Found while choosing a board for the instrument spike (schematic survey, 2026-09-04).
cmods/micropython/ports/esp32/machine_i2s.c:378 hardcodes .mclk = I2S_GPIO_UNUSED with no path that substitutes a user pin, so the mck= argument extmod/machine_i2s.c accepts is accepted and discarded on every esp32-family target, S3 and P4 alike. No error, no warning.
Four of our board configs pass one as though it worked:
pydevices/board_configs/busdisplay/spi/t-embed/board_peripherals.py:84pydevices/board_configs/busdisplay/spi/m5stack-cores3/board_peripherals.py:75pydevices/board_configs/fbdisplay/m5stack_tab5_st7123/board_peripherals.py:62,104pydevices/board_configs/fbdisplay/m5stack_tab5_ili9881c/board_peripherals.py:62,104
We already know the workaround and wrote it down once: board_configs/fbdisplay/esp32-p4-wifi6-touch-lcd-4b/board_peripherals.py:20 says "Firmware has no I2S mck= — PWM supplies MCLK". That note never propagated to the other four.
Anything with an ES8311, ES7210 or AW88298 needs a master clock from somewhere, so these configs are either silently working via some other path or quietly wrong — a bench check on a CoreS3 or Tab5 decides which. Parts that self-clock (MAX98357A, UDA1334A, PCM5102A, NS4168) are unaffected, which is why the spike shopping list favours them.
Two candidate fixes, not exclusive: drop the dead argument and carry the PWM-MCLK trick in the configs that need it; and/or make the port honour mck= and send that upstream, since silently discarding a documented argument is an upstream bug in its own right.
Found while choosing a board for the instrument spike (schematic survey, 2026-09-04).
cmods/micropython/ports/esp32/machine_i2s.c:378hardcodes.mclk = I2S_GPIO_UNUSEDwith no path that substitutes a user pin, so themck=argumentextmod/machine_i2s.caccepts is accepted and discarded on every esp32-family target, S3 and P4 alike. No error, no warning.Four of our board configs pass one as though it worked:
pydevices/board_configs/busdisplay/spi/t-embed/board_peripherals.py:84pydevices/board_configs/busdisplay/spi/m5stack-cores3/board_peripherals.py:75pydevices/board_configs/fbdisplay/m5stack_tab5_st7123/board_peripherals.py:62,104pydevices/board_configs/fbdisplay/m5stack_tab5_ili9881c/board_peripherals.py:62,104We already know the workaround and wrote it down once:
board_configs/fbdisplay/esp32-p4-wifi6-touch-lcd-4b/board_peripherals.py:20says "Firmware has no I2S mck= — PWM supplies MCLK". That note never propagated to the other four.Anything with an ES8311, ES7210 or AW88298 needs a master clock from somewhere, so these configs are either silently working via some other path or quietly wrong — a bench check on a CoreS3 or Tab5 decides which. Parts that self-clock (MAX98357A, UDA1334A, PCM5102A, NS4168) are unaffected, which is why the spike shopping list favours them.
Two candidate fixes, not exclusive: drop the dead argument and carry the PWM-MCLK trick in the configs that need it; and/or make the port honour
mck=and send that upstream, since silently discarding a documented argument is an upstream bug in its own right.