Summary

Fixes #2373 (hardware PTT into radio doesn't drive forward RF when mic_selection=PC) and the duplicate #2200 (Jerry KE9U's external PTT on FLEX-8400M).

When the user keyed the radio's hardware PTT (mic-line PTT, footswitch into ACC, RCA TXREQ), the carrier came up but no PC mic audio modulated it. The gate at RadioModel.cpp:3382 forced TransmitModel::setTransmitting(false) whenever m_txRequested was false — even when the radio reported our client handle as the TX owner. moxChanged(false) propagated to AudioEngine::setTransmitting(false), gating the mic capture path.

Root cause

if (!m_txOwnedByUs || (!m_txRequested && !m_cwKeyActive && !m_cwxActive && !m_transmitModel.isTuning())) {
    m_transmitModel.setTransmitting(false);
    if (m_txAudioGate) { m_txAudioGate = false; emit txAudioGateChanged(false); }
}

The bypasses (m_cwKeyActive, m_cwxActive, isTuning()) cover the existing non-m_txRequested TX paths, but hardware-keyed PTT had no such bypass. Effectively a latent bug: the PC-mic-with-hardware-PTT combination has likely never worked through this branch.

Fix

Parse the radio's interlock source= field (SW/MIC/ACC/RCA/TUNE per FlexLib v4.2.18 ParsePTTSource), persist it across status updates that omit it, and add MIC|ACC|RCA to the bypass list:

const bool hardwarePtt = (m_lastInterlockSource == "MIC"
                          || m_lastInterlockSource == "ACC"
                          || m_lastInterlockSource == "RCA");
if (!m_txOwnedByUs ||
    (!m_txRequested && !m_cwKeyActive && !m_cwxActive
     && !m_transmitModel.isTuning() && !hardwarePtt)) {
    // force off
}

SW source still requires m_txRequested, so the optimistic local-unkey behaviour from docs/TX_SYNC_FIX_REPORT.md is preserved — a stale state=TRANSMITTING after setTransmit(false) on SW source still falls through to the force-off branch (no SSB unkey-tail regression).

m_lastInterlockSource is cleared on:

• Disconnect (alongside the other TX flags)
• Interlock confirming a non-TX state (RECEIVE/READY/NOT_READY) — so a stale hardware source can't outlive the actual key release if a later status update omits source=.

Validation against FlexLib v4.2.18

Validated against /Users/patj/aether/FlexLib_API_v4.2.18.41174 reference docs:

• source= field exists: Radio.cs:7932-7944 ParsePTTSource parses exactly these five values. Confirmed.
• FlexLib's MOX gate: Radio.cs:7577 is IsInterlockMox(state) && ShouldUpdateMoxOrTuneState(handle). IsInterlockMox accepts Transmitting | PTTRequested | UnkeyRequested. FlexLib has no m_txRequested analogue — it doesn't do AetherSDR's optimistic-off, so it doesn't need to distinguish SW from hardware sources for the gate.
• Why we still need the source distinction: AetherSDR diverges from FlexLib intentionally (per TX_SYNC_FIX_REPORT.md) for SSB unkey responsiveness. The m_txRequested flag is what guards optimistic-off — we can't drop it. The source field is exactly the right primitive to apply optimistic-off only to SW source while accepting interlock as authoritative for hardware sources.

Timing / ordering preserved

The user explicitly asked that this not adversely impact panadapter/waterfall timing. Verified:

• radioTransmittingChanged is emitted at RadioModel.cpp:3380 unconditionally, before any gate logic. This is what drives waterfall freeze, S-Meter mode, and the TX status indicator at MainWindow.cpp:2006-2024. The fix touches only the gate at line 3382 and below — radioTransmittingChanged emit timing is byte-identical to pre-fix behaviour.
• moxChanged → AudioEngine::setTransmitting (MainWindow.cpp:1965-1986) now correctly fires true when hardware PTT keys the radio, so PC mic audio enables. This is the intended new behaviour.
• txAudioGateChanged remains a TX-OFF-only fallback at MainWindow.cpp:1991-2000 (the report explicitly warns against reintroducing TX-on wait-for-interlock for SSB).

Acceptance criteria coverage (from issue #2373)

• TransmitModel::isTransmitting() returns true when radio reports our handle + TRANSMITTING with hardware source
• AetherialAudioStrip TX indicator and Chain applet TX endpoint pulse (driven off moxChanged)
• Forward RF appears (audio gate opens via moxChanged → AudioEngine::setTransmitting(true))
• MOX path unchanged (source=SW + m_txRequested=true → existing accept path)
• Multi-Flex correctness (!m_txOwnedByUs outer check unchanged, hardwarePtt only relaxes the inner condition)
• SSB local-unkey race protection preserved (SW source still gated on m_txRequested)

Files changed

• src/models/RadioModel.h — add m_lastInterlockSource member
• src/models/RadioModel.cpp — parse source=, add hardwarePtt bypass, clear on non-TX/disconnect

40 insertions, 1 deletion. No protocol changes, no new dependencies.

Test plan

• Hardware mic PTT with mic_selection=PC produces forward RF and lit Aetherial Audio TX indicator (primary acceptance — issues #2200, #2373)
• Footswitch via ACC TXREQ produces forward RF + audio
• RCA TXREQ produces forward RF + audio
• MOX button still works (no regression on SW path)
• CW key, CWX, TUNE still work (no regression on existing bypasses)
• SSB local unkey tail still immediate (no regression on optimistic-off)
• When another client owns TX (!m_txOwnedByUs), local indicators stay dim, no audio flows (Multi-Flex)
• Waterfall freeze/unfreeze timing unchanged on local MOX (sanity check that Fix waterfall TX trail by gating on radio interlock state (#1927) #2368 wasn't disturbed)

🤖 Generated with Claude Code