fix: handle missing USB string descriptors in linux_list_serial_ports

[StefCoene]

Problem

USB devices are not required to provide manufacturer, product, or serial number string descriptors (indicated by index 0 in the USB device descriptor). When these descriptors are absent, the corresponding sysfs files (serial, manufacturer, product) do not exist under /sys/bus/usb/devices/<device>/.

Previously, these fields were read directly with .read_text() inside the SerialPortInfo(...) constructor call, outside the try/except OSError block. A missing sysfs file raises FileNotFoundError (a subclass of OSError), which was caught by the outer except OSError: continue, causing the entire device to be silently skipped.

Affected device

Velleman VMB1USB (VID: 10cf, PID: 0b1b) — a CDC-ACM USB serial interface without a USB serial number descriptor. The device appears correctly in /dev/serial/by-id/, /sys/class/tty/, and is detected by pyserial, but was invisible to serialx.list_serial_ports().

This caused the device to not appear in the Home Assistant Velbus integration USB port selector, which uses serialx via homeassistant.components.usb.async_scan_serial_ports().

Fix

Add a _read_optional_sysfs() helper that returns None on OSError instead of raising. Use it for the three optional USB string descriptor fields (serial_number, manufacturer, product) in both the usb-serial and usb (CDC-ACM) subsystem branches.

Mandatory fields (idVendor, idProduct, bcdDevice, bInterfaceNumber) remain in the existing try/except block that handles devices disappearing during iteration.

This mirrors the existing pattern already used for interface_description, which was already handled with an explicit .exists() check.

[puddly]

Thanks!

Could you run umockdev --all and add the output to a file tests/data/linux? It snapshots your system state in a way that can be "replayed" for unit tests. I don't believe the file contains any sensitive info but double check to be sure.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.52%. Comparing base (b068974) to head (e530e3a).

Additional details and impacted files

@@            Coverage Diff             @@
##              dev      #95      +/-   ##
==========================================
+ Coverage   91.46%   91.52%   +0.05%     
==========================================
  Files          22       22              
  Lines        3599     3610      +11     
==========================================
+ Hits         3292     3304      +12     
+ Misses        307      306       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[StefCoene]

Could you run umockdev --all and add the output to a file tests/data/linux? It snapshots your system state in a way that can be "replayed" for unit tests. I don't believe the file contains any sensitive info but double check to be sure.

If I understand correctly, umockdev should be run on HA?
That's a bit tricky since I don't have ssh access, only console access.

Stef

[StefCoene]

Could you run umockdev --all and add the output to a file tests/data/linux? It snapshots your system state in a way that can be "replayed" for unit tests. I don't believe the file contains any sensitive info but double check to be sure.

If I understand correctly, umockdev should be run on HA? That's a bit tricky since I don't have ssh access, only console access.
Update to my own reply, I can attach the device to an other VM (HA is running as VM on QNAP and execute the umockdev on that VM.
I will do that this evening and post the result.

Stef

[puddly]

Ah, I didn't realize this was with Home Assistant OS. Ignore me 😄