> you beat me to it Yes, we're obviously working in parallel.
I've also spotted that loaddevstr is very wrong. First it reads into a 128-byte buffer, when the spec says strings are up to 256 bytes (including type and length codes). Then it ignores the length code at the beginning of the string and uses the number of bytes returned by the Rgetdesc, which would explain why usb/probe often shows garbage trailing characters in vendor and device identifiers.
