sendDone error codes

[Badisso]

Hi All,
It seems there is a mismatch in the error codes returned by the sendDone event (at least for TOSSIM). Thus, the AMSend and Send interfaces specify that sendDone returns FAIL if a transmission has failed (https://github.com/tinyos/tinyos-main/blob/master/tos/interfaces/AMSend.nc#L104). However, TossimPacketModelC.nc for example returns an EBUSY code for such a case (https://github.com/tinyos/tinyos-main/blob/master/tos/lib/tossim/TossimPacketModelC.nc#L245)
I am not sure if it is the case with other platforms, but a quick look to the CC2420TransmitP.nc (https://github.com/tinyos/tinyos-main/blob/master/tos/chips/cc2420/transmit/CC2420TransmitP.nc) seems to not return a FAIL code.
Please share your opinions on the appropriate fix

[gnawali]

It is a bug in TOSSIM. EBUSY should be returned only when you call send when there is an outstanding send call, i.e., send was successful but sendDone is not signaled yet.

[Badisso]

Yes, as you mentioned EBUSY is returned by the send command. The sendDone as documented in the AMSend and Send interfaces returns SUCCESS, FAIL or ECANCEL. In TOSSIM there might be a possibility for EOFF (https://github.com/tinyos/tinyos-main/blob/master/tos/lib/tossim/TossimPacketModelC.nc#L285). In the CC2420TransmitP it seems another code ERETRY can go up to both Send.sendDone and AMSend.sendDone (https://github.com/tinyos/tinyos-main/blob/master/tos/chips/cc2420/transmit/CC2420TransmitP.nc#L531)

[gnawali]

I think we should update the documentation because EOFF and ERETRY are possible return values.

[Badisso]

I agree about EOFF, but I think ERETRY should not be returned to upper layers which are more interested to know whether the transmission resulted in a SUCCESS or FAIL at the end.

[gnawali]

many other places in the code use ERETRY so the question is in the radio stack -- what is the highest level to which it is appropriate to send ERETRY.