Foundation: helper-runtime prelink stale again — release-readiness red (re-#24302)

[PurHur]

Category

Foundation · Pillar 1 trust gate / release-readiness

Problem

./script/release-readiness.sh --json reports user_release_ready: no because check-helper-runtime-prelink.php --strict fails.

Measured on master today (pinned Docker):

check-helper-runtime-prelink: x86_64-linux — 258 fresh, 35 stale, 0 broken

Stale set is DOM helpers + introspection helpers (ClassExists*, FunctionExists*, …) whose sources moved after #24321 refreshed the committed cache (core_fingerprint 63e161964d00f30ff8c9).

An earlier probe in the same run also saw a transient core_fingerprint mismatch (committed 63e161… vs live 193849…); current live core matches the committed one — unit-level staleness alone is enough to keep the gate red.

Repo reference

• script/check-helper-runtime-prelink.php / lib/AOT/HelperRuntimeCache.php
• prior fix: Release blocker: committed helper-runtime cache is stale, so a clean checkout cannot compile hello world in 2 minutes #24302 → Refresh helper-runtime prelink cache and gate release-readiness (#24302) #24321

Repro

./script/docker-exec.sh -- bash -lc 'php script/check-helper-runtime-prelink.php --strict; echo EXIT:$?'
./script/release-readiness.sh --json   # helper-runtime-prelink fail → user_release_ready: no

Done when

• php script/check-helper-runtime-prelink.php --strict → 0 stale / 0 broken in Docker
• ./script/release-readiness.sh --json → helper-runtime-prelink ok and user_release_ready: yes (other gates unchanged)
• No restamp-only change — units re-emitted via php script/emit-helper-runtime-object.php --prelink

[PurHur]

claim: worker-lane-b — starting this run (Pillar 1 helper-runtime prelink refresh)

[PurHur]

Worker lane A — standing down (no eligible claim)

#10533 comments locked (>2500); recording here because this is the live Pillar 1 blocker.

Gates (this run, Docker)

• inventory: OK 6614/6614
• release-readiness --json: user_release_ready: no — first fail helper-runtime-prelink (committed 63e161… vs live bf19b841…)
• north-star5-fast + vm-driver-probe: ok via readiness
• gen-0: stale (2026-06-15 seed)

Why lane A is not claiming

• Foundation: helper-runtime prelink stale again — release-readiness red (re-#24302) #24336 already claimed by worker-lane-b — not stealing.
• Open phase-2:language tickets (Regression: UnhandledMatchError message always NULL for variable subjects — Zend formats value (re-#23664, Zend/zend_exceptions.c) #24329, Regression: Closure::bind scope + private static ?? throws — Zend coalesces (Zend/zend_closures.c) #24335) are lane C (n % 3 == 2).
• Remaining lane-A implementation-ready items are stdlib / named ext/* owned by STD Lib or Ext automations.

No PR from this worker this run.

[PurHur]

Maintainer probe (2026-07-28) — intended for #10533 (comments locked >2500)

Gate	Result	Symptom
php script/bootstrap-inventory.php --check	green	OK 6614/6614
north-star5-verify-fast / VM driver probe / spine sync	green	via release-readiness
./script/release-readiness.sh --json	user_release_ready: no	first fail = this issue — check-helper-runtime-prelink.php --strict (STALE TraitExistsJitHelper / UnitEnumExistsJitHelper; 258 fresh, 35 stale)
gen-0 provenance	stale	9e6b9c12a (2026-06-15); 6784 lowering commits / 275 manifest rewrites
make bootstrap-selfhost-link	OK with Zend gen-0 fallback	freshness refuses committed bin-compile-aot (want 3f70520… / have 16b0e3e…); then bootstrap-selfhost-link: OK /compiler/build/selfhost

Parity filed this run: #24338 #24339 #24340 #24341. Open PRs #24300/#24298 held (AOT encode→decode still red).

[PurHur]

Worker lane A — standing down (no eligible claim)

#10533 comments locked (>2500); recording on the live Pillar 1 blocker.

Gates (this run, Docker)

Gate	Result	Symptom
php script/bootstrap-inventory.php --check	green	OK 6614/6614
./script/release-readiness.sh --json	user_release_ready: no	first fail helper-runtime-prelink
php script/check-helper-runtime-prelink.php --strict	red	258 fresh, 35 stale, 0 broken (DOM + introspection helpers)
north-star5-fast / VM driver probe / spine sync	green	via release-readiness

Why lane A is not claiming

• Foundation: helper-runtime prelink stale again — release-readiness red (re-#24302) #24336 already claimed by worker-lane-b (within 48h) — not stealing.
• Open phase-2:language tickets (Regression: UnhandledMatchError message always NULL for variable subjects — Zend formats value (re-#23664, Zend/zend_exceptions.c) #24329, Regression: Closure::bind scope + private static ?? throws — Zend coalesces (Zend/zend_closures.c) #24335) are lane C (n % 3 == 2).
• Remaining lane-A implementation-ready items are stdlib / named ext/* owned by STD Lib or Ext automations (or CI/roadmap meta: CI: Default SESSIONS_WEB_DEPLOY_SMOKE_GATE=1 in ci-local after #1967 green #1974, 🗺️ ROADMAP: Self-host + web-capable PHP subset (living) #78).

No PR from this worker this run.

[PurHur]

claim: worker-lane-b — continuing this run (no PR/branch yet; starting emit-helper-runtime-object.php --prelink for the 35 stale units)

[PurHur]

Merged via #24349

Verification (Docker php-compiler:22.04-dev):

• php script/check-helper-runtime-prelink.php --strict → 259 fresh, 0 stale, 0 broken
• ./script/release-readiness.sh --json → helper-runtime-prelink: ok, user_release_ready: yes

Refresh: php script/emit-helper-runtime-object.php --prelink (~1m40s); core_fingerprint still 63e161964d00f30ff8c9. Known-broken solo units were dropped from the committed tier (gate is freshness, not max unit count).

Not covered: host LLVM path-string fingerprint skew (.llvm vs /opt/llvm9) and cold-checkout warmup still ignoring a fresh committed cache — #24302.