test(gc): re-pin ratchet after accepted collector shifts

[proggeramlug]

Closes #7843

What changed

• Re-pin the internal Perry-vs-Perry GC ratchet on the quiet M1 mini at measured source commit 98e9ecdb5 (perry 0.5.1484).
• Record the full benchmark suite and the exact compiler/runtime/stdlib hashes in the generated artifact.
• Explain the accepted shifts in the artifact notes instead of weakening tolerances or removing gating rows.

The artifact was measured before rebasing this artifact-only branch over #7907/#7912, so its recorded commit and code-tree provenance remain the exact tree that produced the numbers.

Reproduction and acceptance rationale

The old a1fd5a7e8 pin fails on frozen main in two independent seven-repeat runs with byte-identical shared-CI counters:

• 03_cross_gen_writes: two additional copied objects, copied_bytes 590,688 -> 656,256 (+65,568), while retained heap and RSS both fall.
• 14_grow_then_churn: 19 minors / 20 steps, 13,013 copied objects, 403,514 promoted objects, and 151,713,056 promoted bytes in both runs. heapUsed falls 91.4%; pinned-host RSS rises about 4x.

These are consequences of already reviewed and merged collector changes, not measurement noise:

• fix(gc): report live arena bytes after collection (#7879) #7886 replaced arena high-water heapUsed with live-byte accounting, explaining the large retention improvements.
• perf(gc): promote a fully-live young generation without tracing it — retain −33.6%, deeplist −43% #7888 deliberately allows one bounded untraced promotion before a retain-to-churn phase flip is measured and documented the same adversarial footprint tradeoff.
• perf: shapes 0.139 -> 0.061 s (beats node) and pipeline 0.240 -> 0.175 s — a 2000-element array was born immortal, and this.vals[i]=v had no inline arm #7895 moved medium array backing stores into the nursery and re-derived the untraced-promotion threshold, changing the work performed by this exact grow-then-churn shape.
• perf(gc): describe a promoted block's page-object list instead of storing it #7914 removes permanent per-object promoted-page index storage and is included in the measured source tree.

No tolerance or probe definition changes. The new artifact passes both gate profiles against the fresh measurement.

Validation

• PYTHONPATH="$PWD" python3 tests/test_gc_ratchet.py — 91 passed
• gc_ratchet.py validate — passed
• gc_ratchet.py check --profile shared_ci — passed against the new pin
• gc_ratchet.py check --profile pinned_host — passed against the new pin
• full quiet-host pin driver — exit 0; all Node oracles passed
• git diff --check — passed
• bash scripts/check_file_size.sh — passed

Adjacent observation

The pin driver records the full suite with --warn-only. It reported existing/current timing regressions for Fibonacci (+23%) and bench_numeric_array_numeric (+1000%) while all correctness checks passed. This PR does not open a separate issue or mix those unrelated performance investigations into the GC ratchet re-pin.

Summary by CodeRabbit

• Documentation
  • Updated the changelog to document revised garbage-collection performance baselines.
  • Clarified accounting for live bytes, nursery allocation, promotion, pinned memory, and grow-then-churn behavior.
  • Existing probes and tolerances remain unchanged.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 8bda66d7-5f09-42b3-9b3b-6ce28f2bd9f7

📥 Commits

Reviewing files that changed from the base of the PR and between c109b08 and 7f22b8b.

📒 Files selected for processing (2)

• benchmarks/gc_ratchet/baseline/gc-ratchet-v1.json
• changelog.d/7921-gc-ratchet-repin.md

---

📝 Walkthrough

Walkthrough

The PR adds a changelog entry for the re-pinned GC ratchet baseline. The entry records updated live-byte accounting, nursery allocation, bounded untraced promotion, grow-then-churn behavior, promotion work, and pinned-host footprint.

Changes

GC ratchet documentation

Layer / File(s)	Summary
Baseline re-pin changelog changelog.d/7921-gc-ratchet-repin.md	Documents the re-pinned GC ratchet baseline and its recorded collector behavior and constraints.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• PerryTS/perry#7783: Introduced the grow-then-churn probe and related GC accounting updates.
• PerryTS/perry#7432: Added adaptive tenuring and live-byte accounting changes associated with GC ratchet baseline updates.
• PerryTS/perry#7557: Modified GC ratchet baselines, tolerances, and probe accounting.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	The changeset contains only a changelog entry and does not include the generated ratchet artifact required to re-pin the baseline.	Add the generated GC ratchet artifact and required baseline metadata so the PR performs the re-pin without changing probes or tolerances.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the GC ratchet re-pin and its cause: accepted collector shifts.
Description check	✅ Passed	The description explains the change, linked issue, rationale, validation results, and unrelated observations in sufficient detail.
Out of Scope Changes check	✅ Passed	The changelog entry documents the GC ratchet re-pin and accepted collector shifts, which are within the linked issue objectives.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/7843-gc-ratchet

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}