> > > usb_submit_urb should really have well defined semantics for failures: > > > > > > 1. the completion handler is _not called_ and the result of > > > usb_submit_urb indicates _failure_ (the way uhci handles it now), > > > ... > > > 3. the completion handler is _called_ and the result of usb_submit_urb > > > indicates _success_ (simulating the asynchronous case). > > But should errors in root hub operations be indicated by the result of > usb_submit_urb (completion handler not called) or by the completion > handler (usb_submit_urb returns 0)? I would prefer 3 over 1 as it would > completely hide the fact that the root hub is really a special case of > hub.
I see no reason to define that behavior in an interface spec. Callers have to handle the faults they see reported, even the less common failure modes. What's important is that the faults be reported ONCE through clearly defined channels (1 & 3 above, vs the deleted #2 which as I recall reported them twice). - Dave _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
