DRAFT: Use automatic light sleep for ESP32 alarm sleep

[EternityForest]

Follows up on the conversation in #9463 and in Discord.

Draft PR, should be working but not tested in every possible way, posting for discussion purposes.

What this patch does:

• On ports without native USB, don't consider the console connected once there have been no new bytes for a while
• Configure UART wakeup(Note: the initial few bytes will always be lost with the hardware UART in sleep)
• Enables tickless idle but does not actually use it in normal operation
• When the user does a sleep until alarms, it allows the ESP-IDF to manage sleep/wake/CPU speed for that period

[tannewt]

Thanks for the initial PR!

Any idea of what functionality will need to disable this light sleep and can you add the "disable" calls for that? Adding it here will make it easier when we find other things impacted by this change.

[tannewt]

Please breakout the 60 * 1024 * 5 into a macro that is documented with what it means.

[EternityForest]

I'm not exactly sure what would need to disable it, but I think it should be handled already by the ESP-IDF's wake locks API, the ESP won't sleep if one of those is held

…

On Tue, Aug 27, 2024, 2:56 PM Scott Shawcroft ***@***.***> wrote: ***@***.**** requested changes on this pull request. Thanks for the initial PR! Any idea of what functionality will need to disable this light sleep and can you add the "disable" calls for that? Adding it here will make it easier when we find other things impacted by this change. ------------------------------ In supervisor/shared/serial.c <#9559 (comment)> : > @@ -175,7 +180,9 @@ bool serial_connected(void) { #endif #if CIRCUITPY_CONSOLE_UART - return true; + if (_console_uart_rx_timestamp && (_console_uart_rx_timestamp + (60 * 1024 * 5) > port_get_raw_ticks(NULL))) { Please breakout the 60 * 1024 * 5 into a macro that is documented with what it means. — Reply to this email directly, view it on GitHub <#9559 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFZCH55QQ73PD2FN6GWX3LZTTRXNAVCNFSM6AAAAABNA32CRSVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDENRUGQZTMNJUGU> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[tannewt]

I'm not exactly sure what would need to disable it, but I think it should be handled already by the ESP-IDF's wake locks API, the ESP won't sleep if one of those is held

Ah ok. So we rely on the IDF to prevent sleep if we have things going on? Sounds good to me.