ci: add Windows and macOS smoke lanes for cross-platform regression

[xinhuagu]

Step 4 of #357: CI scaffolding

What's added

New platform-smoke job in ci.yml with:

• windows-latest runner
• macos-latest runner
• fail-fast: false (one platform failure doesn't block the other)
• Runs: ./gradlew build (compile + unit tests)
• Excludes: replay, benchmark, quality gates (Linux/bash-only)

Test isolation

Unix-only tests already annotated — no new test changes needed:

• @DisabledOnOs(OS.WINDOWS) on CommandHookExecutorTest, HookIntegrationTest
• @DisabledOnOs(OS.WINDOWS) on BashExecToolTest shell-dependent tests

What to watch

This is the first time Windows CI runs. Possible first-run failures:

• Path separator issues in test fixtures
• Temp file cleanup differences (Windows file locking)
• Gradle daemon behavior on Windows

Risk

• Low: Linux pre-merge-check unchanged. New jobs are additive.
• If Windows smoke fails, it does NOT block the Linux gate (yet — Step 7 will promote it)

Summary by CodeRabbit

• Tests
  • Normalize Windows line endings in a renderer test for consistent trailing-newline assertions.
  • Conditionally skip several integration test suites on Windows to avoid platform-specific runs.
  • Reduce flakiness by increasing timing budgets and delays in a plan-executor test.
  • Compare filesystem-equivalent paths in a shell-execution test to handle OS path differences.
• Chores
  • Set a fixed maximum heap size for forked test JVMs.

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

Walkthrough

Normalize CRLF in one CLI unit test before trimming/counting trailing newlines; change a Bash exec test to compare filesystem identity; increase timing in a planner test to reduce flakiness; add class-level JUnit OS guards to skip several daemon integration tests on Windows; set Gradle test JVM max heap to 1g.

Changes

Cohort / File(s)	Summary
CLI: trailing-newline normalization aceclaw-cli/src/test/java/dev/aceclaw/cli/TerminalMarkdownRendererTest.java	Normalize \r\n → \n before stripTrailing() and when computing trailing-newlines in renderedTable_doesNotEndWithExcessiveTrailingNewlines.
Tools: working-directory identity check aceclaw-tools/src/test/java/dev/aceclaw/tools/BashExecToolTest.java	Parse command output into a Path and assert Files.isSameFile(outputPath, workDir) instead of substring matching.
Core: timing stabilization aceclaw-core/src/test/java/dev/aceclaw/core/planner/SequentialPlanExecutorReplanTest.java	Increase total wall-clock budget and per-step watchdog cap (≈80ms → 200ms) and lengthen post-step listener sleep to reduce flakiness; update comments.
Daemon: Windows-skip integration tests aceclaw-daemon/src/test/java/dev/aceclaw/daemon/BackgroundTaskIntegrationTest.java, aceclaw-daemon/src/test/java/dev/aceclaw/daemon/DaemonIntegrationTest.java, aceclaw-daemon/src/test/java/dev/aceclaw/daemon/SkillIntegrationTest.java, aceclaw-daemon/src/test/java/dev/aceclaw/daemon/SubAgentIntegrationTest.java	Add @DisabledOnOs(OS.WINDOWS) and required imports to skip these integration test classes on Windows.
Build: test JVM heap size build.gradle.kts	Set maxHeapSize = "1g" for all Gradle Test tasks in subprojects.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related issues

• [EPIC] Windows Compatibility: Native Support Without Regressing macOS/Linux #371 — Matches Windows-compatibility work: OS guards and CRLF/path normalization.
• ci(platform): Windows matrix and macOS/Linux non-regression gates #370 — Relates to isolating Unix-only tests and normalizing Windows line endings/paths.

Possibly related PRs

• feat(tools): add Windows shell support to BashExecTool #2 — Both touch BashExecTool tests and working-directory verification (path identity).
• fix: strip trailing blank lines from cron/deferred output (#185) #186 — Related to TerminalMarkdownRenderer trailing-newline handling and CRLF normalization.
• fix(watchdog): add per-step wall-clock reset for multi-step plans #158 — Related adjustments to SequentialPlanExecutorReplanTest timing and watchdog behavior.

🚥 Pre-merge checks | ✅ 5 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 16.67% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'ci: add Windows and macOS smoke lanes for cross-platform regression' accurately summarizes the main change in the PR: adding CI infrastructure (smoke test lanes) for Windows and macOS platforms.
Block Major Correctness And Security Risks	✅ Passed	PR introduces no major correctness or security risks. All changes are defensive improvements for cross-platform test support with proper thread-safe synchronization and exception handling.
Require Test Coverage For New Logic	✅ Passed	All 8 modified files are test files or build configuration; no src/main business logic changes were introduced.
No Api Breaking Changes Without Version Bump	✅ Passed	The pull request contains no public API signature changes. All modifications are limited to test files and build configuration. No version bump is required.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/357-windows-step4-ci

---

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 and usage tips.}

[greptile-apps]

Your free trial has ended. If you'd like to continue receiving code reviews, you can add a payment method here.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@aceclaw-tools/src/test/java/dev/aceclaw/tools/BashExecToolTest.java`:
- Around line 109-112: The test compares paths asymmetrically: outputPath is
built from result.output().trim().toAbsolutePath().normalize() while
expectedPath uses workDir.toRealPath(), causing macOS symlink mismatches; change
both sides to use toRealPath() (and remove the redundant toAbsolutePath() on
expectedPath) so build outputPath from
Path.of(result.output().trim()).toRealPath() and expectedPath from
workDir.toRealPath() to ensure consistent symlink resolution (update references
in BashExecToolTest where outputPath and expectedPath are created).

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 062410aa-dc1c-4d5c-b42f-d88dba2ee5b4

📥 Commits

Reviewing files that changed from the base of the PR and between d7c37ad and e9152d8.

⛔ Files ignored due to path filters (1)

• .github/workflows/ci.yml is excluded by !.github/**

📒 Files selected for processing (2)

• aceclaw-core/src/test/java/dev/aceclaw/core/planner/SequentialPlanExecutorReplanTest.java
• aceclaw-tools/src/test/java/dev/aceclaw/tools/BashExecToolTest.java