Add unit tests for cansdo module

[Copilot]

No tests existed for the CanSdo SDO protocol implementation. This adds 26 test cases covering the full request/reply surface.

New files

• test/test_cansdo.cpp — test suite using the same UnitTest/CanStub pattern as test_canmap.cpp
• test/test-include/errormessage_prj.h — minimal project header (ERROR_MESSAGE_LIST empty, ERROR_BUF_SIZE=10) required to compile errormessage.cpp in the host test environment

Makefile

Added test_cansdo.o cansdo.o errormessage.o printf.o to OBJS.

Test coverage

Area	Tests
Parameter read/write via index 0x2000	value round-trip, out-of-range → SDO_ERR_RANGE, invalid index → SDO_ERR_INVIDX
Parameter access by UID (0x21xx)	read/write, unknown UID → abort
CAN map add TX/RX (0x3000/0x3001)	full 3-step flow; invalid COB-ID and unknown UID error paths
CAN map read (0x3100/0x3180)	COB-ID, signal position, gain/offset; empty map → abort
CAN map delete (0x3100)	delete entry, confirm removal
Error objects (0x5003/0x5004)	read returns reply; write → abort
User-space SDO	unknown index populates pending frame; SendSdoReply clears it
Node ID	wrong ID silently ignored; SetNodeId activates new ID
String upload (0x5001)	GetPrintRequest() set on read

Include ordering workaround

The libopeninv printf.h declares printf/sprintf without extern "C", which conflicts with system stdio.h (pulled in via <iostream>) on modern host compilers. The test file resolves this by including <cstdio> and forward-declaring IPutChar before cansdo.h, then setting PRINTF_H_INCLUDED to prevent the problematic header from being re-processed:

#include <cstdio>
class IPutChar { public: virtual void PutChar(char c) = 0; };
#define PRINTF_H_INCLUDED

#include "cansdo.h"
// ... remaining includes

RX map verification

Because the test stub stores only the last-registered CanCallback (CanSdo), delivering a raw CAN frame via canStub->HandleRx would not invoke CanMap's receive path. RX mapping acceptance is therefore verified with canMap->FindMap() instead.

[jsphuebner]

@davefiddes What do you think about these?

[davefiddes]

Not bad. I've not found AI generated tests up till now very good. It usually goes off and generates fragile nonsense. These look OK. Not massively deep but a solid basis for any future regressions.

[jsphuebner]

Alright, thanks for checking!