fix(gc): stop the from-space scan reporting dead relocation stubs as offenders

[proggeramlug]

Summary

PERRY_GC_FROMSPACE_SCAN skips owners that are themselves in from-space — they are about to be reclaimed, so a dead object legitimately still points at its dead peers. It did not skip owners carrying GC_FLAG_FORWARDED, which are dead for exactly the same reason one step further on, and which — unlike a from-space object — can sit in old-gen where the space check never reaches them. Old-gen defrag and array growth both leave such stubs behind, so every one of them was reported as an offender.

Measured on a Perry-compiled zod workload, these were the single largest population in the residue after #7179: all at 2^k - 1 element indices of repeatedly-grown arrays (the last element written before each capacity doubling), i.e. the stale copy of a growing array, reported once per growth generation. #7154's triage has been reading these counts.

The skip is counted, not silent

[gc-fromspace-scan OFFENDERS] objects=… words=… fwd_owners_skipped=N missing_rewrites=… dangling=…

A filter that shrinks the offender count without saying so reads exactly like progress, which is the failure mode this instrument exists to prevent (#6942, #7024). The count is in the report line so a suppressed population can never be mistaken for a fixed one.

Test

fromspace_scan_skips_but_counts_forwarded_owners_7154 asserts both directions, in one test, against the same planted reference:

1. planted reference is reported while the holder is live (the premise — otherwise the test could pass vacuously),
2. the identical reference stops being reported once the holder is marked forwarded,
3. forwarded_owners_skipped moved.

Sabotage (predicate removed): FAILED — "a FORWARDED owner is a dead relocation stub — its stale payload must not be reported as a missing rewrite". Green with the fix. cargo test -p perry-runtime --lib gc::tests::fromspace_scan: 4 passed, 0 failed.

Instrument only — no collector behaviour changes.

Refs #7154.

Summary by CodeRabbit

• Bug Fixes

  • Improved garbage-collection scan reporting by excluding stale references from objects already marked as forwarded.
  • Added a separate count for skipped forwarded objects in scan results.

• Tests

  • Added regression coverage confirming forwarded objects are ignored during scanning while still being counted correctly.

[coderabbitai]

📝 Walkthrough

Walkthrough

The from-space scanner now skips forwarded owners because they are dead relocation stubs. It counts skipped owners in FromSpaceScanReport and adds regression coverage for stale old-to-young references.

Changes

From-space scan instrumentation

Layer / File(s)	Summary
Forwarded-owner filtering and reporting crates/perry-runtime/src/gc/fromspace_scan.rs, changelog.d/7182-fromspace-scan-forwarded-owners.md	The scanner skips forwarded owners before scanning their payloads. The report includes forwarded_owners_skipped. The changelog documents the instrumentation-only behavior.
Forwarded-owner regression coverage crates/perry-runtime/src/gc/tests/fromspace_scan.rs	The test verifies that stale references from forwarded owners are excluded and that the skip count increases. Cleanup restores the test state.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• PerryTS/perry#7041: Introduced the fromspace_scan implementation and tests extended by this change.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the main change: preventing the from-space scan from reporting dead relocation stubs as offenders.
Description check	✅ Passed	The description clearly explains the problem, solution, metrics, regression test, verification command, issue reference, and scope.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/7154-residue-offender

---

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.}

[coderabbitai]

Actionable comments posted: 1

🤖 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 `@crates/perry-runtime/src/gc/fromspace_scan.rs`:
- Around line 25-29: Update the documentation around scan_object to distinguish
the two skip paths: from-space owners return before the forwarded-owner check
and are not included in fwd_owners_skipped, while that counter only tracks
owners skipped because they carry GC_FLAG_FORWARDED.

🪄 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: b88f6f48-8251-433d-a1be-a37a39deb076

📥 Commits

Reviewing files that changed from the base of the PR and between 0bd77ca and 5b11ab6.

📒 Files selected for processing (3)

• changelog.d/7182-fromspace-scan-forwarded-owners.md
• crates/perry-runtime/src/gc/fromspace_scan.rs
• crates/perry-runtime/src/gc/tests/fromspace_scan.rs

[coderabbitai]

📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win

Correct the scan-count description.

Lines 25-29 say both skip populations are counted by fwd_owners_skipped. scan_object returns for is_from_space(owner_space) before it evaluates GC_FLAG_FORWARDED. The counter only represents owners skipped by the forwarded-owner predicate. State that distinction.

🤖 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 `@crates/perry-runtime/src/gc/fromspace_scan.rs` around lines 25 - 29, Update
the documentation around scan_object to distinguish the two skip paths:
from-space owners return before the forwarded-owner check and are not included
in fwd_owners_skipped, while that counter only tracks owners skipped because
they carry GC_FLAG_FORWARDED.