ADR-0036: The Cover Screen Sleeps When Nothing Has Played for Five Minutes, and Wakes for Music

Context

The cover screen is lit all the time. With nothing playing it shows a drawn record all night, to nobody. Its owner asked first for it to sleep when the turntable’s power, iTunes and Spotify are all gone, and to come back when one of them is.

The screen knows no source, by decision (ADR-0020, ADR-0025), and its broker account reads two topics. nowplayd, though, already decides who is there. Its arbiter picks only among sources that are online and not switched off:

• the turntable’s driver is online whenever its sound card is, but while its beacon says the outlet is off it is not heard (ADR-0023);
• iTunes is online while its driver holds a session with the player;
• Spotify is online while Spotify answers its driver.

When none is left, current is published as stopped with no source. With one, source is always named.

The first build slept on exactly that, and did, on the board. Then iTunes was played and paused, and the screen stayed lit: a paused iTunes is still there, and after fifteen seconds current says stopped with iTunes still named, so that play can be pressed (ADR-0024). Its owner expected it to sleep, and widened the rule: nothing playing, not only nobody there.

The board’s backlight is a plain pin, and the ILI9341 has a sleep command that keeps what it was showing.

Options considered

The screen reads the presence topics itself: the beacon’s and each driver’s status. It teaches the firmware the sources’ names, widens its broker account, and puts in the screen a second copy of the arbiter’s rule about switched-off turntables. And it would not see a pause.

nowplayd says it outright, with a new field in current or a topic of its own. Clearer to read, and a contract change to say what current already says in state and source.

Nobody there: current names no source. Shipped first, and too narrow: a paused player keeps it lit.

Nothing playing: current does not say playing, or names no source. Both are in the one message the screen already reads.

Decision

The cover screen turns its panel and backlight off after five minutes in which current has not said playing with a source named, and turns them on again when it does.

• Five minutes, counted from the last thing that kept it awake: the music stopping or pausing, the source going, a touch, or the broker coming back. Paused, then stopped by nowplayd, then nobody there, is one quiet, not three: none starts the wait over. A side turned over on the turntable is well inside it.
• Playing wakes it, and so does a source named where there was none: the turntable switched on in a room where nothing was there.
• A touch wakes it, and only wakes it. The cover, or the record, comes back without the buttons, and the finger has to lift before the next touch means anything, as with the first touch on a lit screen. The next touch raises the buttons as before, and play is pressed from there.
• No network or no broker wakes it, and keeps it awake, with its corner marked as before. A dark screen has to mean asleep, not broken; a screen that cannot hear current cannot know that nothing is playing.
• Until it has heard current, it is awake. A screen just switched on shows that it works.
• What arrives while it sleeps is kept and drawn on waking, the picture before the light, as at power-on.
• Asleep is not off. WiFi and the broker stay up, which is how it hears that it should wake, and how an update reaches it (ADR-0035).
• The rule is pure C beside what a touch means (app/ui/ui.c), host tested; the board’s half is two calls into the panel.

Consequences

• The screen still knows no source, and nothing on the bus changed. It reads state, which it already did, and whether source is there.
• A pause is five minutes of light, then dark. Play pressed on the Mac or the phone wakes it; on the screen, a touch and then play.
• A turntable switched on in a room where iTunes is paused does not wake it: the source changes and nothing plays. The record starting does.
• Spotify does not keep it awake unless something plays, though its driver is online whenever Spotify answers. It is off (CHG-0024).
• A screen with older firmware stays lit, as before.
• The stand is not this. Its dark and dim are lightd’s, from the look (ADR-0028); the screen’s sleep is its own.

What would settle it

Tested on the host: it sleeps five minutes after the music stops, pauses or the source goes, and not before, and none of those after another starts the wait over; playing, a source named, a touch, or losing the broker wakes it; a touch that wakes it neither raises the buttons nor sends anything.

On the board, the first build (CHG-0026):

• With iTunes unreachable and the turntable off: screen.asleep five minutes after it connected, still on the broker; dark to the eye.
• A song started in iTunes: screen.awake, the cover shown.

On the build with the wider rule:

• iTunes paused: screen.asleep five minutes after the screen heard current, iTunes still named.
• A touch on the dark screen lit it without buttons; the next touch raised them and play and pause worked; dark again five minutes after the last touch.
• Play pressed away from the screen, with it dark: screen.awake and the cover, with no command.sent from the screen before it.