No MAC address in boot_out.txt

[Jerzeek]

CircuitPython version and board name

We made a custom .uf2 file for our board (https://circuitpython.org/board/studiolab_picoexpander/) but when we upload this .uf2 file to the board the MAC address in the boot_out.txt prints 00:00:00:00:00:00, when I upload the Pico 2W .uf2 file to the board, it does print the right MAC address.

Code/REPL

print("hello world")

Behavior

hello world

Description

No response

Additional information

No response

[dhalbert]

Please point us to the repo code for your board, so we can compare the board-specific files.

[Jerzeek]

Hi Dan,
Thanks for the quick reply, here are the board specific files:
https://github.com/adafruit/circuitpython/tree/main/ports/raspberrypi/boards/studiolab_picoexpander
It uses a complete PICO 2W.

The PCB files are published here: https://github.com/id-studiolab/PicoExpander

[dhalbert]

If I comment out

//#define MICROPY_HW_NEOPIXEL (&pin_GPIO2)

in mpconfigboard.h, then it works (!). Investigating further.

[Jerzeek]

Fascinating,
I did ask Gemini, and it also suggested this fix, and for my it was so wild that I didn't even try it, i thought it was hallucinating.
When I run code.py to print the MAC address it works fine however.
Interesting behaviour.

[Jerzeek]

Let me know if you want me to test anything on the device.

[dhalbert]

I did ask Gemini, and it also suggested this fix, and for my it was so wild that I didn't even try it, i thought it was hallucinating.

Well, I did it without AI, but as something that varied between the working and non-working boards. That is my usual debugging strategy in cases like this: see what's different, and then try undoing the differences. Did Gemini say why?

Thinking about this further: using the NeoPixel causes some delays, and it also uses PIO to send the data to the NeoPixel. If I remember right, the communication with the CYW43 is also via partly via PIO, to simulate an SPI peripheral, due to the pins that were chosen that did not allow use of a regular PIO peripheral. There may not be proper mediation between these uses.

[Jerzeek]

Makes sense,
This was the response:

I will modify circuitpython/ports/raspberrypi/boards/studiolab_picoexpander/mpconfigboard.h to comment out the MICROPY_HW_NEOPIXEL definition, as this is the most likely cause of the WiFi driver failure due to potential PIO resource conflicts. I will leave the SPI definitions untouched for now.

So it seems like a PIO problem. I guess there is no direct solution for this, we will have the students grab the MAC address from code.
Thanks anyways for the deep dive.

[dhalbert]

I think it's fixable -- if we cannot manage the access to the PIO we may just need to delay using the NeoPixel until after the CYW43 is set up and the initial data (like MAC address) is read from it. Or maybe we can just set up one of them on a different PIO.