Send BTP acks more frequently; large memory reduction in responder

[ivmarkov]

Subject says it all. Really two unrelated fixes, but since the memory fix in the respond module was a one-liner, I combined it with the BTP timeout fix.

EDIT 1: Also back-ported an mDNS fix from this: sysgrok/edge-net#104
EDIT 2: The memory fix in respond (a port of sysgrok/edge-net#105) might have far-reaching consequences, like completely getting rid of the bump allocator in downstream crates.

[Copilot]

🔵 Needs a closer look

The respond change relies on subtle unsafe Pin invariants for async futures that warrant a human sign-off despite appearing sound.

Pull request overview

This PR bundles two independent fixes for rs-matter. First, it corrects the BTP (Bluetooth Transport Protocol) stand-alone ACK cadence: the responder was scheduling ACKs using the 15s receive-side ACK deadline (BTP_CONN_IDLE_TIMEOUT_SECS / 2), which risks the peer tearing down the session during long-running responses (e.g. ConnectNetwork). It introduces a dedicated, smaller BTP_ACK_SEND_TIMEOUT_SECS (2s) for the send side. Second, it reduces the memory footprint of the respond module by pinning the handlers vector in place instead of via pin!, avoiding a duplicate copy of the (large) handlers array in the generated future.

Changes:

• Add BTP_ACK_SEND_TIMEOUT_SECS = 2 and use it as the BTP send-side ACK timeout everywhere BTP_ACK_TIMEOUT_SECS was previously used (the old constant is retained for documentation with #[allow(dead_code)]).
• Replace pin!(handlers) + map_unchecked_mut with Pin::new_unchecked(handlers.as_mut_slice()) to avoid storing the handlers twice in the future.
• Document both fixes in CHANGELOG.md.

File summaries

File	Description
rs-matter/src/transport/network/btp/session.rs	Adds the new BTP_ACK_SEND_TIMEOUT_SECS constant and explanatory docs; keeps the old constant for reference.
rs-matter/src/transport/network/btp.rs	Switches the default/reset ACK timeout and import to the new send-side constant.
rs-matter/src/respond.rs	Pins the handlers slice in place to halve the future's memory usage; imports Pin.
CHANGELOG.md	Documents the BTP ACK and respond memory fixes under a new [Unreleased] section.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 1
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

PR #554: Size comparison from b6a708f to 9176001

Full report (8 builds for (core), dimmable-light, onoff-light, onoff-light-bt, speaker)

platform	target	config	section	b6a708f	9176001	change	% change
(core)	riscv32imac-unknown-none-elf	infodefmt-optz-ltofat	FLASH	419706	419782	76	0.0
			RAM	72864	72864	0	0.0
	thumbv6m-none-eabi	infodefmt-optz-ltofat	FLASH	341280	341440	160	0.0
			RAM	68412	68412	0	0.0
	thumbv7em-none-eabi	infodefmt-optz-ltofat	FLASH	323048	323192	144	0.0
			RAM	68180	68180	0	0.0
	x86_64-unknown-linux-gnu	infologs-optz-ltofat	FLASH	843059	843387	328	0.0
			RAM	73810	73810	0	0.0
dimmable-light	x86_64-unknown-linux-gnu	infologs-optz-ltofat	FLASH	2057968	2058184	216	0.0
			RAM	5472	5472	0	0.0
onoff-light	x86_64-unknown-linux-gnu	infologs-optz-ltofat	FLASH	2017272	2017496	224	0.0
			RAM	5456	5456	0	0.0
onoff-light-bt	x86_64-unknown-linux-gnu	infologs-optz-ltofat	FLASH	3461456	3462320	864	0.0
			RAM	5740	5740	0	0.0
speaker	x86_64-unknown-linux-gnu	infologs-optz-ltofat	FLASH	2020440	2020640	200	0.0
			RAM	5456	5456	0	0.0