[GH-ISSUE #4136] [Bug] Windows packaged verifier loses its result on transient install-root cleanup failure #1632

New issue

Closed

opened 2026-10-16 21:08:12 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-16 21:08:12 +00:00

Owner

Originally created by @N0zoM1z0 on GitHub (Oct 15, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/4136

After getting past the independent Windows package-verifier blockers #4125, #4128, and #4134 in a test-only copy, the real ordinary-user scripts/verify-packaged-wasm-artifact.mjs lane installed the npm artifact but failed while closing its PrivateRuntimeRoot.

PrivateRuntimeRoot.close() received Win32 32 while opening an installed dependency file relative to its retained directory handle:

prefix\\node_modules\\rea-agents\\node_modules\\@babel\\parser\\CHANGELOG.md

The npm child had exited and the owning Windows Job settled at active-process count zero. The close failure replaced any earlier child result, left the owned package root on disk, and the verifier offers no retry/retention report to the caller.

Reproduced from PR #4117 head ecd04eea840385caeb7b595c827845b6ff8c6c88 on Windows 11 x64 with Node 24.18.0 and npm 11.16.0. The ordinary-user lane ran with two logical CPUs, BelowNormal priority, and a 2 GiB Job limit: 170.40-second Job wall, 95.30 sampled process-family CPU seconds, 492,584,960-byte peak Job commit, 489,078,784-byte sampled peak working set, 10,838,581,248-byte minimum host free memory, and Job0.

Expected behavior: the package verifier must retain cleanup ownership and retry transient sharing violations with a bounded policy after child settlement. If the operation also failed, report both causes rather than replacing the operation error. Inspect the installation-root ownership boundary and affected verifier consumers instead of weakening Windows native sharing.

Local receipts and the retained owned root are under /tmp/rea-windows-e2e-20261013/results/issue4109-fix/; no package or binary is published.

Originally created by @N0zoM1z0 on GitHub (Oct 15, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/4136 After getting past the independent Windows package-verifier blockers #4125, #4128, and #4134 in a test-only copy, the real ordinary-user `scripts/verify-packaged-wasm-artifact.mjs` lane installed the npm artifact but failed while closing its `PrivateRuntimeRoot`. `PrivateRuntimeRoot.close()` received Win32 32 while opening an installed dependency file relative to its retained directory handle: `prefix\\node_modules\\rea-agents\\node_modules\\@babel\\parser\\CHANGELOG.md` The npm child had exited and the owning Windows Job settled at active-process count zero. The close failure replaced any earlier child result, left the owned package root on disk, and the verifier offers no retry/retention report to the caller. Reproduced from PR #4117 head `ecd04eea840385caeb7b595c827845b6ff8c6c88` on Windows 11 x64 with Node 24.18.0 and npm 11.16.0. The ordinary-user lane ran with two logical CPUs, BelowNormal priority, and a 2 GiB Job limit: 170.40-second Job wall, 95.30 sampled process-family CPU seconds, 492,584,960-byte peak Job commit, 489,078,784-byte sampled peak working set, 10,838,581,248-byte minimum host free memory, and Job0. Expected behavior: the package verifier must retain cleanup ownership and retry transient sharing violations with a bounded policy after child settlement. If the operation also failed, report both causes rather than replacing the operation error. Inspect the installation-root ownership boundary and affected verifier consumers instead of weakening Windows native sharing. Local receipts and the retained owned root are under `/tmp/rea-windows-e2e-20261013/results/issue4109-fix/`; no package or binary is published.

bsakaguchi closed this issue

2026-10-16 21:08:13 +00:00

bsakaguchi commented

2026-10-16 21:08:14 +00:00

Author

Owner

@morluto commented on GitHub (Oct 15, 2026):

Addressed structurally on main in 0679a7972. Bounded Win32 32/33 close retries live at WindowsPrivateRuntime, which retains the same native owner through partial cleanup. Creation admission remains separate. Persistent conflicts preserve the exact final cause and registered owner for a later explicit close; permanent/unproven errors are not retried. Packaged-WASM, packet and signal verifiers now aggregate cleanup without losing their primary failure.

The old-source retry regressions failed; 9 source controls passed after the fix, including native code 5, missing/string codes, nested metadata and concurrent closes. Main CI passed all36 jobs (https://github.com/morluto/rea/actions/runs/38792163880); the Windows lane reports cleanupFailureRetainsOwnershipForRetry=true and cleanupTransientConflictRetries=true. The full WABT packaged lane was not run locally, and this does not promise removal while an external lock persists. Closing as addressed.

@morluto commented on GitHub (Oct 15, 2026): Addressed structurally on main in 0679a7972. Bounded Win32 32/33 close retries live at WindowsPrivateRuntime, which retains the same native owner through partial cleanup. Creation admission remains separate. Persistent conflicts preserve the exact final cause and registered owner for a later explicit close; permanent/unproven errors are not retried. Packaged-WASM, packet and signal verifiers now aggregate cleanup without losing their primary failure. The old-source retry regressions failed; 9 source controls passed after the fix, including native code 5, missing/string codes, nested metadata and concurrent closes. Main CI passed all36 jobs (https://github.com/morluto/rea/actions/runs/38792163880); the Windows lane reports cleanupFailureRetainsOwnershipForRetry=true and cleanupTransientConflictRetries=true. The full WABT packaged lane was not run locally, and this does not promise removal while an external lock persists. Closing as addressed.

bsakaguchi referenced this issue

2026-10-31 00:54:13 +00:00

[PR #1627] [MERGED] fix(onboarding): isolate client homes and clarify npm release support #4710

bsakaguchi referenced this issue

2026-10-31 00:55:06 +00:00

[PR #1749] [MERGED] fix(mcp): keep advertised inputs within client reference budgets #4791

bsakaguchi referenced this issue

2026-10-31 01:11:45 +00:00

[PR #4147] [CLOSED] fix(mcp): advertise tool inputs without shared $defs #6087

bsakaguchi referenced this issue

2026-11-03 21:43:16 +00:00

[GH-ISSUE #1587] [Feature] Improve task-centered REA onboarding and recovery in Codex skills #535