Release v0.3.0-beta · WD9N/Haven-FSK

v0.3.0-beta — July 2026 (current)

Safety / on-air behavior

PTT release failure is now loud: txOff() verifies the unkey command,
retries up to 3 times, and raises a modal "YOUR RADIO MAY STILL BE
TRANSMITTING" alarm if the rig can't be confirmed unkeyed — including
when the rig-control connection dropped mid-TX (previously the unkey
was silently skipped). A failed key-up now aborts the transmission
instead of playing audio into an unkeyed rig.
End-of-transmission key click eliminated: one raised-cosine amplitude
envelope over the whole transmission (ramp-up at start, ramp-down at
end) — the frame previously ended mid-sine at full amplitude, which
splattered on adjacent frequencies.
rigctld replies are now framed properly: stale bytes from timed-out
commands are drained before each new command, and reads continue until
the reply is structurally complete — a stale "RPRT 0" from a frequency
poll can no longer be mistaken for a successful PTT command.

Stability

Fixed a cross-thread use-after-free: moving the TX gain fader in the
final moments of a transmission could call into a just-deleted audio
device object. TX gain is now an atomic value owned by AudioEngine.

Logging (ADIF/QSO)

ADIF exports use MODE=MFSK / SUBMODE=HAVEN-FSK (and MODE=PSK /
SUBMODE=PSK31) instead of the invalid MODE=DIGITAL, so LoTW/QRZ/POTA
uploads are accepted. Legacy DIGITAL rows are mapped at export time.
PSK31 contacts are now logged as PSK31, not HAVEN-FSK.

Architecture / infrastructure

src/dsp is Qt-free again (ADR-003): new DspLog shim replaces QDebug in
Frame/MfskModem; DspPipeline (deliberately Qt-facing) moved to
src/pipeline/.
Debug self-tests actually run now — QT_DEBUG was never defined under
CMake, so every previous "Debug build" silently compiled them out.
haven_debug.log is actually written now — the log path was resolved
before QApplication existed, so the file never opened.
FEC matrices built once per Frame instead of per assemble()/parse()
call; assorted log/comment accuracy fixes.