[devin-ai-integration]

Related Issue(s)	nasa#6166
Has Unit Tests (y/n)	y
Documentation Included (y/n)	y
Generative AI was used in this contribution (y/n)	AI

---

Change Description

GenericHub::send_data no longer asserts when allocate returns an invalid buffer (null data, or smaller than the framed message). It now:

• drops the message (an undersized non-null buffer is returned via deallocate; bufferIn buffers still go back on bufferInReturn);
• increments std::atomic<U32> m_numAllocationFailures on every drop, published as telemetry NumAllocationFailures from a new sync input port schedIn: Svc.Sched;
• reports once per failure episode: the drop that wins std::atomic<bool> m_reportingFailure.exchange(true) emits event BufferAllocationFailed(msgType: HubType, port, size) (WARNING_HI) and one Fw::Logger line; later drops are only counted until the next successful send clears the flag;
• for a dropped cmdDispIn command, replies EXECUTION_ERROR on the matching cmdRespOut so the originator is not left waiting.

The hub message type is now an FPP enum declared in the component (enum HubType : U32 { HUB_TYPE_PORT, HUB_TYPE_BUFFER, HUB_TYPE_EVENT, HUB_TYPE_CHANNEL, HUB_TYPE_CMD_DISP, HUB_TYPE_CMD_RESP }, same values and wire format as before) so the event argument is typed in the dictionary. API note for projects: the C++ Svc::GenericHub::HubType enum is replaced by the generated Svc::GenericHub_HubType (HUB_TYPE_MAX becomes GenericHub_HubType::NUM_CONSTANTS).

New special ports Log/LogText/Time/Tlm. Telemetry is written only from schedIn (not on the failure path) because the hub is often its own event/telemetry sink and emitting synchronously would re-enter it while the pool is exhausted; existing deployments must wire schedIn to a rate group to see the counter. The Fw::Logger line is for the side of the link whose own event cannot get out through the exhausted hub.

Documentation: the GenericHub SDD gains a "Transport formations" section that makes hub.toBufferDriver -> Svc.ComQueue -> Svc.ComStub -> driver the recommended hub wiring (bounded buffering across a link outage, outside the passive hub), with the transport requirements (message boundaries: UDP natively, TCP/UART through F´ framing plus a receive-side adapter the framework does not provide; UDP send errors stall the queued wiring until reconnect) and the sizing rule (D >= N, Q >= N + 8). hub-pattern.md presents the same wiring; the ComQueue SDD notes the hub as a bufferQueueIn producer. The design behind this (hub stays passive, no in-hub queue, Svc.ComRetry excluded) was settled with the maintainers before implementation.

Rationale

CCB on nasa#6166: "don't assert, add event and counter". A pool exhaustion killed the FSW; now it is counted, evented and survivable, and the documented topology gives projects buffering across link outages without adding state to the hub.

Testing/Review Recommendations

• Svc/GenericHub UTs: 14 pass (null / undersized allocation for every message type, deallocate of undersized buffer, one event + one logger line per episode re-armed by a successful send, counter via schedIn, recovery, cmdDispIn fail-back incl. malformed packet and a CmdSplitter-consumed buffer, hub as its own event sink with an empty pool — bounded re-entry — and with the pool recovering, buffer returned from another thread after the send). Full fprime UT suite 135/135 on current devel.
• fprime-generic-hub-reference rebuilt against this branch with both deployments on the recommended ComQueue + ComStub wiring: integration suite 8/8 (round trips, remote command, NumAllocationFailures == 0, ComQueue depth telemetry, no BufferAllocationFailed/QueueOverflow). Companion PR: JPL-Devin/fprime-generic-hub-reference (linked below once open).
• Review focus: send_data failure branch and the cmdDispIn_handler fail-back; the HubType enum move (same values, same framing); SDD "Allocation failure" and "Transport formations".

Future Work

None planned: the queued hub topic was closed in favour of the documented topology; UDP recovery after a send error and a framing adapter for stream transports are documented limitations.

Contributor Checklist

• I have the rights to contribute this change set under the Apache-2.0 license.
• I agree to be bound by the contributor guidelines.

AI Usage (see policy)

Implementation, FPP, unit tests and documentation were written by an AI agent following the component-development skills, under maintainer direction (requirements, once-per-episode reporting, std::atomic, Fw::Logger, sched-port telemetry, cmdDispIn fail-back, the FPP enum typing and the recommended-wiring decision were maintainer decisions). Validated with the GenericHub UTs, the full UT suite, and the fprime-generic-hub-reference GDS integration tests; reviewed locally with the F´ multi-agent review pipeline.

• This contribution follows the F´ AI development skills (entered through AGENTS.md).

IAMAI

Requested by: @thomas-bc-autobot

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access that start with 'Devin'.

• Disable automatic comment, CI, and merge conflict monitoring

---

Original prompt from thomas.boyer.chammard

@playbook:playbook-c60a4c0b400a4b3fa08f897a831c74d6 nasa#6166