Backend adds cyclical distributions ByHour (hour-of-day), ByDOW (weekday), ByDOM
(day-of-month) and ByMonthYr (calendar month), all UTC. The Activity-over-time
card gains a selector: Timeline keeps the chronological month trend; Day/Week/
Month/Year show the hour/weekday/day/month distribution as bars.
Adds Stats.ByBandCategory (per band: CW/phone/data counts, band-plan order) via a
modeCategory() bucketing in stats.go, and a StackedBandBars chart in the Stats
panel showing the split per band with a legend.
MSHV/WSJT-X often stamp a personal STATION_CALLSIGN (e.g. F4BPO) that differs
from the special-event/contest profile you're running in OpsLog (e.g. TM74TFR).
The QSO then failed the extsvc station-callsign guard and was never uploaded to
the (matching) Club Log logbook. Live UDP logging now stamps the active profile's
callsign when the base call genuinely differs (portable variants left alone).
publishLiveStatus only ran on QSO log or the 15s heartbeat, and the widget polled
every 15s, so an operator's band/freq (and their appearance) could trail by up to
a minute. Now ReportLiveActivity debounce-publishes ~1.5s after a freq/band/mode
change, and the widget polls every 5s (and 1.5s after qso:logged).
The 15s poll effect depended on the dbConn OBJECT, which gets a fresh reference
on several events — each reset the interval before it fired, so the widget could
stay stale for over a minute. Depend on the is-mysql boolean instead, and also
refresh ~2s after qso:logged so our own row appears promptly like the ON AIR
badge.
- Removes the RX-audio CW decoder: deletes app_cw.go + internal/cwdecode, the
cwDecoder/cwPitchHz/cwMu/cwStop App fields, and all frontend state/effects, the
header Ear button, the Tools menu item and the decoded-text strip. The CW
KEYER (WinKeyer/Icom/Flex) is untouched.
- Fix: typing a callsign while a macro is keying now aborts the CURRENT engine
(Flex CWX / Icom / WinKeyer) on the first character, not just WinKeyer during
an auto-call loop — via a shared stopKeyerTx() helper.
Implements the SPE Application Programmer's Guide protocol: packets
0x55 0x55 0x55|CNT|DATA|CHK (sum%256), OPERATE key 0x0D (toggles STANDBY/OPERATE)
and STATUS request 0x90 whose 0xAA-framed reply is a 19-field CSV (mode, RX/TX,
band, power level, output W, SWR, V/I, temp, warnings, alarms). internal/spe
polls status ~1/s over USB serial (go.bug.st/serial, 8N1) or TCP (RS232-to-
Ethernet bridge) — same codec, different transport.
Wired via startPGXL (starts the SPE client for spe* types), bindings GetSPEStatus
/ SPESetOperate, and a live status card + OPERATE/STANDBY toggle in the Amplifier
settings panel. Only the two example-anchored commands are sent (safe); other
keystroke codes were ambiguous in the guide's table. Untested on hardware.
Renames the Hardware section to Amplifier and adds an amplifier-type selector
(4O3A PowerGenius XL, SPE Expert 1.3K-FA / 1.5K-FA / 2K-FA) plus a transport
choice for the SPE amps: USB (serial COM + baud) or Network (RS232-to-Ethernet
IP:port). PGXLSettings gains Type/Transport/ComPort/Baud (keys keep the pgxl.*
prefix for back-compat). PowerGenius still drives over TCP; SPE settings are
stored but control is not wired yet (needs the SPE serial protocol).
- Denkovi (FT245 D2XX) now selectable as 4 or 8 relays (a 4-relay board just uses
the low nibble); NewDenkovi takes a count, driven by the device's Channels.
- New 'usbrelay' backend: cheap CH340/LCUS USB-serial boards over a COM port
using the common A0 protocol ([0xA0][relay][state][checksum]); write-only, so
Status is a shadow. Channel count configurable (1/2/4/8/16).
- StationDevice gains Channels; deviceRelayCount() drives label/relay counts for
the variable-size types. Device editor: channel selector + COM-port picker
(usbrelay) / FTDI-serial (denkovi).
Both untested on hardware; the A0 command set / Denkovi bit order may need a tweak
after a live test.
Third relay device alongside WebSwitch and KMTronic. Despite enumerating as a
COM port, this board is driven via FTDI D2XX synchronous bit-bang (one byte = 8
relays, bit 0 = relay 1), not serial ASCII — so we load ftd2xx.dll at runtime
(no CGO) and call FT_OpenEx/FT_SetBitMode/FT_Write, addressing the board by its
FTDI serial (e.g. DAE0006K), exactly like the vendor tool. Windows-only (stub
elsewhere); ListDenkoviDevices enumerates connected serials for the picker.
Wired into relaydev (Count/Status/Set), app.go (deviceDriver/relayCountFor/type
whitelist + ListDenkoviDevices binding), and the Station device editor (new type
+ FTDI-serial field with Detect). Untested on hardware; bit order / on-polarity
may need a flip after a live test.
Completes the CWX type-ahead loop: with send-on-type the keyer-panel CW text
already streams each typed char to the radio's CWX buffer (which keys in order,
so you can keep typing while it sends); this adds the matching un-send. Route
wkBackspace to FlexBackspaceCW -> cwx erase N, so backspacing a mistyped char in
send-on-type removes it from the buffer before the radio keys it.
The ESC abort keyed WinkeyerStop unconditionally, so it never aborted the Icom
CI-V or Flex CWX keyer. Route the stop through cwSource like the panel Stop
button does.
PortBtn was defined inside AntGeniusPanel, so every render produced a new
component type and React unmounted/remounted every port button. Normally
re-renders are rare so it's invisible, but the CW decoder floods the app with
cw:text/cw:status events (re-render many times a second), tearing the buttons
down between mousedown and mouseup — the click never completed and the antenna
never switched. Move PortBtn to module scope (stable identity) and pass
onActivate/t as props.
Adds a Rename button to Settings -> Database: renames the current SQLite database
(e.g. opslog.db -> F4BPO.db) while preserving every setting/profile, unlike New
database which starts empty. Implemented as a VACUUM INTO copy + pointer switch;
the old file (open and locked now) is scheduled via config.json delete_pending
and removed, with its -wal/-shm sidecars, on the next launch.
Adds ExportAwardForCatalog(code, version): writes the selected award as a
catalog-format JSON ({"def":…,"references":…}) stamped with a version and with
user_edited cleared, ready to paste over internal/award/catalog/<code>.json. A
new release then propagates the change to the whole team via mergeCatalog —
unedited copies auto-upgrade, edited ones are offered the update.
UI: a version input + "Publish to catalog…" button in the award editor footer,
defaulting to one past the award's current version.
DDFM now derives the French department from the address postal code, not only
from a note typed by hand — the catalog def gains the address OR-rule (\d{2}\d{3}
-> Dxx), France+Corsica scope, and version 2. On next launch every operator whose
DDFM is unedited auto-upgrades to it (user-edited copies are offered it); the
materialised award_refs then recompute so ~500+ French QSOs gain their department.
Also: a catalog definition update now clears the award_refs backfill flag, so any
future catalog bump self-heals the stored columns without a manual flag bump.
The one-shot award_refs backfill never revisited rows after it ran, so a QSO
materialized before an award gained a rule kept stale values — e.g. DDFM's
postal-code OR-rule matched nothing on the 500+ French QSOs whose department
comes from the address, only the handful with it typed in notes. Two fixes:
bump the backfill flag (awards.materialized.v2) so every operator re-materializes
once on next launch with their current definitions, and make RescanAwards (the
Awards tab Rescan button) also recompute award_refs for an on-demand refresh.
The File menu's Exit item was hard-disabled and its action had no case in the
menu dispatcher, so it did nothing. Enable it and route file.exit to QuitApp,
which runs the same graceful beforeClose shutdown as the window close button.
Live award detection merges pickable refs into award_refs per call, but nothing
dropped them when the call was cleared, so a previous call's ref lingered onto
the next (IT9AOT's ref still showing after typing F4BPO, both listed in the F3
Awards tab). Clear award_refs the moment the callsign field is emptied.
Award refs are now computed once and stored on the QSO as a JSON column
(award_refs, e.g. {"DDFM":"74","WAJA":"12"}) instead of recomputing the whole
award engine on every grid page load. Written on log/edit/UDP-import and
bulk-recomputed when an award definition or reference list changes; a one-time
per-logbook backfill materializes pre-existing QSOs. The grid reads the column
directly like any other field, so it's fast and available everywhere (including
the shared MySQL logbook).
Also fix per-profile grid column layout: the DB copy was already per-profile,
but the localStorage cache used one global namespace and shadowed it, so a
profile switch kept the previous profile's columns/widths. gridPrefs now scopes
the cache by active profile id and the grids remount on profile:changed.
The 7610-only FreqA gate broke CAT on the 7610 when FreqA was momentarily 0.
Match DXHunter/WSJT-X: read FreqA first for all rigs, fall back to the generic
Freq (IC-9100 etc.), then FreqB. OmniRig's generic Freq maps to VFO B on the
7610, which is why keying off FreqA is correct.
Live status: when an operator goes offline (no QSO in the window) OpsLog now
DELETEs its live_status row instead of just flipping online=0, so a status page
that lists present/recent rows shows them as gone without having to read the
online column — the whole point of the feature. The row reappears on the next
QSO.
Recent QSOs columns: toggling an award column (e.g. DDFM) rebuilt columnDefs and
made AG Grid re-apply every colDef hide default, so hidden non-award columns
(QTH/Grid) came back and restoringRef stayed stuck true (saving off). The
restore-saved-state effect now also runs on awardShown changes, so the other
columns keep the user's visibility.
Rate meter: on a shared MySQL logbook it now shows two lines — OP (the active
operator, accent) and Team (all operators, foreground) — each with the 10/60-min
QSOs/hour; single-op keeps the one-line display.
Live-stations widget: only stations currently on air are listed (offline ones
are hidden rather than greyed).
Recent QSOs footer: 'Showing X of Y' now reflects the AG-Grid COLUMN filters
(the funnel icons) — the grid reports its displayed row count, so filtering a
column shows how many QSOs remain instead of always the full total.
GetQSORate now returns both the active operator's rate (their own performance)
and the whole station's rate (all operators combined). RecentRateBreakdown
computes both in one scan of the recent rows (per-operator + all-operators),
replacing RecentRate.
The new build downloaded and OpsLog quit, but never came back. Two Windows
causes: the freshly written exe carried the internet Zone.Identifier mark, so
SmartScreen wanted to prompt "are you sure you want to open this?" — invisibly,
since we launch it programmatically — and silently blocked the launch; and
starting the new exe while the old one was still exiting raced the single-
instance mutex.
Now the swapped exe's Zone.Identifier stream is removed, and the relaunch is
done by a detached, hidden PowerShell that Wait-Process's on our PID (so we're
fully gone and the mutex is free) before Start-Process'ing the new exe.
Closing OpsLog dropped the CAT, whose frequency went to 0; the auto-control
read that as out of range and switched the relay OFF, then ON again at the next
launch. Now a rule is skipped entirely when the frequency/band is unknown (CAT
off), so relays keep their state across a close. And on the first evaluation
after launch/save it reads the boards' LIVE state and only switches a relay
that isn't already in the wanted position — no more clunk when it's already
correct.
At launch nothing is in memory yet, so the badge showed offline even with a
recent QSO. liveLastQSOTime is now authoritative: it takes the most recent of
the in-memory stamp AND the DB (this operator's last QSO — covers a contact
from the shared logbook or one logged before launch), used by both the
published status and the badge. The badge polls the backend every 5 s (down
from 15) and on each qso:logged, so it shows on air within a few seconds and
flips online instantly on a new contact.
The targeted extras write used SET extras, but the column is extras_json, so
sending an OpsLog QSL card (and the recording stamps) failed with an unknown-
column error. Use the real column name.
ON AIR badge (bottom status bar, shown when live-status publishing is on): a
blinking red light + "On air" when a QSO was logged in the last 5 min, dim
"Offline" otherwise. Seeded at launch from LiveLastQSOAgeSec so it's right
immediately, flips online on qso:logged, expires via a 10s tick.
Auto-updater UI: the update check now also runs every 10 minutes, and the
update notification downloads and installs in-app — a live progress bar
(update:progress) then OpsLog restarts on the new build. Falls back to opening
the release page when there's no auto-download asset or on error (retry).
Regenerated bindings for DownloadAndApplyUpdate / LiveLastQSOAgeSec.
CheckForUpdate now also resolves the release's downloadable asset (portable
.exe, or a .zip to unpack). DownloadAndApplyUpdate streams it with progress
events (update:progress 0-100), swaps it in for the running exe (rename the
current one to .old — allowed for a running image on Windows — move the new one
into place, roll back on failure), then relaunches with --post-update and quits.
main: a --post-update relaunch retries the single-instance mutex for ~20 s so
the fresh process waits for the old one to exit and free it, then clears the
.old exe left behind. Falls back to opening the release page when a release has
no auto-download asset.
The multi-op live status heartbeated every 15 s, so an operator who left the
log open but stopped working still showed on air. Now it publishes an explicit
online flag: online when a new contact was logged within the last 5 minutes,
offline otherwise. live_status gains `online` + `last_qso_at` columns (added to
existing tables via ALTER). Logging a QSO (manual, UDP, ADIF monitor) stamps the
time and republishes at once, so they flip back online instantly.
At launch the last-QSO time is seeded from the DB (Repo.LastQSOTime), so an
operator who worked someone just before starting OpsLog is shown online right
away rather than offline until their next contact. LiveLastQSOAgeSec exposes it
to the UI badge.
Duplicate logs: `saving` is React state (async), so a burst of Enter presses /
clicks fired before the re-render each ran a full AddQSO and logged the same
QSO several times when a slow lookup made logging take a moment. A synchronous
savingRef guard now ignores the repeat immediately.
ESC: clearing the entry (ESC) left the "looking up" spinner turning and let a
still in-flight lookup re-populate the field afterwards. resetEntry now cancels
the pending lookup debounce, hides the spinner at once, and bumps a generation
counter so a late lookup result is discarded instead of applied.
Status clobber: sending an OpsLog QSL card (or stamping a recording) read the
QSO, did slow work (e-mail send), then wrote the WHOLE row back from that now-
stale copy — silently reverting clublog/qrz_qso_upload_status from Y to R that a
concurrent auto-upload had just set. That's why QSOs "showed R again, but only
with QSL sending on", intermittently. New Repo.SetExtra does a targeted
`UPDATE qso SET extras = ?` that touches only the extras column, so an app-field
stamp can never clobber another column. Used by SendEQSL, markRecordingSent and
the recording-path stamp.
Logging speed: the log path no longer runs a callsign lookup at all — the e-mail
already arrives on the QSO from the entry lookup, so the second one was
redundant and stalled logging on a slow/not-found QRZ. And the entry lookup
(LookupCallsign) is now bounded to ~2 s: providers get 2 s, then it falls
through to cty.dat instead of spinning 10 s+ on a call that isn't in QRZ.
The single-worker upload queue introduced head-of-line blocking: one slow
upload (LoTW signing via TQSL, or a service on a 30 s timeout) stalled the
whole queue, so every following QSO's Club Log upload waited and the rows sat
at "R". Operators on the older, concurrent build had no such problem — the
tell that this was a fresh regression.
Back to one goroutine per upload (never serialised), so a slow upload never
holds up the rest, while keeping the transient-failure retry with back-off
(concurrent, so it can't block anything).
OmniRig: on the IC-7610 the generic Freq property reports the wrong VFO (its
Main/Sub model confuses the stock ini), so OpsLog showed VFO B. Detect the rig
by RigType and, in simplex, read VFO A explicitly — matching Log4OM. Only the
7610 is affected; other rigs keep using the generic Freq.
Icom network: when the rig tears the session down (control/CI-V 0x05) OpsLog
only logged it and kept the half-dead link until the 6 s liveness timeout
expired. Mark the link dead on 0x05 so Alive() fails on the next poll and the
manager reconnects cleanly right away.
The RST comboboxes allowed free text, so a report typed by hand that isn't in
the mode's list (e.g. "600") was committed. Drop allowFreeText on the four RST
fields (QSO entry + edit modal) and make commit-on-type push a value only when
it's actually in the list; a leftover invalid entry reverts to the last valid
value on blur. Programmatically filled values (FT8 SNR over UDP, presets) are
unaffected — only manual typing is constrained.
Headline KPI tiles are now tinted cards: a soft colour wash, a corner glow and
a left accent bar, each tile its own hue (QSOs blue, calls teal, DXCC violet,
continents orange, confirmed green) — the number itself stays in foreground ink
for full contrast on every theme. Chart cards gain a soft shadow and a coloured
dot in the header matching the section.
Charts are no longer all-blue: BY MODE colours each bar by the categorical
palette (modes are real categories, like the continents donut), and the other
horizontal-bar charts take their card's accent (top entities green, operators
violet, stations orange) instead of a single blue. Single-hue-per-series
discipline is kept; only genuine categories get per-bar colour.
A QSO logged via UDP / ADIF-import fired `go m.upload` per QSO, so a pileup or
an ADIF-import burst hit a service with dozens of concurrent requests at once.
Club Log's nginx answers that burst with 403; the upload counts as failed and
the QSO stays at "R" while a lucky few go through — the real cause behind the
"it's uploaded but shows R" report (not the earlier display race).
Route all immediate/delayed auto-uploads through a single serialised worker
(one at a time, 250ms apart) so no burst ever trips a per-IP rate limiter.
upload() now reports whether a failure is transient (HTTP/rate-limit) vs
permanent (not eligible, wrong station callsign, no record); transient ones
are retried with exponential back-off (up to 4 attempts, 1s/2s/4s), so a QSO
that hit a momentary 403 ends up marked instead of stuck at R. A full queue
falls back to a goroutine rather than dropping an upload.
Automatically switches the Station Control relay boards from the rig's
current frequency / band. Each relay carries one rule: off (manual), a
frequency window (ON inside [lo,hi] kHz, OFF outside), or a set of bands
(ON on those bands, OFF elsewhere). Evaluated on every CAT frequency/band
change; a relay is only switched when its desired state actually changed,
so tuning within a range doesn't hammer the board.
A cached atomic flag keeps the CAT hot path a no-op when the feature is off
(important during FT8 slice churn). Saving re-applies from the live
frequency so a changed rule takes effect immediately.
New Settings → Hardware → Relay auto-control section: master enable plus a
per-relay mode (Off / Frequency / Band) with kHz range inputs or band
chips, per configured relay board. i18n EN + FR. Azimuth/Time modes (the
other two PstRotator tabs) are left for later.
rouggy