Since the logbook was split into its own SQLite file, backup.Run was still
snapshotting a.db (the settings/config database) — so the scheduled and manual
backups silently stopped including the QSOs. The operator was backing up config
and thinking it was their log.
Track the resolved logbook file path (a.logDbPath, set in connectLogbook) and
route the backup through a new runConfiguredBackup: the CONTACTS become the
primary "opslog-*" backup (the logbook file on SQLite, an ADIF export on MySQL),
and the settings/config db is snapshotted separately as "opslogcfg-*" so nothing
is lost. backup.Run takes a name prefix; the two sets rotate independently.
A K3 (and other Kenwood-dialect rigs) answers "?;" to an IF; status poll while
it is transmitting. ask() latched that as "IF unsupported" and ReadState
returned an error, so the Manager dropped the whole CAT link mid-transmit — the
log shows connected=false / "IF rejected" the instant JTDX sent TX;. That tore
down the shared rigctld CAT the digital-mode software keys through.
Two fixes: (1) track PTT (SetPTT) and skip the wire poll while it's held —
ReadState hands back the cached last state, capped at 30 s so a missed
SetPTT(false) can't freeze it; (2) never latch IF/ID as unsupported — those are
universal on Kenwood/Elecraft, so a "?;" to them is a transient busy (mid-TX),
not a missing command.
The follow filter correctly leaves the antenna put on an un-ticked band, but the
TX-inhibit loop ran independently: it gagged FlexRadio transmit whenever the
antenna reported "moving" (a stray polled Moving flag, or a move finishing from
the previous band), even on a band OpsLog was deliberately not managing. An
operator working 15 m on another antenna had their FT8 cut mid-transmission by
an "antenna moving" interlock.
motorTXInhibitLoop now takes the covered-band set and forces moving=false when
the current rig band isn't in it — un-ticking a band means hands off entirely:
no tune AND no inhibit. Also correct the RCU-01 references to RCU-06 (the actual
controller behind the short 11-byte status frame).
New internal/rotator/dcu1 client speaking the DCU-1 command set (AP1nnn / AM1
to go-to, AI1 to read bearing), over a serial COM port (4800 baud default) or a
raw TCP serial-over-IP bridge. Azimuth-only; Stop re-commands the current
bearing since the base set has no stop opcode.
Wired through app.go (normRotorType, default port 4001, GoTo/Heading/Stop/test
switches, park returns the same "PstRotator only" note as ARCO) and exposed in
the rotor settings as "Hy-Gain DCU-1 (RotorCard DXA, Rotor-EZ, Green Heron)"
with the ARCO's serial/TCP transport chooser and a bilingual hint. Protocol is
implemented from the standard DCU-1 spec and still needs field confirmation
against a real RotorCard DXA.
Network relay boards (WebSwitch/KMTronic/Dingtian) were always reached at
http://<host>, so a board on the LAN behind a reverse proxy — the common remote
setup where the operator's 80/443 already go to Nginx Proxy Manager — couldn't be
reached from outside. relayBase() now keeps a scheme the operator supplies
(https://relay.example.com, optional sub-path) and only defaults to http:// for a
bare host/host:port. UI hint + test added.
Confirmed live: "setup fanmode=CONTEST" is accepted (reply code 0) while the
bare "fanmode=CONTEST" we switched to earlier is rejected (0x50000015), so the
fan mode snapped back and never changed on the amp. Restored the "setup " verb
and now check the reply code, returning an error if the amp rejects the set.
An older controller behind an RS232-to-Ethernet bridge returns the status frame
without its trailing FreqMax byte (11 bytes, not 12). The packet checksum is
still valid, so it's a real short frame, not a fragment. queryStatus now accepts
>= 10 bytes and defaults the missing FreqMin/FreqMax rather than rejecting it as
"too short", which reconnect-looped the link. Whether the byte LAYOUT is
otherwise identical across controller versions still needs confirming from the
protocol reference.
New DX Cluster setting (Settings → DX Cluster). Off (today's behaviour) a call
worked anywhere reads as already worked; on, the WORKED-call flag needs the same
band AND mode. It folds through the digital-mode grouping (Settings → General):
grouped, a 20m FT8 contact also marks a 20m FT4 spot worked; ungrouped they are
separate slots. Backend: qso.WorkedCallSlotKeys builds CALL|band and
CALL|band|mode(grouped) keys; ClusterSpotStatuses uses them when the option is on
(a spot with no inferable mode falls back to same-band).
Five operator-requested items:
- Alt+W clears the QSO entry. Handled before the `typing` guard and above
the keyer's key routing, so there is always one key that clears whatever
else is running — Esc is not that key when the CW keyer reserves it.
- The Grid box no longer pops outside the entry panel on a narrow window.
Row 2 needed 300+130+76+gaps = 538 px inside a panel whose min-width is
520, so Grid was pushed out and clipped at every narrow width, not just
extreme ones. QTH's min-width drops to 80 and the row wraps rather than
overflowing if it ever still can't fit.
- Selecting a QSO in the log now drives the Stats (F1) matrix. Uses its own
WorkedBefore call into separate state, NOT runWorkedBefore: that one owns
the entry form's wbRef and can trigger a field backfill, which browsing
the log must never do. The entry form wins whenever it holds a call.
- Clicking a coloured band/mode square lists the contacts behind it. Returns
the exact callsign AND the rest of the entity, because that pair is what
the cell's colour encodes; the call's own QSOs are bolded. The DXCC arm
matches the stored dxcc column only — reconstructing it from the callsign
here would disagree with the matrix above, which is built from that column.
- The Awards DXCC list shows each entity's primary prefix in its own sortable
column. Derived live from cty.dat into Ref.Group rather than stored on the
reference row, so an existing installation needs no re-seed.
Switching an SPE amp off left the already-open connection reading as connected
(the amp stops answering once off, but the stale open handle kept the status at
"connected"), so the panel never showed OFF and PowerOn — which only wakes a
DISCONNECTED amp, and must NOT pulse RTS/DTR on a running one — did nothing
until the operator restarted OpsLog. PowerOff now drops the connection and marks
the amp offline, exactly as a fresh start would; the poll's plain reopen never
wakes a switched-off amp, so it stays off until the ON button's RTS/DTR edge.
JTDX polls get_ptt DURING transmit to confirm the rig is keyed (WSJT-X does not,
which is why it worked and JTDX cut after ~2 s). We answered a blanket "0" (RX),
so JTDX concluded PTT had failed and dropped the over. get_ptt now echoes the
last PTT state commanded via set_ptt — always consistent with the client's own
command. Trace showed T 1 → three t→"0" polls → T 0.
The optimistic get_freq echo added in 0.23.5 (to stop the "Fake It" dial creep)
stopped JTDX/WSJT-X from going into transmit: the client reads frequency back
during its TX sequence, and echoing the commanded value instead of the rig's
live value interfered with it. No-TX is far worse than a 0.5 kHz drift, so revert
to reporting the rig's live frequency. get_freq/get_split_freq return s.rig.Freq()
again; the echo state, noteSetFreq, reportedFreq and the drift test are removed.
The antenna does NOT echo our sequence number — its replies carry their own
counter — so the previous seq-matching drained every reply and stalled status
updates. Revert to reading one reply per command, but flush any bytes left in
the stream before each command (drainStale): a reply left by a timed-out command
is discarded so the next read stays 1:1. readPacket also resyncs to the next STX.
This fixes the intermittent disconnects, phantom frequency jumps and wrong
element-length readings without depending on the seq.
Also: report motion for a short window right after a commanded move, so the
"moving" indicator and the Flex TX-inhibit fire the instant a band/pattern is
clicked instead of a poll (~2 s) later; the real motor state takes over once
polled.
sendCommand discarded the reply's sequence number and accepted whatever frame
came back. On a slow remote link a command that timed out left its late reply in
the stream, and the NEXT command read it as its own — crossing STATUS with
READ_BANDS/PROGRESS. That surfaced as phantom frequency jumps (a spurious
follow-loop re-tune), intermittent "reply too short" disconnects, and wrong
element-length readings (READ_BANDS getting the status frame).
Now every reply is matched to its request seq and stale/malformed frames are
drained (bounded), with readPacket resyncing to the next STX. Tested with a
net.Pipe that injects a stale reply ahead of the real one.
A digital app (JTDX/WSJT-X) sharing OpsLog's rig in "Fake It" split follows
the dial by polling get_freq. Freq() is the last polled value and lags a
set_freq by a poll cycle, so right after a transmission the client read the
still-shifted frequency, mistook it for a manual QSY and adopted it — the dial
crept down every over and never came back. get_freq now echoes the last
commanded frequency until the rig confirms it (or a short deadline passes),
closing the race. Backend-agnostic, so it fixes every rig, not just Kenwood.
Changing Normal/180/bi over a remote connection reverted to the old pattern
after ~4s: the fixed 4s optimistic window expired while the motors were still
flipping the elements, and the lagging remote status poll then overrode the UI.
Hold the commanded direction while the motors are moving, then a short grace
window after they stop for the confirmation poll to land (SteppIR already used
a generous 45s hold, which is why only Ultrabeam-remote hit this).
No single Kenwood mode digit fits every rig for a data mode: a K3/K4 wants
DATA (MD6), a TS-590SG/TS-990S data mode is a USB modifier set on the rig,
and MD6 on a plain Kenwood is FSK/RTTY. So the operator now chooses in
Settings → CAT: USB (default), DATA (MD6), or leave the rig's mode unchanged.
Only soundcard/data modes (FT8, PSK, JT…) follow this; RTTY/FSK stays its own
native mode, and voice/CW are untouched (isKenwoodDataMode + test).
The frontend engine loader had no "kenwood" case, so a saved engine of
"kenwood" fell through to the WinKeyer default — the Kenwood/Elecraft
CW-over-CAT engine never ran even though the setting showed it selected.
Also clarify the Kenwood data-mode comment: there is no single mode digit
right for both a K3 (DATA=MD6) and a TS-590SG (MD6=FSK), so data stays on
USB as the safe common base; a rig-specific DATA mode is a follow-up.
- PGXL: remote AUTH password; direct-link meter fallback (VITA-first) with
250 ms poll + peak-hold; fan mode uses bare "fanmode=" (was ignored).
- TCI: spot colour sent as decimal ARGB (ExpertSDR dropped the hex string);
spot click handled via CLICKED_ON_SPOT / RX_CLICKED_ON_SPOT to fill the entry.
- FlexRadio: clicking an OpsLog spot on the panadapter now applies the mode too.
- Filter presets: the trash icon deletes again (Radix pointer-down intercept).
- PTT hotkey: keys over CAT when a backend is active (was hijacked by a stale
Audio-tab RTS/DTR serial setting); AltGr keys allowed; presses logged.
- CW macro <LOGQSO>: logs after the preceding CW is sent, not on the first letter.
Settings -> Rotator is now a list like the amplifiers: add/remove rotors,
mix PstRotator / Rotator Genius / ARCO, and a Rotator Genius can drive both
its ports (one entry = two rotors). The compass gains a rotor selector, and
a per-rotor motorized-antenna flag so the boom/pattern paths show only for
the rotor carrying the Ultrabeam/SteppIR.
Also in this batch: a keyboard PTT hotkey (hold-to-talk or toggle, reusing
the audio PTT method with a CAT fallback), and TCI spot push to the
panadapter with click-to-fill of the callsign.
Both defaults break someone. 0.22.7 lowered the lines on Yaesu and
Kenwood and silenced a TS-990 whose interface needs RTS raised to
transmit; 0.22.8 stopped lowering them and an FT8 station reported its
output power wandering, cured by closing OpsLog — Windows raises both on
open and the interface reads one as PTT.
So it is now a checkbox on those two backends, off by default: that is
how they behaved before any of this. Xiegu keeps lowering them
unconditionally (its own report, its own explicit keying setting), and
the CI-V backend keeps the behaviour it has always had.
Reported by one in the field. Kenwood's documentation gives 024 for that
radio; both are mapped rather than betting on which is right, and the log
no longer reads "Kenwood (022)".
d4bfd30 fixed a Xiegu G90 behind a DE-19 that keyed on connect, and I
applied the same line-lowering to Kenwood and Yaesu with no evidence that
either needed it. It silenced them: many USB-serial interfaces will not
transmit with RTS low — hardware flow control, or an output stage the
line enables. A TS-990 on COM3 opened cleanly and answered nothing, which
reached us as "CAT stopped working after the update".
Xiegu keeps the behaviour: that is where the fault was reported, and that
backend now has an explicit setting for which line keys the rig. The CI-V
backend keeps its own, which predates all of this.
Kenwood also distinguishes a silent port from one sending data that never
answers, and quotes what arrived — "the rig is not answering" sent an
operator checking the power switch on a radio whose frames were visibly
on the wire.
A TS-480 connected, answered ID, then never reported a state:
discarding " 000000000010000000;" while waiting for IF
connected on COM3 @ 115200 baud
cat:state → connected=false freq=0
A headless frame tail can only come from bytes that were thrown away.
ask() buffered into a local slice, so everything left after the matched
frame — including the head of a frame still arriving — went with it, and
the link could sit one answer behind its questions. The buffer now lives
on the backend, and Connect drains whatever the rig said before AI0 took
effect rather than inheriting it.
Also: the cat:state diagnostic prints RigState.Error. It read
"connected=false freq=0" and said nothing about why, so every report of a
CAT failure arrived without the one fact that explains it — the reason
only ever reached a tooltip.
New work moves to 0.22.8; 0.22.7 is released.
The cluster already answers this question for a spot; the entry panel did
not answer it for the station actually being worked. A county hunter
needs to know before the QSO ends — by the time the county shows up in
the awards table the station is long gone.
qso.CountyWorked asks about one county instead of loading the whole
worked set, narrowed to the state so it stays a few dozen rows on a
remote MySQL. It compares through award.USCountyKey on both sides rather
than in SQL, because logged spellings vary ("Los Angeles" against
"LOS ANGELES, CA") and that key function is what resolves it everywhere
else.
An empty or non-US county is not "new" but unknown, and shows nothing: a
badge on a guess is worse than no badge.
Six equal columns gave a US county name the same width as a two-digit CQ
zone: the county truncated while the zones, the prefix and the DXCC
number sat mostly empty. The row is twelve columns now, spans set per
field.
The bearing and the two distances are read-only and never exceed
"12345 km", so they move to a flex row at a fixed width and the address —
the one field here that was never wide enough — takes the remainder.
Two DARC DOKs are both called "Gießen" — two clubs share the town. Exact
matching finds both and both are right as far as the matcher can tell, so
it kept both and the operator got a QSO credited to a DOK that may well
have been the other one.
Picking one is not an option: it writes a reference into the log with no
evidence behind it. So a Def can now be marked one reference per QSO, and
an ambiguous match on such an award assigns none, names the candidates in
the trace and leaves the contact under missing references — where the
operator's choice now sticks (06e3437).
Off by default, and deliberately NOT a revival of the old `Multi` flag: a
field holding several references (an n-fer POTA activation, a VUCC grid
line) is several references, correctly, and stays that way.
Reported on the DARC DOK award, with DL2FDM: a German address matches two or
three DOKs, the operator deletes the wrong ones and assigns the right one, and
the next recompute puts the wrong ones straight back.
The manual reference was APPENDED to what the matcher found. Storage already
kept exactly one manual entry per award — setOverrideRef drops the previous one
— so it was an override everywhere except at the one place it mattered. The
operator was arguing with the matcher and could not win.
It now replaces. The trace records what was superseded, so a corrected QSO can
still be explained rather than being merely different. A QSO with no correction
keeps every automatic reference, which the test also pins: an override must not
quietly disable matching for everyone else.
This does not address the over-matching itself — an address containing "Berlin
Mitte, 10115 Berlin" genuinely matches two DOK descriptions. That is the next
question, and a separate one.
Reported on a Xiegu G90. With the blue programming cable — three wires, no
modem lines — CAT works perfectly. Behind a DE-19, which carries audio, CAT and
PTT, the radio went into transmit the moment OpsLog connected and stayed there.
Windows raises DTR and RTS when a serial port is opened, and the DE-19 reads
them as PTT; Xiegu's own documentation asks for both low. The CI-V backend has
dropped them since it was written, for the same reason on Icom rigs with USB
SEND mapped to a line. Xiegu, Yaesu and Kenwood did not. They do now.
The second half of the report: WSJT-X through OpsLog's rigctld decoded fine and
never transmitted. Nothing was broken in that chain — set_ptt reaches the
backend, which sends the CI-V PTT command, which a G90 ignores. That is why
Xiegu keys on a hardware line instead. The backend can now do that: Settings →
CAT → Xiegu → how the rig is keyed (CI-V / RTS / DTR).
Untested here — no G90 in reach. The operator who reported it offered to try.
Reported by an operator whose node sends little: nothing for two minutes, then a
reconnect — losing the login, the filters, and any spot that arrived during the
gap.
The read had a 120 s deadline and ANY error ended the session, a timeout
included. That reads a silence as a verdict on the link. It is not: a cluster on
a dead band legitimately says nothing for minutes.
Only a real error ends a session now. A genuinely dead peer is still caught, and
by the mechanism meant for it — TCP keepalive probes an idle connection and its
failure surfaces as an error on Read, not as a timeout. The dialer sets it
explicitly rather than inheriting a default, since it is now what the design
depends on.
One trap the advice this came from did not mention: ReadString returns the bytes
it HAD read along with the timeout. Discarding them would lose the first half of
any spot line straddling a deadline, so the partial is carried over.
Tested: a listener that goes silent past the deadline and then sends a spot —
the spot arrives on the ORIGINAL connection, and the listener never accepts a
second one. The tick is a var so that test runs in four seconds instead of
sixty; a slow test is a test that gets skipped.
The operator split it in one test: WAV is fine, MP3 is not. The MP3 path is the
only one that resamples — 16 kHz up to 32 kHz, because the encoder emits broken
frames at MPEG-2 rates — and it did so by averaging neighbouring samples.
That is an upsampler in name only. Every component at f is mirrored to
16 kHz − f, and the test measures the mirror just 7 dB below the wanted signal:
a second copy of the whole spectrum. On speech it adds brightness; on receiver
noise, which is broadband, it lays a second noise floor over the top of the
band. Hence "a hiss that is not the QRM, louder than the voice" — and hence its
absence from the WAV, which resamples nothing.
Zero-stuffing plus a 63-tap windowed sinc at the old Nyquist puts the image
56.8 dB down instead of 7, at a cost measured in microseconds against the MP3
encode that follows. Two tests pin it: the image suppression, and that the level
is unchanged — a fix that quietly halved every recording would be a second
surprise on top of the one being repaired.
A station with two radios has two sets of audio endpoints, and the recorder will
happily capture the one nobody is listening to — a Flex DAX channel while the
operator works a Yaesu over USB. The file is then full of hiss with no relation
to what they hear, unaffected by the rig's AGC, and nothing anywhere said which
receiver it came from.
The device names are now logged, resolved to their friendly names, once when
capture starts.
Reported precisely: on the air the voice is well clear of the noise; in the
recording the noise is louder than the voice. Same audio, so the fault is in how
it is captured, not in what is captured.
AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY was set alongside AUTOCONVERTPCM. The
name reads like a sensible default and is not: it selects the CHEAP sample-rate
converter, for cases where quality does not matter. Taking a rig's 48 kHz stream
down to our 16 kHz, it folds everything above 8 kHz back into the audio band —
and receiver hiss is mostly high frequencies. The voice keeps its level, the
noise arrives twice.
Dropped, so Windows uses its normal converter, which filters before decimating.
This affects every capture and render path in the app, the voice keyer included.
Padlocks. LockBtn was declared INSIDE the App component, so every render gave
React a new component type and the button was unmounted and rebuilt — about four
times a second while the CAT polls. It flickered under the pointer and swallowed
clicks, because the node being clicked no longer existed when the click landed.
The padlock now lives at module level and is passed its state.
From-radio level. It reached the QSO recorder and nothing else, so an operator
listening through OpsLog heard no difference between 10% and 150% — the control
was not weak, it was disconnected. It now scales the monitor too, applied on the
captured chunk so the network-fed path keeps its own levels, and a running
monitor picks up a change immediately: making someone restart the monitor to
hear the slider is how a working control gets reported as broken.
Callsign field widened to w-56, the room coming from the RST pair which never
needs more than five characters. It is the one field always typed into, it
carries the REC badges, and a long portable call has to stay readable.
Reported precisely: press play again and it stops the current message, but you
then have to wait the whole length of the message before it will play — otherwise
the radio is keyed and the PTT released at once.
Play only SIGNALLED the previous playback to stop and started a new one straight
away. The old goroutine still held the WASAPI render client for a moment, so the
new client could not start; it returned immediately and the PTT tail was all that
was left. The log now shows both halves of that: a play line, then a release
120 ms later.
Stopping now waits for the goroutine to have returned — the point of stopping
before starting again — bounded at 1.5 s so a wedged device cannot freeze the
operator's click, and saying so in the log if it ever comes to that.
From an operator's log, playing a recording on the air repeatedly:
18:52:40.148 play → 18:52:48.420 PTT released 8.2 s, the whole take
18:52:57.759 play → 18:52:57.879 PTT released 0.12 s, nothing came out
Those 120 ms are the PTT release tail alone: the render device refused to start
and playPCM returned an error that was thrown away. The call reported success
every time, which is what "it plays once and then never again" looks like from
the outside — and why the search went to the recorder, which was innocent.
The error is now logged with the device name.
Reported: the recording plays once and never again, "as if it threw the sound
away after playing it". PeekQSO is what the playback reads and it is meant to
copy — TakeQSO is the one that clears — so this pins that, and that the copy is
independent of the recorder's own buffer.
It passes, which is the useful part: the take survives, so the one-shot failure
is downstream of it and the search moves to the file and the player.
All / CW / Phone / Digital, stacked on top of the worked/confirmed filter. The
question an operator actually has — which DXCC entities have I worked on CW but
not confirmed on CW — needs both at once; answering only one of them is what
sends people to a spreadsheet.
References carried bands but not modes, so the award computation now aggregates
mode CLASSES per reference, worked and confirmed separately. Classes, not ADIF
modes: nobody chasing CW cares whether the digital side was FT8 or RTTY, and a
list of twenty mode names would not answer the question. An unrecognised mode
counts as data rather than as nothing, so a new digital mode does not vanish
from the filter the day it appears.
Two rules the filter needs to be honest:
- Not-confirmed means not confirmed ON THIS MODE. An entity worked on CW and
confirmed on SSB is still a CW entity to chase, and that is exactly what the
filter is for.
- "Not worked" plus a mode is a contradiction — an unworked reference has no
mode — so the mode filter stands aside there instead of emptying the list.
The operator's own fetch settles it. One record, both fields:
<app_qrzlog_status:1>N <qrzcom_qso_download_status:1>Y
QRZ says the QSO is NOT confirmed and sets the download field anyway — it marks
what was handed back, not what was confirmed. OpsLog read it as a confirmation,
so eighteen QSOs came back "UPDATED" and green, none of them confirmed on QRZ.
The site's own logbook page shows them without a star.
Only app_qrzlog_status = C counts now. The test that asserted otherwise is
corrected and carries the real record.
And the download now takes back what it wrongly gave: when QRZ explicitly says
not-confirmed and our log says confirmed, the flag is cleared and reported on
its own line. These QSOs count towards award slots, so leaving them green until
someone noticed was not an option — and nothing else in the app would ever have
undone them. Only on an explicit no: a record with no app_qrzlog_status is QRZ
saying nothing, and silence is not a retraction.
Both faults reported on the air, both reproduced in synthetic signals first —
the existing tests all passed because each starts a fresh decoder and sends
textbook spacing, which is neither of the cases that hurt.
Speed change. A 32 wpm station followed by a 14 wpm reply decoded as
"TTT TTTT TTTTT TT…" for most of the over. Not the element classifier: the
CHARACTER-gap threshold is 2.2 dits, so at the stale fast estimate it stood at
81 ms while the newcomer's element gaps were 86 ms. Every single element was
flushed as its own character, and a lone element with no contrast reads as a
dah. A silence long enough to end an over now drops the estimate back to the
seed, which is exactly the state a freshly started decoder is in — and that case
was always fine. The threshold scales with the current estimate (12 dits, floor
600 ms) because a fixed one cannot serve both 10 and 40 wpm.
Callsign split. The word boundary sits at 4.6 dits, the geometric mean of a
3-dit letter gap and a 7-dit word gap. It assumes textbook spacing; a fist
leaving 5 dits between letters had every letter turned into a word, so a
callsign arrived as "O Y 1 C T". The boundary now also follows the letter gaps
this operator actually sends, whichever is larger. A wide sender's words may run
together — a far smaller price than a callsign in pieces.
The operator asked the right question: why wait 1.5 s at all, when the mode is
known from CAT?
Because the wait is not the problem — what happened during it was. The drift
guard, meant for two sound cards running on their own clocks, saw a starved
source as a one-second skew and trimmed the radio audio to match. So the opening
of every CW recording was thrown away although it had been captured.
The guard now applies only while BOTH sources are actually delivering. Waiting
then costs nothing: everything buffered is written whole the moment the silent
source is written off, which a test now pins with three seconds of audio.
Using the CAT mode to skip the wait would have been a second mechanism for a
case this one already covers — and it would not cover the others: a device
switched off, unplugged, or held by another program.
The log read:
recorder: the second audio source is silent — recording the radio alone
recorder: no audio at all from "DAX Mic" after 3 s — nothing is streaming
Both true, and together they read as a fault while the recording was going
perfectly well — it was saved seconds later. On CW the mic channel legitimately
delivers nothing; the mixer already said what it was doing about it.
The watchdog now mentions the mic only when the radio is silent too, which is
the case where the recording really is empty.
The operator said it from the start: in CW there is no sidetone, only reception.
My previous fix did not cover it — it wrote off a source that STOPPED, but a
source that never starts has no last-delivery time, so it stayed forever "not
yet dry" and held the mixer at min(A,B)=0. Three empty recordings while the
radio audio streamed perfectly.
A source with no delivery is now measured from when capture started. Tested both
ways: nothing is written off in the first moments, and a source that has never
spoken is written off once the grace period has passed.
Third empty recording, same station, and the sequence was as plain as it gets:
type the call, click record, log the QSO. The device opened and never streamed.
Endpoint ids are what gets stored, and a DAX channel that is reconfigured,
disabled or recreated comes back with a DIFFERENT id. The stale one opens
without complaint on some drivers and simply never delivers — indistinguishable
from a quiet band until the file turns out to be empty.
The recorder now checks the id against the current endpoint list when it starts,
which takes milliseconds, and says outright that the configured input is gone —
on screen, not only in the log. If it IS still there, the answer is equally
useful: the device exists and the fault is upstream, in whatever should be
feeding it.
rouggy