chore(release): prepare v6.2.0

CHANGELOG.md

@@ -6,6 +6,28 @@ The revived semantic-versioning baseline starts at `v5.0.0`.
 Historical releases that predate this file remain available in Git tags such as
 `v1.0.0`, `v3.0.0`, `v4.0.0-beta`, and `h10`.
 
+## v6.2.0 - 2026-07-10
+
+### Features
+
+- Add variable-length TCP orders, diagnostic forwarding, and flush throttling
+  - `TcpOrderStreamParser` now supports variable-length orders: when an order's `get_size()` returns 0, all remaining stream data is passed to `parse()`, and the post-parse size is used for byte consumption. A `reset_for_receive()` hook is called after consumption so orders can reset state for the next packet.
+  - `Order` gains a virtual `reset_for_receive()` no-op that subclasses can override.
+  - `Socket` constructor no longer panics on `ERR_RTE` during `tcp_connect()` — sets `pending_connection_reset` instead, letting the connection retry when the route becomes available.
+  - `OrderProtocolDiagnosticSink` now accepts a target `OrderProtocol*` via its constructor, so diagnostic records are sent to a single socket instead of broadcasting to all open sockets.
+  - `DiagnosticTransportOrder` now parses incoming diagnostic orders (IDs 1555, 2555, 3000) and re-publishes them via `Hub::publish_runtime_*()`, enabling forwarding of remote board diagnostics to the control station.
+  - `Diagnostics::Hub::flush_pending` throttles records to at most one per 100 ms per priority level (urgent and non-urgent tracked separately) to prevent flooding sinks.
+  - Default TCP keepalive values tightened: 30 ms idle, 20 ms interval, 2 probes (commented out for now since it can't keep connection with the backend).
+  - `lwipopts.h`: `MEMP_NUM_SYS_TIMEOUT` increased by 10 to avoid pool exhaustion with SNTP and whatever else it may happen, `TCP_TMR_INTERVAL` reduced to 10 ms for faster keepalive response.
+  - Added `Diagnostics::install_ethernet_sink(OrderProtocol*)` to register the Ethernet sink with a specific target socket.
+- Makes the watchdog reset check block the system and have a visual indication similar to hardfault
+
+### Fixes
+
+- Removes the default SNTP server from Ethernet configuration, defaulting to nullptr instead
+- Fix a warning about symbols in HardfaultTrace (doesn't change any behaviour)
+- Adds a check to ensure return values from Scheduler::get_global_tick() are monotonic (never get decreasing values)
+
 ## v6.1.3 - 2026-06-02
 
 ### Fixes

VERSION

@@ -1 +1 @@
-6.1.3
+6.2.0