Two Serial ports show up for pico in offline mode along with other usb weirdness (Windows11 offline beta2)

[tyeth]

Flash nuke a pico1 with adalogger picowbell attached, install beta2.
notice the drive appear, and two com ports in device manager.
connect to both, one gives you the messages saying setup secrets and reboot.

Edit secrets and reboot.
Never see the drive again, and messages don't appear in the serial, although both ports do reappear and reconnect.

[tyeth]

Video of behaviour, use vlc to play
2025-04-28 21-23-11.mkv.zip

[tyeth]

two serial ports may be expected, there is a usb_cdc begin call here: https://github.com/adafruit/Adafruit_Wippersnapper_Arduino/blob/offline-mode/src/provisioning/tinyusb/Wippersnapper_FS.cpp#L111

[tyeth]

Official flash nuke has issues clearing the user data partition, see raspberrypi/pico-examples#642

[brentru]

@tyeth Okay, I am attempting to replicate this but falling flat.

notice the drive appear, and two com ports in device manager.

Yep, Windows shows both COM ports (CDC and MSC). WipperSnapper has code to prevent macOS from indexing both. One is the CDC port, showing the serial data, and one is the MSC port which does not show the data.

Edit secrets and reboot.

When I edit secrets and click save, Windows Explorer shows the reset but it doesn't unmount the drive. It does update the contents of the secrets file.

I think this is expected behavior caused by refreshMassStorage(). It's more "visible" on Windows as it sends a notification to the OS

Edit secrets and reboot. Never see the drive again, and messages don't appear in the serial, although both ports do reappear and reconnect.

I'm unable to replicate this part of your bug. I can see the messages in the serial and the ports.

Here's how it looks for me:

• Edit secrets
• Save
• Press RESET button on device
• I see the drive connect and then reset (this is OK - it's writing the config file to the FS)
• Connect to COM4/CDC port
• I still see the drive
• Wait 30s
• Data from the BME280 is sent to the CDC serial

Did you have a sensor attached to your Pico when testing?
I am using a Feather RP2040 Adalogger and a BME280.

[tyeth]

and one is the MSC port which does not show the data.

I don't believe that can't be the case, they are serial ports not Mass Storage Class devices, those are different device classes, and drivers, and sections in Device Manager.

I think this is expected behavior caused by refreshMassStorage(). It's more "visible" on Windows as it sends a notification to the OS

Agreed, but it's very undesirable, as a single interaction with the device causes it to disappear (and potentially close the active application window - very frustrating).
VsCode is tolerant of disconnects, but most editors are not.
I wonder if there is a way to delay that refresh, or avoid entirely sometimes, based on when we have actually changed the disk from the MCU rather than always refreshing based on host "changes" (even a file read by vscode can cause the reconnect - it's probably altering metadata like last accessed timestamp).
I get the impression only our auto-config changes should require refreshing the hosts view of the disk, and only at boot.

Edit secrets and reboot. Never see the drive again, and messages don't appear in the serial, although both ports do reappear and reconnect.

I'm unable to replicate this part of your bug. I can see the messages in the serial and the ports.

Fair enough, probably fixed by your secrets change, otherwise I'm curious why we (a user and I) see different behaviour to yourself.
I did have sensors connected, also trying without in theory.

On a related note I've tried the two PRs with a Metro S3 (SD attached) and see no serial output with either. I tested a tinyusb example and do see serial (maybe a different class of device as normal serial not CDC). I wonder if my serial port driver is different and may need switching.

[brentru]

I wonder if there is a way to delay that refresh, or avoid entirely sometimes, based on when we have actually changed the disk from the MCU rather than always refreshing based on host "changes" (even a file read by vscode can cause the reconnect - it's probably altering metadata like last accessed timestamp).

We could remove the new call to refreshMassStorage()

On a related note I've tried the two PRs with a Metro S3 (SD attached) and see no serial output with either. I tested a tinyusb example and do see serial (maybe a different class of device as normal serial not CDC). I wonder if my serial port driver is different and may need switching.

Could you try on a different computer?

[tyeth]

closed by #748