On Wednesday 12 September 2007 05:12, Graeme Gill wrote: > Your patch is fine as a "proof of concept", but I really think your > time would be better spent figuring out why Libusb isn't working > with the HCFR on your system. I'm not sure I can be much help here, > since it seems to work fine on my system. > > Problems with your patch as it stands are: > > There's no way to select the instrument. So if something else > was plugged into the system and occupied /dev/ttyACM0, or > if there were two HCFR's plugged in, you wouldn't be able > to use it/them both. > > It breaks HCFR support on MSWindows and OS X. MSWindows doesn't > have /dev/ttyACM0, and on OS X the IO manager interface needs > to be used to locate a device name. > > It breaks the system dependent encapsulation, by putting > system dependent code into the instrument drivers. Currently > this is almost entirely concealed behind the icom object. > > To do this sort of thing properly would entail adding a third > communication type to icom, a "USB connected pseudo-serial device", > and on each system type the mechanism to locate the serial interface > from the USB identification would be needed, and then the appropriate > read/write routines would be needed in icom to use that interface. > > Since there are currently no other instruments that appear as USB > communication devices, and the HCFR is at least capable of > working directly via libusb as it currently stands, I don't > think this is the right direction to go, unless there is some > truly insurmountable obstacle. Thanks for you comments. My goal is not to add such support of HCFR via /dev/ttyACM0 in officiel branch of ArgyllCMS, but just to make it work for me (of for other people having the same problems), under Linux. As I just ran dispcal for tests, I would just like to know if my patch works fine, or if I broke something, not in the design, but in the calibration process. BTW, as I said, I had similar problems with a scanner, under sane. It was taking a long time to setup the scanner, with the same strace errors (ressource not available), and then, once initialised, all was working fine. So, I wonder if I put a longer timeout during initilisation phase... I'll make a test. Thanks, -- Frédéric http://www.gbiloba.org