Le dimanche, 8 juillet 2012 17.03:40, Cyril Brulebois a écrit : > doesn't look insane to me, but I'd appreciate a tiny briefing on the > uPr=WCDMA bit: Care to explain why that fixes a bug? (I suspect it makes > sure some extra properties are checked, instead of just relying on > vendorid+productid.) Your guessing is correct. The more detailed answer can be found in ./usb_modeswitch.tcl from src:usb-modeswitch from line 171 on. The process is the following: 1) ConfigGet conflist: get an ordered list of available configuration files matching the plugged in vendorId:productId pair (from both the tarball shipped in usb-modeswitch-data and from /etc/usb_modeswitch.d/ ). The pair is given as argument to the tcl program from the udev rule. 2) MatchDevice: matches the actual device against a set of specifiers: set match(sVe) scsi(vendor) set match(sMo) scsi(model) set match(sRe) scsi(rev) set match(uMa) usb(manufacturer) set match(uPr) usb(product) set match(uSe) usb(serial) So in that case, the device was previously thought to be uniquely identified by its 19d2:0083 vendorID:productID pair but it turned out that there is another device, not needing modeswitching, that has the same vendorID:productID pair. By specifying the usb "Product" (uPr) attribute (WCDMA), then only the device that needs modeswitching gets the red pill. (This is because most modem-dongles vendors are on crack and use identical ProductId's for different products. Don't go hunt for sanity there.) > Bonus points if you can include that explanation in your changelog, but > anyway: please upload to unstable. Will do shortly, with an enhanced changelog of course. OdyX
Description: This is a digitally signed message part.