fix(test): make the native-windows test suite build and run

[caveman99]

pio test -e native-windows failed all 53 suites at the build stage. Five independent causes, all Windows-only, plus one Unity configuration gap that turned out not to be Windows-specific at all.

Causes

lstat() in test/TestUtil.cpp. MinGW-w64 does not provide it. The state-checkpoint walk added in #11322 is fenced with ARCH_PORTDUINO, which native-windows also satisfies, so the file failed to compile and took every suite with it. This is the regression that broke the Windows runner; the others were masked behind it. Routed through a stat() shim on _WIN32.

Missing setUp/tearDown in four suites. test_default, test_http_content_handler, test_meshpacket_serializer and test_serial define neither and relied on the weak defaults PlatformIO emits in its generated unity_config.c. Those defaults are enabled on MinGW, but GCC lowers a weak definition on PE-COFF to a weak external:

0000000000000000 T .weak.setUp.unityOutputStart
                 w setUp

setUp stays undefined (w) with a renamed alternate, rather than ELF's weak definition (W), so it does not satisfy unity.c's strong reference and the link fails. Defined explicitly, as the other 49 suites already do.

<arpa/inet.h> in test_mqtt. Absent on MinGW; included only for htonl(). Uses winsock2.h there.

UNITY_INCLUDE_DOUBLE was never defined anywhere. Unity omits double support unless asked, so TEST_ASSERT_DOUBLE_* compiles into an unconditional "Unity Double Precision Disabled" failure. This is not Windows-specific. Verified on Debian with gcc, against the Linux env's own Unity 2.6.1 and PlatformIO's generated native unity_config:

UNITY_INCLUDE_DOUBLE : NOT defined
UNITY_EXCLUDE_DOUBLE : defined
test_double_within:FAIL: Unity Double Precision Disabled

The macro appears nowhere in the repo, the ini files, test_native.yml, or PlatformIO's Unity runner, which adds only UNITY_INCLUDE_CONFIG_H. It is therefore defined in portduino_base.build_flags_common, which every native env resolves: native, native-tft, native-fb, native-tft-debug, coverage, coverage-event-policy, native-macos, native-windows, native-wasm.

test_getfiles_rejects_overlong_path. Excluded on _WIN32. Overrunning the 228-byte file_name needs at least 229 bytes below the portduino root, and that root is already about 34 bytes, so every qualifying path passes the 260-byte MAX_PATH. The nested mkdir() fails, the file is never created, and getFiles() has nothing to drop. No component layout satisfies both limits.

Verification

Full pio test -e native-windows on Windows 11 / UCRT64 gcc 16.1.0, before and after:

	Before	After
Suites built and run	46 of 53	53 of 53
Test cases	839	938
Failed	2, plus 7 suites that never built	1

Scope

Every guard is _WIN32-only except UNITY_INCLUDE_DOUBLE, which now applies to all native envs and does change Linux and macOS: TEST_ASSERT_DOUBLE_* becomes a real comparison instead of a stub that always fails. test_gps_update_scheduling is the only suite in the repo using those macros, and its arithmetic is integer-based and bit-identical across platforms, so it should pass wherever it runs.

Worth flagging for review: that suite currently reports PASSED on CI in 0.03 seconds while emitting no Unity output at all, so its assertions appear never to execute there. That is a separate issue and is not addressed here.

No CI job currently compiles the test tree for native-windows: test_native.yml runs on Ubuntu against -e coverage, and build_windows_bin.yml runs pio run, never pio test. That is why the lstat regression stood for a week.

Known limitation

test_fscommon_getfiles still fails in a full run, for a cause outside this change: VFSImpl::rmdir() in framework-portduino calls unlink(), which cannot remove a directory, so the tree its depth-limit case creates survives setUp()'s rmDir() and makes getFiles() report a spurious depth truncation. Fixed by meshtastic/framework-portduino#77; once that lands and the platform pin picks it up, the suite goes green on Windows with no further change here.

Summary by CodeRabbit

• Bug Fixes

  • Improved cross-platform directory traversal and network byte-order handling for Windows builds.
  • Excluded unsupported long-path tests on Windows.

• Tests

  • Improved test compatibility with MinGW by adding required test lifecycle hooks.
  • Enabled double-precision assertions for GPS scheduling tests.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 2d7c6fc8-efbc-45b3-98b5-f34d0d44af79

📥 Commits

Reviewing files that changed from the base of the PR and between c9b3cc9 and 1191f87.

📒 Files selected for processing (1)

• variants/native/portduino.ini

---

📝 Walkthrough

Walkthrough

The changes improve native test compatibility across Windows, non-Windows platforms, and MinGW. They add platform-specific filesystem and networking handling, Unity lifecycle hooks, Windows-specific test guards, and Unity double-precision support.

Changes

Native test compatibility

Layer / File(s)	Summary
Cross-platform filesystem traversal test/TestUtil.cpp	Directory traversal uses stat on Windows and lstat on other platforms.
MinGW Unity lifecycle hooks test/test_default/test_main.cpp, test/test_http_content_handler/test_main.cpp, test/test_meshpacket_serializer/test_serializer.cpp, test/test_serial/SerialModule.cpp	Test targets define empty setUp and tearDown callbacks for MinGW linking.
Windows test conditions and headers test/test_mqtt/MQTT.cpp, test/test_fscommon_getfiles/test_main.cpp, variants/native/portduino.ini	Windows uses winsock2.h, skips the unsupported overlong-path test, and enables Unity double-precision support.

Estimated code review effort: 2 (Simple) | ~10 minutes

Mergeability Score: ⚪ Minimal · up to 1191f

This change is limited to making the native-Windows test suite build and run, with no actionable merge-blocking risk remaining after normal checks and review.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
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.
Title check	✅ Passed	The title clearly and concisely describes the primary change: making the native Windows test suite build and run.
Description check	✅ Passed	The description is detailed and relevant, covering causes, fixes, verification results, scope, and the known remaining limitation.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix-native-windows-tests

---

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.}

[github-actions]

⚡ Try this PR in the Web Flasher

Note

Building this pull request… the flash button, badges and supported-board
list will appear here automatically once CI finishes.

[caveman99]

The test_fscommon_getfiles limitation described above is fixed by meshtastic/framework-portduino#77: VFSImpl::rmdir() called unlink(), which cannot remove a directory (EISDIR on Linux, EACCES on Windows), so rmDir() never cleared the directory tree its depth-limit case creates and the leftovers made a later run report a spurious depth truncation.

Once that lands and the platform pin picks it up, the suite goes green on Windows with no further change here.