Wrong definition of COAP_MAX_TRANSMIT_WAIT

[vuzunov]

The definition of COAP_MAX_TRANSMIT_WAIT in er-coap.h is wrong.
The caret is a binary XOR and not an operator for power calculation. Probably a pow() function shall be used instead.

[jvermillard]

in the CoAP RFC:
ACK_TIMEOUT * ((2 ** (MAX_RETRANSMIT + 1)) - 1) * ACK_RANDOM_FACTOR

in the wakaama code:

#define COAP_MAX_TRANSMIT_WAIT ((COAP_RESPONSE_TIMEOUT * ((2 ^ (COAP_MAX_RETRANSMIT + 1)) - 1) * COAP_ACK_RANDOM_FACTOR))

** was translated into ^ (XOR in C)
I suppose ** mean powerof

[vuzunov]

Indeed, it is 2 to the power of (MAX_RETRANSMIT + 1).

It doesn't really have much of an impact on the correct behavior, since it is only used to guard against registration failure due to datagram loss.

[jvermillard]

Can you review the fix here: #222

[vuzunov]

Thats not quite correct. It has to be something like

#define COAP_MAX_TRANSMIT_WAIT ((COAP_RESPONSE_TIMEOUT * ( pow(2, (COAP_MAX_RETRANSMIT + 1) ) - 1) * COAP_ACK_RANDOM_FACTOR))

And then of course the function pow needs the math.h header included.

[jvermillard]

I did it this way to avoid a dependency on math.h
using (COAP_MAX_RETRANSMIT + 1) * (COAP_MAX_RETRANSMIT + 1) in place of pow(2, (COAP_MAX_RETRANSMIT + 1))

[vuzunov]

That's what I thought too, but its not the same.
(COAP_MAX_RETRANSMIT + 1) * (COAP_MAX_RETRANSMIT + 1) is the same as (COAP_MAX_RETRANSMIT + 1) to the power of 2 and not 2 to the power of (COAP_MAX_RETRANSMIT + 1)

Alternatively a small pow function implementation might be added.

[jvermillard]

try 2 using bit shifting

[vuzunov]

That is definitely the more elegant solution 👍

[dnav]

Nice catch @vuzunov. How could I do that ?..
Thanks for the fix @jvermillard

[dnav]

Fixed by #222