Disallow SOF in messages to handle mid-message death about tinyframe HOT 6 OPEN

It's a good idea, I'm only concerned about back compat. If you add it, please make it configurable, so it can be switched off.

from tinyframe.

@grantramsay Sounds reasonable to me, this resyncing will work. I'd gladly merge that.

(I don't currently use TF in any active project, so I don't see implementing this myself soon..)

from tinyframe.

That would actually be quite an elegant solution, I might try implementing that

from tinyframe.

I might be missing something, but wouldn't a simple back compatible way be to:

• buffer entire packet when receiving, need to increase buffer size to handle full packet (header + payload + checksum/crc)
• if checksum/crc fails or len > TF_MAX_PAYLOAD, memmove buffer to next SOF (if present) and reset rx index

This way the state machine will resynchronise on the next valid checksum/crc with no change to the message structure

Edit:
Also TF_ResetParser would reset the rx index, so reliable links without checksum/crc (e.g. TCP) could call it to resync if connection is restarted.
Additionally a timeout could be added to move to the next SOF. For the case where the link fails with a large but not invalid message length, and the next message is short. The state machine would need to receive extra rx data to resync, which may not be timely enough for the application

from tinyframe.

@mon - I had a little go at this the other day.
It got a little more complicated than I'd like, but there might be something helpful here (definitely a "work in progress"):
master...grantramsay:wip/feat-16/sof-resync

from tinyframe.

Hmmm, certainly a little more complex than I expected. I'll experiment a bit over the next few weeks and see how reliable it is.

from tinyframe.