Skip to content

test(ci): stabilize JADX recovery and ignore generated reports - #1668

Merged
morluto merged 1 commit into
morluto:mainfrom
Y0UW1SH:fix/ci-jadx-cleanup-and-test-reports
Oct 11, 2026
Merged

morluto merged 1 commit into
morluto:mainfrom
Y0UW1SH:fix/ci-jadx-cleanup-and-test-reports

Conversation

@Y0UW1SH

@Y0UW1SH Y0UW1SH commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Stabilize the synthetic JADX cleanup-recovery test and keep generated Vitest blob reports out of source formatting.

These two CI fixes were verified during #1612 maintenance. That browser proposal remains closed following the maintainer's #1608 scope decision. The runtime catalog/deadline fixes are already proposed in #1363.

Problem and expected behavior

After the injected ownership-verification failure clears, the child can move between live and exited snapshots. A single immediate close can reject while the retained cleanup owner can still recover. Require verified cleanup within a bounded recovery window.

Vitest writes generated JSON into .vitest/blob/; source formatting should ignore those reports alongside existing coverage and report directories.

Change and scope

  • Retry provider.close() for at most five seconds only during the test's intentional recovery and final teardown. Preserve the initial shared failure, retained workspace, verified process/workspace removal, and single-launch assertions.
  • Add .vitest/ to .gitignore.

Exactly two files change: +11/-4. Head: 5226656a2f550c3a40503ab5cfdf03706fa1b371; branch parent: 7bb4ac219c16884fcadd946eed4015ace2a2419f; verified tree: 891b3e099bbf3a965fa601782538e9b6bca7e5e5.

Contract and boundary impact

  • Semantic owner and earliest changed stage: test recovery and generated-report eligibility.
  • CLI and MCP/tool-catalog contract: none.
  • Provider, bridge, target-format, or platform compatibility: none.
  • Evidence, artifact, provenance, or reconstruction contract: none.
  • Process execution, authorization, cleanup, or containment impact: production behavior unchanged; test recovery still requires verified ownership and removal.
  • Generated metadata, package, or installation impact: none; generated reports stay untracked.

Evidence and regression coverage

  • Updated the existing recovery test; neighboring retention and launch-blocking cases remain unchanged.
  • Fresh CI for this head passed all 31 jobs, including the required gate, all four test shards, and aggregate coverage: 7,242 tests passed, 89 skipped; line coverage 88.37% with the normal thresholds enabled.
  • The synthetic JADX integration suite passed 33/33 cases. CI tested virtual merge 8451051a4fd5f84e042e1bd655c92150abb7f3fe against upstream 79569349651b5089f04d466b955f5f1ac59a6a1d.
  • User-visible CLI/MCP output: not applicable.
  • Local host limitation: this container's incompatible ps prevents owned-process tests from launching; the fresh GitHub-hosted suite provides that verification. These JADX tests use the repository's synthetic MCP fixture, not a real JADX engine.

Validation performed

  • npm run check:ci — 8/8 tasks and architecture guards passed across 1,101 source files; duplication 2.45% under the unchanged 2.6% gate. A cold parallel run hit the container memory limit; isolated typechecking passed, then the normal command passed.
  • npm run build:cached — 3/3 tasks passed.
  • npm run docs:check — 6/6 tasks passed.
  • Source formatting passed with an actual generated Vitest blob report still present.
  • Commit hooks, git diff --check, and independent review of the final two-file diff passed.

Compatibility, safety, and release

  • Breaking changes or migration steps: none.
  • Fresh CI passed the real-browser lane and Linux x64/arm64, macOS x64/arm64, and Windows package lanes. This test-only change makes no new real JADX-engine verification claim.
  • Reviewed subsequent upstream changes through f88c99021b88ce67e513025901f57478d8107d0c: the JADX provider/test and report ignore paths are unchanged, and the shared process-identity extension preserves existing calls.
  • Package or release metadata impact: none.
  • Security, privacy, process, or containment review: production ownership checks remain unchanged.

Review checklist

  • This PR addresses concrete CI failures.
  • One focused maintenance outcome and conventional title.
  • Related CI history and the existing runtime proposal are linked.
  • Existing meaningful failure and recovery assertions remain.
  • Owning docs, contracts, and generated metadata: no changes required.
  • User-visible CLI/MCP output: not applicable.
  • Final diff checked for secrets, unrelated changes, and unsupported claims.

Published in draft and promoted only after the fresh full CI run passed. Ready for review; unmerged.

@Y0UW1SH
Y0UW1SH marked this pull request as ready for review October 11, 2026 02:06
@chatgpt-codex-connector

Copy link
Copy Markdown

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Repo admins can enable using credits for code reviews in their settings.

@CassieFox23

Copy link
Copy Markdown
Contributor

TL;DR: I could not make the flaky test fail on Linux, before or after this change. All 178 runs of the JADX test file passed, 33 of 33 tests each time. So I have no proof the retry fixes anything here, and the change did no harm.

Results

Each run is npm run test:focused -- tests/boundary/process/jadxIntegration.test.ts. Each cell shows passes out of runs.

Tree Node 22.23.1 Node 24.18.0 Node 26.11.1 Node 24.18.0, busy CPU
base 7956934 10/10 30/30 10/10 14/14
head 5226656 10/10 30/30 10/10 14/14
main 17f7040 with the head test file 10/10 30/30 10/10 not run
  • Failures: 0 in every cell. The shutdown cleanup recovery test passed every time.
  • Busy CPU means 20 busy loops at normal priority on 24 threads. The load average (runnable tasks, out of 24 threads) was about 27 to 29. A run took 26 to 49 seconds instead of 14 to 16.
  • The busy-CPU phase had a 10 minute cap, so it got 14 runs per tree, not 20.
  • The no-load rows ran base, head and main side by side at low priority, so they were not on an idle machine.
  • .gitignore: with a real Vitest blob report on disk, oxfmt --check . failed with base's .gitignore and passed with head's. The report was written to .vitest/blob/blob.json, which sits under the new .vitest/ line.
  • This test uses the repo's synthetic MCP fixture, tests/fixtures/android/jadx-mcp.mjs, not a real JADX. No JADX was needed.

Not run

  • Real JADX engine.
  • Windows and macOS.
  • Busy-CPU runs on Node 22 and Node 26, and on main.
  • A box with the odd ps from the description. Here ps works. I do not know if that is why the flake did not show.
  • Full test suite, npm run check:ci, docs check.

Thanks, Y0UW1SH, for the focused fix and the clear description.

Test setup
  • Box: Nobara 44 (Fedora 44 base), Linux 7.2.9
  • CPU: AMD Ryzen 9 5900X, 12 cores / 24 threads
  • RAM: 32 GB DDR4-3600
  • Node: 22.23.1, 24.18.0, 26.11.1 (installed once under 24.18.0, then run under each)
  • Vitest: 5.0.3
  • Head: 5226656
  • Base: 7956934
  • Main: 17f7040

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

@morluto
morluto merged commit 0a6b46d into morluto:main Oct 11, 2026
31 checks passed
@github-actions github-actions Bot mentioned this pull request Oct 11, 2026
3 tasks done
@morluto

morluto commented Oct 11, 2026

Copy link
Copy Markdown
Owner

Appreciate the targeted CI cleanup, @Y0UW1SH. Your bounded retry lets the JADX recovery fixture verify eventual owner cleanup through snapshot churn, and ignoring .vitest/ keeps generated blob reports out of source formatting. Both changes remove noise without weakening the ownership assertion.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area: tests Test suite and verification

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants