as $$) into each filename.
Update the assignments where these temp files are created from $REPORT, $KNOWN,
and the comm command output, and ensure the wc -l command references the correct
updated variable name for /tmp/gap_all_fails.txt.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 301df9eb-f909-4158-8e87-a5913c9fbe0e
📒 Files selected for processing (3)
.github/workflows/test.ymlCLAUDE.mdscripts/run_gap_tests.sh
| - uses: actions/checkout@v6 | ||
|
|
||
| - name: Install Rust toolchain | ||
| uses: dtolnay/rust-toolchain@stable | ||
|
|
||
| - uses: Swatinem/rust-cache@v2 | ||
| with: | ||
| shared-key: "${{ runner.os }}-perry" | ||
| save-if: ${{ github.ref == 'refs/heads/main' }} | ||
|
|
||
| - name: Setup Node.js | ||
| uses: actions/setup-node@v6 |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify unpinned action refs in workflow files.
# Expected after fix: no matches.
rg -nP '^\s*-\s*uses:\s*[^@\s]+@(?![0-9a-fA-F]{40}\b).+$' .github/workflows/*.ymlRepository: PerryTS/perry
Length of output: 4338
🏁 Script executed:
# Extract the job context around lines 279-300 to verify if this is a new smoke-parity job
sed -n '279,300p' .github/workflows/test.yml | cat -n
# Also check if all actions in the snippet (284-295) are unpinned
echo "--- Checking lines 284-295 for unpinned actions ---"
sed -n '284,295p' .github/workflows/test.yml | cat -nRepository: PerryTS/perry
Length of output: 1241
Pin all GitHub Actions in the smoke-parity job to commit SHAs.
Four actions use mutable version tags instead of commit SHAs:
actions/checkout@v6(line 284)dtolnay/rust-toolchain@stable(line 289)Swatinem/rust-cache@v2(line 291)actions/setup-node@v6(line 294)
Pinning to full commit SHAs is required to prevent supply-chain drift and ensure policy compliance.
🧰 Tools
🪛 zizmor (1.25.2)
[warning] 284-284: credential persistence through GitHub Actions artifacts (artipacked): does not set persist-credentials: false
(artipacked)
[error] 284-284: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)
(unpinned-uses)
[error] 287-287: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)
(unpinned-uses)
[error] 289-289: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)
(unpinned-uses)
[error] 295-295: unpinned action reference (unpinned-uses): action is not pinned to a hash (required by blanket policy)
(unpinned-uses)
[error] 289-289: runtime artifacts potentially vulnerable to a cache poisoning attack (cache-poisoning): enables caching by default
(cache-poisoning)
[error] 295-295: runtime artifacts potentially vulnerable to a cache poisoning attack (cache-poisoning): enables caching by default
(cache-poisoning)
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In @.github/workflows/test.yml around lines 284 - 295, Replace all mutable
version tags with full commit SHAs for the four GitHub Actions used in the
smoke-parity job in the test.yml workflow file. Update actions/checkout from `@v6`
to its commit SHA, dtolnay/rust-toolchain from `@stable` to its commit SHA,
Swatinem/rust-cache from `@v2` to its commit SHA, and actions/setup-node from `@v6`
to its commit SHA. Each action should be specified in the format
owner/repo@<full-commit-sha> to ensure reproducibility and prevent supply-chain
drift.
Source: Linters/SAST tools
There was a problem hiding this comment.
♻️ Duplicate comments (1)
.github/workflows/test.yml (1)
124-126:⚠️ Potential issue | 🟠 Major | ⚡ Quick winPin
mozilla-actions/sccache-actionto a full commit SHA.Line 125, Line 206, and Line 402 use a mutable tag (
@v0.0.10). For CI supply-chain integrity and reproducibility, this needsowner/repo@<40-char-sha>pinning.Suggested change
- - name: Start sccache - uses: mozilla-actions/sccache-action@v0.0.10 + - name: Start sccache + uses: mozilla-actions/sccache-action@<full-commit-sha>Also applies to: 205-207, 401-403
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/test.yml around lines 124 - 126, The mozilla-actions/sccache-action action is pinned to a mutable version tag (`@v0.0.10`) at three locations in the workflow file, which compromises reproducibility and supply-chain security. Replace the mutable tag `@v0.0.10` with a full 40-character commit SHA for the mozilla-actions/sccache-action action at all three affected locations: .github/workflows/test.yml lines 125, 205-207, and 401-403. Use the same commit SHA across all three locations to ensure consistent behavior.Source: Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Duplicate comments:
In @.github/workflows/test.yml:
- Around line 124-126: The mozilla-actions/sccache-action action is pinned to a
mutable version tag (`@v0.0.10`) at three locations in the workflow file, which
compromises reproducibility and supply-chain security. Replace the mutable tag
`@v0.0.10` with a full 40-character commit SHA for the
mozilla-actions/sccache-action action at all three affected locations:
.github/workflows/test.yml lines 125, 205-207, and 401-403. Use the same commit
SHA across all three locations to ensure consistent behavior.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 7070c0c8-861a-47aa-ba29-e1f0a32ed83b
📒 Files selected for processing (2)
.github/workflows/test.ymlCLAUDE.md
✅ Files skipped from review due to trivial changes (1)
- CLAUDE.md
|
Good catch — fixed in the latest commit. |
run_gap_tests.sh wrote its failure lists to fixed /tmp/gap_*.txt names, so concurrent runs (a second PR, local + CI on the same box, or the upcoming node-suite-guard alongside) could clobber each other and produce a false gate result. Allocate a run-scoped dir with mktemp -d and rm -rf it on EXIT.
1 participant
What
scripts/run_gap_tests.shfor the 235-test gap suite (test-files/test_gap_*.ts), replacing the out-of-repo/tmp/run_gap_tests.shthatCLAUDE.mdpointed at.smoke-parityCI job that runs it. Informational (continue-on-error) for now.CLAUDE.mdparity-status line (28 → 235 tests;/tmp→scripts/run_gap_tests.sh).Why
The gap suite — AOT-compile each
test_gap_*.tsand diff byte-for-byte againstnode --experimental-strip-types— is the highest-signal, most deterministic test Perry has, yet it had no committed runner (CLAUDE.mdpointed at/tmp) and no CI gate. The full parity job is opt-in (tag/label only) and runs node 22, so a single-feature regression can merge green.Design
run_gap_tests.shis a thin wrapper overrun_parity_tests.sh --filter test_gap_, so it reuses the one canonical normalizer + skip-list + per-test output cap (seed of the single-normalizer cleanup) instead of inventing a third differ.test-parity/known_failures.json. (run_parity_tests.sh's own exit code only trips below 80% aggregate parity — far too loose to catch a single-feature regression.)actions/checkout,dtolnay/rust-toolchain,Swatinem/rust-cache,actions/setup-node) — no sccache — so it is not blocked by the org action allow-list.Rollout (intentionally staged)
continue-on-error(informational). The first runs tell us which gap tests currently fail on the Linux image under node 26.known_failures.json(or fix them).continue-on-errorand addsmoke-parityto the branch-protection required checks.Part of a tiered-CI cost-reduction effort. No source changes; no existing job altered.
Summary by CodeRabbit
Release Notes
Tests
Documentation