Record RS232 diagnostics validation and baseline RAK RAM failure

docs/test-results/index.md

@@ -11,6 +11,7 @@ Completed validation reports, experiments, and test guides are collected here. O
 
 ## Radio and RF
 
+- [RS-232 bridge diagnostics: host/build checks and existing RAK Full RAM failure, 2026-10-22](rs232_bridge_diagnostics_2026-10-22.md)
 - [ESP-NOW bridge diagnostics: V4 and V4 R8 TFT Full build/host validation, 2026-10-22](espnow_bridge_diagnostics_2026-10-22.md)
 - [V3/V4/T096 radio2 with two fixed witnesses, matched SF7/BW62.5 at preamble 32, 2026-10-22](radio2_simultaneous_receivers_2026-10-22.md)
 - [SolarXiao Room Server power profiles and history build/host validation, 2026-10-21](solarxiao_room_2026-10-21.md)

docs/test-results/rs232_bridge_diagnostics_2026-10-22.md

@@ -0,0 +1,88 @@
+<!-- meshcore-hosted-doc-link:start -->
+<p class="meshcore-hosted-doc-link"><a href="https://mikecarper.github.io/MeshCore/test-results/rs232_bridge_diagnostics_2026-10-22/">View this page on MeshCore Docs</a>.</p>
+<!-- meshcore-hosted-doc-link:end -->
+
+# RS-232 bridge diagnostics, 22 October 2026
+
+Status: **host tests and ESP32 Full build PASS**. The RAK4631 Reduced
+sensor/LoRa OTA build passes, but its Full sensor build fails the RAM gate.
+The same Full failure occurs on the previous source, before these diagnostics.
+The required nRF52 Full/Reduced pair remains incomplete and is not a qualified
+release. No radio was flashed or contacted in this work.
+
+The feature implementation is
+[`281f1a519`](https://github.com/mikecarper/MeshCore/commit/281f1a519),
+with additional host UART platform adapters in
+[`df08877c1`](https://github.com/mikecarper/MeshCore/commit/df08877c1).
+The ESP32 build used the implementation source with those test-only adapters
+already present locally. The nRF52 build used committed `df08877c1`.
+
+## Behavior
+
+The existing `get rs232.status` heartbeat/automatic-standby report remains.
+`get rs232.transport` adds UART state, baud, RX/TX pins, packet source,
+partial/pending frame state and historical errors. Separate RX, TX, drop and
+test counter pages distinguish framing/checksum failures, pool pressure,
+blocked/short writes, filters and duplicates. Statistics occupy 120 bytes
+inside the bridge object, with no new persistent preferences.
+
+`rs232 test [8 hex digits]` queues a reserved, one-way UART control frame.
+The receiver consumes it without allocating a mesh packet, updating duplicate
+tracking, establishing heartbeat liveness, replying or forwarding onto LoRa.
+Confirm receipt using the other updated board's increased probe `rx` count
+and matching `rxid`; local UART write completion does not prove delivery.
+
+Fault testing found that cancelling a partially emitted control frame could
+confuse an older receiver's following frame. The implementation cancels an
+entirely unsent probe at its deadline, but retains and drains an already-started
+frame's original suffix. It never retransmits the prefix or interleaves another
+frame. A timeout records one failure; a later local completion can also count
+as sent. Each loop's diagnostic work remains bounded.
+
+Both `usb.logging on` and `usb.debug on` enable bounded `RS232:` summaries in
+ordinary bridge images, at most once per second when diagnostic state changes.
+CLI pages and test probes require neither logging gate. See the
+[setup and diagnostic guide](../rs232_bridge_setup.md) for exact commands.
+
+## Embedded build checks
+
+Canonical builds used version `v1.18.0.1-halo-keymind-cascade-dev`, Cascade
+runtime defaults and the USA Cascade preset. These were compile/package
+checks, with no physical UART, LoRa, USB capture or OTA operation.
+
+| Target/profile | Available internal bytes | Required heap bytes | Result |
+| --- | ---: | ---: | --- |
+| `Heltec_v3_repeater`, selected Full MQTT/UART/ESP-NOW image | 209,472 | 203,920 | PASS |
+| `RAK_4631_repeater`, Full sensors + LoRa OTA | 82,324 | 85,152 | FAIL |
+| `RAK_4631_repeater`, Reduced sensors + LoRa OTA | 85,336 | 85,152 | PASS |
+| Previous `e7b863494`, same RAK Full recipe | 82,324 | 85,152 | FAIL |
+
+The ESP32 Full image passes its application/merged package and linked
+capability checks with no feature reductions. Its largest internal region is
+176,704 bytes against a 21,184-byte contiguous requirement. These are linked
+capacity estimates, not running free-heap or maximum-load measurements.
+
+The RAK Full shortfall is exactly 2,828 bytes on both sources. No memory guard
+or reserve was lowered. The new nRF52 bridge allocation is 2,656 bytes, within
+the unchanged 2,816-byte source assertion. Reduced is a separate supported
+profile; its pass does not qualify Full or replace the failed required pair.
+
+## Host validation
+
+- Eight diagnostic methods passed across five sanitizer builds: ESP32 ordinary
+  and native USB, nRF52, RP2040 and STM32. Cases include invalid control frames,
+  both legacy parsers, pool exhaustion, liveness isolation, every partial probe
+  prefix, timeout recovery, saturation, reset and bounded runtime logging.
+- Three CLI methods passed across seven sanitizer configurations, including
+  local-only actions, strict grammar, reply bounds, transport combinations and
+  unsupported roles. Existing ESP-NOW CLI checks also passed.
+- Bridge-only native USB formatter checks passed on nRF52 and ESP32 with
+  compile-time debug flags off. Existing heartbeat, UART lifecycle, independent
+  MQTT/UART runtime, power settings/ownership, filter wiring, bridge recovery,
+  ingress and ESP-NOW diagnostic regressions passed.
+- Allocation-budget boundary checks, hosted links and strict documentation
+  rendering passed.
+
+Physical cable/line-driver behavior, actual peer probe receipt, UART sleep
+rendezvous, USB debug capture and end-to-end LoRa delivery remain untested
+for this change.