Enable unusedFunction checks

[dzid26]

Find unusedFunctions for Panda program. Allows to:

• treat F4 and H7 variants as a part of same program
• differentiate between functions used in Panda and e.g. PandaJungle

Links: #1954 #1794 #1878

[dzid26]

• I enabled --project compile_commands.json for Misra and unusedFunctions now.
  • it uses scons, so it can check macros we are interested in.
  • for CI I think we may run twice. With and without RELEASE env
  • it checks all main() files, so it found 104 violations (to be fixed)
• I still kept per build checks for unusedFunctions but only for Panda and bootstubs builds - I think this best practise and I think this similar will be need to be achieved for misra 8.x rules.

[dzid26]

@adeebshihadeh is SPI ever used for Panda F4 besides the CI test?
I removed the stm32f4/llspi.h header since normal scons job doesn't use it and mutation test was failing sometimes.

I can:

• remove building with ENABLE_SPI from CI test
• or add ENABLE_SPI in misra tests

[adeebshihadeh]

@adeebshihadeh is SPI ever used for Panda F4 besides the CI test? I removed the stm32f4/llspi.h header since normal scons job doesn't use it and so mutation test was failing sometimes.

I can:

• remove building with ENABLE_SPI from CI test
• or add ENABLE_SPI in misra tests

Let's add it to the tests. We don't use it yet, but might want to in the future.

[dzid26]

I noticed that cppcheck doesn't report system level misra violation when used with --project= compile_commands.json argument.
I will see if I can fix it, because otherwise, it is the most robust way to call whole program analysis.

[dzid26]

--project=compile_commands.json is definitely the way to go in the future, but I needed to give up on it until cppcheck bugs (resulting in false negatives on system level misra checks) get fixed.

In the meantime, I figured out a way to check different macro configurations together using --force and force-included dummy/helper header which specifies possible combinations for F4 & H7. This should make checking for unused macros (rule 2.5) fairly straightforward as well.

[adeebshihadeh]

why did all of this get removed?

[dzid26]

The diff is a bit weird - the second variant used to start at line 440 and now is removed.
Basically now, instead of checking F4 and H7 separately, they are checked together using cppcheck configuration combination capability (which is enforced with... --force argument ).

[dzid26]

Converting to draft.
It seems that Cppcheck bails out if it encounters [unknownMacro] violation. But since I had --suppress=unknownMacro it wouldn't report any remaining violations. https://sourceforge.net/p/cppcheck/discussion/general/thread/769381a303/#f0dc

We need to add --safety flag to avoid quiet bail-outs like this. danmar/cppcheck#5777 (comment)

Underlying issue is related to how --force option works:
https://sourceforge.net/p/cppcheck/discussion/general/thread/5c740643e6/?limit=25#7278