Hey @jensenpat — thanks for working on this, the async log writer is a nice direction. 👍

I took a look at the CI failure on b4d6dc6. Quick read on what's happening:

What failed

• ✅ check-windows — passed
• ❌ build (Linux) — failed at the Build step
• ❌ analyze (cpp) (CodeQL, Linux) — failed at the same Build step

The Linux-only failure pattern, plus a Configure step that succeeded, points at a link-time issue rather than a code issue.

Likely cause: missing pthread link on the four test executables

src/core/AsyncLogWriter.cpp uses std::thread, std::mutex, and std::condition_variable. On Linux with GCC, those need -pthread at compile and link time — without it you get undefined reference to pthread_create / pthread_join etc. On Windows the Win32 thread implementation is built in, so MSVC links cleanly, which matches what we're seeing.

In the CMakeLists.txt change, AsyncLogWriter.cpp was added as a source to four test targets:

• ole_compound_file_test (line ~1275)
• meter_model_test (line ~1330)
• transmit_model_apd_test (line ~1348)
• transmit_model_test (line ~1386)

…but none of those four targets link pthread. The main AetherSDR target already does (it picks up pthread via the DFNR/MQTT branches around CMakeLists.txt:1050 and :1107), and iambic_keyer_test shows the established pattern for tests:

if(UNIX)
    target_link_libraries(iambic_keyer_test PRIVATE pthread)
endif()

Suggested fix

Add the same if(UNIX) block after each of the four target_link_libraries(...) calls — e.g. for meter_model_test:

target_link_libraries(meter_model_test PRIVATE Qt6::Core)
if(UNIX)
    target_link_libraries(meter_model_test PRIVATE pthread)
endif()

…and the same for ole_compound_file_test, transmit_model_apd_test, and transmit_model_test. (Using Threads::Threads via find_package(Threads) would be more portable, but matching the existing in-repo style is probably the right call here.)

Notes

• This is purely a build-system gap from adding a new threaded source to existing test targets, not a problem with the AsyncLogWriter design or your main.cpp refactor — both look clean to me on a read-through.
• Because the failure was at link time on the test targets, the main AetherSDR binary itself almost certainly compiled fine; it's just that the test stage falls in the same CI step.
• No Copilot review comments visible on the PR yet, so nothing to cross-reference there.

Once you push the pthread fix, the Linux build and CodeQL analyze should both go green. Holler if I've misread the symptom and the actual error is something else — I was working from the check-run summary rather than the raw build log.

🤖 Generated with Claude Code

Claude here — really clean foundation milestone, jensenpat. Producer-consumer split with the high-priority reserve + drop-summary coalescing is exactly the production-grade pattern (spdlog/glog shape), and the lifetime hardening (heap-allocated singleton + explicit shutdown ordering with qInstallMessageHandler(nullptr) before shutdownLogging) addresses the late-teardown reentrancy that the old code was already hedging against with the heap-allocated regex objects. The two redaction tightenings (IPv4 negative-lookbehind for ver= prefixes, colon-separator MAC pattern) caught real cases the previous regex would miss. Format preservation means existing log analysis tooling keeps working unchanged. CI ran fully on Windows too. Two follow-ups filed (#2497 unit tests, #2498 rotation) — both are tracker hygiene, not blockers. Merged.

73, Jeremy KK7GWY & Claude (AI dev partner)