[reubenmiller]

The ESP32-C6 Modbus tedge-full image didn't offer c8y_Command. The shell-diagnostics extra that the S3 builds use costs 15.9 KB of RAM and 120 KB of flash on the C6, which puts the image at 82.4 % of its slot (budget 80 %).

What changes

• release/extras/shell-lean.conf: goes after shell-diagnostics.conf and trims the shell to what a cloud command uses. It drops history, completion, VT100, the shell log backend, the device and log shells, stack painting, and the network and Wi-Fi shells. The shell thread gets 1.5 KB, and the output buffers are cut to 256 B, because a result is cut at about 220 bytes anyway.
• tedge net (lib/common/net.c, CONFIG_APP_NET_SHELL_CMD, off by default): one short answer with Wi-Fi state, SSID, channel, RSSI, address/prefix, gateway, DNS server, and gateway ping count. It replaces net iface and wifi status, which print more than a result holds.
• Allow-list on the C6: kernel uptime, kernel version, kernel heap, tedge net, tedge params list, tedge diag, tedge identify. kernel heap shows the kernel heap the Wi-Fi driver and the tunnel allocate from.
• release/devices.yml: the C6 Modbus full build uses both extras and is now a pr: true build.
• Docs: DEVICES.md has the measurements, and README.md and cumulocity/README.md list the C6 commands.

Cost (against 0.11.1)

Image	Flash	RAM (sram0)	malloc arena
modbus tedge-full (0.11.1)	959,692 B	435,312 B	74,128 B
+ full shell extra	1,080,508 B (82.4 %)	+15,928 B	63,280 B
+ lean shell extra (this PR)	966,203 B (73.7 %)	+9,360 B	64,768 B

The largest single item is the command thread's 4 KB stack, tiny-edge's unmeasured default. The Wi-Fi driver allocates from the fixed kernel heap (about 55 KB), not the malloc arena the shell's RAM comes from. Other release builds are unchanged: the new option is off unless the lean extra sets it.

Testing

• scripts/release/build.sh modbus-server-tedge-full-esp32c6-devkitc --write-baseline passes the size checks, and the baseline is updated for that build only.
• scripts/release/matrix.py validates, and scripts/release/tests pass (43).
• On the board (eu-latest, ESP32-C6): the test image 0.11.1-shelltest.1 (same configuration and code) installed over the air and confirmed, and the device lists c8y_Command.
• Not yet done: running the commands from the Shell tab, a refused command, and a 30-minute run with a firmware download and a tunnel. That run is where the full extra failed on the C6 agent image in September. DEVICES.md notes it as pending.

🤖 Generated with Claude Code