Thanks for this, @jensenpat — this is a substantial, well-organized feature. The reusable MapView/MapMarkerItem/MapPathItem split is clean, the OSM tile-policy compliance (app User-Agent, disk cache honoring HTTP headers, viewport-only fetch, mandatory attribution, patching out the fake-MSIE UA + restored TLS verification) is genuinely careful, and the MainWindow/menu wiring mirrors the existing Ax25HfPacketDecodeDialog pattern (QPointer member, m_persistentDialogs, frameless honoring) exactly. CI is fully green across macOS/Windows/Linux + CodeQL. A few notes, none blocking:

1. New flat AppSettings keys — Constitution Principle V. The dialog adds two new flat keys, PskReporterShowPaths and PskReporterUpdateIntervalMs (kIntervalKey). Principle V asks new feature config to live as a single nested-JSON blob under one root key rather than new flat keys (existing flat keys can stay). Suggest consolidating both into one object under e.g. PskReporterMap ({ "showPaths": ..., "updateIntervalMs": ... }). Minor, but it's an explicit project convention and easy to do now while the surface is small.

2. Nice latent-bug fix in the MQTT TLS change. Worth calling out that the old const char* ca = caFile.isEmpty() ? nullptr : caFile.toUtf8().constData(); was also a use-after-free — the toUtf8() temporary died at the end of that statement, so ca dangled before mosquitto_tls_set() read it. The new code passes caFile.toUtf8().constData() directly as a call argument, so the temporary lives through the call. The MOSQ_OPT_TLS_USE_OS_CERTS path is the right fix for the all-NULL MOSQ_ERR_INVAL rejection. 👍

3. Per-message marker rebuild in live mode (perf). In MQTT mode every incoming spot runs appendSpot + pruneOldSpots (an O(n) remove_if scan) and emits spotsUpdated(), which drives rebuildMarkers() → setMarkers() → full teardown/recreate of every marker and path QGraphics item (up to kMaxSpots = 2000) plus a full great-circle path recompute. Own-callsign spot rate is usually low so this is fine in practice, but on a busy/contest channel the bursts could churn the scene hard. Consider a short coalescing/debounce timer on spotsUpdated → rebuildMarkers so a burst collapses to one rebuild. Follow-up-able, not a merge blocker.

4. Hardcoded appcontact. poll() sends appcontact=ki6bcj@aethersdr.com to the PSK Reporter query API. Just confirm that's the intended maintained project contact (vs. a personal/testing leftover) — it's the address PSK Reporter ops would use to reach the app author.

Overall this is solid work and reads like the surrounding codebase. The Principle V keys (#1) are the only thing I'd nudge before merge; the rest are optional follow-ups.

_{🤖 aethersdr-agent · cost: $7.2874 · model: claude-opus-4-8}

Reviewed the ~1,900 lines of AetherSDR code (skimmed the vendored QGeoView only for the patch claims). This is a clean, well-structured feature — the MapView widget has a genuinely reusable shape, ownership/RAII is careful throughout (QPointer dialog members, manual removeItem+delete for the non-QObject QGV items, QSaveFile for atomic cache writes), and the system-boundary error handling is solid: reply-error checks, XML parse-error logging, the m_fetchInFlight guard, and the hard 5-minute poll floor. All six CI checks are green. Thanks for the thorough PR writeup and the OSM-policy diligence.

A couple of nice incidental wins worth calling out: the MqttClient TLS fix also removes a latent dangling-pointer bug — the old const char* ca = caFile.toUtf8().constData() pointed into a temporary QByteArray that was destroyed before mosquitto_tls_set() ran; passing .toUtf8().constData() directly as an argument now keeps the temporary alive for the call. Good catch on MOSQ_OPT_TLS_USE_OS_CERTS.

A few findings, none blocking:

1. New flat AppSettings keys (Principle V). PskReporterMapDialog adds two new top-level flat keys — PskReporterUpdateIntervalMs and PskReporterShowPaths. Constitution Principle V wants new settings grouped as a nested-JSON blob under a single key (cf. AudioEngine's AetherialTubePreampTx and HidEncoderManager's RC28Mapping). Suggest collapsing these into one PskReporter key holding { updateIntervalMs, showPaths }. The sibling feature PRs (#3527, #3530) explicitly carry the "no new flat-key AppSettings" checklist item, so it'll likely be flagged otherwise.

2. Map can stay permanently idle if opened before a callsign is known. showEvent() calls restartClient() once (guarded by m_started). If no radio is connected at that point, PskReporterClient::start() hits the empty-callsign early-return and never runs. There's no callsignChanged signal on RadioModel, and the only later hook — gpsStatusChanged → updateHomeFromRadio() — updates the home marker but never restarts the client. So opening the dialog before connecting (then connecting) leaves the map fetching nothing until the user reopens it or nudges the interval selector. Consider re-running restartClient() when the callsign first becomes non-empty (e.g. inside the gpsStatusChanged handler, or have RadioModel emit on callsign change).

3. Minor i18n. The new viewMenu->addAction("PSK Reporter...") string isn't wrapped in tr(), unlike the adjacent tr("FreeDV Reporter..."). Worth wrapping for consistency.

One observation, not a request: appendSpot() dedupes by receiverCallsign alone, so a station that hears you on multiple bands collapses to its latest spot — meaning the band/mode filters can hide a station that genuinely heard you on the filtered band. That matches the documented "one marker per reporting station" intent, so just noting it in case multi-band-per-station is desirable later.

Nice work overall — happy to see the reusable widget land ahead of the APRS tab reuse.

_{🤖 aethersdr-agent · cost: $8.6710 · model: claude-opus-4-8}