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.
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.
Club Log: on a failed batch, log the callsign#id of every QSO in it so a
per-record rejection (e.g. a field value nginx's WAF blocks with a 403) can
actually be located instead of hiding behind "batch FAILED".
FCC ULS: catch the maintenance bounce before following it. data.fcc.gov
redirects to www.fcc.gov/system-maintenance during maintenance windows, and
that page then HTTP/2-stream-errors — which surfaced as a cryptic
INTERNAL_ERROR. Detect the redirect and return "try again later".
Opt-in via Settings→General (portable pref opslog.showQsoRate). Shows the
contest-style QSO rate in QSOs/hour, projected from the trailing 10-minute
(count ×6) and 60-minute windows, between the widget icons and propagation.
Backend: qso.RecentRate counts QSOs whose start time falls in each trailing
window, scanning only the last 400 rows (cheap on a large log); App.GetQSORate
exposes the 10/60-min counts. Frontend refreshes on qso:logged and a 30s tick.
The meter shares the propagation grid cell — the header is a fixed 6-column
grid, so adding it as its own child pushed profile/band-map/compact onto a
second row. i18n EN + FR.
New Operator picker next to the contest picker, populated with the operators
actually present in the log. Selecting one recomputes every statistic for that
operator — QSO/DXCC totals, top countries, continent split, by band/mode, and
the rate/best-60. "— Station owner —" buckets QSOs logged with no OPERATOR set.
Shown only when the log has more than one operator.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
rouggy
