[lxsaah]

Description

PR 3 of 3 into feat/352-log-defmt for #352. Since PR 2 (#359), the log_*! macros forward to defmt. A hand-written #[cfg(feature = "defmt")] defmt::…! next to a log_*! line therefore logs the same event twice on the device.

Removed: the 7 real pairs (defmt mirror right after a log_*! call, same event and arguments):

File	Event
aimdb-core/src/router.rs (3)	ingest failed; route matched but not produced; no route
aimdb-core/src/session/client.rs (1)	mirror gap (updates lost)
aimdb-mqtt-connector/src/embedded/mod.rs (2)	inbound/outbound route asks for qos=2, capped at 1
aimdb-mqtt-connector/src/embedded/session_loop.rs (1)	oversized publish skipped

Some of the mirrors had shortened the text. The device now shows the full log_*! text, which costs no flash at runtime because defmt interns its strings.

Kept on purpose: the standalone defmt calls with no log_*! counterpart, which are the only output for their events: the MQTT connector's embedded backend (connect, TLS, SNTP, session errors, …) and the Embassy adapter (runtime logger, slot error). #352 estimated "30 mirrors", but that counted every direct defmt call; only these 7 are duplicates. Turning the standalone calls into log_*! (so they'd also show up in host builds of the embedded backend) would be a separate change.

Live on the NUCLEO-H563ZI (feature branch with PR 1–3; temporarily with a TCP server on port 0 and a record without a free slot). Everything that used to be invisible shows up in the device's defmt log:

ERROR accept failed: Io; retrying, further failures are summarized
WARN  accept still failing: 10 attempts in a row, last: Io
WARN  accept still failing: 100 attempts in a row, last: Io
WARN  accept still failing: 10000 attempts in a row, last: Io
WARN  accept still failing: 100000 attempts in a row, last: Io
INFO  new handshake on a live connection: ending 1 subscription(s) of the previous client
ERROR AimDB: SpmcRing subscriber slot exhausted (max SUBS=2) …
WARN  subscription 2 ended: its source closed

The summary for 1,000 attempts doesn't appear. The unit test from #351 covers that logic; the likely cause is defmt-rtt dropping lines when its buffer fills during the startup burst. Afterwards the board was flashed back to the clean demo.

Verified locally: all 58 Makefile commands for aimdb-core and aimdb-mqtt-connector (tests, clippy, every thumbv7em check/clippy, including defmt and TLS).

Afterwards: changelogs on the feature branch, then the PR to main, the first full CI run over #352.

Related Issue

• Part of [FEAT] Core log macros reach defmt: forward log_*! to defmt and drop the hand-written mirrors #352

🤖 Generated with Claude Code

https://claude.ai/code/session_01YPDjQ947yKjLLAq3Xzgu3e