On Mon, Apr 30, 2012 at 08:51:28PM +0200, Benjamin Tissoires wrote:

On Mon, Apr 30, 2012 at 20:30, Henrik Rydberg [off-list ref] wrote:

quoted

quoted

I remarked a small regression in this patch: the detection of the
serial protocol is not handled anymore. I was indeed relying on the
fact that the parameter "id" in mt_probe was null to know that the
device was not already in the list of known devices.

I see. Will this suffice?

Yes, but I'd rather do the test against either id->product or
id->driver_data as we may use HID_ANY_ID for vendors specific
handling. For instance, I had the confirmation from EETI (eGalax) that
they will always use the same protocol for their usb products, thus
the same quirks could be handled by HID_DEVICE(BUS_USB,
HID_GROUP_MULTITOUCH, USB_VENDOR_ID_DWAV, HID_ANY_ID).

Sounds good. I am sending out a fourth version of the patch which will
enable this possibility as well.

On a different note, I wonder why the various slot allocations, sysfs
entries and post-parse settings are done after the device has been
started?

Thanks for testing,
Henrik