FlexRadio overhaul: meters HUD + auto-discovery + one-tap connect (and QTH crash fix)#329
Open
patrickrb wants to merge 19 commits into
Open
FlexRadio overhaul: meters HUD + auto-discovery + one-tap connect (and QTH crash fix)#329patrickrb wants to merge 19 commits into
patrickrb wants to merge 19 commits into
Conversation
afterDecode() runs on the FT8 decode thread and submits a QTH-lookup task to getQTHThreadPool after every decode pass. That pool (and sendWaveDataThreadPool) is shut down in MainViewModel.onCleared() when the ViewModel is destroyed on app close or connection reset/reconnect. A decode result landing in the same instant the pool is shut down reaches ExecutorService.execute() on a terminated pool and throws RejectedExecutionException on the decode thread, crashing the app (observed on a FlexRadio network reconnect). Add ExecutorUtils.safeExecute(), which skips submission when the pool is shut down and swallows the residual RejectedExecutionException race, dropping the follow-up task instead of crashing. Route all three submit sites (QTH lookup + two network/CAT TX-audio sends) through it. Covered by ExecutorUtilsTest. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The FlexRadio network picker only refreshed its discovered-radio list via a one-shot FlexRadioFactory event listener that fired on the first add. Discovery typically takes a few seconds, so a radio that surfaced shortly after the dialog opened never appeared — the user had to close and reselect Network to get a fresh seed. Replace the listener with a 1s poll of FlexRadioFactory.flexRadios while the dialog is open, rebuilding the displayed list only when the discovered set actually changes (FlexPickerLogic.identityKey/listChanged) so the list doesn't recompose / reset scroll every tick. Covered by FlexPickerLogicTest. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A top-anchored HUD, opened by a swipe-down from the top edge anywhere in the app, showing the two meters every supported CAT rig reports back over the existing link: ALC and SWR. These are only measurable while keyed, so the readout is live during TX and labelled "LAST TX" otherwise (and "no data" until the rig has ever reported, distinguished from a legitimate 0 reading via a new meterDataReceived flag on MeterProtectionController). Reuses the existing meter plumbing: MetersSheet observes lastAlc/lastSwr and reuses normalizedSwrToRatio for the readout; the ALC target window and SWR halt threshold from GeneralVariables drive the green/amber/red zones so the HUD and the protection controller agree on what "good" means. Decision/geometry logic (bar fraction, ALC/SWR zones, freshness, edge-open commit rule) is extracted into pure functions in MetersDisplay.kt and unit tested; the top-sheet scaffold mirrors FT8AFBottomSheet (slide from top, drag-up/scrim/Back to dismiss) without touching the tested bottom sheet. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The HUD only read MeterProtectionController.lastAlc/lastSwr, which is fed
solely by the serial-CAT meter path (Yaesu RM4/RM6, Kenwood, Icom, serial
Xiegu). On a network rig the HUD sat blank: a FlexRadio streams its meters
over UDP into FlexConnector.mutableMeterList, and Xiegu network into
X6100Radio.mutableMeters — neither ever reached that controller.
Make the HUD source-agnostic and adaptive:
- rememberRigMeterSamples observes whichever stream the connected rig
produces — Flex (SWR ratio, ALC dB, power W, S-meter dBm, PA temp),
Xiegu (SWR, ALC, power, S-meter, voltage), or the serial controller
(ALC, SWR) — and returns the rig's available meters.
- Pure per-source converters (MetersDisplay.kt) normalize each rig's
native units into a common MeterSample (bar fraction + readout + zone),
so the HUD renders uniformly. Flex ALC is dB (low = good) vs the serial
0-255 window; SWR converts from ratio or the normalized scale.
- A configurable meter set: Settings → Meters toggles each meter (SWR +
ALC on by default, power/S-meter/voltage/temp off). The HUD shows the
intersection of enabled and rig-available meters ("adapt per rig"), so
an enabled meter the rig doesn't report is dropped, not shown empty.
All converters + the enabled/available filter are unit-tested.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Network rigs set TX power via flexMaxRfPower (default 10 W), pushed to the rig at connect — but the only screen that ever showed or changed it (FlexRadioInfoFragment's seekbar) is part of the legacy Java UI the Compose app no longer reaches, so there was no way to see or adjust it. Add an adjustable TX-power row at the top of the meters HUD, shown when the connected rig supports it (FlexRadio; Xiegu uses a different scale and is left for later). The label tracks the slider live; the value is pushed to the radio (FlexConnector.setMaxRfPower → RFPOWER) and persisted to config on release, not on every drag tick, to avoid spamming the network. Default the PWR (output watts) meter on, so set-power and live output read together on the rigs that report power; serial rigs don't report it, so "adapt per rig" still hides it there. The 0..100 W clamp is a pure, unit-tested helper. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Selecting Network + FlexRadio dropped the user on a manual "IP Address" field. Discovery code existed (FlexRadioFactory listens for the Flex UDP broadcast on 4992) and the picker even rendered found radios — but the list was always empty, so typing the IP was the only thing that worked. Root cause: Android's Wi-Fi chip filters out incoming broadcast/multicast packets unless an app holds a WifiManager.MulticastLock, and we never took one (nor held the CHANGE_WIFI_MULTICAST_STATE permission needed for it). So the discovery socket received nothing. - Add CHANGE_WIFI_MULTICAST_STATE and acquire a MulticastLock while the Flex picker is open (released on dismiss, so no battery cost otherwise). - Lead the picker with discovery: a "Searching…" state, the found radios as the primary list, and manual IP entry collapsed behind a link as a fallback (different subnet / VPN / broadcast blocked). - Auto-connect when discovery finds exactly one radio and the user hasn't started typing — connect to your Flex with zero input. The decision (shouldAutoConnectFlex) is a pure, unit-tested one-shot. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…session Field report: the picker "closes the instant you open it" and the rig stops connecting. Cause: FlexRadioFactory is a singleton, so once a radio is cached, reopening the picker seeds the list with it and the auto-connect LaunchedEffect fired immediately — dismissing the dialog and, worse, reconnecting an already-connected rig, which tears down the working session. Gate auto-connect so it only fires on a genuinely fresh discovery: - startedEmpty: the picker must have opened with no cached radios, so we never auto-connect to a stale singleton entry. - rigAlreadyConnected: never reconnect over a live session. (plus the existing single-radio / not-typing / one-shot guards.) Also add debug.log diagnostics (multicastLock.held, seeded count, each radio added, and connect path) so we can confirm whether discovery actually receives the UDP broadcasts on the user's network. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
"If my Flex settings are set, I should just click the CAT button near CQ and have it connect." Two gaps stopped that: the Flex address was never persisted, and reconnectRig() no-ops when no connector exists yet (a cold start) — so the CAT chip did nothing until you'd already connected once via the picker. - Persist the Flex address (flexLastIp) whenever we connect — discovered, tapped, or typed — and load it at startup. - Route the CAT chip tap (catTapAction): reconnect an existing connector if one's live; otherwise, on a cold start with a saved network Flex, connect straight to the remembered address; with nothing saved, point the user to setup. Pure routing (catTapAction) is unit-tested. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Field test on a real FLEX-6400 exposed two bugs behind "it never found the radio / didn't connect until I hit CAT": 1. Discovery worked (MulticastLock fine — debug.log showed the radio found in 0.4s) but the picker kept spinning. FlexRadioFactory fires OnFlexRadioAdded BEFORE appending to its list, so the listener re-read an empty list and dropped the radio (logged "total=0"). It also meant the discoveredRadios count stayed 0, so auto-connect never fired. Now we include the radio handed to the callback (mergeDiscovered, deduped by address) — so it shows and auto-connect can fire. 2. The Flex connect never reported success: FlexConnector.onConnectSuccess didn't notify the rig state, so onConnected() (which sets CONNECTED + toasts) never ran — the CAT chip stayed idle and there was no confirmation. Added an OnConnectionResult callback; connectFlexRadioRig now sets CONNECTING up front and CONNECTED + a "connected" toast on success (ERROR on failure). mergeDiscovered is pure + unit-tested. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The open gesture was an invisible 24dp strip at the very top — undiscoverable,
and it fought Android's notification-shade gesture, which owns the screen's top
edge, so the swipe often just opened the system shade ("can't get it to slide
down").
Replace it with a small visible "METERS ▾" tab centered at the top, below the
status bar (statusBarsPadding, so the system doesn't steal the gesture). It
opens the HUD on a tap or a short downward drag — whichever the user reaches
for. Verified on device: tab is visible and opens the sheet.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
… into feat/flex-meters-all
…into feat/flex-meters-all # Conflicts: # ft8af/app/src/main/kotlin/radio/ks3ckc/ft8af/ui/settings/ConnectionDialogs.kt
This was referenced Jun 23, 2026
…opulates Reported: the meters HUD shows the TX-power slider (so the Flex is connected) but "No meter data yet" — the meter rows never populate. Two causes: 1. The meter subscription is only sent when the METER_LIST *response* arrives at connect (FlexConnector's response handler) — an easily-missed path. The old legacy meter screen used to force it directly. Add FlexConnector.requestMeterStream() (unconditional meter list + sub) and call it from the HUD when it opens. 2. The Flex posts the SAME FlexMeterList instance on every update, which observeAsState dedupes via structural equality — so even once data flowed the bars wouldn't update. Poll the live (UDP-thread-mutated) instance ~4x/sec instead, and gate the empty state on a new FlexConnector.meterDataReceived flag (set on the first packet, also logged for diagnosis). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The meters appeared but didn't move. The 250ms poll tick was read *inside* the per-rig sampler @composable, so only that child recomposed — its fresh values never propagated to the bars rendered in MetersSheet's scope (a returned List doesn't update the caller unless the caller recomposes). Move the poll to MetersSheet: a produceState frame (250ms, only while the sheet is visible) is read in the render scope and passed into rememberRigMeterSamples, forcing it to re-run and re-read each rig's live, in-place-mutated meter values. The per-rig readers are now plain functions so their results propagate normally. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…d gate) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A FlexRadio (network) was transmitting 0 W and the operator had to listen to RX audio acoustically. Root cause: GeneralVariables.connectMode was left as USB_CABLE (from a previously-attached USB sound card, even after unplugging), and both the transmit path and onTransmitByWifi (the DAX sender) — plus the RX source selection — are gated on connectMode == NETWORK. So TX audio went to the stale USB output (radio transmitted silence) and RX fell back to the phone mic. Connecting to a Flex always means network/DAX audio, so connectFlexRadioRig now forces connectMode = NETWORK (persisted) and clears any stale USB-audio output override. TX then streams to the Flex over DAX (real power, meters move) and RX comes from DAX instead of the mic. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Meters intermittently showed "No meter data yet" on a connected Flex. The subscription (requestMeterStream) only fired once, when the rig connected — and if that raced the link coming up, it never retried, so meters never started. Key the subscription on the HUD's visibility (active) as well as the connector, so opening the meters HUD always (re)requests the stream. Restores a light debug.log diagnostic (requestMeterStream call + first packet) to confirm. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fixes the Flex meter mapping and connection handling, and lays the groundwork for DAX transmit. Meters (ALC/temp/S-meter), connection reliability, and slice ownership are working and verified on-device; DAX transmit now reaches the modulator but does not yet make full power (paused — see below). Meters / mapping (FlexMeterInfos, FlexConnector): - Resolve the SWR/ALC/PWR/S-meter/temp meter ids deterministically from the full meter table (lowest id, exact-name match), so duplicate per-slice/per-TX-block meters can't latch ALC/S-meter onto an inactive duplicate. Idempotent across meter-list re-requests. Connection reliability (FlexConnector, MainViewModel): - Run the client/gui + slice/stream setup exactly once per logical connection (slicesConfigured guard, re-armed in connect()); a re-fired onConnectSuccess no longer re-mints the client and orphans the slice. - reconnectRig() no-ops while CONNECTING/CONNECTED so repeated CAT-chip taps can't spawn parallel sessions. Guard connectFlexRadioRig against a duplicate connect to the same IP. Slice ownership (FlexRadio, FlexConnector, FlexNetworkRig): - Capture the radio-assigned slice index from the "slice create" response instead of hardcoding 0; configure THAT slice (DIGU/filter/DAX bind) and claim it as the transmit slice (slice s <n> tx=1). Band/frequency changes retune the owned slice. DAX transmit audio (FlexRadio, VITA): - Parse DAX TX stream id via long (was overflowing Integer.parseInt -> 0). - Tag TX audio with the DAX-audio class id (0x534C03E3), TSI_UTC, and the correct VITA packet_size, matching the radio's own RX audio format. - Send TX audio to the radio's VITA data port (4991). - Enable DAX transmit (dax tx 1 + transmit set dax=1). This gets audio to the modulator (ALC moves off the floor), but output power is still low — likely the SmartSDR DAX-TX gain (a GUI/profile setting with no API command). Transmit is paused here pending a packet capture vs a working DAX app. Test: FlexResponseSliceTest covers the slice-index parse. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Single PR consolidating three open PRs as requested — supersedes and closes #327, #325, #324.
1. Meters HUD (pull-down)
2. FlexRadio network connection
CHANGE_WIFI_MULTICAST_STATE+ aMulticastLockheld while the picker is open (Android was dropping the discovery broadcasts without it).FlexPickerLogic) to avoid scroll resets.3. QTH executor crash fix (from #324)
ExecutorUtils+MainViewModel). Unrelated to the Flex/meters work — bundled here at the author's request.Testing
clampTxPowerWatts,shouldAutoConnectFlex,FlexPickerLogic,catTapAction,ExecutorUtils. All green.🤖 Generated with Claude Code