test(ios): cover purge restore timeout, commit and reset orderings

[sunnylqm]

Summary

Follow-up to #646: turn the three suggested timing regressions into deterministic native tests. This PR changes tests and CI only; it does not change production activation policy or public APIs.

Coverage

Case	Controlled ordering	Required outcome
late_commit	Close the restore window before releasing an already-started request's response	Keep the packaged bundle selected; allow response/metadata caching for JS, but do not activate B or label it purgeRestore
commit_before_signal	Commit B, hold the completion signal, report a launch-wait timeout, then deliver the signal	Re-resolve and launch B, keeping running identity and first-load protection consistent
reset_wins	Pause reset while holding the production state lock, start the stale commit, then release reset	Reject all stale-generation writes: activation, version metadata and response cache
commit_wins	Pause the commit while holding the production state lock, start reset, then release the commit	Reset clears the successful commit's state, metadata and response cache; the late signal cannot undo reset
complete_in_time	Commit and completion signal both arrive within the window	Preserve the ordinary successful restore path

Implementation

• Build a small macOS Foundation-only Objective-C++ test executable. No React Native installation, CocoaPods, simulator, network requests or sleeps.
• Extract the actual, unchanged RCTPushy.mm method/helper/global definitions at build time and compile them with the real state_core.cpp. Generated implementation copies are not checked in. The extractor rejects missing or ambiguous definitions and preserves source locations.
• Inject only collaborators such as network delivery, app paths, defaults storage, post-reset cleanup and the launch wait/completion boundary. The production lock and state-changing bodies execute unchanged. Defaults writes assert ownership of the real os_unfair_lock.
• Use real GCD workers and semaphore handshakes; five-second waits are deadlock guards, not timing assumptions.
• Add an ASan/UBSan-enabled macOS job to the existing test workflow.
• Add three generated-only negative controls: remove the activation veto, skip post-timeout re-resolution, or remove the reset-generation guard. Each must compile and fail its specific regression assertion (unrelated failures do not count).

Validation

• Extractor unit tests: 8 passed locally and in macOS CI.
• Python syntax, shell syntax and workflow YAML structure checked locally.
• All 5 native Foundation ordering tests passed in macOS CI with ASan/UBSan.
• All 3 negative controls compiled and failed the exact expected assertions, confirming that the original regressions are detected.

Verified native job for head e72197353e55cedca5eb929e9adfa6ab522822c8: Native purge-restore ordering — successful run and logs.

[PASS] late_commit
[PASS] commit_before_signal
[PASS] reset_wins
[PASS] commit_wins
[PASS] complete_in_time
5 native purge-restore ordering tests passed
[PASS] negative control: late-activation is detected by late_commit
[PASS] negative control: skip-reresolve is detected by commit_before_signal
[PASS] negative control: ignore-reset is detected by reset_wins

Run on macOS:

SANITIZE=1 VERIFY_REGRESSIONS=1 bash scripts/test-ios-purge-restore.sh

The editing environment is Linux; native execution was validated through the actual macOS CI job above, not claimed as a local run. See scripts/tests/ios-purge-restore/README.md for individual case commands, test seams and limitations. These are orchestration regression tests, not a replacement for tvOS device E2E, NSURLSession behavior, cache-purge or download/diff validation.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

Summary by CodeRabbit

• Tests

  • Added automated macOS checks for iOS purge-and-restore behavior, including timing, ordering, state persistence, and reset scenarios.
  • Added optional memory and undefined-behavior checks, along with regression checks for known failure cases and unsupported test interactions.
  • Added checks for deadlocks and unexpected test outcomes.

• Documentation

  • Added guidance for running the test suite, its coverage, and its limitations.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: b67aead9-b132-4188-81d4-2065cd41f9a8

📥 Commits

Reviewing files that changed from the base of the PR and between e721973 and ebfa1c2.

📒 Files selected for processing (3)

• scripts/test-ios-purge-restore.sh
• scripts/tests/ios-purge-restore/README.md
• scripts/tests/ios-purge-restore/purge_restore_test.mm

🚧 Files skipped from review as they are similar to previous changes (1)

• scripts/tests/ios-purge-restore/README.md

Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 2 remain after this review.

---

📝 Walkthrough

Walkthrough

The pull request adds a host-based Objective-C++ suite for purge-restore ordering tests. It extracts selected production code, checks five ordering cases, and supports sanitizer and regression-mutation runs in a macOS workflow.

Changes

Purge-Restore Test Suite

Layer / File(s)	Summary
Extract production code for the test host scripts/tests/ios-purge-restore/extract.py, scripts/tests/ios-purge-restore/test_extract.py, scripts/tests/ios-purge-restore/README.md	The extractor generates source-mapped includes from selected production definitions and declarations. Unit tests cover extraction behavior, and the README describes the extraction process.
Exercise purge-restore ordering scripts/tests/ios-purge-restore/purge_restore_test.mm, scripts/tests/ios-purge-restore/README.md	The native test host uses test collaborators and coordination controls to exercise five orderings. Assertions check commit behavior, reset state, bundle selection, and timing.
Run baseline and mutation checks scripts/test-ios-purge-restore.sh, .github/workflows/test.yml, scripts/tests/ios-purge-restore/README.md	The runner executes extraction tests and native tests, with optional sanitizer and mutation checks. The macOS workflow enables sanitizers and regression verification.

Priority: ⬇️ Low

Estimated code review effort: 3 (Moderate) | ~25 minutes

Change: Other

Merge Risk: ⚪ Minimal · up to ebfa1

The added regression tests and CI checks are mergeable after normal checks; no concrete issue requiring a pre-merge fix was identified.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 15.79% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 19 functions across 3 files. (2 skipped: …	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the primary change: adding iOS purge/restore tests for timeout, commit, and reset orderings. It is concise and specific.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

Docstring coverage is 15.79% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 19 functions across 3 files. (2 skipped: 2 unsupported.)

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[sunnylqm]

已在 ebfa1c2 修正这两项测试基础设施问题。

1. 不再把普通产品代码警告升级为测试失败

移除全局 -Werror，保留 -Wall -Wextra。仅将 objc-method-access 设为错误，用于发现测试宿主没有实现的 Objective-C 接口，而不是给抽取出的产品代码增加一套更严的通用告警门槛。新增编译探针，验证普通 warning 确实不导致构建失败。

2. TestDefaults 不再继承 NSUserDefaults

改为 NSObject + 私有、每实例独立的内存字典，显式声明并实现当前需要的六个访问器；写入仍经过 ObserveWrite()，保留真实状态锁的 owner 断言和时序控制。

仅在测试 translation unit 中、Foundation 导入后进行 NSUserDefaults → TestDefaults 类型替换，因此产品方法正文不用修改。将来使用 boolForKey: / setBool:forKey: 等尚未实现的接口时会明确编译失败，不会继承到真实 preferences 实现。即使通过 id 动态派发绕过静态类型，doesNotRecognizeSelector: 也会直接报告测试失败，不能被产品代码的异常处理吞掉。

新增实例隔离/快照独立性检查，以及未覆盖读写访问器的编译期、运行期反向验证。

验证

已经读取 macOS CI 的成功日志，运行 SANITIZE=1 VERIFY_REGRESSIONS=1 bash scripts/test-ios-purge-restore.sh：

• 8 个提取器测试通过。
• 原有 5 个原生时序测试和 3 个故障注入反向验证通过。
• defaults_isolation 通过。
• 普通 warning 保持非致命。
• 未实现的 boolForKey:、setBool:forKey: 分别在编译期和运行期被预期诊断拦截。

上述原生测试启用 ASan/UBSan。本次只改测试脚本、测试宿主与说明文档，没有修改生产代码；这里确认的是该原生 job，不代表其他仍在执行的工作流已全部通过。