morluto/rea

Fork 0

mirror of https://github.com/morluto/rea.git synced 2026-11-05 06:45:22 +00:00

[GH-ISSUE #4610] [Bug] main CI red since `e65c3955f`: binaryTargetAdmissionCancellation and managedHashCancellation fail #2395

New issue

Closed

opened 2026-10-17 21:04:07 +00:00 by bsakaguchi · 5 comments

bsakaguchi commented

2026-10-17 21:04:07 +00:00

Owner

Originally created by @Nitin13-git on GitHub (Oct 16, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/4610

Before submitting

I searched the open and closed issues for a related report.
I removed credentials, tokens, private paths, and sensitive target data.

Area

Evidence, artifacts, or reconstruction

Observed behavior

Two tests fail on main since e65c3955f ("fix(filesystem): preserve digest failures through cleanup cancellation"); main's CI is red from that commit on (cd9252bc6, 67384b23e, 1b56237ca), and unrelated PRs (for example #4602) fail the same two tests.

tests/process-global/binaryTargetAdmissionCancellation.test.ts, "retains the cleanup failure and its file owner when cancelled admission cannot close". expect(failure.cause).toBe(reason) fails: failure.cause is a RegularFileOperationCancelled (with reason: Error("caller cancelled")), while the test expects the caller's Error("caller cancelled") itself.
tests/boundary/providers/managed/managedHashCancellation.test.ts, "cancels a cached managed source hash and recovers its exact owner (close denied: false, reason: cleanup)". The retry fails with ProviderCleanupError: "Artifact cleanup remains incomplete" caused by "File handle was invalidated before this owner confirmed its close" (OwnedFileHandle.close, fd: -1) from retryRegularFileCleanup in RegularFileRead.ts.

Reproduction

Executed locally on macOS arm64, Node 24.18.0, locked dependencies:

git checkout --detach <commit>
npm run test:focused -- tests/process-global/binaryTargetAdmissionCancellation.test.ts tests/boundary/providers/managed/managedHashCancellation.test.ts

0d2247428: 17 passed.
e65c3955f: 2 failed, 15 passed (the two tests above).
Later commits: the first test fails on cd9252bc6, the second fails on 1b56237ca (three of three runs), and both fail together on 36062643.

CI shows the same failures (shard 1/5 for the first, shard 5/5 for the second).

Expected behavior

main is green. Either the code at e65c3955f restores what these tests assert (cause identity and retry of a failed descriptor close), or the tests are updated to the new contract if the wrapper type and handle invalidation are intended.

Impact and scope

Observed: CI is red on main and on PRs that do not touch this code, so they cannot reach CI required. Not established: which side (code or tests) should change, whether other tests are affected, or whether Windows behaves the same. I did not inspect the commit's design beyond the files it changes (src/application/RegularFileRead.ts, src/artifacts/ArtifactFileOperation.ts, and others).

REA version

6.5.0 at e65c3955f through 36062643.

Runtime and platform

macOS arm64, Node 24.18.0 (CI: GitHub Linux runners)

Target and artifact context

NOT_ENOUGH_INFO: test-only reproduction; no target artifact involved.

Interface

Other

Earliest failure stage

Cleanup, cancellation, or lifecycle

Relevant output or evidence

First test: AssertionError: expected RegularFileOperationCancelled: Regular fi… { …(2) } to be Error: caller cancelled // Object.is equality at binaryTargetAdmissionCancellation.test.ts:158.

Second test: Error: Provider rea-dotnet-static adapter failed during inspect_managed_members, caused by ArtifactReaderFailure: Artifact cleanup remains incomplete: .../other-input.dll, caused by Error: File handle was invalidated before this owner confirmed its close (OwnedFileHandle.close, called from ArtifactResourceScope.release via retryRegularFileCleanup).

Analysis or suspected cause

Hypothesis, not verified: e65c3955f changed how cancellation and descriptor cleanup are projected in RegularFileRead.ts and ArtifactFileOperation.ts; the tests were written against the earlier shape. The bisect above narrows the break to that commit's parent and itself, but I have not read it closely enough to say whether code or tests are wrong. Possibly related, not claimed: open PR #4438 also edits withRegularFileOwned.

Originally created by @Nitin13-git on GitHub (Oct 16, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/4610 ### Before submitting - [x] I searched the open and closed issues for a related report. - [x] I removed credentials, tokens, private paths, and sensitive target data. ### Area Evidence, artifacts, or reconstruction ### Observed behavior Two tests fail on `main` since `e65c3955f` ("fix(filesystem): preserve digest failures through cleanup cancellation"); `main`'s CI is red from that commit on (`cd9252bc6`, `67384b23e`, `1b56237ca`), and unrelated PRs (for example #4602) fail the same two tests. 1. `tests/process-global/binaryTargetAdmissionCancellation.test.ts`, "retains the cleanup failure and its file owner when cancelled admission cannot close". `expect(failure.cause).toBe(reason)` fails: `failure.cause` is a `RegularFileOperationCancelled` (with `reason: Error("caller cancelled")`), while the test expects the caller's `Error("caller cancelled")` itself. 2. `tests/boundary/providers/managed/managedHashCancellation.test.ts`, "cancels a cached managed source hash and recovers its exact owner (close denied: false, reason: cleanup)". The retry fails with `ProviderCleanupError`: "Artifact cleanup remains incomplete" caused by "File handle was invalidated before this owner confirmed its close" (`OwnedFileHandle.close`, `fd: -1`) from `retryRegularFileCleanup` in `RegularFileRead.ts`. ### Reproduction Executed locally on macOS arm64, Node 24.18.0, locked dependencies: ```sh git checkout --detach <commit> npm run test:focused -- tests/process-global/binaryTargetAdmissionCancellation.test.ts tests/boundary/providers/managed/managedHashCancellation.test.ts ``` - `0d2247428`: 17 passed. - `e65c3955f`: 2 failed, 15 passed (the two tests above). - Later commits: the first test fails on `cd9252bc6`, the second fails on `1b56237ca` (three of three runs), and both fail together on `36062643`. CI shows the same failures (shard 1/5 for the first, shard 5/5 for the second). ### Expected behavior `main` is green. Either the code at `e65c3955f` restores what these tests assert (cause identity and retry of a failed descriptor close), or the tests are updated to the new contract if the wrapper type and handle invalidation are intended. ### Impact and scope Observed: CI is red on `main` and on PRs that do not touch this code, so they cannot reach `CI required`. Not established: which side (code or tests) should change, whether other tests are affected, or whether Windows behaves the same. I did not inspect the commit's design beyond the files it changes (`src/application/RegularFileRead.ts`, `src/artifacts/ArtifactFileOperation.ts`, and others). ### REA version 6.5.0 at `e65c3955f` through `36062643`. ### Runtime and platform macOS arm64, Node 24.18.0 (CI: GitHub Linux runners) ### Target and artifact context NOT_ENOUGH_INFO: test-only reproduction; no target artifact involved. ### Interface Other ### Earliest failure stage Cleanup, cancellation, or lifecycle ### Relevant output or evidence First test: `AssertionError: expected RegularFileOperationCancelled: Regular fi… { …(2) } to be Error: caller cancelled // Object.is equality` at `binaryTargetAdmissionCancellation.test.ts:158`. Second test: `Error: Provider rea-dotnet-static adapter failed during inspect_managed_members`, caused by `ArtifactReaderFailure: Artifact cleanup remains incomplete: .../other-input.dll`, caused by `Error: File handle was invalidated before this owner confirmed its close` (`OwnedFileHandle.close`, called from `ArtifactResourceScope.release` via `retryRegularFileCleanup`). ### Analysis or suspected cause Hypothesis, not verified: `e65c3955f` changed how cancellation and descriptor cleanup are projected in `RegularFileRead.ts` and `ArtifactFileOperation.ts`; the tests were written against the earlier shape. The bisect above narrows the break to that commit's parent and itself, but I have not read it closely enough to say whether code or tests are wrong. Possibly related, not claimed: open PR #4438 also edits `withRegularFileOwned`.

bsakaguchi commented

2026-10-17 21:04:09 +00:00

Author

Owner

@morluto commented on GitHub (Oct 17, 2026):

Reproduced both reported failures on Node 24.18.0 at 0c7b6d834: 2 failed, 15 passed.

Two distinct causes:

The admission test expected a raw caller reason after the existing boundary changed to an explicit RegularFileOperationCancelled checkpoint. 3bd9aa348 asserts that wrapper and exact original reason while retaining the descriptor/cleanup checks.
withRegularFile interpreted an opaque caller reason shaped like RegularFileCleanupFailure as this operation's cleanup record and retried a foreign descriptor. 746a044e4 preserves the exact reason before class interpretation, including checkpoint-shaped reasons after a successful close retry. Signal-aware reference-source and capture-export cleanup consumers now carry the same context.

Validation: 156 focused tests passed across 14 files, covering the two reported files, opaque reason identity, foreign descriptor ownership, cleanup retry, nested cancellation, physical failure precedence, and affected ADB/JADX/Kuna/reference-source/export consumers. These are local production-boundary tests with controlled faults; no native provider behavior is claimed. The heavy Qodo/profiler test remained stopped. Full CI results remain separate.

The filesystem owner primitive remains unchanged: signal equality is not used to replace a physical failure with cancellation.

@morluto commented on GitHub (Oct 17, 2026): Reproduced both reported failures on Node 24.18.0 at `0c7b6d834`: 2 failed, 15 passed. Two distinct causes: - The admission test expected a raw caller reason after the existing boundary changed to an explicit `RegularFileOperationCancelled` checkpoint. `3bd9aa348` asserts that wrapper and exact original reason while retaining the descriptor/cleanup checks. - `withRegularFile` interpreted an opaque caller reason shaped like `RegularFileCleanupFailure` as this operation's cleanup record and retried a foreign descriptor. `746a044e4` preserves the exact reason before class interpretation, including checkpoint-shaped reasons after a successful close retry. Signal-aware reference-source and capture-export cleanup consumers now carry the same context. Validation: 156 focused tests passed across 14 files, covering the two reported files, opaque reason identity, foreign descriptor ownership, cleanup retry, nested cancellation, physical failure precedence, and affected ADB/JADX/Kuna/reference-source/export consumers. These are local production-boundary tests with controlled faults; no native provider behavior is claimed. The heavy Qodo/profiler test remained stopped. Full CI results remain separate. The filesystem owner primitive remains unchanged: signal equality is not used to replace a physical failure with cancellation.

bsakaguchi commented

2026-10-17 21:04:09 +00:00

Author

Owner

@Nitin13-git commented on GitHub (Oct 17, 2026):

@morluto thanks for the fixes. One more failure from the same commit range: with 3bd9aa348 and 746a044e4 the two reported tests pass, but main CI still fails tests/boundary/providers/kuna/kunaProvider.test.ts ("waits for an aborted pre-session hash before retrying its late close owner").

It expects error: { _tag: "ProviderAdapterError", cleanupIncomplete: true } but receives an AnalysisCancelledError (with cleanupIncomplete: true).

Bisect on a local checkout (macOS, Node 24.18.0, npm run test:focused -- tests/boundary/providers/kuna/kunaProvider.test.ts):

passes at 0d2247428 (31 passed);
fails at e65c3955f, and at cd9252bc6, 360626434, d5be402fb, 0c7b6d834, 3bd9aa348, 746a044e4 and bfb7ed0bd (current origin/main).

The test file is unchanged between 0d2247428 and main, while e65c3955f also changed src/kuna/* (KunaCommand, KunaInstallation, KunaProjectCache, KunaProjectExport, KunaProvider, KunaSessionClient). I have not checked whether the test or the code should change; the new result may be intended given the cancellation changes in that commit.

@Nitin13-git commented on GitHub (Oct 17, 2026): @morluto thanks for the fixes. One more failure from the same commit range: with `3bd9aa348` and `746a044e4` the two reported tests pass, but `main` CI still fails `tests/boundary/providers/kuna/kunaProvider.test.ts` ("waits for an aborted pre-session hash before retrying its late close owner"). It expects `error: { _tag: "ProviderAdapterError", cleanupIncomplete: true }` but receives an `AnalysisCancelledError` (with `cleanupIncomplete: true`). Bisect on a local checkout (macOS, Node 24.18.0, `npm run test:focused -- tests/boundary/providers/kuna/kunaProvider.test.ts`): - passes at `0d2247428` (31 passed); - fails at `e65c3955f`, and at `cd9252bc6`, `360626434`, `d5be402fb`, `0c7b6d834`, `3bd9aa348`, `746a044e4` and `bfb7ed0bd` (current `origin/main`). The test file is unchanged between `0d2247428` and `main`, while `e65c3955f` also changed `src/kuna/*` (`KunaCommand`, `KunaInstallation`, `KunaProjectCache`, `KunaProjectExport`, `KunaProvider`, `KunaSessionClient`). I have not checked whether the test or the code should change; the new result may be intended given the cancellation changes in that commit.

bsakaguchi commented

2026-10-17 21:04:09 +00:00

Author

Owner

@Nitin13-git commented on GitHub (Oct 17, 2026):

Follow-up with CI evidence for the Kuna failure above:

Run at 746a044e (includes 3bd9aa348 and 746a044e4): https://github.com/morluto/rea/actions/runs/38986298607. Test shard (4/5) and Portable tests fail only on tests/boundary/providers/kuna/kunaProvider.test.ts > waits for an aborted pre-session hash before retrying its late close owner (no other FAIL line in those jobs).
Run at 0c7b6d83 (before those two fixes): https://github.com/morluto/rea/actions/runs/38981672194. The same Kuna test fails there too, together with managedHashCancellation, binaryTargetAdmissionCancellation, and tests/boundary/mcp/serverIdentity.test.ts > live server identity over MCP > compares the caller's digest against the selected compact tools/list catalog (I did not look into that last one).
Assertion, from the CI log (kunaProvider.test.ts:154): expected { ok: false, error: { _tag: "ProviderAdapterError", cleanupIncomplete: true } }, received { ok: false, error: AnalysisCancelledError { _tag: "AnalysisCancelledError", cleanupIncomplete: true } }. The test aborts the signal while the profile close is held, releases the close, then awaits previewTarget.

I have not identified which of the src/kuna/* changes in e65c3955f produces the cancellation.

@Nitin13-git commented on GitHub (Oct 17, 2026): Follow-up with CI evidence for the Kuna failure above: - Run at `746a044e` (includes `3bd9aa348` and `746a044e4`): https://github.com/morluto/rea/actions/runs/38986298607. `Test shard (4/5)` and `Portable tests` fail only on `tests/boundary/providers/kuna/kunaProvider.test.ts > waits for an aborted pre-session hash before retrying its late close owner` (no other `FAIL` line in those jobs). - Run at `0c7b6d83` (before those two fixes): https://github.com/morluto/rea/actions/runs/38981672194. The same Kuna test fails there too, together with `managedHashCancellation`, `binaryTargetAdmissionCancellation`, and `tests/boundary/mcp/serverIdentity.test.ts > live server identity over MCP > compares the caller's digest against the selected compact tools/list catalog` (I did not look into that last one). - Assertion, from the CI log (`kunaProvider.test.ts:154`): expected `{ ok: false, error: { _tag: "ProviderAdapterError", cleanupIncomplete: true } }`, received `{ ok: false, error: AnalysisCancelledError { _tag: "AnalysisCancelledError", cleanupIncomplete: true } }`. The test aborts the signal while the profile close is held, releases the close, then awaits `previewTarget`. I have not identified which of the `src/kuna/*` changes in `e65c3955f` produces the cancellation.

bsakaguchi referenced this issue

2026-10-31 00:29:41 +00:00

[GH-ISSUE #2383] [Tracker] Audit analysis and debugging coverage for the TIOBE top 100 languages (October 2026) #794

bsakaguchi closed this issue

2026-10-31 00:33:25 +00:00

bsakaguchi commented

2026-10-31 00:33:26 +00:00

Author

Owner

@CassieFox23 commented on GitHub (Oct 18, 2026):

Checked on Linux: part of this is fixed on main by 3bd9aa348 and 746a044e4. Their messages say "Refs #4610", which is why GitHub left this issue open.

tests/process-global/binaryTargetAdmissionCancellation.test.ts and tests/boundary/filesystem/regularFileCancellation.test.ts pass on main 36d156ff0 on Node 22.23.1, 24.18.0 and 26.11.1 (49 tests).
With 746a044e4 reverted on main, 5 tests fail. With 3bd9aa348 reverted, 1 fails.

The issue wants a green main. At 36d156ff0 main CI was red on a different test, tests/boundary/mcp/nativeUiFinalizationMcp.test.ts, reported on #5056. So part of this may still be open.

Thanks, Nitin13-git, for the clear report.

🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

@CassieFox23 commented on GitHub (Oct 18, 2026): Checked on Linux: part of this is fixed on main by 3bd9aa348 and 746a044e4. Their messages say "Refs #4610", which is why GitHub left this issue open. - `tests/process-global/binaryTargetAdmissionCancellation.test.ts` and `tests/boundary/filesystem/regularFileCancellation.test.ts` pass on main 36d156ff0 on Node 22.23.1, 24.18.0 and 26.11.1 (49 tests). - With 746a044e4 reverted on main, 5 tests fail. With 3bd9aa348 reverted, 1 fails. The issue wants a green main. At 36d156ff0 main CI was red on a different test, `tests/boundary/mcp/nativeUiFinalizationMcp.test.ts`, reported on #5056. So part of this may still be open. Thanks, Nitin13-git, for the clear report. 🤖 Generated with Claude Code (Sonnet 5.5, verified by Opus 5.5) 🦊

bsakaguchi commented

2026-10-31 00:33:26 +00:00

Author

Owner

@morluto commented on GitHub (Oct 18, 2026):

Closing as fixed on main after reconciling this report with the landed owner/consumer changes.

746a044e4: fix(filesystem): preserve opaque caller reasons through cleanup. Validation: 156 focused tests across 14 files passed; no native provider execution. Refs #4610
3bd9aa348: test(filesystem): assert admission cancellation provenance.

Evidence scope: this closure uses the landed source, regression coverage and validation already recorded in the linked commits/thread. No new local tests or native-provider reproduction were run during this backlog pass; previously documented platform and real-provider gaps still apply.

@morluto commented on GitHub (Oct 18, 2026):  Closing as fixed on `main` after reconciling this report with the landed owner/consumer changes. - [746a044e4](https://github.com/morluto/rea/commit/746a044e42c9eeff76a26d24c7dc89ae0eeff88d): fix(filesystem): preserve opaque caller reasons through cleanup. Validation: 156 focused tests across 14 files passed; no native provider execution. Refs #4610 - [3bd9aa348](https://github.com/morluto/rea/commit/3bd9aa348478eebfd339219d8e1ecee07f635333): test(filesystem): assert admission cancellation provenance. Evidence scope: this closure uses the landed source, regression coverage and validation already recorded in the linked commits/thread. No new local tests or native-provider reproduction were run during this backlog pass; previously documented platform and real-provider gaps still apply.

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#2395

No description provided.

Rows
Columns

[GH-ISSUE #4610] [Bug] main CI red since e65c3955f: binaryTargetAdmissionCancellation and managedHashCancellation fail #2395

Before submitting

Area

Observed behavior

Reproduction

Expected behavior

Impact and scope

REA version

Runtime and platform

Target and artifact context

Interface

Earliest failure stage

Relevant output or evidence

Analysis or suspected cause

[GH-ISSUE #4610] [Bug] main CI red since `e65c3955f`: binaryTargetAdmissionCancellation and managedHashCancellation fail #2395