CHG-0008: The Stand Reports Its Colors


Date	2026-10-11
Change type	Upgrade
Status	Complete 2026-10-11
Systems	`lp-stand-01`, over USB from the development Mac
How	`make -C firmware/lp-stand build flash PORT=/dev/cu.usbserial-0001`
Risk	Low. Most likely to go wrong: a flash that also erased the stand’s credentials, which would leave it off the network until provisioned again
Decisions	ADR-0005
Related	CHG-0001

Summary

When the stand applies a scene it reports the effect, how many colors and the brightness. It does not say which colors. The firmware that does has been on main since it was written for a dashboard panel that was not kept, and has never been on the stand.

After this change scene.applied also carries the speed and the palette, as #c90027 65%, #ac0030 13%, the form lightd logs it in. The log store then holds what each stand was sent, in the stand’s own words.

Goal

The stand runs firmware built from this commit.
It rejoins the network and the broker without being provisioned again.
Its next scene.applied event carries speed and palette.
It still applies scenes.

Scope

In scope: the application image on lp-stand-01.

Out of scope: its credentials, which are in their own partition and are not written by a flash; the effects; the LEDs, which are still not attached.

Risk and impact

Risk	Where	Guard
The flash erases the credentials partition	the stand, off the network	`idf.py flash` writes the bootloader, the partition table and the application, each at its own offset; `creds` is none of them. If it happens, `make provision` puts them back
The larger event does not fit	every scene would go unreported	host tests hold the largest scene, eight colors, to fitting in one event
The stand is dark while it restarts	the room	it has no LEDs attached

Procedure

. ~/.espressif/tools/activate_idf_v6.1.sh
make -C firmware/lp-stand test
make -C firmware/lp-stand build flash PORT=/dev/cu.usbserial-0001

Verification

make -C firmware/lp-stand logs shows system.boot, then the network and broker events, then scene.applied with a palette.
A new album in iTunes produces another, with that cover’s colors.

Undo

Flash the commit before the palette was added. The credentials are untouched either way.

Outcome

When (UTC)	What happened
19:06	Host tests passed. Built and flashed over USB. The flash wrote the OTA data, the application and one small block before it; the `creds` partition was not among them.
19:06:27	The stand started (`reason=power`), joined `DVNTM-IOT` on the first attempt and reached the broker 3.1 s after boot, with nothing provisioned again.
19:06:27	It applied the retained scene and reported it: `effect=breathe colors=3 brightness=0.8 speed=0.4 palette=#2a2a81 80%, #ec1800 16%, #f0da85 4%`.

All of the goal’s checks hold. The colors are the ones lightd sent for the album loaded in iTunes at the time.

Departures from the plan

The firmware names itself c9e3286-dirty. It was built with this record written and not yet committed, so the tree was dirty by one documentation file. The code is the commit’s.
The same deployment window put a playing-time counter on the console (below), which is not this change and shares its pull request.

Follow-ups

The strip is still not attached.
breathe lights the dominant color alone. With the palette now in each event, the other colors are visibly unused by the default effect.