Auto-sync the clock by default (one-slot acquisition), drop callers who gave up, swap the TX over on a late reply

[patrickrb]

Three field complaints from the same activation, all in the QSO timing loop.

1. Auto-sync still needed the manual "Apply suggested correction"

ClockSelfSync stopped acting once |median DT| ≤ 0.30 s and required ≥ 4 decodes per slot — so on a quiet POTA band it never fired, and on a busy one it parked the clock ~0.25 s off, which is exactly the residual the operator kept applying by hand.

• Deadband 0.30 s → 0.15 s (two DT measurement steps; still doesn't chase noise).
• A slot with 1–3 surviving decodes now counts, but needs three agreeing slots instead of two (a lone station's DT is also that station's clock error).
• The Suggestion card no longer offers the manual apply while auto-sync (or GPS discipline) owns the clock; it shows the last auto-correction and its time instead (GeneralVariables.mutableSelfSyncApplied).

2. Queued callers who gave up were still called back

Nothing ever pruned the caller queue, and it was only fed on cycles where the partner went silent (the scan lived behind the no-reply branch).

• QueuedCaller.lastHeardUtc tracks the slot of their latest call; every pass (fast and evidence-only) refreshes it via refreshCallerQueue().
• CallerQueueOrdering.pruneGivenUp() drops anyone not heard in the most recent receive slot (one full cycle) before the queue is worked — on QSO completion, on give-up, and while idling on CQ. Nobody fresh → back to CQ, or answer whoever is calling right now. Strongest-first picks among live callers only.
• Operator-tapped callers (dequeueSpecificCaller) are never pruned.
• The QUEUE chips now reflect who's actually still calling.

3. Re-sending the report instead of RR73

When the partner's R-report decoded just after key-up, #704's mid-cycle TX swap was supposed to fix the over — but it could never fire: both decode paths in MainViewModel.afterDecode stashed anything that landed while transmitting, and the drain refused to run while transmitting, so the reply was replayed a full cycle later (our own-slot fast pass ~14 s after key-up).

• Late decodes are now parsed immediately as evidence-only (same mode the replay used), so the order advances at once and shouldRestartForNewOrder can swap the over inside the audio slack. Past the slack the advanced order keys the next cycle — no stash needed.
• The TX worker also picks up a swap requested during the PTT settle before the first sample plays; otherwise playFT8Signal() resets the cancel flags, plays the old over in full, and replays the new one ~13 s in where the clip math strips it.
• PendingSequencerDecodes removed; FastPassDisposition.Action.STASH → EVIDENCE_ONLY.

Tests

New/updated: ClockSelfSyncTest (sparse slots, tighter deadband, convergence), CallerQueueOrderingTest (pruning), CallerQueueLivenessTest (sequencer-level: queue fed on smooth QSOs, stale callers pruned, completion goes to CQ / works the live caller / strongest-first ignores a loud caller who left), FastPassDispositionTest, TimeCorrectionTest.showSuggestionApply.

Full unit suite: 3396 tests, 0 failures.

On-air check (please)

• Time Sync: with auto-sync on, the suggestion card should now read "Auto-sync is on…" and the DT residual should settle under ±0.15 s within a few RX slots, even with 1–2 stations heard.
• debug.log: look for QSO: TX restart … / TX message updated before audio start after a fast decode landed after key-up line, and dropped queued caller(s) no longer calling.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 8.47458% with 54 lines in your changes missing coverage. Please review.
✅ Project coverage is 41.65%. Comparing base (d24b408) to head (075dab8).

Files with missing lines	Patch %	Lines
...radio/ks3ckc/ft8af/ui/settings/TimeSyncSettings.kt	0.00%	33 Missing ⚠️
...tlin/radio/ks3ckc/ft8af/ui/components/ClockSync.kt	19.04%	17 Missing ⚠️
...n/radio/ks3ckc/ft8af/ui/components/SlotTimerBar.kt	0.00%	4 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##                dev     #758      +/-   ##
============================================
- Coverage     41.69%   41.65%   -0.05%     
- Complexity      284      290       +6     
============================================
  Files           278      278              
  Lines         32960    33166     +206     
  Branches       3940     3994      +54     
============================================
+ Hits          13742    13814      +72     
- Misses        18868    18996     +128     
- Partials        350      356       +6     

Flag	Coverage Δ
android	17.94% <8.47%> (+0.18%)	⬆️
native	9.93% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...n/radio/ks3ckc/ft8af/ui/settings/TimeCorrection.kt	100.00% <100.00%> (ø)
...n/radio/ks3ckc/ft8af/ui/components/SlotTimerBar.kt	21.81% <0.00%> (-1.26%)	⬇️
...tlin/radio/ks3ckc/ft8af/ui/components/ClockSync.kt	22.66% <19.04%> (-0.97%)	⬇️
...radio/ks3ckc/ft8af/ui/settings/TimeSyncSettings.kt	0.00% <0.00%> (ø)

... and 6 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[patrickrb]

Second commit: auto-sync is now on by default and acquires in one slot. A well-populated slot (≥4 decodes) whose median DT is >0.5 s out is corrected in full immediately; inside 0.5 s it tracks with two-slot confirmation and half-steps. Sparse slots (1–3 decodes) need three agreeing slots but take the full step when plainly off. The DT pill in the slot bar shows an AUTO badge (or GPS) so you can see it's being handled. Cold start on a busy band: 1.5 s off → in deadband after one slot. Full suite: 3410 tests, 0 failures.

[Copilot]

🔵 Needs a closer look

It changes timing-sensitive TX/decoder interactions and default clock-discipline behavior in core operating paths, which warrants final human review and on-air validation.

Pull request overview

This PR tightens the QSO timing loop by (1) making decode-based clock self-sync effective on sparse bands and enabled by default, (2) keeping the caller queue “live” by pruning stations that stopped calling, and (3) ensuring late decodes are acted on immediately as evidence so the sequencer can swap the on-air message within the audio slack.

Changes:

• Enable and surface decode-based auto clock sync by default, with a tighter deadband and acquisition/track behavior for sparse vs busy slots.
• Add caller-queue liveness tracking (lastHeardUtc) plus pruning of callers who gave up, and refresh the queue on every sequencer pass.
• Remove decode stashing and parse late decodes immediately as evidence-only to allow mid-cycle TX restarts/swaps.

File summaries

File	Description
ft8af/app/src/main/java/com/k1af/ft8af/timer/ClockSelfSync.java	Implements acquisition vs tracking, tighter deadband, and sparse-slot confirmation rules.
ft8af/app/src/test/java/com/k1af/ft8af/timer/ClockSelfSyncTest.java	Expands estimator unit tests for sparse slots, acquisition, tracking, and convergence.
ft8af/app/src/main/java/com/k1af/ft8af/GeneralVariables.java	Turns auto-sync on by default and adds runtime status fields for UI (“last step”, “last applied”).
ft8af/app/src/test/java/com/k1af/ft8af/database/DatabaseOprConfigHydrationTest.java	Updates hydration expectations to reflect the new default-on auto-sync behavior.
ft8af/app/src/main/kotlin/radio/ks3ckc/ft8af/ui/settings/TimeCorrection.kt	Adds showSuggestionApply() gate so manual apply is hidden when an automatic source owns the clock.
ft8af/app/src/test/kotlin/radio/ks3ckc/ft8af/ui/settings/TimeCorrectionTest.kt	Tests the manual-apply visibility logic for manual vs auto-sync vs GPS discipline.
ft8af/app/src/main/kotlin/radio/ks3ckc/ft8af/ui/settings/TimeSyncSettings.kt	Updates the Suggestion card to show auto-sync status and last auto-correction instead of a manual action.
ft8af/app/src/main/res/values/strings_compose.xml	Adds strings for auto-sync status messaging and clock-sync accessibility text/badges.
ft8af/app/src/main/kotlin/radio/ks3ckc/ft8af/ui/components/ClockSync.kt	Adds AUTO/GPS badge selection + accessibility string composition for the DT pill.
ft8af/app/src/test/kotlin/radio/ks3ckc/ft8af/ui/components/ClockSyncTest.kt	Tests AUTO/GPS badge selection behavior.
ft8af/app/src/main/kotlin/radio/ks3ckc/ft8af/ui/components/SlotTimerBar.kt	Wires auto-sync/GPS flags into the DT indicator so the badge reflects current clock ownership.
ft8af/app/src/main/java/com/k1af/ft8af/ft8transmit/QueuedCaller.java	Adds lastHeardUtc to track caller freshness for pruning.
ft8af/app/src/main/java/com/k1af/ft8af/ft8transmit/CallerQueueOrdering.java	Adds “given up” detection and in-place pruning helpers.
ft8af/app/src/test/java/com/k1af/ft8af/ft8transmit/CallerQueueOrderingTest.java	Tests given-up detection, pruning, FT4/FT8 scaling, and strongest-first behavior post-prune.
ft8af/app/src/main/java/com/k1af/ft8af/ft8transmit/FT8TransmitSignal.java	Refreshes caller queue every parse pass, prunes given-up callers, and supports pre-audio TX message updates on restart.
ft8af/app/src/test/java/com/k1af/ft8af/ft8transmit/CallerQueueLivenessTest.java	Adds sequencer-level Robolectric coverage for queue refresh + pruning across QSO completion/CQ.
ft8af/app/src/test/kotlin/radio/ks3ckc/ft8af/ui/components/CallerQueueDisplayTest.kt	Updates QueuedCaller construction to include lastHeardUtc.
ft8af/app/src/main/java/com/k1af/ft8af/ft8listener/FastPassDisposition.java	Changes fast-pass handling from PARSE/STASH to PARSE/EVIDENCE_ONLY.
ft8af/app/src/test/java/com/k1af/ft8af/ft8listener/FastPassDispositionTest.java	Updates tests to reflect “never dropped, never deferred” fast-pass behavior.
ft8af/app/src/main/java/com/k1af/ft8af/MainViewModel.java	Removes pending-decode stashing and parses late decodes immediately as evidence; posts auto-sync UI status.
ft8af/app/src/main/java/com/k1af/ft8af/PendingSequencerDecodes.java	Deleted: no longer needed after switching late decodes to immediate evidence-only parsing.
ft8af/app/src/test/java/com/k1af/ft8af/PendingSequencerDecodesTest.java	Deleted along with the stashing mechanism.

Review details

• Files reviewed: 22/22 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.