Model and test frame handling

[Jookia]

Right now we have a fairly incomplete picture of how the frame handling works. Writing a simulation in Python and then a tester in C that can run on the printer and validate assumptions is a good start.

The current state machine model I'm going to assume the ACE has is:

• It discards bytes until seeing 0xFF
• It reads 0xAA
• It reads the length
• It reads length bytes and calculates the CRC
• It reads the CRC and checks it
• It reads 0xFE

If any of these fail, it will restart the entire process. The following tests should help validate this behaviour:

• Writing a stream of 0x00 through 0xFE should do nothing at all.
• Writing 0xFF then any byte except 0xAA should do nothing at all.
• Writing a 0x0000 length should have an empty payload.
• Writing a 0xFFFF length should have a full payload.
• Embedding 0xFF or 0xFE in the payload should get a response.
• Writing an invalid CRC should do nothing at all.
• Writing a non-0xFE end should do nothing at all.
• Writing a valid packet should get a response.

The test would work by sending a valid packet, then a bad packet, then a valid packet. The test succeeds if the bad packet does not influence the second response.

[wiltwong]

Probably easier to compile a serial to net proxy for the Kobra 3 and use socat on a remote computer to do this kinda testing before going to make a static binary. I'll try to compile and get binaries for https://github.com/cminyard/ser2net or something equivalent this week.

[Jookia]

After thinking through #8 , I'm genuinely curious if the input buffer is capped at 64 bytes (one CDC packet). It would be a good idea to test larger packets.