bug: NINAFW adaptors broken when acting as peripheral

[deadprogram]

Since commit 457af75 is appears that the NINAFW adaptors are broken. They advertise without a problem, however on connecting to them they do not correctly respond. Instead they get into a death spiral of bad messages probably due to losing their place.

[jsoriano]

Hey @deadprogram,

I was trying to use this library with an arduino-nano33 (Arduino Nano 33 IoT), and adapter.Enable() already fails. I can confirm the issue starts happening on 457af75.

This line looks suspicious to me:
457af75#diff-1acd23db7ad68fc2a873aff592c71288609ca6c78895f3c3ac3249791a1e2ae8R19

c := sz + 4 - (sz % 4)

When sz is 4, c will be 8, is this intended? Could it be reading parts of the next event when sz is already a multiple of 4?

I have tried to modify this line so it keeps multiples of 4 as they are, and it seems to work better.

c := sz + (4 - (sz % 4)) % 4

What is the intention of this line? To get the closest multiple of 4, or the next one?

Would it make sense to revert to the previous code based on ReadByte()?

[deadprogram]

Hi @jsoriano thanks for the correction, I added it to PR #304 just now.