Repository navigation
fix: pr1048-ci-integration-build - #1057
Closed
0xPlayerOne wants to merge 5 commits into
Closed
0xPlayerOne wants to merge 5 commits into
0xPlayerOne wants to merge 5 commits into
Conversation
Executable evidence for M17.02.2 (#1025), proof-only — no source changes. - tests/profile-recovery-crash-restore.test.mjs: crash/restart through the real LocalControlPlaneComposition (lease abandonment, fencing, single recovery at attempt 2, terminal outcomes never reclaimed), portable backup/restore with active-work and digest fail-closed paths, physical snapshot total-loss restore, and the hosted/self-hosted fail-closed binding matrix (no silent Hosted→local fallback). The hosted-server Postgres total-loss restore is integration-gated (RUN_DATABASE_INTEGRATION). - tests/profile-recovery-fenced-rollback.test.mjs: one attempt owner via WorkflowJobStore lease tokens, stale channel-generation rejection and heartbeat fencing, evaluateRuntimeEligibility reconnect-grant revalidation, exact rollback input binding with typed parent mismatch, interrupted/repeated operations with stored-command replay, and the interrupted local→hosted-server import rollback through applyPortableImport. - tests/profile-qualification.test.mjs: the qualification facts — profile mapping, exact wake dispatcher binding, topology/authority guard fail-closed instances, CLOUD_PROFILE_UNQUALIFIED, no implicit fallback. - docs/profile-recovery-acceptance-map.md: clause→test→result map with the per-profile supported/unproven matrix and explicit remainders. Registered additively in the smoke lane (tests/repository.test.mjs and scripts/run-bun-test-group.mjs). Ordinary sessions stay Pi-independent; no Pi-owned surfaces, recovery engines, or cross-harness transplants.
…1025) - Register the self-hosted-server PostgreSQL restore as an integration test (tests/profile-recovery-postgres-restore.integration.test.mjs) in the unsharded integration runner, the shard-3 root group, and the repository integration inventory. Disposal of every provisioned database runs after partial setup or body failure, and disposal failures surface. - Dispose every LocalControlPlaneComposition before its directory is removed, and surface failures from every release step. - Add a local graceful-drain proof: close() waits for an in-flight scheduled reconciliation pass before storage closes. - Route the local backup/restore proof through the composition's persistence. - Consolidate the adapter-level rollback fence and managed-Pi rebind proofs from PR #1046 (tests/profile-recovery-rebind-fencing.test.mjs). Its restore-binding test duplicates packages/profile-adapters and is not carried. - Remove fixture-only duplicates already pinned by package tests: tests/profile-qualification.test.mjs, lease-fencing, cancellation, import-rollback, topology, and hosted fail-closed cases. - Correct docs/profile-recovery-acceptance-map.md to the exact tests and results; drain and upgrade remain partial or unproven. Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The drain proof held a reconciliation pass open and released it only after its assertions. A failing assertion left the pass pending and the close unsettled. Release the gate in finally and settle the drain before the assertions after it run. Verified: a deliberately failing assertion now fails the test in about 2s instead of hanging. Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The Validation integration lane runs `turbo run test:integration`, whose tasks depend only on their dependencies' builds (^build). The profile recovery restore file runs as a root-level bun test afterwards and imports @control-plane/profile-portability through its package exports (dist/), but nothing builds that package in this lane, so the import fails with "Cannot find module" and the lane goes red. Build the consumed package explicitly in the unsharded runner before the root-level files. The remote lifecycle expectation rises from three to four bun invocations to account for the build, and a new test pins that the build precedes the root-level file.
The J1 evidence poller parses process-evidence.jsonl while live child processes append to it. The CI failure for run 37985178251 reports JSON.parse rejecting an unterminated fragment at column 11, which fits a poll that reads a row before its newline is written. Only newline-terminated rows are complete, so parse those and leave any fragment for the next poll, the same way the stdout reader already buffers lines. This is a hardening change. The partial-row mechanism was not reproduced locally; the J1 concurrent-recovery flake that does reproduce is a different failure mode and is not addressed by this commit.
Member
Author
|
Closing as superseded. All five commits on this branch are ancestors of PR #1048 head 98bde39 (0 commits missing, verified by git ancestry). The Integration profile-portability build step is present on #1048 (scripts/run-integration-tests.mjs). The J1 evidence-reader change was superseded on #1048 by 2120fda. The profile-recovery tests and acceptance map are byte-identical. #1048 remains draft. This PR does not close or merge any issue. Branch fix/pr1048-ci-integration-build is retained. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Scope and issue linkage
Integration-lane fix for the
Validation / Test / Integrationfailure reported on PR #1048 (Cannot find module '@control-plane/profile-portability'from the root-level integration files).98bde39821f9b24c2237409b28f4c91393d47aad, so this PR's change is already carried by test(profiles): prove profile recovery and fenced rollback evidence #1048. Recommended action for root: close this draft as superseded once test(profiles): prove profile recovery and fenced rollback evidence #1048 owns the change. This PR is not being closed or marked ready here.Change
scripts/run-integration-tests.mjs: builds@control-plane/profile-portabilitythrough turbo before the unsharded root-level integration files run.tests/integration-runner-lifecycle.test.mjs: remote-lane bun invocation count 3 → 4 (the added build), plus a new test asserting the build runs before the root-level file.tests/j1-unfinished-child-process.fixture.mjs: evidence poller parses only newline-terminated rows.Changed paths vs PR base: 3 files, +38 / −2.
Tested head
bd16941c3bdf14fcc52ffcfd1d2741dcab3eb48a(fix/pr1048-ci-integration-build). Clean worktree at validation time.Commands and results (this head, local, Bun 1.4.2)
bun install --frozen-lockfilebun run buildbun run format:checkbun run lint--deny-warnings tests: no issues in 1034 files; boundaries: no issues in 45 packages; canonical ordering: pass. Existing lint warnings incontractsandlocal-control-planeare not failures.bun run type-checkbun run test:integrationdatabase208 tests across 23 files passed. Root-level files: 34 pass, 0 fail across 3 files. Cloud and Postgres drills completed.tests/repository.test.mjs+tests/integration-shards.test.mjstests/j1-unfinished-child-process-restart.test.mjstests/integration-runner-lifecycle.test.mjsorigin/mainthe same file fails 2 of 2 runs under the same load, so this is pre-existing on this host, not introduced here.Earlier evidence on the same tree (not repeated here):
bun testunit group 3126 pass / 0 fail across 369 files;test:group:e2e320 pass.Known gaps and dependencies
mainunder the same load.mainon this host.